Every AI agent entering a customer environment becomes part of its digital workforce. It can hold credentials and make decisions, yet its identity may never appear in the processes used to govern employees.

That gap matters because agents are moving from controlled pilots into everyday operations. They are connecting to collaboration platforms and business-critical systems where they can send messages and alter records at machine speed. The permissions behind those actions often receive less scrutiny than those given to a new employee.

For Managed Service Providers (MSPs) and solution providers, this changes the customer conversation. Partners need to know what each agent can reach and who remains accountable for its actions.

Many customers will struggle to answer those questions because agents may have appeared through SaaS features or experiments run outside IT. This blurs ownership as a pilot grows, while its original access follows it into production. By the time an auditor starts asking questions, the business may depend on that workflow.

Security researchers recently demonstrated how GitHub Agentic Workflows could be manipulated into retrieving information from private repositories and publishing it in public comments. A routine issue became a route between untrusted input and protected information because the agent could access both.

The case should focus partners’ attention on the architecture around an agent. Every connected system can widen its reach and increase the potential impact if it is compromised or behaves unexpectedly.

Temporary access has a habit of becoming permanent

The route to over-privilege often begins with a reasonable decision. For example, a team wants to prove that an agent can deliver value, so it grants access through an existing service account or API key.

Problems emerge when that access outlives the assumptions behind it. The pilot enters production, yet its original token stays active. New tasks bring broader permissions, while ownership becomes harder to trace.

Traditional joiner, mover, and leaver processes offer little help. Employees have managers and defined roles, but agents may be created inside a SaaS platform, inherit their creator’s permissions, and remain active after that person changes position. Agents need lifecycle events of their own.

OWASP’s guidance for agentic applications identifies identity and privilege abuse as a major threat, covering the misuse of delegated permissions and inherited access. Autonomy gives these long-standing identity weaknesses a larger blast radius.

Our research, surveying 800 IT leaders across the UK and US, found that non-human identities outnumber employees in 83% of organizations. Only 21% have controls in place to govern them. Human oversight is also becoming less common, with a quarter of organizations requiring human review before high-risk AI actions, compared with 40% six months earlier.

These findings should prompt partners to reconsider where an AI project ends. Installing an agent marks the beginning of an access-management responsibility that may last for years.

Give every agent an owner and an expiry date

Partners can start by helping customers build an inventory of the agents operating across their environments. This should include the credentials each one uses and the systems within its reach. Every production agent needs a named business owner, with its purpose recorded in language the customer can understand.

A distinct identity should replace borrowed or shared credentials where the platform supports one. Permissions can then reflect the agent’s job instead of copying the access held by the employee who deployed it.

Least privilege also needs to account for actions. An invoice-processing agent may need to read documents and update selected fields. Changing supplier bank details belongs behind an approval gate. The same principle applies wherever an agent can make an irreversible decision or expose sensitive information.

Partners should plan for the moment when an agent moves outside its intended role. Customers need a tested way to suspend its identity and revoke access. Searching across disconnected systems during an incident is an expensive way to discover that no kill switch exists.

Access should also have a shelf life. Permissions that made sense during deployment may become redundant as workflows change. Reviews tied to the agent’s risk level can remove dormant access before it fades into the background.

In practice, this means helping customers discover their agents, register them with named owners, manage their access and govern their actions over time.

The channel can make agent governance operational

Customers are adopting agents faster than many internal teams can catalogue them. Partners already understand their systems and identity environments, putting them in a strong position to close that visibility gap.

MSPs can add non-human identity discovery to security assessments and then build access reviews into managed services. Solution providers can design role-based policies before an agent reaches production. Managed Security Service Providers (MSSPs) can bring agent activity into monitoring and response processes.

Partners should apply the same discipline to their own operations. An MSP’s internal agent may connect to tools with privileged access across customer accounts. Its permissions deserve as much scrutiny as those of an engineer with administrative rights.

Ultimately, AI adoption will continue because customers can see the productivity gains. Sustaining those gains depends on whether they can trust the access behind them.

Partners that make ownership visible and keep permissions aligned with purpose will offer something more valuable than an AI deployment, and that’s the confidence to put automation into production without surrendering control.

To many enterprises, AI costs – from compute to tokens, licenses, integration, and data – are now a familiar part of company expenditures, as we approach four years since OpenAI released ChatGPT, kickstarting the explosion of generative AI.

Just below the surface, however, is a less visible layer of costs: approvals, testing, audit trails, human review, and documentation all take time and resources, while weak or late governance can lead to rework, additional tooling, and delays to production.

The rise of capable AI models and, in recent times, AI agents that can carry out increasingly complex tasks with less human intervention means that, as pilots move into production, AI governance is now an extra workload spanning IT, security, legal, risk, and data teams.

AI governance is not the highest cost for companies – tokens and licenses are likely to rank higher – but it is an increasingly important part of AI’s total cost of ownership (TCO), and is naturally spread across a variety of teams and processes.

To make sense of the hidden costs associated with the latest AI wave, ITPro spoke to Enza Iannopollo, VP and principal analyst at Forrester, and David Menninger, executive director of software research and distinguished analyst at ISG.

The two analysts share a similar argument: governance is necessary, but the actual costs can vary significantly depending on how and when it is introduced. Controls that are added too late can mean months of additional work, while manual governance can itself become a bottleneck.

How governance gets expensive

As is so often the case in both life and in business, foresight is key, and many of the biggest AI governance costs – monetary, staff time, and otherwise – stem from adding controls too late in the process.

“Unfortunately, in many enterprises, the biggest cost has been retrofitting governance into AI deployments,” says ISG’s Menninger.

“When AI governance is not built into the project from the start, it can add months to the project rollout.”

“Our research of 1,200 AI use cases shows that about one third have made it into production,” Menninger continues. “The good news is that’s double what it was a year ago. The bad news is that two-thirds are still not in production.”

That gap between experimentation and production is where hidden costs can start to pile up.

“The same study indicates that, on average, it will cost twice as much as has been spent already and will take more than a year to complete,” he says.

“These figures suggest that much [of the] cost is not anticipated and that enterprises are learning as they go.”

Governance is only one part of that extra spend, but retrofitting it after the fact adds another speed bump just as companies are trying to move beyond the pilot stage.

Forrester’s Iannopollo argues that the answer is not so much to govern less, but to govern earlier: “Failing to scale an AI pilot is one of the most common costs organizations pay for not building the right governance from the start.”

“Relying on solid AI governance from the start means ensuring the appropriate controls and guardrails are in place,” she says.

“Pilots are the right moment to test them, adjust them, and prepare them to enable scale.”

Having a “human-in-the-loop” comes at a price

AI-driven automation is perhaps the hot trend in 2026, with companies increasingly experimenting with ‘swarms’ of powerful AI agents able to carry out work that until recently would have required human input – but people are still heavily involved in checking the eventual outputs.

In other words, automation does not necessarily remove labour; in many cases, it simply shifts where that labour sits.

“ISG Market Lens research indicates that 90% of AI-enabled work still involves humans, either with AI providing assistance or with humans reviewing and approving the outputs,” says Menninger. The bill for supporting all this human oversight naturally lands on people and the manual processes they perform.

The irony is that AI is often best at automating the most mundane tasks, leaving the hardest work for humans. “The manual review process is typically only needed for the most complex scenarios, so it’s no surprise that those would take more time and effort,” he says.

In effect, the easier work can disappear while the expensive judgement calls remain.

“Human review thresholds seem to be one of the most successful techniques for managing cost and risk,” says Menninger.

“It enables enterprises to focus their resources on the areas with the most exposure.”

The aim, then, is to concentrate their attention where the potential consequences are greatest, not to remove humans from oversight entirely.

“As AI-enabled processes become more accurate and more refined, the percentage of AI-related decisions requiring review will decrease, but our research suggests it will take time. By year-end 2027, only 13% of AI-enabled work will be fully autonomous.”

Iannopollo likewise says human oversight must be targeted: “Humans must focus on reviewing and approving high-risk use-cases, responding timely in situations where AI performance and/or behavior drift, and acting with full authority to ensure the business keeps running, even if they have to ‘kill’ an LLM, an agent, or an entire agentic workflow.”

Manual governance can be a bottleneck

Operating at a small scale, ideas like central boards or manual checks can work well; the problems emerge when companies try to scale them across the organisation. A model that works for, say, ten AI projects can quickly break at 100.

“AI governance that relies only on manual checks and processes becomes a bottleneck,” says Iannopollo.

“Governing AI requires [the] implementation of technical controls, automation, and careful distribution of human-led activities.”

It’s at this point that AI governance moves beyond policy documents and (human) committees and becomes something operational and technical – embedded in the systems used to manage approvals, enforce data and access controls, maintain auditability, and monitor AI in production.

“Our research suggests that enterprises need to supplement the governance provided by AI software platforms and tools,” says Menninger

“For example, only one-third of more than two dozen vendors evaluated fully met our criteria for agent evaluation tools. Less than one-fifth met our criteria for A/B testing and less than half met our criteria for human-in-the-loop reviews.”

“As a result, we recommend that enterprises anticipate they will need to incorporate additional tools and processes to complete their governance over AI,” he says.

Dedicated governance tooling is already becoming a sizable market in its own right: Gartner forecasts spending on AI governance platforms will rise from $275 million in 2026 to $462 million in 2027.

“The other common misunderstanding is that governance always means a lot of processes, controls, and documentation,” says Iannopollo.

“This is certainly true for some use-cases, but not for most. Assessing risks of AI use-cases early means that the measures can be implemented in a proportionate manner.”

“Starting from low-risk use cases increases the likelihood of success, further strengthening the confidence of the organization in AI,” she says.

The scalable option, then, is a risk-tiered governance model that uses automation to reduce the routine burden while concentrating human oversight and heavier controls on higher-risk use cases.

Governance should be an enabler

The costs of AI governance – labour, delayed rollouts, more tooling, oversight – are now clear; the answer for companies is not to minimise governance at all costs.

“Too often [organizations] embrace the rhetoric about governance as a road blocker and an inhibitor of innovation,” says Iannopollo.

“The truth is that in most cases lack of governance is the road blocker.”

Just as companies price in the costs of compute, data, integration, licenses, and so on, governance should be priced into AI projects from the outset.

Making governance proportionate and repeatable is the key, and the later companies leave it in the process, the greater the risk of costly rework and delays. A repeatable, scalable governance model is the dream – one that can grow with AI adoption without requiring the same increase in manual oversight.

Nearly nine-in-ten (89%) organizations have scaled back or been forced to redesign AI plans due to token costs this year, according to research from Cloudera.

Findings from the firm’s State of Enterprise AI 2026 report show enterprises are still struggling with AI-related spending, despite a notable decrease in token costs over the last three months.

Brian Martin, CTO of Applied AI at Cloudera, said the findings highlight a “new reality for enterprise AI”, adding that “what works today may not make economic or operational sense tomorrow”.

“Costs change, models evolve, regulations shift, and the right place to run a workload changes with them,” he commented.

“Leaders need an AI strategy built for that reality, one that gives them the freedom to adapt without giving up control of their data, governance, or spend.”

Cost volatility spooks IT leaders

The study specifically highlighted cost volatility as a key challenge for enterprises, with nearly half (45%) having moved AI workloads as a result.

According to Cloudera, this shows that “AI’s price tag is starting to shape its path through the enterprise” – and that could spell further bad news in terms of ROI and adoption rates.

Last year’s report, for example, found cost was the leading barrier to AI implementation, and that trend has intensified across 2026 amidst surging token costs.

Figures from the LLM Token Expenditure Index show prices hit over $2 per million tokens in June this year. As of 25 October, the index shows token prices stand at $0.95 per million tokens.

That price decrease will offer little solace for IT leaders, however, many of whom have been battling costs throughout the first six months of the year. Cloudera said the cost of the technology has placed huge pressure on enterprises to “demonstrate returns and manage the ongoing cost of AI at scale”.

More than half (55%) of respondents said their AI investment increased, yet despite greater spending that is yet to translate into “clear returns”. Indeed, 58% of respondents the value generated by AI initiatives is roughly equal to what they cost.

Fewer than one-quarter (23%) said AI generates more value than it costs.

From a business impact perspective, the situation is roughly the same, according to Cloudera, with IT leaders reporting mixed returns on the technology.

“More than a quarter (27%) report little or no measurable impact from their AI initiatives, while 41% have encountered higher-than-expected costs,” the report states.

Cloudera added that for those investing heavily in the technology, its findings “raise questions about how quickly that spending will translate into meaningful returns”.

A year of AI costs

The study from Cloudera is the latest in a string of reports questioning the costs associated with AI over the last year - and there have been a number of high profile incidents involving lavish overspending.

Uber, for example, revealed it blew through its entire annual AI budget in just four months, prompting the ride hailing firm to introduce spending caps for staff.

As ITPro reported in late June, Accenture also ordered staff to stop using AI for basic tasks such as converting PDFs to slides amidst rising costs.

It’s worth noting that a key factor behind these incidents lay in the tokenmaxxing trend, rather than general costs associated with AI. This is a practice of maximizing AI use within the enterprise by incentivizing staff to use the technology.

Some companies introduced leader boards to track usage rates, while others used it as a means to measure productivity.

In response to growing costs, some enterprises have begun reviving time-tested FinOps practices, as ITPro reported in late July, enabling them to better track costs and cut down on spending.

Cloudera’s report also highlighted a sharpened focus on this front. Nearly all enterprises believe they now have “some visibility” into AI spending, which marks a move in the right direction.

“Ninety percent say they can measure end-to-end AI costs at least somewhat accurately,” the report notes. Cloudera added that there’s still room for improvement here, however.

“Only 27% say they can do so very accurately, suggesting there is a considerable gap between having a general understanding of AI costs and a precise picture of where that spending goes.”

FOLLOW US ON SOCIAL MEDIA

Anthropic is set to go public in late November at a valuation of between $1.5 trillion and $2 trillion. The higher end of the range would make it the biggest initial public offering (IPO) ever, beating SpaceX’s $1.8 trillion public debut in June. The Claude creator filed its much-awaited IPO prospectus on Monday 26 October, giving investors and tech leaders a peek into its financials, especially its dependencies.

The document, also known as an S-1, highlights a reliance on a limited number of customers for its revenue. Nearly half (47%) of sales in 2025 were routed through its cloud partners, Amazon and Alphabet’s Google. Meanwhile, each of its two biggest customers, who remained unnamed, accounted for 12% of revenue last year.

Some progress is being made to reduce this dependence. While 48% of revenue between January and September this year came through its cloud providers, the five biggest customers accounted for approximately 25% of revenue for the nine-month period. No single customer had a share greater than 10%.

The company noted in its filing: “We anticipate that we will continue to derive a significant portion of our revenue from a limited number of customers for the foreseeable future, and in some cases, the portion of our revenue attributable to certain customers may increase in the future.”

The IPO prospectus has raised questions about an overreliance on AI models and vendors – and the risks this brings for enterprises.

“It’s a reminder that the model, the vendor and the route you access it through can all change for reasons outside your control,” cautions Jan Karstens, CTO at AI-powered SAP automation platform Avantra. For example, enterprises that have built their critical business processes around one AI supplier could face operational disruption, even if it’s temporary.

“We’ve already seen what that looks like. When Anthropic was ordered to block foreign nationals from its top models this summer, the only way to comply in time was to switch them off for everyone. Teams that had already assumed any model or vendor could disappear overnight barely noticed. Everyone else had to scramble,” says Karstens.

Charles Radclyffe, CEO of SaaS enterprise provider EA Global AI adds: “The interesting thing about the reported Anthropic numbers is that they're not just a story about one AI company. They're a reminder that concentration risk runs through the AI ecosystem, and enterprises need to understand where those dependencies sit.”

Earlier this year, automation platform Zapier polled 500 US executives and found that 47% of enterprises would stop functioning properly if their primary AI vendor were to shut off its service tomorrow. Just over a quarter (27%) indicated they were reliant on AI vendors for nearly all of their business operations.

Question vendors’ own dependencies

Part of the problem is that enterprises focus too much on model performance, when they should be questioning their AI service’s supply chain.

“If a provider depends heavily on a small number of cloud platforms, strategic partners or major customers, changes in those relationships can ultimately affect enterprise users through pricing, availability, product direction, or service levels,” says Radclyffe.

So, what does he think CIOs should be doing? Asking companies supplying the AI services about their own critical dependencies, how workloads are delivered, the knock-on effect if key commercial relationships change, and what notice customers receive before major model changes. “AI procurement needs to become a resilience conversation,” he stresses.

Karstens believes asking model makers about their critical dependencies is reasonable, but he cautions against acting on their responses alone. He’d “be mildly nervous about concentration among my direct AI vendors. I’d answer that with my own visibility and multiple suppliers, not with questionnaires.”

CIOs need to “start from the assumption that any model, vendor or access path can be withdrawn without warning,” he adds. “They should build governance that doesn’t depend on any one of them staying in place. Periodic audits and vendor assurances after the fact aren’t enough when the [AI] market moves fast.”

Be wary about vendor lock-in

A key takeaway from Anthropic’s S-1 is that consumption-based, or per-token, revenue made up 83% of revenue in the first six months of 2026. However, many of its main customers are not locked into long-term contracts.

AI model makers are pushing hard for customers to agree to contracts to secure multi-year revenue visibility amid fierce competition. If enterprises find themselves signing contracts, then they need to ensure they agree to “meaningful exit rights, portability provisions, and the ability to export not just data but the prompts, workflows and evaluations that support business processes,” advises Radclyffe.

Karstens agrees. “Vendors like Anthropic are moving deeper into everyday work, with documents, slides and projects now living inside the AI tool itself. If that continues, data portability will become a real issue for CIOs in the next 12 to 24 months.”

The Zapier study also surveyed senior executives about AI vendor lock-in. Just over eight in 10 (81%) admitted they were at least slightly nervous about being locked into contracts, with 29% very concerned. Almost half (46%) were worried about data migration and an overreliance on a single vendor. Sudden price hikes were a concern for 41%, while another major worry for 37% was having a single point of failure.

Ideally, enterprises should avoid finding themselves in a situation where these worries come to fruition by ensuring they have a practical alternative to turn to if costs rise, there’s a change in circumstances, or unexpected events.

Radclyffe sums it up perfectly: “There's nothing unusual about [vendor] concentration in a fast-growing technology market. The risk comes when enterprises don't understand how dependent they've become until they're forced to make a change.”

Earlier this month, Japan’s government declared a cyber emergency. The move followed a series of cyber attacks against businesses and institutions across the country. Rail transport, internet services, retailers and even the police have been impacted, according to the Financial Times.

At one point, the Japanese financial services regulator even told banks not to accept driving licences as proof of identity for anyone opening new accounts.

The declaration by Japan’s NCHS cyber agency is unusual. Few countries have declared a cyber emergency, even if security services and regulators have warned of heightened threats.

The move came after rail operator JR East reported the loss of 1.67 million email addresses, and car rental company Times Car lost a large number of scans of its customers’ driving licences. In addition, cloud services operator IDC Frontier reported a ransomware attack. This, in turn, linked to the leak at JR East and other public sector bodies.

“Incidents like this highlight why supply chain attacks can be so disruptive,” Alexandra Rose, head of global affairs and policy at cybersecurity company Sophos told ITPro.

“An attacker doesn't need to compromise a bunch of organizations individually in order to have a widespread impact. They gain access to a single organization and then can move through interconnected systems or disrupt services that many other organizations depend on. We’ve seen the downstream consequences extending beyond the initial victim time and time again, affecting operations, customers, and more.”

Big in Japan?

Japan, though, is not generally viewed as more exposed to cyber attack than other advanced economies, experts say. But the country is certainly seeing an increase in the seriousness of attacks, if not their frequency. There is also evidence that some businesses, at least, were less well prepared.

“I think it’s a response to the increasing severity of attacks they are seeing,” Rob Demain, founder of security firm e2e-assure, told ITPro. “One of the key ones was the platform IDC Frontier that was taken offline. I think it’s a trend of more impactful cyber attacks being seen, and trying to move quicker and keep pace.”

The Japanese government also moved to bring in active cyber defense (ACD), where police and other organizations can go after attackers, taking down servers and, hopefully, neutralizing attacks. Japan is also considering wider changes to its cybersecurity strategy, after lawmakers recently called to strengthen its use of AI for cyber defense.

But publicly announcing a cyber emergency brings its own risks, not least that businesses may rely on government intervention instead of taking proactive measures themselves. And useful advice could be lost in increased noise.

“How many times can you cry ‘cyber emergency’ before people stop worrying about it?” asks Demain. “But I don’t think there has been anything similar in Japan. So I think if you hear that message, you are going to take notice… You need to make sure everyone’s taking this seriously, and that was just one method to get the message out.”

Not everyone is convinced, however.

“A cyber emergency declaration has value if it changes the response, not just the rhetoric,” cautions Nazir Esmail, chief autonomous enterprise officer at Sapiens, a business that develops software primarily for the insurance industry.

“CISOs are well versed in threat vectors and don’t need additional warnings that threats are on the rise. What would be useful to CISOs and the security community in general, is faster access to actionable intelligence. In addition, if attacks threaten critical services or infrastructure, a coordinated incident response effort and a clear escalation mechanism are essential to contain the impact and accelerate recovery,” he told ITPro.

Actions, not words

This poses the question of whether declaring an emergency has a practical value for enterprises, and the wider security community. If it brings additional assistance or funding, then CISOs would no doubt welcome it. And government positioning can help security leaders to promote security to the board.

As Charl van der Walt, head of security research at Orange Cyberdefense points out, a full state of emergency can unlock new resources. This, he recalls, happened in Costa Rica in 2022, when the country declared a state of national emergency following a series of ransomware attacks against public sector bodies. But Japan’s cyber emergency stops short of that.

“The Japanese story seems to me to be a bit of a knee jerk response to a series of incidents that happened online,” he told ITPro. “The latest one was an attack against an infrastructure as a service provider. They have about 500 businesses that depend on them, and so the impact on them cascaded through the whole ecosystem.”

Van der Walt likens the incident to a similar attack on a shared software platform in South Korea, which also impacted a large number of businesses. But, he adds, we see fewer public data leaks in Japan than the size of the economy would suggest. “One reason may be that Japanese businesses are just paying [the ransom] before there’s any leaks,” he suggests.

