Three-quarters of companies have rolled back or shut down an AI agent used in customer service, according to new research.

That high rate of failure in using AI agents for customer communications comes via a survey commissioned by Sinch, which polled more than 2,500 leaders in the industry.

The survey found nearly two-thirds of respondents already had AI agents up and running in their organizations, showing that they were getting beyond pilots into real use cases in customer service. Indeed, 88% are set to have agents in full production over the next year.

But they're having issues with those deployments, Sinch found. 74% said they had shut down or rolled back AI customer communications agents due to governance failures – that echoes analysis from Gartner last year that suggested companies were backtracking on AI customer service and rehiring human workers.

According to the Sinch report, the leading factors behind roll-backs lay in concerns about customer data exposure (31%), hallucinations or brand risks (22%), and lack of auditability (16%).

“These aren’t abstract risk categories. [Personal identifiable information] exposure means a customer’s personal data surfaced in an interaction it shouldn’t have. Hallucination means an AI agent said something confidently wrong to a real customer, on a live channel, under your brand’s name,” the report noted.

The Sinch survey follows research via Pega that showed more than half of customers aren't confident that organizations are using AI responsibly in customer service, while a report from 2024 suggested people were searching online for answers to avoid using an automated chatbot.

Governance works

The rate of AI u-turns is even higher for organizations that report having fully mature guardrails in place, with 81% saying they've been forced to rollback an AI agent in customer service, perhaps because better monitoring lets them spot failures more quickly or that are otherwise going missed.

"The industry has assumed that better governance leads to better outcomes," said Daniel Morris, CPO at Sinch. "But that's not enough: If governance was the fix, the most mature teams would roll back less, not more."

So what's happening? "The most advanced organizations aren't failing less; they're seeing failures sooner," said Morris. "Higher rollback rates reflect better monitoring and control, not weaker performance."

Despite the many roll backs of AI agents, 98% of those polled said they were increasing AI investment this year, showing a few missteps isn't enough to dissuade customer service operations from the AI bandwagon.

Trust versus tech

Sinch also argued that companies were spending more time on trust, security, and compliance than they were AI itself. Three-quarters (75%) of those polled reported investment in those areas versus 63% for the technology itself.

Beyond that, 84% of AI engineering teams are spending at least half their time on safety infrastructure.

"Engineering teams are spending most of their time building and maintaining safety systems, a lot of which their communications infrastructure should be providing, instead of focusing on improving the customer experience," said Morris. "That's the guardrail tax that slows organizations down."

The report showed chatbots and email responses were the most popular use cases for AI agents in communication channels, both in use at more than six in ten organisations for those polled.

That was followed by social media, texting, and WhatsApp or similar messaging tools, with half of those polled using AI agents to manage these channels.

The main motivations are to improve customer satisfaction (36%) and boost revenue and conversions (24%), followed by reducing operational costs (16%).

FOLLOW US ON SOCIAL MEDIA

Consumers are growing increasingly concerned about the use of AI in customer service, new research shows, with trust in chatbots dropping and a desire for human interaction building.

A YouGov survey for software firm Pega found that two-thirds (68%) of people were either not at all confident about the way businesses use generative AI when interacting with them.

More than half (54%) said they weren't confident that organizations are using it responsibly.

“AI can be transformational for customer service – but it has to live up to customer expectations,” said Simon Thorpe, director of Pega.

“There’s a simple reason why we’re seeing a lack of consumer trust in the use of AI. There are just too many first-hand examples of businesses deploying these tools in ways that lead to dead ends and frustration."

The main gripe is that AI needs to be more reliable, with half (50%) of respondents saying they rarely or never get successful customer service outcomes when interactions are AI-only.

Almost as many (48%) don’t trust businesses using AI to completely handle their customer service interactions as a result.

Overwhelmingly, consumers prefer dealing with people, with 80% saying they always or often achieve better outcomes when interacting only with a human and two-thirds (65%) saying they prefer human-led support.

By comparison, just 2% said they wanted to interact exclusively with generative AI chatbots.

"Businesses must build back consumer trust by moving past simple chatbots and deploying predictable AI agents that consistently get work done on behalf of customers," said Thorpe.

"If businesses can use AI to make customer service faster and easier, they can drive massive new efficiencies while retaining customer trust.”

