Boards are increasingly expecting results from AI, but technology leaders are still working out which bets to make, where to invest, how to govern them and what readiness looks like when the technology changes every few months.
This tension shaped the latest NashTech Leaders Lab roundtable in London, where 20 senior technology leaders from across the UK shared their experiences.
This article captures just some of the discussion, which may help other CIOs and CTOs navigate the same decisions. Let’s dive into it.
Before each Leaders Lab roundtable, delegates complete a short survey to help shape the discussion. For this event, technology leaders were asked to assess their organisation’s AI maturity. Of the 20 respondents:
Looking 24 months ahead, nobody was expecting to remain at the exploratory stage. Instead:
When asked what the biggest impact of AI would be over the next three years, the most popular answer was ‘enterprise decision-making and intelligence’. The biggest concerns were around AI’s reliability and accuracy of output, a topic that was widely debated in the peer-led discussion.
NashTech’s Senior Technology Advisor, Chris Weston, kicked off the discussions:
“We have moved beyond the question, ‘Should we use AI?’ The questions now are: what are we going to do about it, how does it affect us as a business, how does it affect our individual roles, and how does it change the way customers, employees and suppliers interact with us? Once you think about those questions properly, AI becomes a significant transformation issue.”
Boards are adding pressure. Almost half the people in the room (45%) said AI was treated as a strategic priority and investment area. A further 25% said their board encouraged implementation but wanted clearer evidence of value. For 15%, expectations were growing faster than organisational readiness and another 15% had significant concerns around risk and governance.
This is an unusual position for technology leaders. Instead of fighting to put an emerging technology on the agenda, many are having to moderate their board’s enthusiasm and explain what safe, valuable adoption looks like.
“With cloud, CI/CD and other changes, we spent a lot of energy convincing non-technical people why they mattered. With AI, it is the other way around. You have to say, ‘I love the enthusiasm, but you also need to understand costs, risks and potential unknowns.’”
Alex Davidson, Technology Advisor, Vertech Consulting
“Historically, at the board, you had to fight for every pound of technology investment. Now leaders are saying, ‘I’ve got £100, what can we do with it tomorrow?’ That creates a huge opportunity, but it also changes the challenge: IT leaders need to be at the table, helping the business decide where AI can genuinely create value and what the risks are in doing that."
Paul Stapleton, Interim CIO, Balance 360
“Technology leaders need to be more proactive than ever. We cannot wait for the chair or the board to set an AI agenda based on what they have heard elsewhere. We have to educate the board and help it understand the opportunities, limitations and pace of change.”
George Lynch, Global CIO, NashTech
“We need to be strong on balancing this. You need guardrails, but you also need to encourage people and give them AI literacy. That is on us more than it has been with anything else.”
Andrew McManus, Head of Technology, Bridgepoint
| Key takeaway: CIOs and CTOs need to explain both where AI is worth investing in and where restraint is necessary. They must respond to pressure from the board by showing how AI can create value, while being clear about the data, governance, skills and operational changes needed to achieve it. |
As mentioned earlier, reliability and accuracy were the most widely selected concerns in the survey, with cost and human accountability close behind. The leaders agreed that dependable outcomes rely on the quality of the input, the skill of the user and the controls around the system. It requires trust in the technology, but more so in the people using it.
“It is often less about the output of the tool than the quality of the input. You can give people a really good car, but only a good driver will get the best performance out of it. The AI tools will get better, but people don't always know how to use them properly. Person A and person B might ask the same prompt in different ways and create different outputs.”
James Wehner, Founder of JDW Consulting
“We can both do the same thing and get different answers [from the same AI input]. Once you introduce that doubt, you can lose trust with non-technical people on boards and elsewhere in the organisation.”
Alex Davidson, Technology Advisor, Vertech Consulting
“It is not human in the loop anymore. It is human in the lead. We have to make sure there is genuine human input rather than blindly accepting the output.”
Giles Lindsay, CIO adviser to the Government of Jersey
The leaders agreed that responsibility starts with the education system, but businesses must also give people the right knowledge and skills.
“If people leave education without learning how to work with AI, we’re asking them to do jobs they haven’t been prepared for. They may have been taught how to write software, but they also need to know how to direct AI tools and judge the results. For example, that’s the shift from software engineer to software orchestrator, and education needs to reflect that."
Giles Lindsay, CIO adviser to the Government of Jersey
“The leadership skills may be broadly the same, but now everybody thinks they are an AI developer. Giving people access to these tools can produce excellent proofs of concept, but it can also produce a large amount of technology that has not been governed, tested or designed to operate at scale.”
Andrew McManus, IT Director, Bridgepoint
| Key takeaway: Leaders therefore need to ask more than ‘Can we trust AI?’ to ‘Do our people have the right skills?’ and ‘What combination of data, instructions, expertise, testing and human accountability makes this use case trustworthy enough?’ |
Around the table, the concern extended beyond prompt technique. Leaders questioned how organisations will develop judgement when early-career work is increasingly automated.
“We are going to get cognitive offloading from people doing the job today and end up with cognitive debt tomorrow. If we blindly accept that the outputs from AI models are de facto truth, we are in trouble.”
Giles Lindsay, CIO adviser to the Government of Jersey
“Is it the organisation’s job to train people for this new world of AI, not just software developers but, maybe everyone? No one knows the full answer, and it is going to keep changing, so perhaps our job is to get everybody to a certain level.”
Andrew McManus, Head of Technology, Bridgepoint
“It depends on what organisation you want to be. If your AI maturity is at level four, you’ve got a bigger gap to bridge, and you need to understand what the education system is producing. Hopefully it will evolve, but you can probably guarantee it’ll be much slower than the technology. As business leaders, we have to understand where we want to go and how we’re going to get there. Then, as we recruit talent, find people who are hungry to learn and willing to adopt new standards, while still having the core foundations you know you can’t succeed without.”