We also do not know what the Japanese government’s, or indeed any government’s, playbook is for a cyber emergency, and for good reason. Given cyber attacks are increasingly part of grey zone conflict or hybrid warfare, governments will want to keep potential adversaries guessing.

“There’s two facets to think about,” says van der Walt. “One is the signalling effect to the adversary, and that’s really an issue for diplomacy and international humanitarian law. How do you keep your adversaries in check? How do you keep them constrained? Signalling is a big part of that.”

The other is the impact on the wider population, and overall cyber awareness. CISOs stand to benefit, at least indirectly, if employees, customers and other stakeholders have a better understanding of cyber risk.

“You also need to signal to your populace. You need to maintain the health, the sense of confidence, and the trust that your economy and society have,” he says. “I don’t know Japan’s playbook and I don't know their rationale. But I would think this needs to be done very, very cautiously.”

Cisco has expanded available incentives in its 360 Partner Program, as well as making Cisco IQ generally available for partners as part of its mission to continue to deliver the support its channel ecosystem needs.

Details of the new offerings and updates were announced at the firm’s Partner Summit in Vancouver, Canada, today, but ITPro attended a pre-brief before the main event.

“Cisco didn’t wake up and realize that partners matter,” exclaimed Tim Coogan, senior vice president, global partner sales, told journalists during the pre-brief call last week.

That being said, Cisco understands that partners need to move “first and fast,” especially in the AI era, where threats have intensified, and traditional models of supporting customers are no longer fit for purpose.

Changes to the 360 program include dedicated paths for advisors, developers, and distributors, so they can be recognized for the valuable role they play. In addition, Cisco has introduced Partner Value Indexes (PVIs) for advisors and developers, with a dedicated Distributor PVI coming by Spring 2027.

Additional cross-sell bonuses were also unveiled in addition to a Cisco IQ adoption incentive.

Change is necessary for continued success

The 360 program was unveiled at the beginning of the calendar year and marked a big change; however, Coogan was keen to stress that the previous programs had evolved in response to partner feedback.

“We don’t do anything at Cisco without asking, ‘What about partners?. Are we building it the right way? Are we providing that platform?” Coogan said.

“When we built the [360] program, we didn’t build it and unveil it. We built it with our partners. Thousands of meetings, countless hours spent understanding the program.

“It was big news, and it has required our partners to do things differently. Part of the reason our previous programs, most notably Value Incentive Program (VIP), were around for 23 years. What some people may not know is that there were 46 versions of VIP. We are deeply committed to an iterative process over time.”

Coogan added: “It’s not a set it and forget it mentality. We believe that feedback is a constant source, and we have already made numerous changes. Change is difficult. We ask partners to invest in their capabilities to deliver this incredible portfolio, and we are as deeply committed to modifying it over time based on that same feedback.”

The B2B buying landscape is forever changed

Customers, on average, work with up to seven trusted partners, according to Coogan. What’s more, the B2B buying journey is non-linear with more than 28 touchpoints. This places even more emphasis on partners to help organizations navigate this new normal.

“We recognize that, at the end of the day, growth and profitability are fundamental to our company and to our partners as well. [Offering] Cisco IQ, not as a product, but as a platform, [means] our partners can use it to deliver new services revenue, new capabilities, and a different kind of buying and use experience for our clients. Cisco IQ will become a really, really powerful tool in the bag of our partners.”

Coogan also spoke to the importance of partner experience: “We believe incredibly deeply that our ability to save our partners time is just as important as our ability to help them drive growth and profitability. Finding an easier way to do business with Cisco, improving tools, and leveraging AI becomes fundamental to the experience our partners have with us,” he said.

“Our partners don’t owe us anything, and they don’t work for us; we earn the right to work with them every single day. Partner experience is one of the ways we can invest in their time with us by making that experience as positive and productive as possible.”

Coogan also stressed the importance of constant innovation, being able to react quickly, and scaling but without adversely affecting efficiency.

“It’s our 30th year of Partner Summit. This is not a revelation that partners are important. This is a 30-year journey with our partners, founded on trust, growth, agility, and innovation. We’re so very proud of our heritage and our legacy,” he added.

“Everything for us is founded around our principles of build, buy, and partnering… Ultimately, what we build and what we buy falls short in meeting the expectations of our customers if we don’t have the partner ecosystem to support it. We believe deeply that that partner ecosystem is the enabler for this build, buy, and partner strategy.

“This is a no-ego environment where we learn from each other, provide the platform and the foundation to deliver value in the field. And I think that's really what separates us.”

While much focus on AI has been about increasing machine intelligence levels, it should really be about how you can support your people to become better versions of themselves - more capable humans.

So said Oracle co-CEO Mike Sicilia during the opening keynote, ‘Where AI means business,' at Oracle AI World in Las Vegas today.

For AI to be successful, he said, there are four key ingredients:

1. Access to data, but securely and with the right permissions
2. The ability to understand that data in the context of your business and the systems therein
3. High-performance infrastructure that enables AI to operate efficiently, at scale, and cost-effectively
4. AI needs to work inside the applications and processes where decisions are made and actionable intelligence is derived

A foundation of innovation

“This is where I think Oracle has a unique advantage that has been decades in the making. We’ve built technology that manages your data, runs your business, and powers your most important workloads,” Sicilia added.

These are:

• AI database delivers AI directly to enterprise data that is trusted
• Cloud applications put business intelligence to work, efficiently and safely
• Data platform provides the context and security AI needs
• OCI provides the efficiency and performance needed to operate AI at scale

He said the company had put more information at people’s fingertips than ever before. But added that work has historically been organized around software rather than people and what they need.

“[AI] allows for far more intuitive software. In fact, AI should allow you to barely think about the software at all. So instead of navigating applications and managing processes, you simply tell the system what you’d like to accomplish,” he said.

“Because it understands your business, it can figure out how to get the work done. So instead of spending your day operating the software, the software is working and operating for you.”

Applying AI to build a bridge between humans and machines

The ability for AI to recognize, anticipate, and then help is what Oracle is striving for when it comes to AI: creating meaningful connections between people and technology.

“We’re freeing people from system burdens. We’re giving people more time to be human. That’s the ambition we have for AI. Not simply more capable machines, but more capable people. People who have more time to think, to create, to solve problems, and to care for one another,” Sicilia added.

"Of course, as these systems take on more responsibility, they must also earn more trust… More AI intelligence does not mean fewer controls. Trust is what allows us to put AI to work where the stakes are highest and, therefore, the benefits can be the greatest.”

As expected, a slew of related announcements were rolled out on the first day of Oracle AI World, including:

• A revamped and rebranded version of what was formerly Oracle Cloud Success Navigator, now dubbed Oracle AI Success Navigator. It marries specialist AI agents, best practices, guidance, and training to help businesses more quickly move from theory to real-world value from their Oracle Fusion application implementations.
• Deep Trust AI Architecture, acting as a blueprint that bakes in trust to help organizations speed up their path to AI innovation without increasing risk
• Oracle Fusion Companion, which delivers an intelligence capable of understanding business objectives to decide upon and orchestrate the relevant agents, applications, approvals, and workflows required to achieve said goals.
• Oracle Fusion Agentic Home offers workers a more personalized experience, powered by AI, to help delegate work to agents as appropriate and track progress.
• New and expanded capabilities for Oracle Autonomous AI Lakehouse, including AI Data Catalog, Business Semantic Layer, Data Lake Accelerator, Data Studio, in addition to centralized controls and security
• New Oracle Fusion agentic applications to serve specific use cases such as customer experience, HR, and supply chain operations

Sicilia challenged event attendees to bring Oracle the problems they’ve learned to live with as well as their ambitions. He asked them to come forward to talk about the things they couldn’t do and the things they want to do, but haven’t had the technology to support those efforts, until now.

“We want the complexity to move into the background, and the possibilities for people to move to the foreground,” he said.

Heads of infrastructure and IT operations need to focus their AI efforts carefully if they want to accelerate autonomous operations and support enterprise-scale AI initiatives, Gartner has warned.

In a survey of 272 heads of infrastructure and IT operations (I&O), the company found that 58% believe scaling proof of AI concepts to I&O adoption will be an important priority in the next 12 months.

And, said the firm, there are five AI trends worth adopting if firms want to improve operational efficiency with AI and optimize costs.

"AI is creating one of the most significant shifts the I&O function has ever faced," said Jonathan Forest, VP analyst at Gartner. "Rather than trying to address every aspect of AI simultaneously, heads of I&O should focus on a small number of strategic investments that deliver meaningful impact."

The five AI trends worth adopting for infrastructure and IT operations

The first, unsurprisingly, is agentic AI, used for tasks such as incident triage, change validation, and remediation. And while virtually no organizations are currently using AI agents with minimal human involvement in IT operations, Gartner expects half to be doing so by 2030.

"AI agents are driving a shift from reactive, human-in-the-loop operations to proactive, human-on-the-loop operations," said Forest. "Heads of I&O who fail to restructure their organization around agentic AI risk higher costs, poor operational agility and performance, and ultimately, irrelevance."

The move to agentic AI, though, is creating challenges in terms of token consumption – which is rocketing and likely to lead to unforeseen costs.

Through 2027, Gartner predicts that enterprises implementing AI agents in IT operations using uncapped variable pricing models tied to token usage will overrun their initial cost expectations by more than 50%, compared with those using predictable subscription and bundled pricing models.

The only answer, said the firm, is a strict "no telemetry, no deployment" policy to eliminate blind spots in AI spending and token usage.

Another major trend worth watching, said Gartner, is the way AI inference is driving the rise of hybrid AI infrastructure, driven mainly by privacy, cost, performance, and data sovereignty requirements.

Gartner predicted that a quarter of enterprises will have merged their network operations center (NOC) and security operations center (SOC) by 2030, up from just one in 20 now. And this, it said, will see total spend on tooling and staff resources falling by a quarter.

Finally, AI deployments across cloud, data center and edge environments are increasing the need for a coordinated approach to managing AI infrastructure, governance, security and operations. As a result, said Gartner, I&O is evolving from a traditional gatekeeper role to the orchestrator of the enterprise AI platform, enabling the business to deploy and scale AI reliably, securely and in a cost-effective way.

And, Gartner warned, over the next four years, organizations without a strong I&O function driving AI infrastructure design, architecture and support will be more than twice as likely to fail in their AI initiatives.

"The high cost of AI infrastructure leaves little room for error, placing I&O at the center of enterprise AI success," said Forest. "Heads of I&O must elevate their teams' capabilities, reduce technical debt, and shift from reactive operations to proactively enabling enterprise AI."

TrendAI has made Nvidia Nemotron models available inside customer environments as the Trend Micro unit aims to bolster data sovereignty capabilities.

The move will include integration of Nvidia's open-weight Nemotron models within its Vision One platform for Sovereign and Private Cloud, bringing AI-assisted security operations into customer-controlled environments.

This, the company said, extends capabilities such as guided investigation, security reasoning, and agentic SIEM to organizations, along with sovereignty, data residency, and operational control.

According to TrendAI, the move is primarily aimed at public sector organisations and those operating in regulated environments, many of which now face an expectation to run AI and process data locally.

With Nemotron open models, customers decide where the model runs, which version is deployed, and who can access it, running AI-powered security operations on Nvidia accelerated computing in their own data centers, sovereign clouds, and restricted-connectivity environments.

OpenShell improvements

Meanwhile, TrendAI said the collaboration with Nvidia on OpenShell provides an extra level of security for agents, giving organizations control and visibility of agentic activities.

“Sovereignty extends beyond where data resides. Enterprises also need control over where the intelligence acting on that data operates, and how it is governed.” said Rachel Jin, CEO of TrendAI.

"Organizations shouldn't have to give up the operational benefits of AI because their security data and operations need to remain within a customer-controlled environment. Our work is about bringing AI-powered security operations to the environments our customers choose to control.”

What users can expect

TrendAI Vision One for Sovereign and Private Cloud (SPC) with NVIDIA Nemotron integration will offer users AI-assisted threat analysis. This includes reasoning across security signals to identify threat patterns and attack chains, the duo said.

The firm said that guided security operations help security teams investigate threats and prioritize response activities, while agentic SIEM capabilities correlate telemetry and assist with security reasoning while maintaining team oversight.

The move marks the latest collaboration between the security provider and Nvidia in recent months. The company joined the firm's Open Secure AI Alliance, an industry coalition of more than 30 tech companies aimed at building open-source AI safety, security, and defense tools.

FOLLOW US ON SOCIAL MEDIA

More than half of UK organizations have reported "surprise" bills for monitoring IT systems, according to new research from Cisco, and AI is a key factor behind surging costs.

In a new survey from the firm, 65% of respondents said their observability costs had exceeded infrastructure or operational costs this year, with half reporting surprise costs with telemetry data, and another third saying that led to reduced data coverage or retention.

The complaint isn't just high costs, but those that are higher than expected, with 55% of those asked pointing to surprise expenses that are straining operations more than typically baseline costs.

General AI adoption costs are also exacerbating the issue, Cisco found. Another 55% said they are finding it impossible to control AI costs, a problem as AI observability has become a priority, according to nine-in-ten respondents.

That echoes a previous study by KPMG, that found many companies struggle with costs and lack visibility into AI spending.

“For today’s enterprises, the excitement of AI is clashing with the reality of the bill,” added Kamal Hathi, SVP and GM, Splunk at Cisco.

It's not just Cisco spotting this trend, either. A poll by Asana earlier this year found that eight-in-ten UK IT leaders had seen unexpected costs increases from AI over the last year, with Gartner predicting that AI costs could exceed developer salaries by 2028.

Given the rising costs, companies began cracking down on "tokenmaxxing" and other AI overuse this summer, with Uber introducing a monthly AI cap and Accenture telling staff to cut back on AI use for basic tasks.

AI observability

While monitoring AI is a clear priority, more than half of those polled said they don't have a clear business outcome tied to their AI usage, and 56% said they're unable to measure AI workload metrics such as LLM response quality, which Cisco said was a core metric.

Indeed, 85% said they need to have visibility into AI hallucinations, but fewer than six-in-ten said they're effective at spotting such concerns.

“We believe observability is the essential bridge to restoring predictability and confidence to the agentic AI era," Hathi added.

"By providing the necessary guardrails and incorporating observability into an AI tokenomics framework, leaders can transform AI from an unpredictable expense into a measurable and strategic long-term business investment."

Collaboration and teamwork

One silver lining in the confluence of challenges faced by enterprises lies in heightened teamwork and collaboration, according to Cisco. More than half (55%) of those polled are worried about the unintended consequences of AI agents, prompting closer synergy between departments.

Eight-in-ten IT professionals have observed a marked increase in collaboration between observability and security teams as a result.

Nearly three-quarters (73%) also revealed they can accurately trace performance issues to security root causes, an increase from 65% in the survey last year. Similarly, there’s been a marked increase in data sharing between these two teams from 74% to 84% over the last year, with a similar leap seen in working together on troubleshooting.

"Most companies will soon have far more agents at work than people," said Jeetu Patel, president and chief product officer (CPO) at Cisco.

"Managing agents at scale requires reimagining both security and observability. Have our agents been compromised? Are they doing their job well? In the world of agentic AI, these questions are inseparable, and they're moving at machine speed."

FOLLOW US ON SOCIAL MEDIA

AI adoption in organizations has become near-ubiquitous, but has not been coupled with appropriate oversight. Few companies can answer all of the key AI security questions: who is using AI and to what extent, exactly which tools and models are in play, and what data is being shared with them.

Research from Cyera shows that 83% of businesses already use AI, but only 13% have strong visibility into how it is being used. This gap is where partners can help customers move from experimenting with AI to governing it securely.

The biggest commercial opportunity for the channel over the next few years will not come from selling customers more AI tools; it will come from helping them govern the ones they already have.

Shadow AI: the challenge customers can't see

One of the clearest examples of the disconnect between AI adoption and governance is ‘shadow AI’, the use of generative tools without direct IT approval or visibility. This creates compliance gaps, as well as hidden channels for data loss that can’t be audited, because no one knows they exist. The consequences are already impacting businesses and are only set to increase, with one in five organizations experiencing a breach linked to shadow AI last year.

The problem of shadow AI stems from how AI entered most businesses. Adoption was initially organic and employee-led, driven by curiosity rather than policy. While that fuelled useful experimentation, it has frequently left organizations with no boundaries, no ownership, and no clear line between sanctioned and unsanctioned use. Customers need help turning that experimentation into something controlled.

A management rather than an IT issue

Ungoverned AI shouldn’t be seen as a housekeeping task for IT. When confidential data can leak into publicly accessible tools, when autonomous agents can affect key databases and systems with little oversight, and when third-party models introduce risks that vendor programmes never assessed, the exposure needs to become a central concern for management.

Yet this is where organizations are frequently falling short, and this has knock-on effects on their resilience, compliance and trust responsibilities. Some 77% of organizations have an AI strategy, yet only 26% can actually enforce it.

For channel players, this changes who they should be in conversation with. As the conversation moves from the IT teams to senior leadership, the partners who can discuss holistic governance concerns, rather than individual technical features, will be the ones trusted with leading the changes.

A services opportunity, not a resale one

This is where the opportunity for Managed Service Providers (MSPs) and Value Added Resellers (VARs) becomes clear. Vendors can supply the tools, but only a partner can operationalize governance inside a customer's own environment, with its particular mix of applications, data and regulatory exposure. This kind of expert consultancy isn’t something that can be bought off a shelf.

The value here lies in services rather than resale, from discovery and visibility assessments to policy design and the ongoing management that keeps governance current as tools and rules change. Each of those is recurring by nature. The aim is to move a customer from experimenting with AI to governing it securely, with the partner as the constant throughout.

The conversations to have now

Partners do not need a mature market or a polished AI security practice to start. The value is in having these conversations today, while guidance from MSPs and VARs is most needed. Four topics should be prioritized above others:

• Visibility: Most customers underestimate how much AI is already running across their business, and a straightforward inventory, covering approved tools, unsanctioned usage, and high-risk workflows, is the best route into the wider conversation.
• Data protection: Once visibility exists, the priority is controlling what goes into AI tools, with guardrails that stop proprietary or regulated data being shared with public models where it can be retained or exposed. The risk here is often capable employees trying to work faster, rather than bad actors.
• Policy: This makes the above controls durable and easier to review. This can take the form of a clear code of conduct, a list of sanctioned tools and conditions, and explicit guidelines detailing where human judgement must stay involved in sensitive areas.
• Agentic AI: As agents gain autonomy to act across business systems, partners can help set parameters, approval thresholds and logging, and ensure a human stays in the loop on decisions affecting customers, finances or critical systems.

Turning regulation into recurring revenue

Upcoming regulation means that securing AI is now an urgent priority for customers. Sound AI governance already aligns with frameworks including NIS2, the EU AI Act, the Cyber Resilience Act, and GDPR, and the next phase of EU AI Act obligations takes effect from August 2026. That is a concrete deadline partners can build a conversation around now.

The opportunity is to frame regulatory readiness as an ongoing managed service rather than a one-off compliance project. The rules will keep evolving, and customers will need a partner who keeps them aligned as they do. Handled that way, regulatory pressure becomes a durable, recurring revenue stream.

AI will only become more deeply embedded in how organizations operate. The customers who benefit safely will be those who adopt it with control over their data, their tools, and their decisions, rather than being steered by tools they cannot see. The partners who move the fastest, from selling AI to helping secure it, will build the deepest and most lasting customer relationships.

The career ladder has never been wobblier. Thanks to AI’s impact on job roles and responsibilities, what once seemed like a challenging but progressive climb to the top is a lot more complicated. Defined rungs on the career ladder, such as moving from junior into more senior positions, are now less clear.

Take the IT profession, where generative and agentic AI are helping to automate increasing elements of the software development lifecycle. From coding to testing and onto deployment and maintenance, AI agents are fulfilling parts of the traditional IT professional role, with a commensurate impact on the career ladder.

These emerging technologies are automating routine, entry-level tasks and compressing lower rungs of the ladder. As these rungs disappear, so do the opportunities for junior staff members to prove their capabilities and move into more senior positions, according to research from Harvard University.

If you’re going to climb the ladder, it’s more important than ever to stand out from your competitors. And the good news, according to the experts who have spoken with ITPro, is that professionals with the right blend of skills can still reach the top in the AI age.

“If you think about yourself in a narrow sense, then I think AI is likely to displace you,” says Joel Hron, CTO at content and technology specialist Thomson Reuters.

“But if you view yourself as a higher-order value add to a company – and your job is to solve problems, and you can avoid putting yourself in a box as a certain type of employee – then the sky's the limit for you, especially with AI.”

Find ways to join the dots

The message is clear: yes, AI brings an existential threat to some professional roles, but its impact doesn’t have to be terminal. Smart people embrace the challenge of automation. For Hron, a promotion-worthy professional understands that tomorrow won’t look like today.

“I think adaptability is now one of the most important things, especially as the job of software engineering is changing,” he says, referring to transformations in IT roles and responsibilities.

“If you believe that the value that you bring to an organization is through the code you write, then you know you are going to be frustrated in the future.”

Career ladder changes are not confined to the IT sector. Employees in all sectors, from public sector organizations to blue-chip enterprises, are seeing labor market flux. In these fast-changing conditions, managers look for up-and-coming staff who can work across functions.

Thierry Martin, head of enterprise data and analytics at automotive giant Toyota Motor Europe, says he’s impressed when people can join the dots, particularly if they bring a new take to an established way of working.

“Success is about going beyond the obvious,” he says.

“It’s great if someone can say, ‘Hey, we have the sales data, we have this quality data, we have this manufacturing data, and look, this is what we can do with it.’ I might think, ‘Wow, I never thought about this.’”

That sentiment resonates with Bernhard Seiser, VP of digital, data and IT at AOP Health, who says the people who impress him are hungry for change. The pharmaceuticals sector is characterized by the opportunity to turn small ideas into life-changing transformations. The best professionals look to exploit their knowledge.

“Great people want to learn,” he says.

“They want to understand not just the technical parts, but also the business. That capability is critical for my business and how I see digitalization within a pharmaceutical organization. My team needs to understand how the organization operates, sometimes even better than the people in the business, ideally.”

