The importance of pilots, open source, and consultancy in the new world of AI

Artificial Intelligence (AI) is now part of daily life, and you’d be forgiven for thinking that everyone is already AI-savvy and integrating it into their daily lives and work. On a consumer level, this may be true, but the real work of integrating AI into the enterprise – and even into the mid-market – is only just getting started.

In fact, when we look at the industry today, a large number of AI pilots and proof-of-concepts are failing to pass to production. The challenge is that we have high aspirations for AI, but also that it is complex and rapidly evolving! VARs, SIs, and resellers need to keep their options open and stay agile if they’re to succeed – but let’s take a step back and look at the main issue in more detail for a moment.

Problems with PoCs

As anyone who has used generative AI on a consumer level knows, it is very powerful and helpful, but not perfect. AI tends to hallucinate and produce false results from time to time. Although it might have limited impacts on the individual level, when it’s deployed into the enterprise and business processes, it becomes unacceptable. Many organizations are finding that their AI pilot tests are coming across hallucinations and reliability issues, which are all hindering the rollout.

However, over the last few years, we have seen many successful implementations from other ‘flavours’ of AI, in particular, machine learning. The manufacturing and healthcare sectors have used this kind of AI to great effect, leveraging it for process optimisation and pattern recognition, which is a far more mature application than generative.

AI is also tremendously complex, which tends to hinder effective pilot tests because it’s hard to set success criteria. In many cases, organizations also struggle to scale AI after their pilot tests, which makes ROI hard to achieve. All of this is a significant opportunity for the channel to help consult with customers, guide them through projects with hard-won knowledge and expertise, and provide a fresh perspective on what will and won’t work.

Place your bets: Or keep them open

As I’ve already said, AI is still evolving. No one knows what the landscape will look like when it reaches maturity. It’s also likely that even when it does reach maturity, it’ll continue to change, much like how the internet continued transforming decades after its establishment - from static websites to social media, streaming, cloud services, and beyond. For this reason, it’s heartening to see that many of the models in the top ten AI listings are open-source, not to mention that the performance gap between these open-source models and their proprietary rivals is closing fast. We believe that the future of AI should be built on transparency, accessibility, and collective progress.

The advantage of using open source models for both customers and partners is that it’s easier to change. Open source is cheaper, provides better ROI for customers, better margins for partners, and helps both sides to avoid lock-in.

Furthermore, in the enterprise space, AI integrations will be critical as AI delivers more value when it's connected to existing data owned by companies. But one challenge with integration is that organizations - especially large ones - can still have many legacy systems in place, which require custom solutions to be integrated effectively. Data is the most critical asset for any company, so partners that can help to modernize legacy systems and unlock the data for AI integration will be in high demand.

The agent opportunity

However, despite these challenges, there is one very clear opportunity in AI that is approaching rapidly: agents. These are AI systems that can use the tools that have historically been used by humans, and ‘do the work’ instead. We’ve all heard the stories about AI making restaurant bookings for humans, but in the future, this kind of approach will be the norm.

There’s a tremendous amount of work to do at the back end before this becomes standardized. In many ways, the shift from manual web use to chatbots has been like the shift from cars with manual transmissions to automatic gearboxes – but the shift to agentic AI will be more akin to the shift from automatic cars to autonomous cars.

This is a challenging step, because for many organizations, taking your hands off the wheel is both daunting and uncertain: we don’t know the optimal place for the human to give away control and the agent to take it. But the change is coming, and in a very large number of cases, agents will be the primary users of services, not humans.

The challenge is that this requires a lot of re-engineering. Interfaces that have been designed for people will need to be re-coded for agent use, usually via APIs. This means re-thinking processes, understanding where users fit into the journey, and where the agent comes in. Customers, partners, and vendors will need to work closely together to meet this challenge, and it’s something we’re already looking at very closely with the launch of a new partner program to help organizations get their SaaS ready for AI.

Again, this is an opportunity for partners to guide and consult, offering guidance on how end-users use services, as well as how to optimally engineer experiences for both humans and AIs. However, one of the main challenges for partners is that the AI landscape is still rapidly evolving, and no one quite knows which horse to back, which makes open standards doubly important.

Know (and guide) your customer

With such rapid change, guiding and supporting customers is essential to successful AI projects, creating and maintaining margin for partners. The opportunity that AI agents present is unprecedented, but it’s critical that partners help to set clear objectives, shape successful pilot tests, build scalability into use cases from the start, and measure ROI effectively.

This then helps customers turn AI tests into profitable commercial projects. At the same time, it’s vital that partners stay agile throughout the process – and open-source is a key way of doing this and avoiding lock-in.

Change is coming, and the channel has an enormous opportunity and role to play in making sure that customers are guided to pragmatic ways of rolling out AI, building on firm foundations and establishing reliable, agile systems that will stand the test of time – while also providing strong revenue streams for partners – even when the machines are in charge!