Consumers know AI isn’t going anywhere

Despite a lack of trust and confidence in chatbots, consumers told Pega they’re resigned about the role generative AI will continue to play in modern life.

While nearly half (49%) say they never actively choose to use it in everyday tasks, they still do. Just 14% say they actively choose to use AI less than once a month, with 10% saying they choose to use it every day. And more than a quarter (26%) think they probably use it every day without realizing it.

A recent study from Gladly and Wakefield Research found a similar dislike of generative AI. While 88% of customers told the researchers that their issue was resolved through AI or a hybrid AI-to-human interaction, only 22% said the experience made them prefer the company.

"Customers don't resent AI," the report noted. "They resent wasted effort. When AI loops, blocks access to a human, or forces people to repeat themselves, trust erodes — even when the issue is eventually resolved."

Customer service leaders don't have much choice about implementing AI, according to recent data from Gartner. More than nine-in-ten (91%) told the firm they were under pressure to implement AI this year.

Respondents identified improving customer satisfaction, operational efficiency, and self‑service success as their top priorities for 2026.

FOLLOW US ON SOCIAL MEDIA

Community Fibre is London’s largest alt-net and provides 100% full-fiber broadband in competition with larger players such as Virgin Media O2 (VMO2) and BT. With a goal to provide full-fiber connectivity to 2.2 million businesses and homes by the end of 2024, it had to rapidly expand its network as well as its customer service operations.

With this aim in mind, the firm partnered with AWS to deliver its customer service operations via Amazon Connect, the cloud-based contact center solution.

Community Fibre CIO Chris Williams is bullish on the firm’s performance so far, saying it continues to move at a pace and enjoys a 4.9-star rating on Trustpilot.

“Having that focus on London means that we can probably do things in a slightly different way to our competition, certainly the bigger alt-nets, and the VMO2s, the BTs, that are far more geographically spread,” he tells ITPro.

Opting for AWS against the odds

In the 18 months since he joined Community Fibre, the company has transformed sixteen areas in need of an overhaul, thanks in no small part to assistance from AWS. Despite the partnership bearing fruit, Williams tells ITPro it wasn’t a shoe-in when Community Fibre first began its search for a partner. 

“We did a big scan of the market, if I’m honest. Coming from BT, EE, there’ve been particular vendors I would have traditionally used in that space.”

Prior to his role at Community Fibre, Williams worked for five years as technology director for sales and service at BT, and nine years as director of IT service management at EE. Both roles prepared him for large call center implementations. At BT, one of the last projects he undertook was to deliver work-from-home capabilities to BT’s entire agent workforce, a massive undertaking that required large–scale application of existing technologies.

“I would say, at the start, Amazon Connect was one we'd added to our list because we've been having conversations anyway,” he admits.

“I’m not sure I expected the end of the process to be that we would have chosen, but we had some pretty strong criteria. And what we wanted to do was put the customer really at the heart of what we did, and experience is absolutely everything in terms of what we're doing.”

Embracing new communication channels

When it comes to its customer experience (CX), Community Fibre has been keen to avoid being a “one-trick pony”. Practically, this meant moving beyond voice and adopting an omnichannel approach including email alongside phone calls, and routing customers to the most appropriate agent at all times.

“We understand that particular channels are geared toward answering particular customer query types,” he explains. “And we wanted to make sure that we had the flexibility to be able to put the customer in the channel that was going to serve them best, as well as the one that these customers wanted to be served in as well.”

Going into the partnership, Williams says his team discovered no downsides to a more digital approach, in which customers could derive value from more personalized experiences.

“You still paid for services, of course, but it encouraged you to be able to focus on things like that experience without worrying about how many licenses you bought; how many seats in the contact center; do you need to spend another £10,000 because you need another five agents, et cetera.”

Down the line, Community Fibre is interested in further exploring Amazon Connect’s speech analytics features, available through Amazon Connect Contact Lens. This allows for customer conversations to be transcribed, customer ‘sentiment’ to be analyzed in real-time for instant feedback, and for agents to receive automatic performance evaluations.

A machine learning layer to the service also seeks to derive value from recurring data points across conversations, such as which agent sign-offs make customers the happiest. “I would say, for us, it's core to get things right first,” he says. “We know we have a number of data use cases that are obvious to us and probably a bigger focus than the analytics piece at the moment because we can see from our data that those are things we need to address.”