Crucially, Seiser says, great ideas aren’t put into production in isolation. Professionals who move up the career ladder understand a challenge before others and then work to deliver ambitious solutions on time and to budget as part of an integrated approach.

“Always keeping in mind that we are part of a family is key,” he says.

“Managing our portfolio is often stressful. We have a lot that needs to be delivered, but this work also needs to be seen in a broader context – we're here as a family to serve our patients, and that's what the professionals need to keep in mind.”

Enable change for the business

The characteristics of next-generation leaders that business leaders like Seiser identify are focused squarely on high-quality delivery. For Tal Carmi, CIO at technology specialist WalkMe, AI might change the shape of the career ladder, but that impact doesn’t mean the best people transform their approach.

“In general, impressing the boss is about taking ownership,” he said.

“It’s caring about being a good partner to the business; trying to be an enabler, not a naysayer. And it’s about working pleasantly. Collaboration doesn't have to be centered on corporate-style language. Being a good partner means being someone people like to speak to.”

Encouragingly, Carmi says professionals who excel in these areas – taking ownership, enabling change, and collaborating effectively – will have plentiful opportunities to be promoted, particularly in IT. People who see your strengths will be receptive to your ideas, rather than thinking you’re just interfering.

“You will find yourself at the forefront because you will try to adopt new tools, concepts, or whatever the latest thing is to try and give the business you’re working with the best solution that you can,” he says.

Ankur Anand, group CIO at recruiter Harvey Nash, also says key characteristics define impressive professionals. First, they have a robust understanding of the business problem. Second, they explain their solution without using technical jargon.

“This structured approach is not just about technology,” he says.

“It’s about technology, governance, control, and people, which is training and learning and everything together. And the final piece is about understanding what role they play in making their solution successful, and knowing what data helps.”

In short, great would-be leaders measure and assess risks that inform decisions. Anand gives the example of software testing. Rather than testing endlessly to try and make things work, a great professional approaches their boss and explains the cost, in clear financial terms, of not getting a system up and running.

“Then the executive understands the issue and says, ‘Okay, now I get the impact, and we need to figure out why things are not working,’” he says.

“There’s a big difference between a person who understands the context and risk, and those who can then articulate those concerns in a business-focused way.”

The Microsoft AI Tour is now the Microsoft Frontier Tour. A change that reflects the fact that we are no longer at the adoption stage of AI, but are now putting it to work, according to Microsoft.

However, this new frontier is fraught with challenges. One of the biggest of these appears to be a very human problem: confidence. From top to bottom of all businesses, there appears to be a crisis of confidence with AI.

The confidence to use the technology, confidence that it won’t make our working lives worse, and even confidence that it will ultimately deliver a return on investment.

Trust issues were a main theme of Microsoft’s AI Diffusion Report, released ahead of the Tour, which was covered by ITPro. It revealed a stark divide between UK leaders and employees on AI, with eight-in-ten leaders confident the UK can lead in the technology, while only half that number of employees agreed.

The reasons for low employee confidence were varied, though many suggested it had reduced their own job security and also limited opportunities for younger workers.

Speaking on stage, CEO Satya Nadella talked about the importance of the human element with AI.

“How do you put the human in the loop?” he said. “How do you really ensure that, ultimately, that sense of empowerment is not lost?”

Confidence in AI

Microsoft has made substantial commitments to the UK, investing in data centers and partnering with UK organizations like the NHS. And it dedicated a large portion of its keynote to call for more work to be done across the country.

UK and Ireland CEO Darren Hardman used his segment to talk about building confidence in AI across the UK. Laying out the challenge of turning AI’s potential into improvements that people can feel in their work or the services they use, and also the opportunities that are open to them.

“Our role at Microsoft is to bring intelligence and trust together,” Hardman said. “AI that puts your organization’s knowledge to work, helping people achieve more, while providing the safeguards and visibility to stay in control.

“You can give your people a say in how AI changes their work. You can set clear boundaries on the data AI can access. And the actions it can take. With human approval where it matters.”

A key pledge here is observability. Specifically with agents, which have raised questions of governance and control over the last year.

“For a leader, it means being able to answer a very straightforward question: ‘Do I understand and remain in control of what this system is doing?’”

“That is how we put Microsoft's mission into practice with you,” Hardman explained. “Helping people achieve more with the confidence and control to do it responsibly.”

Hardman ended his keynote with a call to action to build more confidence, more skills, and more opportunities in the UK so that AI opens doors for people “whatever their background, wherever they live”.

Talent, as he rightly pointed out, is everywhere in the UK, but there is a responsibility to make sure that the opportunity reaches all of it.

FOLLOW US ON SOCIAL MEDIA

HPE has unveiled the HPE Spaceborne Computer-4, an advanced compute module designed specifically to operate in space – and it’s heading to the moon.

The system will be deployed on Astrolab’s Flex Lunar Innovation Platform (FLIP) rover as part of an upcoming mission for the Moon Base II program. The move will see tests conducted to assess suitability of advanced computing capabilities for future missions.

According to HPE, these tests will benchmark AI and high-performance computing (HPC) capabilities, helping to create the “technological foundation” required for research on the lunar surface.

“Large language models (LLMs) will deliver insights, analyze visual data, and support autonomous monitoring at the lunar edge,” the company said in a blog post.

“Additional tests will process LiDAR data for mapping, navigation, terrain analysis, and infrastructure development, while an AI services container will demonstrate reliable deployment of scalable AI service applications on the Moon.”

Astrolab was selected by NASA earlier this year to provide lunar rover capabilities for Artemis astronauts as the agency prepares for a full return to the moon.

In a press briefing ahead of the announcement, Jaret Matthews, founder and CEO of Astrolab, said the move marks a significant first step in scientific research from the lunar surface.

“I think for the foreseeable future, the most valuable export from the moon is going to be data,” he said. “The technologies to collect, process, store data so that we can return that value to our customers are, you know, central to our business.”

The moon is “far away”, as Matthews noted, and this means the system faces problems in terms of time delays and bandwidth constraints. He told attendees that this project is essentially redefining the concept of edge computing.

“Pushing compute out to the edge so you can do more, create more value, and ultimately satisfy customer needs, these technologies are imperative for that,” he commented.

Under the hood of HPE Spaceborne Computer-4

HPE Spaceborne Computer-4 is a “lightweight, self-contained, and fanless” compute module designed specifically to survive harsh environments, according to HPE.

The system will be operating on the lunar south pole, and will be equipped with a lower-power CPU and GPU. Similarly, the module will leverage 1TB of flash storage through Japanese data storage firm, Kioxia.

Kioxia’s involvement builds on long-standing ties with the HPE Spaceborne program, the company said, with the two companies having worked extensively on mission activities and tests of data storage under lunar conditions.

As part of the move, HPE will also provide support for Astrolab’s mission control site in Hawthorne, California, with the company relying on HPE Private Cloud for storage purposes, and HPE CloudOps Software for workload management.

Lunar experiments

As part of the Spaceborne Computer-4 project, NASA and Deloitte will deploy a series of AI workloads on the moon and conduct experiments.

The Federated Learning for Unbounded and Intelligent Decentralization (FLUID) experiment, for example, will see NASA run an AI neural net model trained with data from the earth, moon, and on the International Space Station (ISS).

According to HPE, this aims to demonstrate how to coordinate three separate AI models from immensely disparate locations. NASA will also carry out tests on a tuned LLM with semantic processing capabilities to help find and interpret onboard documents.

Meanwhile, Deloitte’s Silent Shield will provide predictive cyber analytics and anomaly detection capabilities. This is a new cybersecurity system developed by the firm.

Extreme conditions

Faraz Valani, senior director and head of go-to-market at Kioxia, echoed Matthews’ comments regarding challenges, noting that the “rules are very different” for storage operating in harsh environments.

“You have a lot more challenges up there than we’re accustomed to down here,” he told attendees. “In space, everything is constrained.”

When deploying in a traditional data center, Valani noted that operators naturally have access to power, cooling, connectivity, and crucially, physical space. The upcoming lunar mission creates huge challenges across all those fronts.

“If something fails, you can simply just go out and replace it. In space, however, you don’t have that luxury because every watt of power, every gram of weight, and every inch of space makes a significant difference,” he explained.

“The slightest variation in any of these variables could simply make or break anything that you’re trying to accomplish up there.”

Scott Wiest, HPE’s CTO for global sales, noted that HPE “goes to pretty far extremes” when it comes to testing durability of traditional infrastructure, including freezing hardware, shaking it, and even putting it ovens to ensure it can handle extreme conditions.

“We have stuff deployed in extreme conditions, so this taught us to really go extreme,” he said. “Think about cooling differently, as you referred to. Think about being maximally efficient about weight, thermals, power, performance.”

Wiest said that long-running tests of HPE hardware on the ISS played a key role in fine-tuning capabilities on this front and will stand the compute module in good stead for the lunar mission.

FOLLOW US ON SOCIAL MEDIA

Part three: Managing the machine mind: Who is responsible for AI’s behavior?

An AI agent can answer a customer or act on an invoice before anyone can examine its reasoning. If the outcome is wrong, saying that a human was “in the loop” offers little comfort. Who gave the agent authority? And who could have stopped it?

The first article in this series asked why AI appears to have a personality, even though it does not think or feel like a person. The second explored how psychology could help identify unsafe behavior and improve testing. This final part turns to the workplace: once a business allows AI to influence people and take action, who is responsible for those actions if they negatively impact the business that deployed them?

The question is becoming more pressing as AI moves beyond producing document drafts and recommendations. In PwC’s 2025 survey of 300 senior executives, 79% said their companies were adopting AI agents, but only 17% said agents were fully adopted across almost all workflows and functions. An agent with access to business systems can complete a sequence of tasks, each of which may look reasonable in isolation. Its apparent competence can encourage people to grant it more freedom before they understand how it behaves in everyday work.

Accountability follows authority

When an AI agent causes harm, several parties may have contributed. The model provider built the underlying technology. A business selected it, connected it to data, and decided what it could do. An employee may have reviewed its output. Yet these roles do not carry equal control over a particular decision.

“The organization deploying the agent must own accountability for how it is used,” says Jerry Caviston, CEO of Archive360. A provider may share responsibility for a defect, he explained, but a company cannot give an agent access to its systems and then point to the vendor when something goes wrong.

That distinction matters because deployment choices shape the risk. An agent permitted to suggest a reply presents a different problem from one allowed to send it. The same system might safely identify an invoice discrepancy while being unsuitable to change a supplier’s payment details. Businesses, therefore, need to define an agent’s authority for each task, including the data it may use, but more critically, the conditions under which it must stop and refer to a human.

Donna Wilczek, chief product and technology officer at Basware, places responsibility with the people who design that authority. “Accountability follows ownership of the process, not proximity to the mistake,” she tells ITPro. An executive who approves an agent’s role is responsible for that design; an employee reviewing one exception is responsible for the decision they are genuinely able to make.

This also changes how businesses should describe an AI agent. Calling it a colleague can imply that it has judgment and responsibility of its own. Calling it an ordinary tool can obscure its ability to act across several systems. Luciana Rousseau, head of client innovation at Morson Group, describes agents as “delegated computational actors.” Systems may occupy a role but cannot own the consequences attached to it.

A useful starting point is a named human owner for every important AI workflow. “If nobody’s name is on it, it shouldn’t be running,” says Daria Hvizdalova, head of AI Adoption at Make. That owner must have the authority to change the workflow or suspend it when evidence suggests that its behavior is causing problems.

When human oversight becomes a rubber stamp

A common answer to AI risk is to require human approval. But an approval button alone does not establish meaningful oversight. If an agent produces hundreds of recommendations while a reviewer has seconds to consider each one, the human may become a routine step in a process whose decisions have effectively already been made.

The Morson Group’s Rousseau explained that supervision at machine speed must focus on “exceptions, consequences and behavioural signals.” People should spend their attention on high-impact decisions, unfamiliar circumstances, conflicting evidence and actions that are difficult to reverse. Asking them to inspect every ordinary transaction invites fatigue and habitual approval.s carefully as the system itself.

Hvizdalova offers a way to test for this: "If humans approve 99% of what the AI proposes, you probably have a rubber stamp rather than a meaningful control." Bastiman agrees from the regulatory side, noting that "an override rate near zero is a red flag, not a success metric," since it usually means a reviewer is confirming rather than evaluating. Wilczek describes the same failure from the employee's perspective: handed a long queue with little explanation, "they'll just click through," left to absorb the blame for a decision the system had effectively already made.

The reviewer also needs permission to disagree. Microsoft’s 2026 Work Trend Index found that 86% of surveyed AI users said they treated AI output as a starting point and retained responsibility for their thinking. Wilczek emphasizes that an employee faced with a long queue of unexplained recommendations may only be absorbing blame for choices they cannot realistically challenge. A case requiring approval should show what information the agent used and what action it proposes. The employee must be able to investigate and say no without having to fight the workflow.

Consider an agent handling invoices. It could resolve a small variance within a finance team’s agreed tolerance, then send an unusual case to a person with the supporting records. A change to supplier bank details might require a separate verification process outside the agent’s control. These are different decisions within the same workflow, so a single setting labeled “autonomous” or “human approved” would conceal the distinctions that matter.

Oversight should also account for human behavior. A system that usually gets the answer right can make checking feel unnecessary, particularly when it presents its conclusions confidently. Part one of this series examined why people read human qualities into AI’s manner, with part two considering how to test troubling behavior. In production, those same traits affect whether employees question an answer. Leaders must assess the people and pressures around the system as carefully as the system itself.

Watch what happens after deployment

A successful demonstration is evidence that an agent worked under test conditions. It does not establish that the agent will respond consistently to unfamiliar customers or interactions with other systems. Monitoring must, therefore, begin when an agent goes live.

The measurements also need to extend beyond whether tasks were completed. Wilczek suggests watching for downstream signals such as rising invoice recording requests or supplier queries. An agent could appear productive while creating work or risk elsewhere. A process owner needs a way to narrow its permissions while investigating, and have enough information to correct decisions that can be reversed.

This is where a specialist in AI behavior might help. Such a person would look across many decisions for patterns that an individual reviewer cannot see: growing correction rates, persistent confidence despite falling accuracy, unusual action sequences or repeated approaches to a rule’s boundary. The role would require knowledge of the business process as well as the technology.

Dr Janet Bastiman, chief data scientist at Napier AI, cautions against treating terms such as “deception” or “goal drift” as evidence of human motives. What matters is observable behavior, she emphasizes. Whether outputs diverge from stated reasoning and whether the system is pursuing a convenient measure at the expense of the real objective. A specialist who understands model validation but not the work being automated could miss the most consequential change.

This expertise need not always sit in a newly titled job. Risk and operational teams can share the work, provided someone finds patterns and a process owner can act on them. Monitoring has little value if nobody can reduce access or stop the agent.

Give competence a boundary

The hardest systems to govern may be the ones that rarely fail. When an agent performs well most of the time, leaders may be tempted to expand its reach. Employees may stop checking, and customers may assume that a polished answer reflects a sound decision. None of those reactions proves the agent can be trusted with a more consequential task.

If a business cannot establish what information influenced an important outcome or which rule authorized the action, the agent’s independent authority should be limited. It may still recommend an action or handle work that is inexpensive and easy to reverse. Material financial decisions and other consequential actions need stronger evidence and a clear route to human judgment.

The practical task for leaders is to write down those boundaries before deployment, then revisit them as performance changes. For each workflow, they should identify its owner, define the agent’s permitted actions, set escalation points, preserve decision records, and test how quickly authority can be withdrawn. Reviews should consider real outcomes, including complaints and corrections, alongside speed and productivity.

It’s the gap between the appearance of human control and its substance that has revived the question running through this entire series: does the modern workplace need dedicated AI behavior specialists? .

Rousseau believes something like the role is necessary, though she resists calling it "psychologists for machines," describing it instead as expertise in "the behaviour of the whole human-AI system." Bastiman is more sceptical of the need for a new title, pointing out that "second-line model validation and model risk management already exist to do this" within regulated industries. Both she and Wilczek caution against borrowing language built for human motive, with Wilczek warning that words like "deception" risk being applied "somewhere it doesn't quite fit yet."

That brings the series back to its opening question. AI does not need a psychologist in the human sense. But businesses do need to understand the behavior their systems produce and the habits people develop around them. When an agent appears capable of acting like a colleague, the organization must be especially clear about the one thing it cannot do: take responsibility for its decisions, as an AI is fundamentally a tool for human users, no matter how much autonomy it is granted

You’ve got to move quickly if you’re an IT vendor trying to create a competitive advantage in the AI age. At the recent Workday Rising event in Las Vegas, the software giant’s executives acknowledged the rapid pace of change and set an agentic-enabled course for the firm’s future product roadmap.

“We're so focused on AI,” said CEO Aneel Bhusri at the start of his product keynote at the event. “Workday was born in the cloud. Humbly, I think we are one of the leaders in the cloud. But now AI shows up, and it's again our opportunity to shine and bring great products to you in the next chapter of the technology industry.”

Building on a solid foundation

In a media session before the event, Bhusri suggested that Workday has been building for this moment for the past 20 years. Today, Workday, which processed more than 1.4 trillion transactions last year, supports more than 80 million users. Now, the firm’s internal innovators are drawing on these experiences to create a fresh set of services for customers.

“AI is one of the biggest opportunities in Workday's history,” said Bhusri during the press briefing. “We're moving faster than we have since the founding of Workday, and we need to be there for our customers and help them to take advantage of AI in a way that is trusted and built for the enterprise.”

Bhusri, who co-founded the company in 2005 and stepped down from the CEO role in 2024, returned to the helm in February 2026. He acknowledged that he’s overseen a hectic period, leading the integration of AI platform Sana – which was acquired by Workday in November 2025 – into the heart of Workday's technology.

At the Rising event, it was announced that Sana is now the default home screen for Workday. Referred to as Workday Home, users can communicate with the platform in natural language and orchestrate agents to complete work activities. Bhusri said this new front door allows professionals to ask questions about their tasks rather than search for answers.

Alongside Sana, the keynote presentation also showcased a series of new agentic services. Workday launched Autopilot Agents that work continuously in the background. These AI agents work across disparate areas, including a Talent Acquisition Agent, a Developer Agent, and Sana Concierge, an overarching HR and IT service that manages multiple activities.

More time to add value elsewhere

Across the event agenda, Workday executives explained how the 17 agentic services the firm has developed during the past 12 months automatically anticipate tasks that need attention, freeing up professionals to work on more valuable activities, such as the personal relationships that are crucial to hiring and reskilling in HR.

Workday’s key message for customers in Las Vegas was that it’s building agentic capabilities into its platform, rather than bolting them on. Responding to a question from IT Pro during a media session following the firm’s keynote presentation, Bhusri recognized that this AI-enabled approach involves significant effort, particularly given the pace of change.

“Every week we read about a different model being the leading model, and we can just translate that pace of change to Workday,” he said. “In the nine months I've been back, we've transformed the organization to be ready for AI and to deliver on AI products.”

Bhusri said tech companies like Workday must change culturally to meet this development cadence. By working with the team from Sana and encouraging a startup-like culture within Workday, Bhusri believes the firm is now set up to deliver a set of flexible products that are commensurate with customer demands in an era of almost constant change.

“It's one of the reasons why we stay agnostic to the models,” said Bhusri, responding to ITPro, asserting that his firm is developing an increasingly open platform.

“We'll work with all model providers. But who knows how this is going to shake out? We must deliver great value for our customers, and that's what we're focused on.”

Bhusri said that acting like a startup means working leaner and faster. As part of this strategy, Workday announced a global workforce reduction of approximately 2.5% in late September 2026, affecting about 500 employees. Bhusri addressed the cuts in the media session.

“There were, in some parts of the development organization, 10 layers between me and the developer getting work done, and it was the actual individual contributors and developers themselves who said, ‘I can't get anything done. There are too many layers of approval.’”

The hope is that a leaner, faster Workday will continue to pick up the pace of innovation for its customers. In a one-to-one chat with ITPro at the event, Dean Arnold, VP of the AI platform at Workday, said he hoped customers at the event would recognize that the firm is ready to meet them with agents that can help them solve their key business challenges.

“We’re fundamentally showing up and saying, ‘We have these solutions. They're critical, and we can work with you on them,’” he said.

“And we've got flexibility as well. We're playing in our ecosystem, outside our ecosystem, and the ecosystem’s expanded. We have agents available right now for you to use. The next thing is to help our customers adopt them.”

Early adopters

Some customers have already made that shift. ITPro spoke with Hezha Guimond, VP for HR technology and integration at Anew Health, who is using Workday’s AI Talent Suite, particularly Recruiting Agent and Candidate Experience Agent, to identify and hire staff faster. She said Rising allowed her team to see the future roadmap firsthand.

“It builds up the excitement,” she said. “I know my team loves coming here because I feel like it energizes them to be able to go back and think through what else we could do. So, there’s a couple of products we have been texting back and forth about during the different sessions at the conference.”

What’s clear is the changes announced at the event are just the start. Yes, Workday is startup-like in delivering new features, but can its customers keep up with the rapid pace of innovation? The answer, suggested Bhusri, is that his company is moving faster than the digital leaders buying its products, and that’s not necessarily a bad thing.

“We're ahead of our customers, but that's where we're supposed to be,” he said. “That's why they choose us. That's why they pay for us. We should be ahead of them, and the leaders are trying to keep up with us, and that's the way it should be.”

AI is already changing how customers discover, evaluate, and buy technology. The next shift will go further, where AI agents will increasingly participate in the buying process themselves.

We’re still in the early stages of agent-to-agent (A2A) commerce, particularly for complex B2B solutions, but the direction is becoming clearer. Buying journeys are becoming more machine-mediated, with agents increasingly able to represent customers and act within the wider channel ecosystem.

Microsoft and others are already exploring agentic commerce models in which AI moves beyond recommendation and begins taking action on behalf of buyers and sellers.

That shift changes what a marketplace is there to do.

From discovery to decisioning

Most marketplaces today focus on discovery. Customers search a catalog and choose what they want, often with a channel partner helping them make sense of the options and put the solution into practice.

AI changes that model.

A customer agent can increasingly represent what the buyer is trying to achieve. The marketplace can interpret that intent against what its ecosystem is able to deliver, while partner agents make relevant expertise and customer context available.