Simon Hall, Head of Product, Tech and Data, Pivotal
John Bovill of Gordon Ramsay Restaurants added another dimension. In his view, effective AI teams need visual design, analytical engineering and storytelling. His concern was that organisations may overvalue the engineering skill set and leave designers and storytellers behind, weakening the customer experience and the ability to differentiate.
“You need three skill sets: visual design, an analytical or engineering brain, and storytelling. My concern is that the engineer will run ahead and we will leave the designers and storytellers behind. If we do, we risk losing our competitive advantage.”
John Bovill, IT Director, Gordon Ramsay Restaurants
“We can teach technical skills, but the future of working with language models may depend on a much wider set of human capabilities. Do we only hire engineers, or do we need linguists, psychologists, designers and people from other disciplines?”
George Lynch, Global CIO, NashTech
Alex Davidson agreed that technology tools change rapidly, but the principles of good engineering, data management and networks remain relevant.
“AI is creating new skills requirements rather than replacing the need for strong professional software engineering foundations.”
Alex Davidson, Technology Advisor, Vertech Consulting
George Lynch summarised the point by saying:
“We have to hold two ideas at once. AI will change the economics of software delivery, but senior engineers still have to come from somewhere. We need enough opportunities for junior people to learn the craft, even while the tools can perform more of the work. People entering the profession will naturally use the tools in front of them, and their trust in those tools may become implicit. If the tools change every six months, the organisation has to keep refreshing its capability and teaching people when not to trust the output.”
George Lynch, Global CIO, NashTech
The speed of model development creates another problem. Organisations can spend months integrating a model only to find that a newer, cheaper or more capable option has arrived before the first implementation reaches scale.
“AI models are advancing every few months, often with significant increases in capability. Yet even with a very practical approach, it takes time to embed that capability into real organisational processes. By the time you have done that, another generation has arrived. The challenge is therefore not simply adopting AI, it is designing organisations that can continually absorb and exploit that pace of change.”
Paul Stapleton, Interim CIO, Balance 360
“For many use cases, the model available now, or even a year ago, is fine. Some organisations get stuck chasing the newest model when ‘good enough’ is already here.”
Alex Davidson, Technology Advisor, Vertech Consulting
“Changing models is like changing the engine from unleaded to diesel while the car is moving. That is a different question from whether you know how to drive the car. The planning element needs much more detail.”
Azhar Sadique, CEO of SCAURUS
“The AI landscape is changing quickly. We are seeing new models and more capable implementations, including models working together and using third-party tools. Anyone who claims to know exactly what this will look like in six months is being brave.”
Chris Weston, Senior Advisor, NashTech
These comments suggest a practical architecture principle:
| Key takeaway: Leaders discussed token prices, data sovereignty and the possibility that access to a model could change for commercial or geopolitical reasons. Portability may never be perfect, but knowing which dependencies are tolerable is now part of business continuity planning. |
One concrete example of AI value came from Simon Harrow, Co-founder and COO of Tilt Insurance. His advice was not to chase the biggest transformation story. He argued that individual and team-level improvements often offer substantial value with lower cost and risk.
“The AI stories that get press are the large, ambitious builds. But in our experience. whatever you build today can be replicated months later, far faster and at a fraction of the cost. If you're relying on AI to build a defensible moat, think carefully."
Simon Harrow, Co-founder and COO of Tilt Insurance
Simon described an underwriting workflow that previously took around two days for a large, complex case. Underwriters had to turn documents and emails into structured data, verify information against external sources and load it into an underwriting platform. According to Simon, an AI-assisted workflow now completes that processing in about two minutes, with 98% accuracy.
“You don't always need the latest frontier model. Established models are often more than capable of the job, and far cheaper. We keep human gates where they matter, but we trust AI with more as it proves itself."
Simon Harrow, Co-founder and COO of Tilt Insurance
He offered a second example from quality assurance. The business previously audited 3% of customer contact calls manually. It now uses AI to assess 100% of calls across channels, accepting that the automated review is not perfect.
“Auditing 100% of cases at 90% accuracy beats auditing 3% at 100%. For us, the trade-off is worth it."
Simon Harrow, Co-founder and COO of Tilt Insurance
|
Key takeaway: This is a useful way to assess AI’s value. Instead of comparing perfect AI with perfect human performance, compare the proposed service with the current process based on cost, coverage, speed and errors. |
Several delegates warned that AI strategy often stops at efficiency. That can leave a much harder question unanswered: if a firm becomes dramatically faster at doing the work that it currently charges for by time, what happens to revenue, differentiation and demand?
“In law, something sold for a hundred years by the billable hour is now being made more efficient. The next question is why: do you want to do more work, or charge more when the client is asking to pay less? Firms are not planning for that third and fourth stage of their business model.”
Azhar Sadique, CEO of SCAURUS
“I see it as not only for the organisation but for our clients too. Clients are becoming more AI-savvy, so we need to ensure we’re able to meet their needs.”
Uliana Hattersley, Tech Operations Director, Buzzacott
| Key takeaway: Technology leaders must take these downstream scenarios into the boardroom earlier. A pilot may improve a task, but at scale the same capability could alter pricing, staffing, routes to market or the basis on which customers choose a provider. |
AI readiness means being able to answer these questions as the technology changes:
AI readiness is ultimately the ability to answer those questions repeatedly as the technology develops. For support mapping the answers to these questions, our NashTech AI advisory team is here to help.
Join the NashTech Leaders Lab community
Leaders Lab brings senior technology and tech-centric business leaders together for candid, practical conversations, small-group events and peer-created resources. Visit the Leaders Lab hub and request to join the community to receive updates and invitations to future sessions.