Finding the right use cases

Williams says in the lead-up to the Amazon Connect implementation, the firm was “frantically writing” some of these use cases to be delivered as soon as the platform was finalized. 

“We feel as though we've got the data points, but we're not yet universally using them to understand even 90% of the reasons the customer calls before they land,” he explains.

The same goes for some of AWS’ other advanced capabilities, such as the firm’s increasing focus on AI. While Community Fibre is interested in AI’s varied use cases, it’s more focused on ethical implementations.

“It’s fair to say that we could probably brainstorm 100 applications for it right now in our business where it could potentially add value. What we’re trying to do is hone down those opportunities to ones that are safe, that provide value to us and to our customers.”

What does a win-state look like for Community Fibre? Williams says using the data derived from its omnichannel contact center to more intelligently resolve customer concerns before they even have to get in touch.

“That's our end game, and we're trying to design our experience programs not to shave seconds off average handle time necessarily, but looking at where the core problem is and trying to resolve that before it gets to the customer. 

“The other thing is when when you're growing the base as fast as we are, a good result for us is the same number of calls comes in next month as this month. Since I joined, we're three-and-a-half times the base that we were, and if we can negate the product-to-consumer impact of the base growth then we're doing a tremendous job.”

8x8 has announced its new partner ecosystem which will allow organizations to embed technologies such as AI into its unified communications platform to enhance business intelligence, insights, and analytics.

The new 8x8 Technology Partner Ecosystem has been designed to democratize next-generation integrations without requiring complex, custom development, or expensive overhead costs, 8x8 said. 

Fresh integrations with more than 20 new partners mean businesses of all sizes can add AI-driven customizable experiences to help with their customer experience goals. 

“Organizations expect their technology solutions and providers to innovate, adapt, and evolve to meet their current and future business and customer needs,” said Hunter Middleton chief product officer at 8x8. 

“We are thrilled to announce the 8x8 Technology Partner Ecosystem, and even more excited to partner with so many industry-leading, innovative CX vendors.

“The program complements our native AI-driven 8x8 XCaaS platform capabilities, giving enterprises the flexibility to build the right solutions to meet and exceed their customer experience goals.”

The move comes as the pace of innovation in the AI space is being rapidly accelerated by “billions of dollars of venture-backed investments in thousands of startups,” according to Zeus Kerravala, founder and principal analyst at ZK Research.

"Getting locked into a single communications provider's AI technology is holding back organizations from truly benefitting from the phenomenal rate of innovation taking place,” he said. 

“8x8's new Technology Partner Ecosystem allows enterprises to easily plug in the right AI solutions that best meet their business needs and transform the contact center and customer engagement."

The ecosystem is built on the 8x8 experience communications as a service (XCaaS) open architecture, an integrated cloud contact center, and a unified communications platform that bundles together contact center, business phone, video meetings, team chat, and SMS capabilities.

8x8 said it now has more than 30 total members, thanks to the latest additions of Cognigy, Avoira, Awaken, Balto, Converse360, Cresta, Engage Hub, LogicDialog, PromptVoice, Roboyo, Skybox Communications, and Syndeo.

The partner network aims to help customers break down data barriers and surface data across a variety of apps and interfaces, as well as ensure conversational context is maintained during any handoffs between the XCaaS platform and partner integrations.

Ultimately, 8x8’s aim is for customers to be able to access next-generation technology, create tailored solutions without expensive custom development, as well as leverage curated, seamless integrations.

Members of the Technology Partner Ecosystem will also be able to collaborate on go-to-market strategies, including promotional support, sales enablement, joint field engagements, and lead sharing.

“Customers want to engage across new channels and through emerging technologies, so we are excited to partner with 8x8 on delivering transformative AI-powered digital and voice CX solutions to their global customer base leveraging Cognigy's industry-leading conversational and generative AI automation platform," said Hardy Myers, SVP of business development and strategy at Cognigy. 

"The 8x8 Technology Partner Ecosystem has already demonstrated remarkable success propelling 8x8 to the forefront of Cognigy's strategic partners, enabling us to penetrate untapped markets and attract new customers.”  