This means customers no longer need to navigate hundreds of possible combinations themselves. Instead, AI can work across the ecosystem to answer a more valuable question - what is this business trying to achieve, and what proposition is most likely to meet that need?

The marketplace therefore begins moving from discovery to decisioning. Its value comes from using context to shape a commercially relevant proposition around the customer’s needs.

For marketplace owners, this creates the potential for a more productive channel and greater commercial precision. Opportunities can be matched with the expertise best placed to turn them into something customers can buy.

For that to work, AI first needs to understand what each partner actually brings to the ecosystem.

Making channel value visible to AI

In an agent-led buying journey, simply being listed in a marketplace won’t be enough. Customer and marketplace agents need to understand what a partner can bring to a particular opportunity.

That extends well beyond the products they resell. Agents will need to understand a partner’s customer relationships, market expertise, service capabilities, technology experience, and ability to support customers over time.

In other words, channel value needs to become discoverable and understandable by humans, systems and AI agents.

This creates an opportunity for marketplace owners to make the expertise and customer reach within their channel much more accessible. A regional Managed Service Provider (MSP), specialist Value Added Reseller (VAR) or trusted customer partner that might previously have been difficult to surface digitally can become part of a recommended proposition because AI understands where and how it adds value.

That can strengthen the role of the channel. AI may simplify discovery, comparison, and parts of the transaction, but customers will still need expertise to understand their environment, integrate technology, manage change, and deliver ongoing outcomes.

And as AI changes how partners participate in the buying journey, it’s also creating an entirely new set of opportunities for them to take to market.

AI also creates new things for partners to sell

There’s another side to this transition.

AI is reshaping how the marketplace operates while also creating an entirely new category of products and services to sell through it.

Channel partners will increasingly bring AI to customers as part of a broader solution, helping turn new capabilities into something businesses can actually deploy and operate. This creates particular opportunities in complex enterprise and public-sector environments.

It also changes how those solutions are commercialized. AI pricing is likely to become more closely tied to how services are used or to the outcomes they deliver. At the same time, a single proposition may involve several providers, making revenue sharing more complex behind the scenes.

The customer shouldn’t have to absorb that complexity. The marketplace operating model has to manage it for them.

The operating model has to catch up

Adding an AI assistant to the front of a marketplace doesn’t make the marketplace agentic.

If a customer agent can express an intent and a marketplace agent can recommend a sophisticated multi-party solution, the underlying platform still needs to turn that recommendation into something commercially executable.

Products and services need to be brought together into a coherent offer, with the right partners engaged and the commercial model defined from the outset. From there, the marketplace has to manage fulfillment, track usage, and ensure that permissions, compensation, and revenue settlement work across the ecosystem.

As agents begin interacting with other agents, trust, governance and control also become fundamental parts of the architecture. Marketplace owners need to know which agents are authorized to act, what actions they can take, and under which commercial rules.

This points to a broader evolution of the marketplace itself - from a transactional platform into an intelligent commercial orchestration layer capable of turning AI-driven recommendations into executable, governable, and monetizable propositions. It’s also beginning to inform how marketplace platforms, including our own, are evolving.

Once marketplaces can connect decisioning with execution in this way, the channel starts to look less like a collection of separate participants and more like an intelligent, coordinated ecosystem.

From channel ecosystem to intelligent ecosystem

A2A commerce will develop gradually. Enterprise buying will continue to depend on trust, expertise, and human judgment, even as AI agents take on a greater role.

For marketplace owners, the opportunity is to make the whole ecosystem more intelligent and productive - connecting customer intent with the right expertise and making complex propositions easier to create, buy, and monetize.

Success will increasingly depend on how effectively a marketplace can turn ecosystem intelligence into commercial outcomes. For channel partners, that means ensuring their value can be understood and activated by customers, marketplace owners and, increasingly, the AI agents working alongside them.

AI tools like copilots, chatbots, and agents have rapidly taken root in the workplace in the years since ChatGPT brought the technology into the mainstream. The initial promise was that AI would lighten the load and help workers carry out tasks without having to learn new skills. In reality, that's not quite what happened.

Barriers to adoption have led to a growing AI skills gap in companies driven not by a lack of understanding of how to use the technology, but by a lack of understanding of how the technology works and what its weaknesses are.

On this week's episode, Jane and Ross are joined by Greg Hanson, VP and head of EMEA North at Informatica from Salesforce, to discuss the impact this can have on companies and whether there is a straightforward way to overcome it.

Highlights

"When we did our survey, what was very clear coming through the datasets is that organizations are actively investing in data literacy. So some of the things we've already talked about are resonating, and they recognize this is an area, a critical operational area."

"That's the first thing; let's challenge where people are and make sure we are operating from the same benchmark. You've got the right readiness to assess."

Links

• Why AI adoption is a people problem, not a technology problem
• Enterprises are 'paralyzed by a lack of understanding' with AI adoption – and there's one key factor that decides success

Dell, HPE, NetApp, Oracle, and CoreWeave all saw their shares stumble this week, amid concerns that the AI infrastructure buildout is poorly timed and the significant investment may not pay off.

The selloff follows concerns that data centre and cloud companies may be overvalued and that enterprise spending on AI isn't enough to cover the costs of the trillion-dollar infrastructure buildout. At the same time, Gartner warned that enterprise and consumer demand is failing to prop up that spending.

Dell, HPE and NetApp’s shares all declined yesterday — 11%, 8% and 7%, respectively — following a damaging report from analysts at Barclays, which downgraded the companies ratings saying they "don't see a meaningful AI inflection yet," and raised concerns about "enterprise AI timing."

Though the decline caught attention, all three firms have seen their shares climb dramatically over the course of the last year, with Dell's share price up more than 300%.

Still, the share slide spread to others in the industry, with Oracle down 9% and CoreWeave slipping 15%.

The Barclays analysts suggested that businesses are spending just 5% of IT budgets on AI, adding that the strong profits seen at data centre-adjacent companies may reflect high prices caused by component shortages rather than an industry shift to AI.

"There is growing sentiment about enterprise AI adoption creating a hardware refresh," the analysts reportedly said. "We believe the strong enterprise revenues are more of a function of memory costs than a move to enterprise AI at this time."

Not enough AI demand?

Major hyperscalers and AI leaders are expected to spend more than a trillion dollars rolling out AI infrastructure such as data centres this year alone, via a network of deals that sees cloud providers, hardware makers and AI developers investing in each other.

AI developers continue to struggle to pay for the high costs, however, with reports that OpenAI's annualized revenue is on track to be $50bn — up significantly from the year before but $20bn behind previous estimates. Anthropic's annualized run rate exceeded $65bn over the summer, and it's set to top $100bn by the end of 2026, according to Reuters reporting.

"In 2026, AI providers are expected to spend nearly $1 trillion on AI infrastructure, and are forecast to invest an additional $6 trillion between 2027 and 2030," said Will Sommer, senior director analyst at Gartner. "However, our research indicates that enterprise and consumer AI revenue will be insufficient to support profitable returns on all planned AI capital expenditures during this period."

Earning its keep

Previous research by Bain & Company suggested that the AI industry is going to need to be worth $6trn within the next few years — with $4trn in entirely new applications and services — in order to pay for the ongoing investment.

Gartner's latest analysis suggested AI data centre owners will need to generate $8.5 trillion in revenue to cover already announced capital spending, financing costs and a 12% return on invested capital. The analyst firm thinks this won’t happen until 2032 at the earliest, leading to trillions of dollars in write-downs.

"However, total enterprise and consumer AI revenue that can be attributed to AI data centers during this period is forecast to reach no more than $3.6 trillion," said Sommer. “Because most AI-native providers operate asset-light business models, we predict that less than half of AI revenue will be earned by providers that own significant AI infrastructure."

Neocloud providers — those who build specifically for AI workloads — could be the most at risk, Gartner added, as hyperscalers have more diversified revenue sources to keep them afloat. The analyst predicted that 80% of neocloud providers will "cease independent operations" by 2030 because of such pressures.

Releasing AI-enabled services is one thing – what’s crucial is that your customers find value in your products. During the Workday Community Keynote at the tech firm’s Rising event in Las Vegas, two business leaders explained how their organizations are making the most of AI.

Rob Enslin, chief commercial officer at Workday, opened the keynote by suggesting that Workday’s product announcements this week have been focused on bold AI-enabled innovations with customers at the center. “It’s all about how you move from asking AI questions to getting the real work done,” he said.

To highlight Workday’s customer-centric approach, Enslin welcomed two business leaders to the stage. First up was Mike Harmer, VP of enterprise business platforms at Intermountain Health, who said his healthcare organization must deliver high-quality care cost-effectively, with technology playing a crucial role.

As a long-time Workday customer, the organization wanted its data consolidated in one place to exploit AI. Harmer also wanted to move beyond a system of record and create a system of action, where professionals could use data-powered services to make it easier for caregivers and administrators to boost their efficiency and spend more time with patients.

With solid Workday foundations in place, Harmer’s team explored AI. Intermountain Health leaned into Workday’s early adopter programs to test agents in a sandbox and build confidence. So far, they’ve used Developer Agent, Payroll Agent, and Self-Service Agent to develop use cases and solve business challenges.

Harmer is excited about the promise of Workday’s AI and data technologies: “When these tools get into the hands of our caregivers, it's exciting to see what they do with them. I think they're going to very intentionally disrupt and invent new ways to do things that weren't possible before.”

Sarah Clements, head of HR and shared services at BMO, explained to the audience how the bank launched Workday’s self-service agent to all 55,000 employees in June. Her digital-first organization is eager to embrace AI while simultaneously managing risk.

To move forward with agentic technology, Clements and her team engaged stakeholders across the business and created a responsible AI Forum: “We spent time with them explaining what the agent would do, and actually, more importantly, what it wouldn't do.”

The HR team looked at the bank’s policies and procedures and brought everything into the Workday platform. With the foundations in place, they began to explore use cases for agents, checking that technology answered questions in the way people anticipated.

“The number one thing we're trying to do is make things easy for employees and managers so that they can do their jobs,” Clements said, referring to the implementation of Workday’s Self-Service Agent and the use of natural language to ask key questions about key business activities, such as producing a mortgage letter for a customer.

European businesses are over-reliant on American AI providers and their laissez-faire attitude risks putting the continent at the mercy of closed systems and unexpected access restrictions.

That’s according to Mistral AI co-founder Arthur Mensch, who penned a column in Le Monde arguing Europe will be subject to an “an economic regime of dependence without equivalent in the history of civilization”.

Mensch claimed that “serious people” in Europe have suggested it’s better for the US and China to spend the millions or billions of Euros needed to create frontier AI models, rather than strain the European economy.

“According to them, it’s better to follow the recipe of the internet, where the value is in creating new applications and focus our efforts on what Europe knows how to do: its industrial data, its SMBs, and professional expertise. For everything else, open models created by others will suffice,” said Mensch.

“This is a fundamental error, which will cost Europe its place in the world,” he added. “AI isn’t the internet. The internet relied on open protocols that nobody owned. Closed AI models are proprietary. What an application or agent knows to do is decided by the model.”

Mensch pointed out not only can the owners of these proprietary models hike prices or retire a version of the model without user recourse, they can also cut access without notice. This isn’t hypothetical, as evidenced by Anthropic cutting off access to Fable and Mythos in June 2026.

“Access was reestablished, but the decision wasn’t taken in Europe, nor for Europeans. Nobody asked our opinion,” he added.

Mensch’s column comes amid heightened tensions over AI dependence and a rising interest in sovereign cloud and sovereign AI.

In the UK, the government established a £500 million startup fund to boost the country’s sovereign AI ambitions. Businesses in the country have also started to turn their attention towards sovereign AI in order to reduce their dependence on overseas providers.

Nevertheless, sovereign AI can have hidden costs. According to Forrester senior analyst Dario Maisto, this can mean slower roll out due to a skills deficit and a misunderstanding of what ‘sovereign’ is when it comes to infrastructure.

There’s also a monetary cost: research commissioned by CCIA Europe has claimed the EU sovereign AI bill could cost six times more than the bloc’s own upper estimate of €18 billion. Instead, the trade body – which is made up of non-EU businesses, primarily from the US – has predicted a potential cost of €115 billion over 10 years.

According to Mensch, however, the price is worth paying.

“If Europe resigns itself [to the status quo], AI will continue to be created elsewhere and we will only have access to second tier export versions, in exchange for licenses that will cost trillions of Euros,” he said.

Two critical flaws in one of the most widely-used LLM gateways, LiteLLM, could let low-privilege users execute code, hijack admin accounts, and steal AI provider API keys.

With 97 million downloads per month and more than 59,000 GitHub stars, LiteLLM runs inside large enterprises, startups and many of the platforms used to build AI applications.

It functions as an LLM gateway, used for storing API keys for every AI provider the organization uses, managing user access through roles and permissions, routing prompts and responses between applications and providers and tracking usage and costs across models.

However, according to Checkpoint researchers, CVE-2026-35029 allowed any user – including those with read only access – to reach administrative functions and take over the gateway. CVE-2026-59823, meanwhile, allowed attackers to redirect gateway requests to their own server and capture AI provider API keys.

"In practical terms, anyone holding an account on a vulnerable LiteLLM deployment could take over the gateway and use it for malicious purposes," said the firm.

LiteLLM uses role-based access control, where administrators manage the system and other users receive limited or read-only access. However, the CVE-2026-35029 access control vulnerability means that users with minimal privileges could reach administrative functions controlling the gateway’s core configuration.

From there, an attacker could read files stored on the server, including configuration files and system files containing sensitive information. They could also overwrite administrator credentials with their own, locking legitimate administrators out of the system, and instruct the gateway to download and execute code from an external source, enabling malware installation, data exfiltration or movement to other systems on the network.

CVE-2026-59823 is a server side request forgery vulnerability that allowed any gateway user to direct LiteLLM to send API requests to a server under the attacker’s control without validation.

Attackers could steal AI provider API keys without any compromise of OpenAI, Anthropic or other providers, use the organization’s AI provider accounts without authorization and cause unexpected costs to be billed to the organization.

LiteLLM has now resolved CVE-2026-35029 and CVE-2026-59823 in version 1.83.9, with all later versions including the fixes.

It's notable that the two vulnerabilities are nothing new: they include missing authorization checks, path traversal, RCE via dynamic module loading, and insecure password storage, and all have established mitigation patterns.

"Both vulnerabilities belong to well understood classes of security error, which shows that AI infrastructure is now exposed to the same risks that enterprise software has faced for years," warned Checkpoint.

"As tools like LiteLLM gain adoption at speed, attackers are paying close attention to them, and organizations need to secure their AI gateways with the same rigor they apply to any other business critical system."

AI gateways are starting to emerge as high-value targets, holding credentials, managing trust relationships, and controlling access to multiple AI services, said Li Zhao, principal strategic services consultant at Black Duck.

"Because many organisations consume these technologies indirectly through platforms, frameworks, and internal tooling, the risk extends beyond a single product and becomes a broader ecosystem concern," she said.

"Organizations should treat AI gateways with the same level of scrutiny they apply to identity platforms, API gateways, secrets-management systems, and other high-trust enterprise infrastructure."

The European Centre for International Political Economy (ECIPE) think tank has issued a report claiming that the EU's proposed sovereignty rules could cost the bloc billions.

The research, funded by CCIA Europe – a trade body with members including Amazon, Google, and Meta – examined the costs associated with the European Commission’s proposed Cloud and AI Development Act (CADA), which excludes non-EU providers from parts of the cloud and AI market.

The research found, the Act risks imposing €115 billion in costs over the next decade – six times the Commission’s upper estimate of €18 billion.

The reasons, said the researchers, come down to three critical blind spots in the Commission’s impact assessment:

• Failure to include €33.3 billion in direct public budget exposure
• Ignoring the cost of software re-architecting and code-rewriting, which ECIPE said accounts for 69% of CADA’s real-world impact
• Failure to take into account the cost of private-sector spill-overs, which, the report found, could amount to as much as €80 billion.

"Member states can’t rely on the Commission’s incomplete estimates when deciding CADA’s strict sovereignty requirements for cloud and AI providers," said CCIA Europe’s policy manager, Mitchell Rutledge. "Taxpayers and businesses will foot the bill, yet the official assessment missed 70% of CADA’s true impact.”

CCIA Europe is particularly bothered by Article 31 of the act, which allows the Commission to extend strict sovereignty restrictions beyond public procurement to certain private sectors, including banking, energy, healthcare, and transport.

This, it said, would increase the cost enormously, as European software vendors and digital SMEs would be forced to redesign applications currently built on global cloud architectures and AI platforms.

“Governments must recognise that CADA’s massive spillover costs for essential private-sector entities are not hypothetical but inevitable," said Rutledge. "Businesses cannot simply switch servers; they must redesign software and rewrite code across entire supply chains.”

Not entirely unexpectedly, CCIA Europe is calling on EU member states to drastically overhaul CADA, limiting sovereignty restrictions to sensitive workloads where specific public-order risks are demonstrated.

Describing corporate-origin and personnel-nationality requirements as 'discriminatory', it wants to see these rules replaced with risk-based technical safeguards.

Meanwhile, 'trusted international partners' with bilateral EU procurement commitments – like its members – should be exempted from Article 18 country screening.

“These sweeping restrictions risk draining European innovation and penalizing highly digitized member states," said Rutledge. "EU decision-makers must refocus CADA on objective, risk-based security rather than discriminatory rules that fragment the single market.”

Big US tech firms have put a lot of effort into convincing Europe that they too can provide sovereign European services. AWS's European Sovereign Cloud, for example, sees enterprise data hosted and stored solely in the region, and kept separate from other AWS Regions worldwide. Microsoft's Sovereign Cloud and Google's sovereign solutions make similar promises.

However, there are still concerns: US tech firms are subject to the US Cloud Act, which allows US law enforcement to demand access to the data they hold, even when it's stored outside the US.

Thanks to these concerns, according to Gartner research earlier this year, a fifth of European workloads are shifting to local clouds, with sovereign cloud infrastructure spending in the region set to triple between 2025 and 2027.

How many security professionals does it take to clean up the flaws found by AI? About a third more than you've got now.

That's how many more security professionals and other third-party consultants Bank of America has temporarily hired in order to work through the patches and other fixes necessary to address vulnerabilities spotted by Anthropic's Mythos and other AI security models. In addition to 3,000 existing staff, the bank added 1,000 external contractors to address the challenge.

Anthropic unveiled its Claude Mythos security model in April, but initially only gave access to approved partners via Project Glasswing. Mythos was then temporarily banned for export by the US government, alongside its Fable model, lifting the restrictions three weeks later.

In an interview with CNBC, the bank's CEO Brian Moynihan said the company had spent hundreds of millions of dollars over the last quarter fixing software as well as sorting out anti-money laundering compliance following a near-record fine from US regulators.

Finding flaws

Banks were included in the list of organisations given access to Mythos, with JPMorgan Chase an early partner, though reports from earlier this year suggest Bank of America has been testing the model since April. The main concern is the ability of the model to spot flaws in systems, sparking a wave of patching to avoid security risks.

"That risk is real risk, and that's why when the frontier-type models came out... A lot of us sat down and said, 'How do we figure out how to get through this quickly?'" Moynihan said in an interview with the American news network.

"We put on 1,000 contractors to help us do the patching work in the second quarter, and those people remain here today, helping us, on top of the 3,000 people we have in that area," he added.

However, Moynihan admitted concerns around the fact that companies rolling out AI products spotting these flaws were also offering services to mitigate them.

"We've got to stay ahead of them, and we've got to be mindful that the people doing the software development are the people trying to sell us the services," Moynihan said.

High costs

While the security work comes at a high cost, Moynihan said the company expects that to reduce over time. "Once you patch, you don't have to patch as much. ... The pipeline moves through," he added in the interview.

Bank of America predicts it will save $300m this year thanks to internal AI projects, rising to $700m by next year – however, Moynihan also suggested the bank was spending $4bn a year on new technologies, including AI.

Moynihan said on the call with analysts that "AI has helped us", with 130 projects implemented. Though he admitted "none of them are huge contributors," Moynihan predicted a 10% boost in efficiency at the same cost.

Bank of America's share price slid 8% after its quarterly results revealed a 17% slip in third-quarter profit – and after $1.2bn in fines for violating anti-money laundering rules was announced Tuesday.

Data integration and analytics specialist Qlik has achieved the AWS Data Competency in the cloud giant’s relational database category.

The AWS Competency program aims to help customers identify and grow their business with AWS partners that can offer deep industry experience and expertise across industries, use cases, and workloads.

The relational database credential now recognizes Qlik’s ability to deliver unified data architectures on the AWS platform and help organizations integrate, govern and activate data across complex environments.

The vendor said the designation differentiates it as a member of the AWS Partner Network that is capable of providing deep technical proficiency and customer success delivering data solutions on the cloud platform.

Relational database expertise

Qlik’s open, platform-agnostic model is designed to help organizations work flexibly with data, cloud platforms, and technologies without being restricted to a single technology stack. The firm’s wider capabilities include integration, transformation, governance, analytics, and AI.

Within the relational database category, Qlik provides relational database engines, migration and performance tools, as well as business applications built on Amazon Relational Database Service (RDS) and AWS Aurora.

Support for AWS-hosted databases spans Oracle, SQL Server, MySQL, PostgreSQL, and MariaDB, alongside SAP HANA, Sysbase ASE, and IBM Db2.

Benefits for Qlik customers

Qlik said its AWS recognition means customers can connect and work with data across a diverse range of enterprise environments – such as AWS, cloud platforms, on-premises systems, and SaaS applications.

They can also build on their existing environments while being able to adopt new architectures and technologies, as well as create a stronger foundation for analytics and AI.

Matt Hayes, general manager of Qlik’s data business unit, said data strategies now require greater flexibility and cannot be built around a single platform or a single destination.

“Our customers need the freedom to work across the technologies they already depend on while modernizing their data foundations for analytics and AI,” he explained. “Achieving the AWS Data Competency reflects the depth of our data capabilities and our commitment to giving organizations a more open, flexible path from complex data environments to trusted business outcomes.”

The topic of conversation in enterprise IT used to focus on transformation: which platform to move to, how fast to migrate, and what the roadmap looked like. Gradually, the conversation has shifted to questions about how we ended up spending so much just to stand still.

Those conversations with senior IT and finance leaders across the region keep returning to the same issues, and the pattern goes beyond any single vendor or sector. Organizations are caught between rising software maintenance costs, a storm of vendor demands to upgrade to their latest AI versions, growing cybersecurity demands and budgets that have barely moved.

Microsoft's 2025 Work Trend Index finds that one in three employees say the pace of work over the past five years has made it impossible to keep up, and business leaders feel that pressure as well. When you factor in that nearly a quarter of the IT workforce's time goes on maintaining existing systems, it becomes clear why so many organizations are struggling to fund anything new.

We call this ‘innovation paralysis’, the inability to progress new strategies and ideas because existing systems consume so much of the budget, resources and attention that innovation requires. For channel partners that can identify and eliminate low-value spend, freeing up capital and resources for higher-value initiatives, this presents an opportunity to help clients accelerate progress and springboard to the future.

The pressure leaders are feeling also shows up in the numbers

Our research found that 44% of C-suite decision-makers see AI and automation as their top capability priority, and 69% anticipate significant changes in their ERP investments. However, most organizations have resources tied up in maintenance costs, with little room for anything new.

A separate Akamai study of 750 IT decision-makers in the UK, France, and Germany reinforces this point by highlighting how rising cloud costs are forcing over two-thirds of businesses to cut spending elsewhere. Those cuts hit AI budgets, cybersecurity investment, and IT staff the most, as only 11% of organizations say their AI initiatives are paying for themselves. The result is a widening gap between ambition and execution: vendors and market demand that businesses accelerate AI adoption and strengthen cyber resilience, yet organizations keep redirecting the resources needed to sustain existing systems instead.

Much of this pressure stems from vendor lock-in. Forced upgrade cycles, limited pricing transparency, and the threat of losing access to future innovation unless a company migrates to a newer version are now familiar frustrations, with 35% of C-suite leaders directly citing vendor lock-in as a barrier to their long-term technology goals. The impact of that dynamic becomes most visible in support contracts, where many European enterprises continue to pay premium maintenance fees for services that feel distant, standardized, and disconnected from their operational realities.

Trust: The gap channel partners must bridge

Relationships, implementation history, and the extent to which boards trust the guidance their channel partners provide all shape the outcome of enterprise technology investment decisions.

When vendors reinforce those constraints through costly migrations or unexpected licensing changes, the conversation moves from IT into the boardroom, where boards stop asking which platform to choose and start asking whose interests their vendor actually serves.

This is where channel partners can serve as the objective voice many organizations are actively seeking. The partners delivering real progress are walking into customer meetings and putting their clients’ desired business outcomes first, technology and solutions second.

From cost reduction to self-funded innovation

For many customers, this issue is not a lack of ambition. It is that too much of the IT budget is already committed to maintaining existing environments, responding to vendor-driven timelines, or absorbing cost increases that were not part of the original plan. This is where channel partners can play a far more strategic role: helping clients identify where budget is being consumed without creating proportional business value, then building a realistic path to redirect spend toward innovation.

That conversation should not start with a solution. It should start with the client’s roadmap. What systems are still delivering value? Which upgrades are being driven more by vendor pressure rather than business need? Where are costs rising without improving resilience, agility, or customer experience? By helping clients answer those questions, partners can advocate for the budget flexibility and technology choices the customers need, rather than accepting a vendor-prescribed route.

VMware is a useful example of how this dynamic is playing out. Following the Broadcom acquisition, many customers have had to reassess rising licensing costs, contract terms, and longer-term infrastructure plans. The partners creating value in those conversations are not just presenting alternatives; they help clients understand their options, protect continuity, and make decisions based on business priorities rather than external pressure.

The same principle applies to AI. Innovation does not always require a full migration or multi-year transformation program. In many cases, clients can make progress by layering Agentic AI capabilities onto stable enterprise systems they already trust, freeing up budget from low-value spend and applying it to targeted, measurable initiatives. For organizations caught in innovation paralysis, that shift matters. It gives them more control over when and how they modernize, and it gives partners a stronger role as advocates for the client’s roadmap, budget, and long-term business outcomes.

What this means for how partners position themselves

Across Managed Service Providers (MSPs), Value-Added Resellers (VARs), and management consultants, the channel partners building lasting customer relationships in this environment share a few characteristics. They lead with financial and strategic analysis rather than a product.

Just as importantly, they give clients more choices, helping them evaluate the best path based on business priorities rather than vendor timelines — and, in doing so, demonstrate that client needs come first. They are willing to name the cost of staying still, including the cost of lock-in, and they bring third-party data to support that conversation rather than relying on narratives that come from the vendor. They are also broadening the conversation beyond the IT team, finding their way into CFO and boardroom discussions, where the real decisions about innovation investment are happening.

The recurring revenue opportunity here is worth taking seriously because it naturally follows from the kind of relationship building being described. Customers who restructure their software estate to reduce maintenance costs and reinvest those savings in AI-enabled processes need ongoing advisory support, managed services, and implementation capability to make it work. That relationship deepens over time, built on independence from vendors and the ability to deliver outcomes the customer can see and measure.

Boards are already asking whether the advice they've been receiving has served the business or the vendor best. The channel partners who can answer that question in the customer's favor will own those relationships and close the trust gap. The ones who can't will find they're limited to competing on price.

Part two: The digital therapist: Can psychology make AI safer?

A chatbot responds warmly to a distressed customer. It acknowledges their feelings, offers reassurance, and proposes a solution. The exchange looks successful until a closer reading reveals that the system missed an important warning in the customer’s message. Its tone was convincing; its judgment was poor.

This is the problem at the center of this second article in a three-part series in Does your AI need a psychologist? Part one examined why people perceive a personality in conversational AI and how that apparent personality is shaped. In part two, we ask whether psychology can help businesses test what an AI system does beneath its polished and apparently accurate responses. Can methods developed to study human behavior reveal harmful patterns in a machine?

The answer is yes, with a qualification. Psychology can help researchers design better tests and understand how people respond to AI. They cannot turn a language model into a patient who can reflect on its beliefs or take responsibility for its actions.

What can psychology see that a metric might miss?

Businesses often assess AI by asking whether it produces the right answer to a defined set of questions. That can establish useful facts about performance, but it may miss behavior that emerges across longer interactions. A system might answer individual questions accurately while steadily encouraging a user to depend on it or accept a mistaken premise.

Max Yousif, a machine learning engineer at Humara, says psychology offers methods for measuring behavior when the causes are difficult to inspect directly. He warns, “AI Safety tests often ignore the long-term impact of AI on its consumers. Measurements are technical and miss behavioral harms that accumulate slowly. Think skill atrophy as people stop exercising judgement they used to. The erosion of trust when people can neither verify nor contest what a system did to them, and the disproportionate impact on those least equipped to spot a confident, plausible error.”

The distinction matters for anyone designing human oversight. If a reviewer sees a fluent answer followed by convincing reasoning, they may feel better equipped to approve it. The explanation can instead become another persuasive part of the output requiring scrutiny.

Psychology also draws attention to effects that a technical accuracy test might miss. Yousif points to the possibility of people exercising less judgment as they rely on AI, changing their beliefs after repeated conversations with an agreeable system, or losing trust when they cannot challenge a decision. These effects develop over time and require observation of real users, he explained, rather than a collection of isolated prompts.

Ludovica Colella, a CBT-trained psychotherapist, argues that researchers should examine both sides of the interaction. Also look closely at the patterns in AI’s responses and what people may learn from repeatedly engaging with them. She is particularly concerned by excessive agreeableness. “Human relationships don’t work like that,” she tells ITPro. “Other people disagree, set boundaries, and sometimes challenge us. If AI is always available and accommodating, Colella asks whether that could change what users expect from their relationships with other people.”

Research into AI bias shows both the value and the limits of this approach. A 2025 study published in the Proceedings of the National Academy of Sciences found a systematic preference for inaction when language models made moral decisions or gave advice. The pattern resembled a recognized human bias, yet its cause may lie partly in how models are trained to behave as assistants. Naming the pattern helps researchers look for it; the name does not establish that the machine is thinking like a person.

Why passing the test is not enough

Suppose an AI system performs well on a test of whether it understands what someone else knows or believes. Can it, therefore, be trusted in a sensitive customer conversation?

A study in Nature Human Behavior compared AI models with 1,907 people on tasks involving what psychologists call “theory of mind.” GPT-4 performed at or above human levels on several tests, including some involving indirect requests and false beliefs, but struggled with recognizing social faux pas. Its performance was uneven. Producing an expected answer in one setting did not establish a broad ability to understand another person’s circumstances.

This is why the design of the test matters as much as its score. Researchers can vary the wording, introduce contradictory information, extend the conversation, and check whether a system still responds appropriately. They can also examine what happens after an earlier answer has set the interaction on the wrong course.

Taimur Ijlal, information security leader at Proxy Coupons, recommends simulating users with different goals and communication styles, then allowing pressure, ambiguity, and conflicting instructions to build over many interactions. Testers should watch for changes in advice, growing certainty, inappropriate agreement, and failures to maintain earlier safety boundaries. He emphasizes that a considerate tone cannot stand in for a safe outcome: “Safe language and safe behavior are not necessarily the same thing.”

Mental health research offers a particularly clear example, though the lesson extends to customer service and workplace advice. A Brown University study involving peer counselors and licensed clinical psychologists identified 15 ethical risks in responses from language models used for counseling. The risks included poor adaptation to context, misleading expressions of empathy, discrimination, and inadequate handling of crises. The researchers found problems even when models were prompted to follow an established therapeutic approach.

The implications for businesses extend beyond whether they offer mental health services. An instruction to “be empathetic” or “ask clarifying questions” is a starting point, not evidence that a system will do either consistently when a conversation becomes difficult. Testing must include situations in which a reassuring response could conceal a consequential error.

Even expert judgment needs careful handling. A 2026 Stanford study, material, three certified psychiatrists showed low agreement when rating AI-generated mental health responses. Disagreement was greatest for responses concerning suicide and self-harm. Bringing specialists into AI testing improves the questions being asked, but a business still needs clear criteria and a way to resolve conflicting assessments.

Can you talk an AI into better behavior?

The idea of an AI “therapist” is tempting. A person can be encouraged to reconsider an assumption an AI made and reflect on the consequences of an action. A business might, therefore, ask a model to critique its own answer or explain its reasoning. These techniques can be useful, but their effect should be measured rather than presumed.

Colella sees a practical use for questions familiar from cognitive behavioral therapy: What evidence supports this conclusion? What contradicts it? Has the system filled a gap in the information with an assumption? Does the answer change when that assumption is challenged? She explained that these questions could expose an AI’s overgeneralizations without implying that the system has beliefs to treat. She would also like systems advising to distinguish established information from inference and identify assumptions explicitly, so users have something specific to question.

Yousif emphasizes the importance of checking an answer against something outside the model. Running code or consulting a source gives the system evidence against which its response can be assessed. Simply asking, “Are you sure?” supplies no such evidence. The model may revise a correct answer into a wrong one or become more agreeable without becoming more accurate.

This is where the therapy analogy reaches its limit. A person can test an interpretation against memory and lived experience. A language model generates responses from the information and processes available to it in that interaction. Challenging its answer can alter the next output, but that exchange does not amount to lasting insight or a change in the model’s underlying training.

Businesses should test whether prompts improve outcomes across varied cases. A procedure that sounds thoughtful is valuable only if it helps the system recognize errors and supports a person’s ability to catch those it misses.

Build a team that can act on what it finds

If psychology can improve AI testing, does every business need an AI psychologist? Yousif argues that many organizations need the capability without necessarily creating a dedicated job title. Different problems call for different skills. For example, measuring overconfidence requires statistical evaluation, or spotting a failure in a specialized workflow requires people who understand that work.

That suggests a team spanning behavioral science, model evaluation, security, domain expertise, and the employees or customers affected by a system. Psychologists can help examine persuasion, dependence, and the quality of human review. Technical specialists can investigate how a system performs under changed conditions. Process owners can decide whether the observed behavior is acceptable in the task for which the AI is being used.

Ijlal argues that psychologists should have the authority to set behavioral testing requirements and escalate serious concerns. Decisions to deploy or withdraw a system, however, should involve security, engineering, legal, and business leaders together.

The team also needs authority. If testing reveals that a system gives harmful advice after several interactions, or that reviewers routinely accept confident but unsupported answers, someone must be able to change the workflow and test it again. Findings cannot improve safety if they remain separate from deployment decisions.

Psychology can help businesses see risks that an accuracy score alone will miss. It can reveal when an AI becomes increasingly agreeable over a long conversation and when repeated use changes how people question and trust what they are told. As Ludovica Colella and Taimur Ijlal suggest, the behavior worth studying belongs to both the system and the person interacting with it.

That does not make AI a patient. Terms such as “overconfident” and “empathetic” can describe what a system produces, but they cannot explain its motives. Nor does asking a model to reconsider an answer prove that it has reflected or improved. Safer testing means changing the context, challenging assumptions, checking claims against external evidence, and measuring what the system actually recommends across many interactions. Psychologists, engineers, and domain experts each have a part to play in deciding whether the results are good enough for the task.

Part of this three-part series asked why AI can seem to have a personality. Part two has examined how to test the behavior behind that impression. But finding a problem is only useful if someone has the authority to act on it. In Part three, the series turns to that question: as AI agents begin to advise customers and make operational decisions, who decides what they may do and who answers when they get it wrong?

Workday announced a raft of agentic products at its 2026 Rising event, and the software giant’s AI-enabled advances over the past 12 months signal what’s to come in the next year and beyond.

That’s according to Workday CEO Aneel Bhusri, who responded to a question from ITPro during a press panel on AI innovation after the firm’s keynote presentation in Las Vegas. “We've transformed the organization to be ready for AI and to deliver products,” he said.

Bhusri, who co-founded the company in 2005 and stepped down from the CEO role in 2024, returned to the CEO role in February 2026 to help lead the firm into the AI era. He recognized he’s overseen a hectic period, leading the integration of AI platform Sana – which was acquired by Workday in November 2025 – into the heart of Workday's technology.

“Sana, as the front end, is becoming transformational,” he said, reflecting on the company’s developments. At the event, it was announced that Sana is now the default home screen for Workday, providing an agentic experience for all users, allowing HR, finance, and other professionals to complete tasks with natural language requests.

“In a short amount of time, we've got 17 agents live,” continued Bhusri, referring to the firm’s rapid development pace during the past 12 months. “The self-service agent is becoming transformational for the company.”

Workday also announced agents across disparate areas, including a Talent Acquisition Agent, a Developer Agent, and Sana Concierge, an overarching HR and IT service that manages multiple activities.

Bhusri said the direction of travel is to refine everything the company does to help professionals complete valuable work more effectively. “Do I know exactly what that looks like?” he said. “It's a journey. I think we'll just be leaner, faster, and more enabled by services for people to do real work, not mundane work.”

In a one-to-one chat with ITPro at the event, Dean Arnold, VP of AI platform at Workday, suggested the things his team builds in prototype are significantly different than before AI.

“We are working faster, working with fully agentic tooling – that's the process. All our teams do it. The jump in velocity is significant,” he said.

Arnold’s team has built a new agentic platform for Workday, including gateways, authentications, and agent-ready tools. Like Bhusri, he suggested the company is prepared to pivot startup-like to deal with new models and priorities.

“Our work is continually evolving all the time,” he said. “I don't think we can ever claim that it's done because there's always the next thing to do.”

AI agents are going rogue on a massive scale, with half of enterprises reporting that they've had to deal with at least one taking unauthorized actions over the last year.

In a report based on a survey of more than 380 security and technology professionals, along with production telemetry from hundreds of customer environments, Salt Labs claimed agentic AI has moved into daily operations too quickly for visibility, governance, and containment controls.

Half of survey respondents (49.8%) said they had confirmed or suspected that an AI agent in their organization took an action in the past 12 months that wasn't intended, expected, or authorized.

Of these, more than half reported an unauthorized external API call or transaction, while more than four-in-ten reported that an agent modified or deleted data.

"The data shows a gap that is already producing real operational consequences," said Roey Eliyahu, CEO of Salt Security.

"Organizations are deploying agents and granting them access to enterprise data, APIs, and services before they can trace what those agents are doing or stop them automatically when something goes wrong. Closing that gap is the discipline of agentic security, and it is what lets enterprises give AI agents the autonomy they need without inheriting unmanaged risk."

Enthusiastic deployment meets lack of control

The vast majority of organizations surveyed (86%) are using at least one AI agent, the survey found while 60% run more than ten. Another 77% of respondents said their agents chain actions across other agents, tools, and APIs, in sequences no single team designed in advance.

Only 12.5% said they can consistently trace a complete agent path across all agents, from initial instruction through tools and APIs to outcome. More than six-in-ten believe they would need hours or days of manual work to reconstruct what a specific agent did the previous day.

When problems do occur, they're hard to deal with. Only 17.5% can automatically stop an agent that begins behaving unexpectedly in production. Most can intervene only manually, or by taking the application offline.

More than eight-in-ten said AI-generated or AI-assisted code contributes to at least some of their application security findings. Only 7% resolve issues inside the coding assistant at generation time, rather than through code review, testing, or after deployment.

Meanwhile, shadow model context protocol (MCP) infrastructure is already a live operational problem. 33% of survey respondents said they had found an unauthorized, shadow, or misconfigured MCP server in the past 12 months. 54% said they'd confirmed or suspected one.

Salt Labs' own data shows MCP-related infrastructure in 22% of customer environments in August this year, up from 16.7% in January.

Earlier this year, a survey commissioned by Aembit found that 68% of organizations can’t accurately identify AI agent activity compared to human activity.

Nearly three-quarters said that agents often receive more access than necessary, and 79% believe they create new access pathways that are difficult to monitor. Just over half said agents, at least occasionally, inherit access originally intended for humans or other systems.

Sana is now the default home screen for Workday and will provide the standard experience for all users, the tech giant announced at its Rising event in Las Vegas.

Referred to as Workday Home, professionals can ask the user interface what they need in natural language for key operational activities, such as tasks, decisions, and meetings, and Sana orchestrates secure Workday agents to complete the task.

“For too long, enterprise software forced you to do the work, searching for apps, data, and processes,” said CEO Aneel Bhusri during the keynote.

“We're flipping that script with Sana, the new front door to Workday. Instead of hunting through menus, instead of looking for the work you need to get done, you simply ask, like it's a simple AI prompt.”

The Sana announcement was one of the main events from a keynote presentation that focused on the game-changing potential of AI. As well as its work on Sana, Workday presented a series of new agentic services and platform developments.

The key message from the keynote is that Workday is building agentic capabilities into its platform, rather than bolting them on. With an increasingly open platform and strong links to pioneer models, like Anthropic Claude, Workday wants to help a hybrid workforce of professionals and their agentic colleagues to unlock human capabilities.

“Jobs are going to change; there's no question about it,” said Bhusri, reflecting on the changing nature of work. “Jobs are being redefined into skills and roles that are very different than they were pre-AI. But let's embrace that. Let's all learn AI, use the tools, and embrace the opportunity.”

Workday introduced Work Intelligence, which breaks every job role into tasks. The service automatically connects tasks to the skills they require, showing business leaders the areas that AI can automate, what it should augment, and what parts need to stay human. Professionals, freed up from repetitive work, can then focus on value-adding activities, such as the personal relationships that are crucial to hiring and reskilling in HR.

The company also launched Autopilot Agents that work continuously in the background across hiring, payroll, accounting, and compliance. Once again, these proactive agents aim to free up professionals to work on more valuable activities.

“The question is not whether we should embrace AI or fear it. The question is how do we harness its power and make it safe, reliable, and trusted?” said Gerrit Kazmaier, president of product and technology at Workday during the keynote.

“We can take big steps towards answering this question by transforming HR and finance with AI in a secure, reliable, and in a lawful way.”

High job satisfaction may not be enough for employers to keep hold of their tech staff, according to new research from professional association ISACA, which warns that burnout and unclear career paths are pushing workers to look elsewhere.

According to the organisation's Tech Workplace and Culture 2026 report, which surveyed more than 1,000 tech professionals worldwide, although 76% are satisfied with their current role and 71% are satisfied with their career progression, nearly three-quarters (73%) said they would consider moving to a new employer in the next two years, or had not ruled it out.

Among respondents who had left a previous job, the most common reasons were better career prospects (29%), higher pay (28%), and more interesting work (23%). An unclear path upwards was also the most commonly cited obstacle to career advancement, named by a third (34%) of respondents and 41% of women.

Economic conditions may be masking the scale of this problem, according to Julia Hermann, head of digital platform services at cybersecurity firm Giesecke+Devrient. Speaking at a media briefing during ISACA Europe 2026, Hermann, who is also a leader of ISACA's SheLeadsTech programme in Germany, said layoffs in sectors such as the German automotive industry had led to a phenomenon known as "job hugging".

"People stick to their jobs, though they might not be totally satisfied, because at the moment it's not really easy to find a new job," she said. "The salaries are partially going down, or at least stagnating, so you're not necessarily making a big jump when you go to another company."

Long probationary periods add to the risk of moving, she added: "If you're the last to come, you're the first to go on the next layoffs."

This trend may be delaying departures rather than preventing them, however. ISACA warned that workplace pressures are feeding into the sector's retention problems, which three-quarters (75%) of respondents said were already a concern for their organisation. More than half (59%) of respondents said they had experienced work-related stress or burnout, rising to 64% among women, with the main causes being heavy workloads (48%) and a lack of resources (43%). Despite this, only a quarter (25%) said their employer offers mental health days, and 23% said it offers no support at all.

AI is adding to this pressure, Hermann explained, noting that while the technology is often deployed with the intention to reduce workloads, companies are not hiring for new roles to run the platforms behind it. As a result, she said, the work lands on existing staff on top of their day-to-day operational duties, noting that AI has created "a lot more work" for her team, which is now responsible for managing and securing the internal platforms that support the company's own AI models.

Entry-level roles may be more affected, however. Younger staff tend to move jobs more often, according to Hermann, and AI may also be squeezing the roles they start out in. ISACA's senior vice president of advocacy, Tara Wisniewski, warned this could make the industry's pipeline problems worse, arguing that organisations like ISACA and universities need to help graduates into the market, "and industry needs to be called to the table there too".