It can be tempting to label every new kind of automated technology tool as something that's entirely never-before-seen. This is especially tempting with innovations in the artificial intelligence (AI) sphere. Although ChatGPT, and other forms of generative AI, may be grabbing the headlines, based on the breadth and depth of use cases, text-generating chatbots have been around for a while and have come to define the customer experience.

Currently, interactions between chatbots and humans mostly happen in the customer service space, and these interactions are mixed. While in some cases, chatbots can helpfully surface the answer to your pesky question that's buried deep within a website's FAQs, a lack of nuance and context defines most interactions, and they serve as a frustrating barrier before the inevitable phone call with a human operator.

ChatGPT, meanwhile, offers far more depth, and possibilities in this space, due to the way the system's been engineered. Yes, it’s still a chatbot and will certainly exhibit some flaws that won’t make it suitable for all use cases, but Open AI’s ChatGPT could revolutionise the interactions between people and machines in all kinds of instances.

When were the first chatbots used?

The history of chatbots can be traced all the way back to Alan Turing in the 1950s. After the advent of his Touring Test, the first chatbot that we can confidently identify, as Greek researchers did in a 2020 literature review, was created in 1966. That bot, named ELIZA, combined pattern matching with pre-written responses and was the first attempt at identifying whether a human audience could tell if a bot speaking to them was, in fact, automated. The notion of AI, labelled as such in the mid-fifties by a group led by researcher John McCarthy, would formalise the field as an area of study.

The first time AI was used within a chatbot environment came in the late 1980s. Although we can’t go back in time and speak to Turing or McCarthy, it’s fair to say that neither probably conceived of tools like Siri and Cortana being ubiquitous in our current moment. OpenAI’s entry into the market in 2015 is remarkably recent, with its technologies leaving significant impressions on enterprises in such a short space of time.

How do ChatGPT and other chatbots work?

To understand the difference between ChatGPT and other types of chatbots it’s important to recognise that, while they’re often grouped together for convenience, the term is not ubiquitous. To further complicate things, tech companies might be unable to adequately define the different chatbots they’ve developed, of varying sophistication and application.

For simplicity's sake, let’s first divide chatbots into two camps. The first, rules-based chatbots, are what they say on the tin: an if-this-then-that system governs their behaviour. They are not, as a rule, adapting or thinking for themselves. One example of a rules-based system would be the pop-up you see if you accept Facebook’s invitation to message a business page and are confronted with prompts asking if you’re looking for information like an establishment’s hours or current array of product listings.

The other broad category is AI-powered chatbots. These interfaces rely on natural language processing (NLP) and machine learning rather than pre-programmed chains of responses. These are smart tools to the relatively analogue chatbots of old. From this point, they’re often characterised by either their use case or where they appear. There are also so-called hybrid bots, which use some elements of both rules-based architecture and AI.

What makes ChatGPT so different to chatbots?

There are so many cautionary tales in this space, including Microsoft’s Tay experiment or Meta’s BlenderBot 3. Where previous attempts at generative AI chatbots quickly devolved into chaos, ChatGPT is programmed in such a way as to reduce the sorts of verbal implosions that led to their downfall. In a few words: it is discerning. According to OpenAI, this has been done through reinforcement learning from human feedback, a process in which the bot’s dataset – in this case, an adapted version of sibling product InstructGPT’s – is reinforced via humans.

“We trained an initial model using supervised fine-tuning: human AI trainers provided conversations in which they played both sides – the user and an AI assistant,” the organisation has previously said in blog post. “We gave the trainers access to model-written suggestions to help them compose their responses. We mixed this new dialogue dataset with the InstructGPT dataset, which we transformed into a dialogue format.”

Microsoft's Tay attempt was almost like leaving a chatbot to learn from an angry pack of wolves with problematic intentions – in this case, social media users. ChatGPT’s initial training was more like a university programme with very defined parameters. The bot, which has its own set of ethical concerns, is currently in a research release phase.

What does ChatGPT mean for the future of human-machine interactions?

When comparing ChatGPT with other chatbots, it’s important to make clear distinctions and not get sucked into what can only be described as a fervent hype machine.

The history of both AI and chatbots is long, winding, and complicated. While generative AI is the new and shiny trend, ChatGPT is just one example of a chatbot that people might interact with now and in future years. This is only one example of an AI-powered textual generation tool, for instance, with OpenAI also working on GPT-4, and other models sure to dazzle the market in future.