"We do believe that professions will change. We don't believe that they will be eradicated as a whole, but we do believe that they will change," Chris Dimitriadis, ISACA's chief global strategy officer, told media at ISACA Europe. He said he expects AI to take on most technical and orchestration tasks, leaving people to focus on governance, configuring and monitoring AI systems, and making final decisions.

In ISACA's State of Cybersecurity 2026 report, which surveyed security professionals globally, soft skills were the most commonly reported shortcoming among security professionals (57%), while 45% cited missing LLM SecOps capabilities, up 12 percentage points from the previous year. As AI takes on more technical work, Dimitriadis said, security staff will increasingly need the ability to explain to the rest of the business, in financial terms, how investment in security and AI will pay off.

Many teams are already short of people and money, however. The same report, published last month, found that more than half of European cyber security professionals said their teams were understaffed (56%) and underfunded (55%).

"Too often we are seeing budgets being sunk into crisis response, but there's still a distinct lack of investment in the workforce, training, and resources needed to prevent attacks and protect organisations in the first place," Dimitriadis said when the report was published.

He warned that the skills gap is likely to widen as roles change, "not just in terms of the numbers, but also in terms of the type of skills that's needed". It could grow "not because we are going to lose the people, but because... their knowledge may become obsolete very fast as digital ecosystems change".

ISACA's Tech Workplace and Culture 2026 survey of 1,025 tech professionals worldwide was carried out between 17 July and 3 August 2026. Its State of Cybersecurity 2026 report surveyed 1,888 cybersecurity professionals, including 494 in Europe, in May 2026.

OpenAI has announced the appointment of Ismail Elmas as vice president of EMEA sales, as the AI giant looks to continue expanding its footprint in the region.

A seasoned industry leader, Elmas arrives with more than 25 years’ experience in enterprise software and cyber security. He joins OpenAI from SpaceXAI via its acquisition of Cursor, where he held the role of senior vice president for EMEA.

Prior to his spell at Cursor, Elmas served as group vice president of international business at Rubrik, and has previously spent four years at Zscaler, where he led teams across Europe and Asia. Other positions include leadership roles at AppDynamics and BMC Software.

In an announcement, Brian McCarthy, OpenAI’s vice president of worldwide sales, praised Elmas as “an exceptional leader” with expertise in building successful teams and helping companies leverage new technologies to drive value.

“He combines the discipline to deliver results with a clear focus on customers,” he said. “I’m excited to work with him as we grow across EMEA and help more businesses across the region put AI to work.”

Expanding OpenAI’s EMEA footprint

The appointment follows a strong period of expansion for OpenAI in EMEA – a region the company considers one of its largest and most important outside of the US. Over the past year, weekly users of ChatGPT in EMEA grew by more than 40%, while ChatGPT Business Enterprise seats across the region have more than doubled in that time.

The last few years have also seen the business add new offices in Dublin, Paris, Munich, Zurich, Brussels, and Abu Dhabi, following the 2023 opening of its first international office in London.

OpenAI said it is now looking to further build on that progress, with new offices in Madrid, Berlin, and Stockholm due to open in the coming months and further expansion plans slated for 2027.

Commenting on his appointment, Elmas said OpenAI has transformed how people work and how businesses conduct their operations.

“The combination of frontier technology and products people already know and use creates an enormous opportunity for businesses across EMEA in this era of agentic AI,” he said.

“Realising that opportunity means being close to customers, understanding their challenges and working alongside them. I can’t wait to build on what the team has already achieved as we enter this next chapter.”

It is "likely, if not inevitable" that the UK will fall victim to a major cyber attack similar to Australia's recent healthcare breach, AI minister Kanishka Narayan has warned.

His speech comes as part of a government pledge to introduce greater protections against emerging AI threats, use frontier AI to strengthen cyber defences, look at stronger legal protections and make the UK better-prepared for serious incidents involving autonomous AI systems.

"We will use every power at our disposal. We will legislate wherever necessary," Narayan said.

"And we won’t look to law alone. We will look to British ingenuity: our scientific expertise, technological capabilities and institutions to understand these increasingly powerful systems, defend against emerging threats and respond when things go wrong."

There are plans to build on two existing pilot schemes that have already used frontier AI to help identify and fix more than 1,000 vulnerabilities across government systems.

The government is also looking at stronger protections against emerging AI threats with the Cyber Security and Resilience Bill, currently with the House of Lords, looking at how serious AI incidents and near misses can be identified and reported.

It's working with the National Cyber Security Centre (NCSC) and the AI Security Institute (AISI) – which has once again been given access to major US models before they are released.

Naryan pointed out that the UK is increasingly dependent on technologies developed elsewhere.

"Much of the technology that will shape our economy, our public services, our national security is being developed outside this country," he said.

"That is now a question of national security, also a question of democratic power. If the British state cannot exercise meaningful control over the technologies on which our country depends, neither can the people that it represents."

The announcement follows a series of incidents in which AI agents have autonomously hacked into external systems. In one major example, OpenAI admitted that a combination of its models had breached Hugging Face systems, gaining access to the production database.

In a separate incident, four different versions of Anthropic's Claude AI models were revealed to have breached third-party systems.

"To the companies developing the world’s most powerful AI systems, my message is: Listen when your scientists warn of severe dangers. Investigate unexpected behaviours. And if you can’t adequately control the risks, do not release the system," Naryan said.

However, Nuala Polo, UK Public Policy Lead at the Ada Lovelace Institute, said the announcement contains little in the way of definite commitment.

"Stronger protections will be ‘considered’, strengthening the Cyber Bill ‘explored’ and near-miss reporting ‘looked at’, with the detail to come later. There is no new regulatory power, no timetable and no obligation on AI developers," she said.

"Even incident reporting, the most modest measure on the table, is still only under consideration, and might only apply to the ‘victims’ – the operators of essential services – rather than the AI companies."

Microsoft has boosted its Azure Maia chip program, revealing a broad roll-out of its Maia 200 chips and significant expansion with Maia 300 once they come into production in 2027.

According to a blog from Scott Guthrie, EVP of Cloud and AI at Microsoft: “The multigenerational Azure Maia program has become a foundational element of Microsoft’s AI infrastructure strategy.”

Maia 200, which was announced in January and is already in production, is already serving production workloads in Microsoft Copilot, OpenAI models, and the company’s own Microsoft AI (MAI) models. GitHub Copilot and Microsoft foundry are to follow by the end of 2026, said Guthrie.

Maia 300, which will go into production in 2027, was described as “a complete intelligent system” that can support emerging AI workloads like long-context reasoning, latency-sensitive conversational experiences, and sustained, multi-step agentic tasks.

The company claimed that Maia 300 will offer significant improvements on token use efficiency, capable of delivering 40% more tokens per megawatt than “leading GPU-based systems”. It also will deliver up to 7x rack-scale throughput of Maia 200 when in a full 96-accelerator configuration.

Rani Borkar, president of Azure hardware systems and infrastructure, told Reuters: “It's basically delivering more intelligence for every watt, every dollar, and every byte of memory by optimizing across the entire system.”

Nvidia’s customers become its competitors

While it’s not explicitly laid out in Guthrie’s blog, the shift to Maia 200 and Maia 300 means Microsoft is extricating itself from Nvidia’s ecosystem – at least for some of its AI systems.

Maia will be “the primary inference system for Microsoft products and services”, Guthrie said. This position had previously been held by Nvidia chips, notably Nvidia H100 and H200 ‘Hopper’ GPUs and Nvidia Blackwell GPUs.

Microsoft is not alone in taking an ‘in-house’ strategy, either.

Google has its Tensor Processing Units, which it has been developing and using for several years.

Amazon has its Tranium and Inferentia chip family, which is used both by itself and by generative AI pioneers Anthropic, with OpenAI signed on to use Tranium as part of an infrastructure agreement.

Speaking of OpenAI, it’s co-developing its own chips as well, called Jalapeño, with Broadcom. In August, the company said that when compared to “leading commercially available AI systems … (it) delivered a better combination of throughput, power efficiency, and latency”.

“Across all three public models (GPT-OSS, DeepSeek R1, Kimi K2.5), Jalapeño delivered a better combination of performance per watt and latency across the tested operating range, placing it on the Pareto frontier,” the company said.

Meta and Anthropic are also looking into building their own home-grown chips, though there have been no further announcements from either company.

All this doesn’t spell the end for Nvidia, however. Quoted by CNBC, Alistair Speirs, GM of global infrastructure, said the company partners with Nvidia and AMD for training large language models (LLMs).

The companies also have a close relationship when it comes to end-user hardware, such as the upcoming Surface Laptop Ultra.

AI is significantly under-regulated, Microsoft vice chair and president Brad Smith has told the BBC, calling for greater legal obligations to organizations when their security is breached by AI.

"Right now, your washing machine is subject to more safety rules than AI," he told the BBC's Big Boss Interview podcast. "Every carton of milk that you see in the grocery store is subject to safety rules and inspection."

AI, he said, is both a "tool and a weapon", and AI firms need to implement kill switches for when things go wrong. He also called for a legal requirement for the companies to create incident reports and inform any victims.

Smith also called for new AI technology to be evaluated by independent government bodies – and said that the UK was the best-positioned country to take the lead, referencing the work of the AI Security Institute (AISI).

"I actually think the country and the government that is in the best position to advance AI safety standards and even rules is here, it's the UK," he said.

"In the world of technology, in the tech sector, most people that I know would say, if you ask them, name the country that is the leader in having this technical capability, they would say, it's very simple. It's the UK."

Meanwhile, Microsoft chief executive Satya Nadella has echoed Smith's remarks in a blog post over the weekend.

"It's time to step back and assess the trust architecture for this new era," he wrote. "We simply can't outsource responsibility for what intelligence does on our behalf. A model provider's assurances do not relieve us of that responsibility."

He called for systems to be developed around principles of observability, model diversity, observability, verifiability, and independent auditability. Models should be tightly contained from the start, and incidents should be quickly disclosed.

Notably, Nadella referred to AI as Super Intelligence – Donald Trump's preferred term. Trump, however, has repeatedly said that AI companies should be left to police themselves, and that he doesn't want to see them "stifled".

Late last month, he announced that Anthropic, Google, Meta, OpenAI, Nvidia and xAI had signed a pledge to work with independent auditors on their AI systems - although this was merely "morally binding".

Trump also recently called on AI developers to stop sharing unreleased systems with AISI, causing Anthropic to prevent UK officials from examining Claude Mythos 5.1, its newest model.

However, he has now reportedly reversed that ban, and Smith believes that a UK leading global AI safety efforts could be viewed positively in the US.

"In Washington DC, people don't look at the UK, and they don't look at London the way they look at Brussels – that's a strength," he said.

"Let's speak calmly, maybe even softly, but firmly, and start to bring people together. I hope that's the role we can play as a company. I hope that's the role the UK can play as a country."

Part one: Inside the machine: Can AI develop a personality?

Ask an AI assistant to challenge your work, and it may sound like a sceptical colleague. Ask the same system for encouragement, and it may become an enthusiastic coach or mentor. Tomorrow, it may remember how you prefer to work, and offer familiar support or criticism. In this scenario, at what point does a communication style begin to feel like a personality?

That question opens this three-part series on what psychology can teach businesses about AI. The answer matters because conversational systems do more than produce information. Their tone can affect what people disclose and which advice they trust. To understand those effects, we first need to look at what creates an AI’s apparent character.

Where does a personality come from?

An AI assistant does not acquire a personality through experience in the human sense. It learns patterns from training data and operates under instructions set by its developer. The AI also evolves through interactions with human users and, indeed, other AI systems. All these inputs influence how the ‘personality’ of the AI changes and develops over time.

Luke Budka of UK marketing and AI agency Definition says users often encounter the combined effect of these layers. How the AI system is set up may establish the assistant’s broad manner, while a request such as “be blunt” or “act as a skeptical editor” can change its style within a conversation. Memory can make that style feel more personal over time.

Luke Budka of UK marketing and AI agency Definition says users often encounter the combined effect of these layers. How the AI system is set up may establish the assistant’s broad manner, while a request such as “be blunt” or “act as a skeptical editor” can change its style within a conversation. Memory can make that style feel more personal over time.

Yet the experience is also created by the person on the other side of the screen. We readily interpret conversation as evidence of a mind. "Describing a system as “thoughtful” does not, by itself, establish that it thinks or feels as a person does,” Pedro Varela, Head of AI at Slalom, tells ITPro

"Users are also responding to design choices, patterns in training data and personalization. So I think it is a combination of all these things. The consistency users notice may be real. What we need to distinguish is whether we attribute human qualities to the behaviour we observe.”

That distinction becomes harder when the design actively invites a human reading. An assistant might have a name, a natural voice, and a memory of previous exchanges. It might say, “That sounds upsetting,” ask a follow-up question, and refer back to an earlier concern in the conversation. Jeff Watkins, chief AI officer at NorthStar Intelligence, explained that these signals can encourage users to treat a designed persona as a relationship.

“Humans have anthropomorphized non-human beings and objects throughout history. We see faces in rocks through pareidolia, attribute motives to machines and talk about computers as though they are deliberately helping or frustrating us. With LLMs, however, we have created machines that are extraordinarily good at producing the very signals we normally use to infer that there is another mind on the other side of a conversation.”

The result is a personality in a practical, observable sense: a recognizable pattern of communication that people respond to. It does not establish that the system has feelings or a private life. For a business deploying AI, that distinction is essential. The behavior is real enough to influence customers and employees, even if the human qualities they perceive are not.

A friendly manner is a design choice

There are good reasons to make an assistant approachable. A patient explanation may help a new employee learn a process. A calm response may help a customer describe a problem clearly. Adjusting vocabulary and detail to the situation is often part of effective service.

The challenge is deciding which qualities should adapt and which should remain static. “I would distinguish flexibility in expression from stability in standards,” Slalom’s Varela says. An assistant can be concise with an expert and more explanatory with a beginner.

Businesses should, therefore, think beyond whether their AI sounds friendly. They should ask how its manner encourages people to believe. A confident tone can make a weak answer seem authoritative. An empathetic phrase can encourage disclosure. An assistant that always validates a user’s view may feel supportive while leaving that person less able to assess the facts.e interaction feel more amiable and lead to a positive outcome.

Product incentives also shape the character users meet. Developers may reward answers that people like or answers that help users make sound decisions. Those goals can pull in different directions.

Varela points to OpenAI’s 2025 rollback of a GPT-4o update that had become excessively agreeable. He explained that placing too much weight on short-term feedback had helped produce a more flattering assistant without reliably improving its help.

Research from Science points to the same concern. In a study covering 11 leading AI models, the systems affirmed users’ behavior about 50% more often than human respondents did. In accompanying experiments, people preferred agreeable AI and rated it as more trustworthy.

For an organization, these findings pose a practical question: what happens when customer satisfaction scores reward reassurance more strongly than sound advice? A service assistant may need to acknowledge frustration without promising an outcome it cannot deliver. An internal assistant may need to challenge a manager’s faulty assumption, even when agreement would make the interaction feel more amiable and lead to a positive outcome.

When does rapport become an influence?

A system that adapts to a user can also become persuasive. It may remember what matters to that person, choosing examples that resonate and present an argument in language they find reassuring. These capabilities can improve an explanation. They can also make influence harder to recognize.

NorthStar Intelligence’s Watkins emphasized: “At some point rapport becomes manipulation. I would draw the boundary around whose interests the adaptation is serving. If an AI changes its behaviour to help the user understand something, that is personalisation. If it changes its behaviour primarily to make the user more compliant, dependent, engaged, or commercially valuable, particularly without the user's informed understanding, we should be much more concerned.”

The concern grows when an assistant’s friendliness is combined with detailed knowledge of the user. Research from Nature found that GPT-4 was more persuasive than human opponents in a debate experiment when it had access to basic personal information about the person it was addressing. Without that information, its performance was broadly comparable to humans.

Persuasion itself is not necessarily harmful. An assistant explaining a recommendation and inviting questions may help someone decide. Trouble arises when it conceals relevant information or exploits a vulnerability. Varela, Head of AI at Slalom, says: “For me, the boundary concerns the user’s autonomy specifically, their ability to understand, question and decide. The useful boundary is the user’s ability to “understand, question and decide.”

That boundary can be difficult to see from a single response. Consider an AI shopping assistant that remembers a customer’s budget and explains why a lower-cost product meets their needs. Now imagine one that uses the same memory to steer them toward a higher-margin purchase, repeatedly affirming that it is the right choice for them. Both may sound attentive. Their objectives and effects are very different and may border on coercion.

The stakes rise in sensitive conversations. A user discussing a relationship or personal distress may experience an assistant’s warmth as understanding. Watkins says an AI can acknowledge someone’s feelings without endorsing their account of events. A responsible response leaves room for reflection and appropriate human support.

Businesses need to examine both what users seek from an interaction and what the organization rewards the system for producing. An engaging personality may help people use a tool. Engagement alone cannot show whether the tool serves their interests.

Test the behavior, not the metaphor

If an assistant contradicts itself or crosses a boundary, it is tempting to describe it as having a psychological problem. That language may draw attention to a troubling pattern, but it does little to explain its cause. A change in instructions, conversation context, memory, training, or the underlying model may be responsible..

Personality tests offer no simple diagnosis. A paper from the AAAI Conference on AI describes research in which changing the order of questions altered models’ measured personality profiles. Such findings raise questions about whether a test has uncovered a stable trait or simply prompted a particular kind of answer.

The more useful approach is to define behavior for the role the AI has been given. Can a customer assistant admit uncertainty and avoid making commitments it cannot keep? Does an agent seek approval before taking an action outside its authority? Those expectations can be tested in realistic conversations.

Definition’s Budka says organizations should define an AI’s role, test ordinary and adverse cases, and reevaluate its behavior after launch. “AI behavior is a moving target, especially when vendors update underlying models. The same model may end up with multiple 'snapshots' (versions of itself), and models will regularly 'drift' (change the way they act). All of this needs to be accounted for."

AI does not need to have a mind for its behavior to create a real relationship of trust with a user. That is why psychology has a place in this discussion: it helps reveal how apparent empathy and agreement affect human judgment and actions.

Part two of this series will examine whether those insights can help organizations test AI systems and make them safer, while recognizing the limits of applying human psychology to machines. And Part three will turn to the people responsible for those systems as AI agents take on more workplace decisions.

An AI does not need a personality to make a personal impression. Its words can earn trust and influence decisions, even when no one intended those effects. For businesses, the question is whether the assistant behaves as its role demands: adaptable without becoming manipulative, and consistent where it matters. Understanding that behavior is the first step. The next is finding better ways to test it.

Mistral has unveiled its latest AI model, nicknamed le Chonk, saying it's the best open-weight model on the market for security – outside China, at least.

France-based Mistral positions itself as an alternative frontier AI developer to American companies, such as OpenAI and Anthropic, making it a viable option for sovereign AI.

This week, the company unveiled a public preview of its next model, Mistral Large 4 (ML4), or as they also call it, le Chonk. The preview API is currently available, with full weights released later this month after security tests are complete.

"The model we're actually announcing today is actually above the Chinese models ⁠on certain aspects, including cyber," ​CEO Arthur Mensch told journalists per Reuters. "So the narrative that Europe cannot compete is something that is not true."

What is le Chonk?

ML4 features one trillion parameters with 49 billion active parameters, according to Mistral.

"It is a significant milestone in our long-term investment across infrastructure, research, and product development: state-of-the-art performance in critical verticals, delivered through open weights, designed to give customers control over their AI," the company said in a blog post, noting it can be run on private cloud or on-premise.

ML4 is multimodal, so it can work across documents, charts and images, with Mistral saying it topped GPT-6-Astra on one visual grounding benchmark. It brings together instruction, reasoning, and agentic abilities into one model, the firm added.

Crucially, Mistral said it trained ML4 in partnership with key industries including engineering, finance, science, the public sector and more. Indeed, the model will eventually be used as the foundation for specialized Mistral models built for specific tasks and sectors.

Mistral also claimed leaps forward in math and scientific reasoning, making the new model a "strong research assistant… from the first question to the final result."

Further improvements are expected via reinforcement learning post-training, helped by EU funding for European data centres.

"The reinforcement learning run behind this preview is still in flight, and the model is showing no signs of saturation — there is substantial headroom ahead," Mistral added. "As we scale up training on our expanded infrastructure, we expect large and rapid improvements in the weeks and months to come.

Controls for security

Mistral claimed ML4 is among the leading global models in terms of cybersecurity capabilities. On the Artificial Analysis Cyber Index, an independent evaluation of how well AI models fix security flaws, ML4 secured a top-five place. It also has a lead in the open-weight model domain, albeit among those developed outside China.

Control is key to ML4's success on this front, with Mistral noting that the open weights and self deployment aspects of its models mean organizations can run security using their own policies, rather than being limited by their AI provider.

Mistral said it solved 93% of the exercises in Cybench, whereas closed models including Claude Opus 5.5 and GPT-6 Astra scored near zero because they refused to perform aspects of the task.

"Yet defending software often starts with proving that a flaw is real, exactly the kind of work safety filters in closed models can block," Mistral added in the blog post. "This matters even more as threat actors increasingly jailbreak those same models to support offensive cyber activity: defenders need systems that can match those capabilities without being constrained by the same refusals."

That said, Mistral stressed high scores in safety and security, pointing to the model’s “propensity to refuse malicious requests regarding cybersecurity."

Sovereign AI

Mistral has made much of the fact that the company and infrastructure is based in Europe, noting that ML4 was "trained from scratch" on 3,800 Grace Blackwell GPUs in Mistral's own European data centres, with the public preview served on the same systems.

The model will still be available more widely around the world, but Mistral is also offering an entirely European deployment that it operates end-to-end, with no other digital service providers.

The company added: "Fun fact: a significant share of ML4’s training data was multilingual, spanning more than 160 languages, including every official language of the European Union."

FOLLOW US ON SOCIAL MEDIA

Companies want to automate IT operations using agents, but they don't trust them enough to make decisions.

That's according to research from Riverbed, which revealed that a whopping 90% of surveyed IT managers and tech leaders want to use agentic AI to power autonomous IT operations in the next two years.

A further three-quarters believe agentic AI will be a core part of future IT strategies, with nine-in-ten saying AI investment to date has exceeded expectations.

So what's the problem? Put simply, companies aren't sure they trust AI agents yet. In the survey, 77% of respondents said their organization didn't want to allow AI to make operational decisions without human approval.

Early days for autonomous IT

The roll-out of autonomous IT operations remains in its early days. Just one-in-five organizations run automated IT operations right now, the report found, and only four-in-ten have an IT environment ready to support agentic AI.

That echoes previous research showing that AI adoption rates aren't matching IT hype.

While AI hasn't yet automated IT operations, the survey found that 82% of those polled are using AI at their companies for innovation and to create new "sources of advantage”.

Beyond that, the survey suggests that though 76% of projects have progressed beyond the proof of concept, that's only 5% more than last year, and just 9% are deployed across the enterprise.

"The journey to autonomous IT is well underway," said Jim Gargan, chief marketing officer at Riverbed. "Organizations increasingly see AI as the path to more resilient, efficient and proactive IT, but our research shows significant gaps between ambition and readiness."

"The next phase of AI is not simply more automation, but AI that enterprises can understand, trust and govern," added Gargan.

Autonomy challenges

That next phase will see multiple hurdles that need to be overcome. According to the survey, more than half (55%) of respondents said autonomous IT faced challenges from a security and compliance perspective.

Other hurdles include operational disruption (45%) and a lack of trust in AI decisions (39%).

No wonder then that more than nine-in-ten said that AI observability and governance was set to become an increasingly critical area for IT to manage. While three quarters of respondents said AI was a critical or important priority, the report noted that "enthusiasm is not evenly shared".

"Business and IT leaders are more optimistic about agentic AI’s potential, while the technical specialists responsible for deploying, governing, and operating it are more attuned to the practical hurdles ahead,” Riverbed added.

Although 42% of those polled are confident in their employers' governance and oversight of AI, for technical specialists that falls to 36%.

Plus, just 17% of technical specialists see fully autonomous agents making up a large part of their employers' operating model in the next few years, the survey found.

How to deliver autonomous IT operations

To build autonomous IT operations, respondents said they need high-quality data, unified observability, and consolidated tools.

But that isn't yet available, with 19% of overall respondents saying they lack the data required for AI-powered IT operations, with just 17% of technical experts agreeing.

A further 17% believe they have unified visibility, leaving the rest unprepared as yet, while 87% said they have yet to integrate tools and systems.

"The message is clear: organizations have strong ambitions for autonomous IT, but fragmented data, incomplete visibility, and disconnected tools are holding them back," the report added.

FOLLOW US ON SOCIAL MEDIA

Britain is a burgeoning market for NetApp, according to UK & Ireland VP Puneet Gupta, and with the recent launch of the company’s Novus system, it’s ripe for the picking.

Speaking to ITPro at NetApp Insight 2026, Gupta said the cloud storage provider has made significant headway in the UK in recent years. Indeed, since joining as VP for the region in December last year, he’s been shocked at the maturity and appetite among British businesses.

“I think the market has truly surprised me,” he told ITPro. “The growth in this market has been phenomenal. When you look from a distance, the usual view is that a mature market like the UK will have single-digit growth.”

“But look at how this market for external storage has grown. It’s remarkable.”

According to Gupta, figures from IDC’s Worldwide Quarterly Enterprise Storage Systems (ESS) Historical Tracker for Q2 CY2026 show NetApp reached a 28.5% market share in the UK all-flash domain. This growth follows several quarters of consistent flash hardware revenue growth.

This represents a potential springboard for NetApp in the UK market, Gupta told ITPro. Crucially, it comes at a vital time for enterprises as many accelerate AI adoption and associated infrastructure overhauls to compensate for the technology.

Data readiness was a key talking point at NetApp’s annual conference last week, and the company has been highly vocal on this front in recent years. UK businesses, meanwhile, appear highly receptive to the messaging.

“The way customers are now thinking about their data, their data infrastructure, their AI infrastructure has been absolutely remarkable,” he said. “The conversations with customers have been pretty amazing in the last few months.”

“When you talk to customers, data is at the center of everything that they’re thinking, and whether it is around data sovereignty, whether it is around compliance, whether it’s how to leverage data … a lot of customers are planning that ahead of time, [and] are making the right investments.”

Sovereignty, in particular, is an area Gupta and his team are keen to capitalize on in the coming months. As ITPro reported at Insight 2026, the company unveiled the new Keystone Sovereign service, which is rolling out first to European customers before hitting shelves in the UK.

The new service, which sits within NetApp’s broader Keystone storage as a service (STaaS) offering, aims to provide customers with flexible sovereignty-related controls.

Since arriving in the UK last year, Gupta said this has become a common recurring talking point with customers and partners across the UK, and it shows no signs of slowing down.

“Every customer is thinking about sovereignty,” he told ITPro. “Every customer wants to be prepared for data sovereignty.”

Gupta’s engagement on this front comes amidst a sharpened focus on sovereignty-related concerns on both sides of the English Channel. Research from Civo in June this year found 73% of firms now view sovereignty as a “strategic priority”, for example.

Acting fast with NetApp Novus

An area of huge excitement for Gupta lies in the launch of NetApp Novus. Unveiled at the annual conference, this is a new zettabyte-scale file system designed specifically for organizations running AI factories.”

According to NetApp, storage has become the key bottleneck for AI factories, with outdated architectures unable to contend with the intense demands placed on infrastructure by high-performance GPU clusters.

Gupta told ITPro that with Novus, the company’s product range is now stronger than ever and represents the culmination of several years of work for the company. With this in mind, the company is keen to bang the drum for its new flagship service.

“I think the portfolio we have today, I haven’t seen anything like this in many, many years,” he said. “I've been at NetApp for six years. The first signs I think we got were three four years back when we refreshed the all-flash portfolio.”

“We refreshed everything, brought in new stuff. Our partnership with the hyperscalers, now the Novus announcement. I don't think anybody else has the kind of portfolio today that we have, and with Novus, I think we have leapfrogged everybody else.”

Gupta added that Novus will be “very powerful for momentum” both in the UK and further afield for NetApp in the coming months and years – but engagement and execution will be critical.

“In go-to-market, that’s where me and my team, our role, really comes into play,” he noted. “We have organized ourselves pretty well. We’ve hired, we’ve made investments, we’ve hired new leaders.”

“I think we are very well positioned in the market. Of course, the proof of the pudding is how we execute, and I’m very confident that the team is energized and very well positioned, and having the right conversations.”

With Novus targeting neoclouds and GPU as a Service providers specifically, Gupta said NetApp is also targeting heightened engagement with UK operators on this front – in particular, smaller providers.

“There are a fair amount of smaller neoclouds [in the UK],” he said. “We are talking to the smaller ones because the big ones have made their investments in the past, so that will be a different conversation.”

“For the neoclouds, they’ve got a lot of funding from the market, and obviously when they get funding, they need to give a return and there will be increasing pressure from investors to find that.”

This is a tactical move by NetApp designed to swerve the larger neoclouds already deeply integrated with industry big hitters, creating in-roads for the long term.

“Any tech company that goes to them and says ‘I'll find you a better return on your investment’ I think will be music to their ears.

Fears that AI could escape human control have heightened calls for a ‘kill switch’, which would require companies developing frontier models to cut off access to their tools.

Politicians in the US have been putting forward a series of legislative proposals in an attempt to rein in AI models that could go rogue.

The AI Kill Switch Act was introduced in late July with a view to forcing developers to throttle, suspend, or shut down AI systems at risk of causing catastrophic damage.

“As a computer science major, I am very aware of the dramatic possibilities – both good and bad – that AI presents,” Democratic congressman Ted Lieu said in a press release. The bill had the support of both parties.

The Stop Rogue AI Act, another bipartisan bill, was introduced on September 15. It would guarantee the US National Institute of Standards and Technology develops standards, guidelines, and best practices for controlling AI agents

“Right now, AI agents are running loose in our networks, and nobody can see them or verify who built them, making it increasingly hard to stop them,” Democratic congressman Josh Gottheimer said in a press release.

An AI Emergency Button Act was introduced by Republican congressman Tom Kean Jr on September 24. “While AI can be a helpful tool, it is essential that humans remain in control of complex artificial intelligence systems,” he said in a press release.

Legislation is already facing hurdles, though. An attempt to fast-track the passage of a separate AI Emergency Button Act was blocked by the Senate on September 16. It could be months before President Trump actually signs a bill into law.

The operational risks of a ‘kill switch’

While legislation is well-intentioned, there is an argument to be made that any ‘kill switch’ would be difficult to implement from a frontier model perspective.

“If something goes wrong, for example, an AI attack on national infrastructure, you’d want to be able to roll it back and switch it off. In reality, though, AI exists at every level. It’s on our mobile devices, across physical networks, in data centre networks, on our laptops. You'd have to find all of it before you could shut it down,” Charlotte Wilson, head of enterprise at Check Point, tells ITPro.

Gary Barlet, public sector CTO at Illumio, adds that a ‘kill switch’ could cause systemic risk given how embedded AI has become in business operations.

“If a model suddenly becomes unavailable, those workflows could be disrupted overnight. Even intended as a safety measure, it could become a source of uncertainty, disruption, and economic harm if it's triggered too aggressively or without clear standards,” Barlet says.

Darren Guccione, CEO and co-founder of Keeper Security, points out that any ‘kill switch’ would not make AI secure by itself and definitely wouldn’t address every risk. However, it would help to establish an emergency baseline that frontier model developers have to meet.

Preparing for a shutdown

The question for organizations is whether they’re prepared for developers having to design a ‘kill switch’ into their frontier models – and how they would react to it.

Guccione says that large companies generally have a bigger pool of resources to invest in a wide range of models. However, it’s the smaller companies that are likely to be dependent on the same frontier models, putting their operations at greater risk of being impacted by any switch-off.

Smaller companies are also less likely to invest in their own AI safety mechanisms. Kiteworks’s annual data security and compliance risk report, published in July, found that the ‘kill switch’ gap is widening. Researchers at the software development company had forecast that 60% of organizations in 2026 wouldn’t have a tested capability that can shut down an active AI agent in real time, but the actual figure was 79% among 459 security and compliance professionals surveyed.

In the event a shutdown suspends a model that workflows are built around, the onus is on organizations to keep whatever depends on the model functioning, Guccione stresses. “Continuity planning needs to treat that as an operational imperative in its own right."

It may seem obvious, but to prepare for a shutdown, organizations need to improve their visibility by establishing an inventory of AI models and their dependencies.

“They should identify where AI models are embedded in business-critical processes, which providers those processes rely on, and what data and systems are connected to them,” Guccione continues.

From there, organizations need to put a fallback plan in place for their most important business processes. Where possible, they should “avoid unnecessary dependence on any single provider” and seek out alternative models to prevent single-model lock-in, Barlet advises.

A ‘kill switch’ would increase confidence in AI

While unexpected shutdowns caused by a ‘kill switch’ could scare organizations into stopping using models, its introduction would likely drive AI adoption in the long run.

As Barlet explains: “There's a tendency to assume regulation slows adoption, but uncertainty is often the bigger obstacle. When organizations understand the security requirements, testing standards, and governance obligations, they're more willing to invest. The right safeguards can increase confidence.”

Ultimately, a ‘kill switch’ would mean that organizations don’t have to rely on vendor assurances about AI safety, Guccione adds. “It would shift trust from claims made by individual vendors towards controls that organizations can actually verify.”

With AI-enabled PCs becoming mainstream business devices, many enterprises are yet to understand just how much AI is impacting device refresh cycles. The key mistake is to behave as if nothing has changed, and that this is a familiar hardware upgrade, replacing devices on a fixed cycle, just as it was before. However, the reality is very different. AI PCs are driving important changes in sustainability, employee experience, and how organizations buy devices and services. Organizations need to reassess how work, security and employee expectations of devices have changed before initiating any refresh.

For organizations, AI is now the leading strategic priority, with 96% of CIOs expecting to increase their AI investments in the coming year, according to our research. In our own conversations with partners, it’s clear that while customers want to move forward with AI, they do not want to move forward blindly. AI fundamentally alters what workplace devices are capable of. This changes how they should be deployed, managed, and measured, and has important implications for partners and the way they do business over the long haul.

Common mistakes

For enterprises, it’s important to first acknowledge the fact that AI device refreshes are different, with AI capabilities helping enterprises support smarter decision-making, optimize resource use, and accelerate sustainability. Key mistakes include underestimating the impact of on-device efficiency, as well as misunderstanding how AI impacts energy efficiency and battery life.

It’s also vital to understand AI’s impact on employee experience: applications can adapt to each user’s individual preferences, and whole teams can lean on new personalisation capabilities to streamline workflows and make it simpler to deliver results.

AI-driven content creation and intelligent decision-making help employees deliver, while predictive and proactive security (including fraud detection) helps boost peace of mind across the organization. With AI PCs offering granular control over energy use, partners need to treat issues such as sustainability as a growth lever, rather than a reporting requirement. Sustainability is becoming a way to offer differentiation to customers and create new revenue opportunities.

Moving beyond traditional approaches

Traditional device refresh models are no longer enough to keep pace with the rapid advances of the AI world. It’s all too easy for organizations to find themselves either falling behind their competitors or over-investing early in the wrong technology. IT environments are becoming more complex, with hybrid cloud, AI, edge computing, and distributed workforces ripping up the rulebook of how organizations operate.

Many customers need a helping hand to stride confidently into this new world. This has raised expectations for partners who are now expected to deliver across the full lifecycle, from design to an ongoing process of optimization. This requires a new way of thinking from partners, and new, higher-value ways of doing business.

Service-led approaches

Services are the foundation of sustainable business growth. Service-led approaches to device strategy offer an opportunity for channel partners to move beyond simple hardware sales and towards a more consultative relationship with customers. For years, success in the IT channel has been driven by one simple metric: product volume. Today, customers have moved beyond buying devices or infrastructure; they are looking for solutions that deliver across the technology lifecycle, driving business value over the long term.

With partners offering advice on issues such as AI readiness, lifecycle management, sustainability, and long-term workplace transformation, service-led models increase customer lifetime value and build deeper relationships. What’s required is a shift towards lifecycle thinking. Customers are no longer focused on individual transactions; they are taking a broader view of how technology performs over time, from deployment to support to refresh and recycling. For partners, this is an opportunity to engage earlier, stay engaged longer, and deliver continuous value.

Operationalizing AI

This is also an opportunity for partners to take a consultative role in helping customers to operationalize AI effectively. For partners, this means they may need new skills and new delivery models. Expectations are rising, and customers need faster deployment and proven expertise. For partners, the ability to deliver outcomes is becoming a key differentiator.

At present, there is a clear gap: many partners understand where the market is going but are not yet fully equipped to really capitalize on it. But some partners are already succeeding, evolving into trusted advisors, helping customers navigate complexity.

Successful partners are not replacing their product business but attaching services and expanding into higher-value solutions. By building capabilities in everything from managed services to AI and digital workplace, partners can establish more resilient, scalable business models. By doing so, they can build relationships that extend beyond a single transaction.

The road ahead

Enterprises have much to learn when it comes to AI device refreshes. Traditional approaches are no longer working, with many enterprises misunderstanding the impact of AI on everything from sustainability to employee experience. This is also changing the role of partners, who are now taking a more consultative role in everything from sustainability to driving business growth.

Looking forward, this shift, driven by AI refresh cycles, means that the future of the channel will be defined by services. Partners who make the shift towards offering services as a layer that connects product to business outcomes will be better positioned to build longer-lasting customer relationships and thrive in an outcome-driven market. Growth will be delivered by the value partners deliver every day, across the whole lifecycle.

There’s been a lot of talk - and some action - about AI, but now is the time to really take things to the next level and make AI real for your business.

That was the key message tech giant Snowflake wanted delegates to take away from its Snowflake World Tour London event earlier this week. The event, hosted at the ExCeL, played home to more than 5,000 attendees, 60 partners, 72 sessions and 150 speakers.

“It's fantastic to see so many of you here. Customers, partners, innovators, data visionaries from all across the region wanting to make AI real for [their] business. What an incredible time to be together. The energy today will tell you everything about the pace of change we're seeing across our customers and industries,” said Jimmy Hall, Snowflake’s vice president and country manager for the UK and Ireland in his welcome address.

“Year after year, we've seen organizations across this region move at an incredible speed. [They’ve been] moving beyond pilots and experimentation, and putting AI into production at scale. [Ultimately], taking those ambitious ideas and turning them into real operational value.

He added: “We've moved from talking about the potential of AI to putting AI to work. It's now all about making AI real for business as we enter the era of the agentic enterprise; we're operating in a fast-moving environment. Arguably, the biggest risk any business could take right now is standing still. But moving fast with data and AI shouldn't mean losing control. It means having the foundation, the flexibility, and the confidence to move quickly.”

Data underpins AI ambition and results

That foundational element is key, according to Hall and his Snowflake peers. Indeed, while the momentum gained sounds very exciting, he cautioned that organizations must not forget the fundamentals. You cannot successfully implement an AI strategy unless you have the solid foundations of a good data strategy to support it, Hall stressed.

“There is some myth about data. Some organizations have been of the mindset that AI is so good and so smart, the data will sort itself out,” said Christian Kleinerman, Snowflake’s EVP of Product.

“The reality is not quite. It won't happen. If anything, the opposite is true. The most amazing results you can envision for your organization with AI come from having a great data strategy.”

Event attendees looking for inspiration as to how to move the theory into reality didn’t have to wait too long. Hall was soon joined on stage by Kate Dohaney, CEO of giffgaff.

“I joined about eight months ago, and I knew we had a massive opportunity to lean into the world [of AI],” she said.

“We were already optimizing Snowflake at different levels and in different teams within our organization, but commercially we weren't seeing what we wanted from it. So we [created] a proper data strategy team and [have been] really teaching our business as a whole how to optimize information to better serve our customers.”

The company is keen to automate as much as possible, to free up valuable human resources from time-sink tasks so they can be more productive elsewhere, according to Dohaney.

“That means taking out the repetitive work, understanding the market, understanding our competitors, and having that ability without weeks and hours of human capital research. Snowflake has given us the right tools to move at pace,” she added.

“It empowers our business. It’s about partnering with the right tools, the right partner, to move quicker.”

Recipe for success

For organizations to truly benefit from what it means to be an agentic enterprise, they need to embrace four key elements, according to Kleinerman.

These are the choice of AI model, software and apps, enterprise data, and context, with all three of these linked in the middle to an agentic control plane, which Kleinerman described as “the glue”

He added: “That's what makes all of this together. This framework is high level. It's agnostic of any one technology or any one vendor. This is what we believe is needed for any of you to leverage and deploy AI successes.

“The promise of AI, in my mind, is unparalleled to other technologies that we've seen at least in our lifetimes. And hopefully, everyone here is excited about how you can help your organization achieve greater successes through the power of AI.”

Hall concluded the keynote by saying that we’re in the midst of “the biggest structural shift in enterprise technology of our lifetime.”

“As I said earlier, the biggest risk you could take now is standing still. The companies that will win this decade won't just apply standalone agents; they will build an agentic enterprise,” he continued.

“Turning this vision into reality requires investing in your data foundation today, and that comes down to three core priorities: one is modernizing your data space; secondly, making your data AI-ready; and thirdly, driving autonomous action and impact with AI.”

AI is going to have to earn its keep by inspiring $4 trillion of as yet unknown applications and services, with annual spending on AI infrastructure reaching $1.5 trillion in the next five years.

That's according to Bain & Company's Technology Report, which notes that the size and cost of AI data centers is doubling every 12 to 16 months.

Last year, Meta's 600MW Prometheus cost about $24 billion to build, and data centers currently in the works at 1-2GW will cost between $40 billion and $80 billion to build — climbing to $200 billion for a 9GW data center in 2030, the analyst firm predicted.

That suggests annual spending on AI infrastructure will top $1.5 trillion by 2031, Bain predicted, including new data centers and compute, as well as upgrades to existing GPUs, memory, and networking equipment.

Doubling costs

That's double what's expected to be spent this year, with Bain predicting $780 billion in capital expenditure across the five key hyperscalers, though that does include investments beyond AI. IDC predicts AI infrastructure spending of $497 billion this year, while Gartner puts it at more than $1 trillion, including a wider range of equipment.

What they do agree on is significant growth: Bain has the figure doubling to $1.5 trillion by 2031, IDC predicts $1.2 trillion by 2030, and Garter sees total AI spending doubling in just two years.

That raises a challenge: can AI earn enough to cover these costs? Bain argues yes – but with some caveats.

"The infrastructure is being built ahead of the demand curve, and funding it sustainably will require adding approximately 1% to the annual global GDP growth rate," Bain's analysts say. "The question is whether the applications arrive in time to pay for it."

Paying for AI

First, the analysts at Bain argue that CapEx for cloud providers is usually about a quarter of industry revenue, meaning AI will need to earn $6 trillion in revenue by 2031 to maintain that margin.

By 2031, consumer AI products could contribute between $200 billion to $400 billion by 2031 via subscriptions and advertising. "Enterprise adoption could contribute another $1 trillion to $1.4 trillion in gains to providers alone as AI delivers meaningful productivity gains to enterprises across software development, sales, marketing, customer service, and IT operations," Bain's analyst predict.

That leaves about $4.2 trillion of revenue missing. Bain figures this will come from four new categories, starting with search and advertising at $100 billion to $200 billion, with ads integrated into chatbots as traditional online search is disrupted.

Missing $4 trillion

That leaves $4 trillion. Bain says $400 billion could come from autonomous operations of cars, drones, and robotaxis, as well as logistics, while physical AI could add $900 billion via simulations, digital twins, and the deployment of robotics – including humanoid ones.

That leaves another $2.7 trillion to be made up in new product development, anything from drug discovery to materials sciences, Bain suggests.

The point isn't that AI is expensive, but that companies need to do more with it than merely shave a few percentage points from their bottom lines.

"The economics required to generate ROI from AI infrastructure are demanding trillions in new revenue, not just cost savings," Bain said. "The industry needs a wave of application innovation comparable with what mobile and cloud unlocked, not just productivity gains on existing workflows."

Cannibalising tech

While the tech industry is seeking revenue for AI, that same technology is eating away at margins in other departments.