It’s too early in the release cycle for ChatGPT to say, for sure, how many businesses might adopt it or integrate its technology into their own tools and services, but a handful of companies have already begun exploring integrating this technology. What’s clear, for now, is the global enterprise landscape is on notice regarding the potential impact of generative AI tools like ChatGPT.

By providing the broadest and most complete set of machine learning (ML) services, AWS is able to meet its customers wherever they’re at in their ML journeys and help them achieve specific objectives.

This eBook shares why hundreds of thousands of organisations use AWS ML to resolve challenges and create new opportunities within their organisations.

Download now to learn how to:

• Improve customer experiences
• Optimise business operations
• Accelerate innovation

Provided by  AWS

Technology continues to be an advancing force, amplified in recent years by the COVID-19 pandemic, bringing greater digital experiences into our daily lives. 

These technological innovations, with smart technology and intuitive devices appearing in our homes, as well as smarter methods of collaboration and communication in our work, have raised the bar when it comes to customer expectations.

As we move forward in this digital era, a positive digital experience is no longer "added value"; businesses should be offering this as a basic standard.  

In a survey conducted by PwC in 2022, 51% of consumers said they would switch brands based on a poor digital experience, and with customers more discerning than ever, this should encourage companies to review their interactions and focus on how to improve their digital engagement.

But what does a good digital experience look like, and why do they matter?

What is digital experience?

A digital customer experience (DCX) refers to how a customer interacts with an organization through its digital touchpoints: websites, mobile apps, or social media platforms. This also includes multiple interactions - for example, where a customer may have ordered a product online, and then spoken to the customer service team via a chatbot regarding the order.

It is very much part of the entire customer experience (CX) and it's important to consider how each individual interaction can contribute to the customer's overall perception of the organization

Why is digital experience important?

In brief, customers are demanding more. They're much more savvy when it comes to prices and products and where to get the best deals, so to stand out from the competition, it's the digital experience that needs to be the reason customers return.

A report from PwC found that 73% of customers rated their digital experience as the most important factor in their buying decisions, surpassing price and product quality, and Smith.AI cites that companies offering better experiences can gain 17% more growth than competitors.

Long-term, better customer experiences equate to higher customer retention, higher lifetime customer value, and greater brand equity, with consumers prepared to pay a 16% premium for a better service.

The costs of bad digital experience

But what happens if you don’t make digital experience a priority?

It’s important not to cut corners when it comes to creating the seamless, personalized experiences every customer expects, particularly when your customers are the whole reason your business exists in the first place.

And in an experience-driven market, you have to compete to stay afloat. According to Invesp, 89% of customers who are leading the way in customer retention and loyalty say it's down to the experience they provide.

And when 63% of customers will stop doing business with brands that don't get it right, the focus needs to be on enhancing those personal digital connections. UK brands that don't risk losing nearly £15 billion a year if they don't.

What are customers looking for in a good digital user experience?

Now that we’ve weighed the advantages and disadvantages of digital customer experience transformation, let’s take a closer look at what you should be trying to achieve with your transformation.

How customers perceive interactions with your brand is everything. You want them to come away feeling that they had an easy, frictionless, and delightful experience.

Interactions need to be consistent across all channels, whether physical or virtual because people perceive these various interactions as one experience. For this reason, it’s also important that each engagement is based on earlier ones rather than siloed, taking into account previous activity and purchases. This personalization is essential, with customers saying that ‘being treated like a person’ is important to winning their business.

Ensure that your interactions are also immediate and conversational. Don’t let too much time lapse after a sign-up, a purchase, or an email before you engage with your consumer in a real way. No one wants to feel like just another number, so make sure you have a comprehensive view of each customer to show them that they matter.

Your digital experience shouldn’t stop at how you interact with customers. Besides personal, seamless experiences to show customers they matter, you might take a look at what your business offers. Customers prefer brands that are constantly innovating, so consider how your company could find new ways to meet customer needs.

Next steps in your digital experience transformation

Despite increased expectations, about half of customers report that these expectations aren’t being met. The opportunity is there for the companies that are willing to jump into transforming their digital user experience and standing out in a crowded marketplace.

Now that we’ve broken down the reasons why you should consider this transformation, take a look at this quick guide on digital transformation quick-wins – simple steps that every business can try to quickly overhaul their operations, and their customers' experience.