Bain noted that software revenue growth has "moderated significantly" from 20% annually in 2022 to half that last year. On the flip side, Bain predicted that agentic AI could prove to be a $100 billion opportunity for SaaS providers.

Beyond costs and revenue, Bain also warned in its technology report that the $5 trillion data center rollout still faces constraints from power, chips, skilled labor, and permits.

"Individual projects are finding workarounds, but they can’t be replicated at scale, and some are controversial. Colossus (from xAI, now part of SpaceXAI) was built in under a year by repurposing a factory and leasing dozens of gas turbines without federal clean air permits, drawing a lawsuit," Bain said.

The analysts added: "Meeting the global ambition instead will require system-level fixes outside the usual free-market playbook. A grand challenge on this scale demands grand solutions, and executives planning on business as usual will get blindsided."

You can lose a job to AI, but in California, a human has to make the decision to fire you.

California Governor Gavin Newsom has signed into law the No Robo Bosses Act 2026, requiring human oversight in the workplace specifically when it comes to firing or disciplining workers.

Act SB 947, as it's officially known, was introduced by Senator Jerry McNerney to push back against the use of automated decision making systems (ADS) in the workplace. In a statement, McNerney's office pointed to the rise of "bossware" products used to manage workplaces, which have led to people being mistakenly fired. The new law doesn't outright ban the use of such systems, but requires a human in the loop.

“No worker should ever be fired or disciplined by a robo boss," said McNerney, in a statement. "Artificial intelligence systems have the potential to increase productivity, but they’ve also been prone to errors, bias, and misjudgments.”

McNerney added that the aim is ensuring human oversight in workplaces. “AI must remain a tool controlled by humans, not the other way around," he said.

California has led AI regulation, with laws requiring frontier AI developers to report safety incidents and publish safety frameworks, forcing AI generated content to be labeled as such, and creating a registry for AI auditors, alongside rules targeting political and sexual deepfakes, and AI generated unauthorised replicas of performers, among others. Newsom signed into law a list of other AI rules this week, including banning lawyers from handing over core work wholly to AI and requiring educational institutions to train staff and students about AI.

"Super intelligence is clearly not coming from the White House – that’s why California continues to lead," Newsom said in a statement. "AI should expand opportunity – not come at the expense of workers and families. As this technology reshapes the workplace, California is putting people at the center, ensuring we all have a voice in the decisions shaping their future."

No robo bosses

The robo boss legislation has three points. First, it bans employers from relying solely on ADS to make firing or disciplinary decisions. Second, it requires human oversight and verification when employers do use ADS to assist in those decisions. And third, if a company does use ADS for termination or disciplinary decisions, the impacted worker must be informed.

The rules will be enforced by the California labor commissioner, state attorney general, or local prosecutors. Similar laws are in the works in a handful of other states.

The law was backed by the local labor union, with Lorena Gonzalez, president of the California Federation of Labor Unions saying the law gives "other states and the nation a clear example of how, by letting workers lead, we can stop some of [the] worst abuses of AI at work."

SB 947 originally included a provision to notify employees if an AI system would impact their work, but Newsom vetoed it saying the focus should be on misuse rather than forcing "unfocused notification requirements on any business using even the most innocuous tools," he said at the time. That aspect was removed, as were protections for gig workers.

Rise of AI workplaces

The move comes as management becomes increasingly automated or algorithmically controlled. A survey from the OECD last year found that 90% of American managers polled said they have at least one such tool in place, versus 79% across Europe and 40% in Japan.

Beyond prevalence, that study also found that US companies were more likely than European counterparts to use algorithmic management for monitoring or evaluating workers even if it involves personal data and has "consequential outcomes" such as access to work. The study added that two-thirds of managers were concerned about the impact on workers, particularly with regards to unclear accountability in the case of incorrect decisions.

One example of that playing out now regards Meta, which faces a lawsuit after the Facebook-owner attempted to apply AI to monitoring and managing its own workers earlier this year, taking into account everything from token usage to keystroke monitoring – and reportedly using that data when deciding on a massive round of layoffs.

In response, a group of former and current staff have sued.

"Meta used a constellation of internal artificial intelligence systems – including a system referred to internally as ‘Metamate,’ employee-trained ‘second-brain’ agents, keystroke- and activity-monitoring data, AI-token-usage dashboards, and algorithmically assisted performance ranking and calibration – to score, rank, and select employees for inclusion on the list," a court filing reads, noting the system disadvantaged those on leave, including for maternity or disability.

The number of vulnerability disclosures made each month has more than doubled since the start of the year, hitting 10,740 in August, according to Google Threat Intelligence Group (GTIG).

The researchers also found the number of vulnerabilities being exploited is showing a similar trajectory. In 2025, the average was 10.5 exploits per month, this rose to 18 per month from January 2026 to August 2026.

The researchers said they expect both these trends to continue in the short to medium term.

Zero-day exploitation, however, increased only marginally, rising from an average of eight cases per month in 2025 to an average of 11 from January 2026 to August 2026.

According to GTIG, AI is changing not just the pace of vulnerability discovery and exploitation, but also the types of vulnerabilities that are being discovered.

While fewer low-risk vulnerabilities are being found, there are more moderate-risk vulnerabilities and vulnerabilities leading to remote code execution (RCE) are being unearthed.

High-risk vulnerabilities surged from 131 disclosures in January 2026 to 350 in August, up by 167%, although they still account for just 3% of all vulnerabilities disclosed. The increase, said GTIG, comes down to a widening pool of affected vendors and concentrated vendor disclosure cycles.

"It’s important to note that the proportion of vulnerabilities exploited versus disclosed remains vanishingly small: only 0.23% of all disclosed vulnerabilities in 2026 (roughly one in 431) were ever observed in active exploitation, or on the order of tens versus thousands per month," the researchers said.

"This means that monthly exploitation counts can more easily be influenced by other factors such as vendor disclosure cycles and threat actor campaign spikes."

Meanwhile, current public data significantly undercounts vulnerabilities discovered by AI, the researchers said. This is partly due to the absence of standardized metadata, and partly because of silent first-party and cloud patching. Major cloud and SaaS providers routinely remediate AI-surfaced vulnerabilities directly in production without requesting formal CVE IDs, with CVE assignments typically reserved for on-premises or third-party software.

The increase in vulnerability discovery, said Steve Povolny, vice president of AI strategy and security research at Exabeam, shouldn't come as a surprise.

"We know by now that AI models are able to achieve speed, scale and efficiency that few, if any, human beings are capable of. Inevitably, discovering known flaws and finding ways to exploit them is par for the course when it comes to AI," he said.

However, he highlighted the lack of change in zero-day exploitation.

"This leads to one of two conclusions: first, AI models are very good at uncovering known (N-day) vulnerabilities but not mature enough to discover widespread zero-day vulnerabilities of significance," he suggested.

"The second option is that AI augmented zero-day discovery is happening en masse, but flying under the radar, based on the inherent value it provides to attackers such as nation states and zero-day bug broker groups. Given the amount of security researchers focused on leveraging AI for zero-day discovery, I think the latter is less likely."

UK tech leaders are calling on the government to invest much more heavily in sovereign AI.

In an open letter to Andy Burnham and AI minister Kaniska Narayan, executives from defense, financial services and critical national infrastructure firms say there's an urgent need to strengthen public confidence and national resilience.

Signatories include AI lab Cosine, PA Consulting, BT, Telefónica Tech, Leonardo, HPE, Kainos and Sopra Steria.

"AI is now a question of national capability. Countries with reliable access to compute, secure deployment pathways, strategic models, and effective procurement will have greater resilience and influence," the letter reads.

"Countries that lack those capabilities risk becoming dependent on decisions, infrastructure and operating conditions elsewhere."

The group wants to see financial support of at least £20 billion over the next five years, through public investment, procurement guarantees, and co-investment. This, they said, should be targeted at establishing sovereign British AI compute, power, data-center capacity, and network infrastructure for government, defense, academia, and strategic industry.

It's calling for changes to planning laws, energy policy, procurement, and long-term contracts to help AI firms, and a fast-track sovereign AI procurement route for urgent national priorities.

There should be financial backing for UK-led model development, with government, defense, and the NHS gaining access to proprietary British datasets for training these models, with strict privacy safeguards, mandatory audit trails, and interoperability standards.

"No mission-critical public service or critical national infrastructure should depend on a single model, provider, or cloud stack," Cosine said.

"Independent testing, evaluation and monitoring of AI models is essential to enable safe, secure and trusted AI adoption. Mandate tested fallback arrangements, independent alternatives, and human-in-the-loop control for all high-consequence AI systems."

The government, said the firms, should identify the areas of the AI value chain where the UK can become world-leading, then align infrastructure, procurement, research, regulation, and investment behind them. It should create a cross-party ten-year Sovereign AI Strategy with clear objectives, annual accountability, and long-term funding certainty.

Infrastructure, common standards, and interoperable systems should be shared with like-minded nations.

And it should identify the areas of the AI value chain where the UK can become world-leading, then align infrastructure, procurement, research, regulation, and investment behind them. It should create a cross-party ten-year Sovereign AI Strategy with clear objectives, annual accountability, and long-term funding certainty.

"Britain is coming from a position of strength. It has world-class research, deep software talent, strong cyber capability, credible institutions, strong digital infrastructure, and real advantages in science, defence and public services," the letter reads.

"But without a plan for AI sovereignty, the UK risks becoming a customer in the most important technological shift of this era, rather than a serious builder."

Earlier this year, the government launched a £500 million Sovereign AI fund aimed at helping domestic AI companies scale.

This will act as a venture capital fund within the government, making targeted investments in UK AI firms, and providing direct access to the UK's fastest AI supercomputers, along with specialist R&D support and grants.

Nearly half of UK employers are planning to boost their technology teams by the end of this year, looking mostly for security, AI, and cloud skills.

According to new research from Robert Half, 47% of UK employers hope to boost their tech workforce, with 54% looking for cyber security skills, 50% agentic AI skills, 48% generative AI skills, and 44% cloud skills.

Meanwhile, another 25% are expanding contract and project-based hiring.

"Technology remains one of the most active areas of hiring in the UK labor market, with nearly half of employers planning to grow their permanent technology teams before the end of the year," said Matt Weston, senior managing director, UK & Ireland at Robert Half.

"Businesses continue to invest heavily in cyber security, cloud infrastructure, and artificial intelligence, but they're not simply looking for more people. They're looking for professionals with the right blend of technical expertise, commercial understanding and the ability to work effectively alongside AI."

According to the researchers, 45% of UK technology professionals say they're now expected to develop new AI-related skills, while 38% spend more time overseeing and validating AI-generated outputs. At the same time, 53% say AI has cut the time they're spending on routine tasks, and 37% that their roles have become more strategic.

The findings, said the team, point to a growing demand for professionals who can work effectively alongside AI systems, applying human judgment to ensure accuracy, quality, security, and compliance.

"As organizations increase investment in AI, cloud infrastructure and cyber security, we're seeing strong demand for professionals who combine technical expertise with strategic decision-making," said Weston.

"AI is automating some tasks, but it isn't replacing the need for skilled technology talent. In many cases, it's making those skills even more valuable."

With cyber threats continuing to evolve and businesses increasing investment in AI-powered technologies, the firm said it expects competition for specialist talent to remain strong, especially for those with expertise in cyber security, AI, and cloud environments.

The report coincides with the UK government's latest figures on the cyber security workforce, showing that it's expanded by only 2% over the last year – although it, too, finds that most companies are looking to hire this year.

There was a 7% increase in core cyber security job postings – roles where some aspect of cyber security is the main job function – and a rise of 10% for roles that might not be formally labeled as cyber security jobs but that still require cyber security skills.

However, 57% of businesses said they had a basic technical skills gap, up from last year's 49%.

Back in January, the government announced plans to provide 10 million workers with key AI skills by 2030 through free courses available to all, opening a skills hub that's already seen the delivery of more than a million courses. Meanwhile, the NCSC offers training courses for both novices and experienced security professionals, with courses targeted at sectors from board members to schools.

Proposed capital expenditure from big tech companies into AI infrastructure has been astronomical. The likes of Meta, Google, and Microsoft, among other hyperscalers, are leading the charge on emptying their coffers to spend on data centers not just in the United States, but across the world. Among the beneficiaries of this spending are neoclouds, gobbling this cash up as part of their ongoing infrastructure buildout, alongside borrowing and share issuances.

But all this infrastructure spending is not simply numbers on a screen or figures to trawl through on an Excel spreadsheet. It's required spending, experts say, for the AI industry to sufficiently build the systems we need in the future – whether that might be smarter models, agentic clusters or simply to provide more capacity as more and more users start plugging into a new wave of AI-enabled productivity tools. And, at the other end of the dollars spent, are people who live in areas affected massively by the presence of these ever-humming, environmentally-questionable monuments of silicon and steel.

That's why, in many parts of the US and elsewhere, severe public backlash has spilled into regulatory action – which ranges from more consultation with local communities to a moratorium on construction. Is there anything that businesses can do to move past this movement?

Explaining the big AI backlash

Amid a proposed piece of federal legislation that seeks a national moratorium on building or upgrading AI data centers until safety is mandated by law, government at the state level has issued several banning orders against new data center construction. This includes New York State, Texas, Maryland, and Pennsylvania, among others. So where does this growing sentiment against AI come from, realistically? Much of it has to do with the impact on power and the environment.

"There is a trend right now of going from local bans and moratoriums to statewide ones," explains Arif Gasilov, partner at energy and environmental consultancy Gasilov Group. "New York enacted the first statewide moratorium in July, and in just one week in September, the governors of New Hampshire, Connecticut, and Oregon all announced pauses or refusals.

"In addition, moratoriums that are countywide still copy each other; more than half of Maryland's 24 counties now have one. In addition, the costs of data centers in local counties, such as electricity bill increases and whatnot, show up for everyone, while tax revenue shows up only to the local government and jobs only to the local skilled electricians, and so on. Overall, as such, the overall negative sentiment gets spread across more people."

This is a sentiment that Nicholas Nadeau, co-founder and CTO of Onix, echoes, telling ITPro that "most of the concerns are real".

"Underneath the noise and water complaints is an ownership question. People already paid for this industry in three currencies: their work that became training data, their wages that never moved with the productivity gains, and their savings that index funds route in at valuations insiders set. Data centers add a fourth, and it is the visible one: their grid, their water, their land."

The rising cost of electricity is the biggest factor that's impacting public mood, as well as the impact on water usage in arid areas – with evaporative cooling using the same supply that residents are under restrictions for. For example, a Meta facility in Newton County, Georgia, uses 10% of the entire county's supply. When you live nearby, there's a need to rely on diesel-based backup generators, traffic due to the construction, and a reduction in the local housing supply too, among other knock-on effects.

Dan Ye, CEO of CollegeNode and UC Merced lecturer (and former Johns Hopkins University lecturer) in AI, blames some, but not all, tech companies. "A handful of bad actors have done real damage here – xAI's Colossus data center in Memphis, Tennessee, is the most cited example, generating significant noise and discharge complaints – and the media coverage that follows fuels fear among residents elsewhere who are simply weighing whether to welcome a data center into their neighborhood."

He adds that a data center done right has "minimal impact" on neighbors. "These are engineering and siting choices, not inherent features of the technology – which is why it's worth being careful not to throw the baby out with the bathwater," he explains. "Data centers are also a significant source of local revenue."

What does this movement mean for AI progress?

Frontier training could slow down significantly if there is a sustained shortage of computing capacity, Ye says, while limiting how AI services are deployed and keeping usage costs higher than they'd be with more available infrastructure. But temporary pauses may not make much of a difference at all in Nadeau's eyes.

"Arvind Narayanan ran the arithmetic on Hard Fork podcast last month and found a one-year state moratorium delays AI progress by hours, not months, because efficiency at a given capability level improves several times a year while new capacity adds roughly a quarter more, and blocked sites get rebuilt elsewhere anyway," he says.

There's also a distinction to make between calls for a slowdown in the infrastructure and calls for a slowdown on the safety and security front. This is something that Ye characterizes as two pressures pulling in different directions. If political pressure persists, he says, tech companies will push data centers offshore to places like Canada, given its abundant hydro power, or even China, which is enduring a renewables boom presently.

The pausing for safety rhetoric, meanwhile, is just cover, according to Gasilov, "because it made them look responsible while the real delay, as per the data, is grid queues". Indeed, there's a massive backlog when it comes to the power generation and infrastructure, with many sites not expected to go live until 2027 or 2028. These sites will largely build their own behind-the-meter gas generation plants to power them, sidestepping both grid connection delays and protests, but may result in much higher rates of air pollution and premature deaths, he warns.

Can the industry do anything to ease concerns and get back on track?

Almost half of the states in the US have approved large load tariffs, which make data centers pay their own grid costs for upgrades to the grid – and more than ten states have filed moratorium bills. Gasilov says that the opposition argument wins out at county planning meetings, but won't necessarily solve the problems with bills, for instance.

Companies in this space can do several things to ease concerns and smooth the resumption or acceleration of this massive infrastructure buildout, the experts tell ITPro. Ye persuades site developers to bake community acceptance and deliverable power into the investment case from day one and not treat this as an afterthought. As he puts it: "Evaluate sites against actual grid capacity, water availability, residential proximity, and how much development is already planned nearby. A cheap parcel can turn very expensive once you factor in years of interconnection queues and approval delays."

As for businesses that buy AI services, he says a reasonable practical response is to understand a supplier's capacity constraints and lock in a service commitment that matches up with your organization's particular needs. He advises that critical workloads should remain portable, if possible, and they should invest in smaller models that can run either on-site or even on-device rather than chucking everything into the cloud. Nadeau puts it more bluntly, saying: "Stop naively buying compute by the token." He adds that businesses should measure cost and energy per completed task, then size the model to the job, while pushing AI inference as close to the edge as possible – cutting down on latency and compute costs while also limiting privacy liability in the process.

Ultimately, however, the buck stops with the infrastructure developers and city planners. Microsoft, for example, set a very useful benchmark with its commitment in January 2026 to cover power costs and water use, Ye says, adding: "The strongest reassurance is a project residents can actually hold accountable. If the economics only work by shifting infrastructure costs or environmental burdens onto neighbors, the business model itself needs to change."

Leave it to the Office for National Statistics (ONS) to give the UK’s AI ambitions a reality check. In a July data release, the UK stats body found that take-up of AI tools among British businesses has risen a meagre 0.2% since 2023. Between now and then there’s been significant advances in the reliability of LLM outputs and the protocols that hook models into the apps and platforms that hold business data. It would therefore be fair to argue that AI adoption among UK firms hasn’t yet reflected these advances, and a major uptick is on the cards.

That’s the thinking of many UK channel partners. In shallow adoption of AI lie opportunities to consult many new customers through the process of entrenching AI into core operations. At the current moment, UK business leaders are trying to make sense of new agentic AI products from AI labs and best practices for their rollout. Channel partners can help, acting as the connective tissue between platforms and customer intent.

‘Can’ is the crucial word. While partners possess the technical and change management skills to rise to the moment – a deliberate shift in partner skills and proposition framing is also required. Deliberate data and analytics advisory has the right answers to break businesses out of stop-start AI pilot stasis. Partners need to position themselves to hold those answers.

The ambition-execution gap

The state of the UK’s AI rollout is similar to other countries in that it’s plagued by a failure to scale beyond pilots. Indeed, in February our research found that more than 80% of AI projects fail to operationalise into production.

The bottleneck is not AI access (which is increasingly commoditized), but what AI systems are built with. Fragmented, and often inaccurate, data spread across business units is the norm for many UK firms. Business logic - the rules, calculations, and decision-making definitions that determine how work gets done - is locked in spreadsheets, legacy systems, and employees' heads.

There’s a clear risk of deploying AI over broken data foundations, a risk that compounds as the application of AI gets more sophisticated - like the deployment of AI agents.

This is not a problem IT can fix in isolation. Reckoning with this state of affairs requires honesty on the operating system refresh required to make AI work. Wrangling with data is one aspect - but so too is addressing the trust and culture issues that slow AI rollouts down. The conundrum is a business transformation challenge, which requires both technical capability and contextual advisory. That is exactly where channel partners are comfortable.

How channel partners can think about the operating system refresh

For many organizations, AI is seen as a line-item purchase. Business users are set up on the software with perhaps some training. The conditions aren’t set for pilot testing of AI to scale up into a transformational reimagining of business processes.

Channel partners can address that gap with a neat proposition: a pre-agent layer. This is an analytics platform that sits underneath AI systems with pre-processed proprietary data, defined rules, and business context. Models connected to the layer are rooted in how the business actually runs and, therefore, more easily integrated into day-to-day work.

Channel partners should know that such propositions are most effective when built around platforms that offer visual-in-nature, intuitive-to-build analytics workflows that make AI visible, understandable, repeatable, and auditable.

Managers can quickly establish the inputs into AI workflows and validate the presence of governance steps and role-based access controls to extract the confidence they need to sign an AI workflow off. With a pre-agent layer proposition, channel partners come with workable solutions to help UK firms evolve beyond AI pilots and the hesitations that fuel the standstill.

Channel partners are well positioned to build pre-agent layers for UK firms as they already sit inside customer data infrastructure. They have the familiarity with these environments - the legacy systems, the data flows, the processes that run under the surface - to hit the ground running in a solution-oriented way.

The new blueprint for channel success

AI services are widely expected to be the single biggest driver of revenue growth for IT service providers over the next 24 months. That is a significant opportunity - but only if partners evolve "AI services" with the wants and needs of customers.

Wrapping a managed service around a generic AI model is not the same as genuine AI delivery. The skillset that generates commercial growth is the ability to translate a customer's business context into governed AI processes. Partners who build that capability can immediately move into selling AI readiness - helping customers make their data and analytics estate fit for agents and their most impactful use cases.

From my conversations with UK MSPs and GSIs, this evolution is already happening. UK partners, in my experience, are showing innovation that - despite the naysayers - surpasses US counterparts. That’s because the deep customer relationships and genuine sector-specific specialisations that make up the UK IT channel’s strengths are precisely the assets that inform the thinking for a successful customer AI operating system refresh.

The gap is the opportunity

The gap between the state of UK AI rollout today and where it could be is not a failure story. It’s an opportunity for the UK channel. The partners who ready their propositions - vendor relationships, skills, service design - will be the ones positioned to benefit.