A practical guide for CEOs and senior leaders who are being asked to set AI strategy without yet having a full grasp of what AI can and cannot do, written by Tim Butler.
There is a quiet truth that comes out in almost every senior leader conversation we have at the moment. The board and investors have requested a clear AI strategy and the leadership team is under pressure to move fast. However, in private, when the door closes, the leader sat opposite me tells me they don't fully understand AI yet. They have used ChatGPT, they have heard about agents, they know it matters but, when it comes to putting a strategy together or approving budget requests, they don't entirely trust their own judgement.
That is a more common position than the conferences and LinkedIn posts would suggest. According to BCG's AI Radar 2026, 72% of CEOs now say they are the main decision-maker on AI in their organisation, and half believe their job stability depends on getting AI strategy right. Among UK CEOs specifically, only 44% are confident AI will pay off, the lowest of any region surveyed. Plenty of pressure to act, less confidence than the headlines might suggest. The reality is that you cannot know everything in this space, because there is simply too much to know. The risk is not in admitting that, it's in pretending otherwise, and then making strategic decisions on top of an understanding that is not yet solid.
The thing about being a CEO or Director is that you are paid to have a view. So when AI lands on the agenda, the temptation is to develop a view quickly, often by reading a few articles, attending a couple of vendor demos, and listening to whoever in the team sounds most confident. That can give you the surface vocabulary. It rarely gives you the judgement to direct a multi-year shift in how your business operates.
What we see consistently is that leaders who try to bridge the understanding gap on their own end up making one of three mistakes. Each of them is expensive, and each of them is avoidable.
Buying AI to Tick a Box
Automating What’s Already Broken
Tool First, Foundations Later
Someone, often a board member or an investor, asks what the business is doing on AI. The leader needs an answer, so a tool gets bought, licences get rolled out, and the strategy becomes, in effect, "we have AI now." Six months later, very little has changed in the business and nobody can quite articulate what the investment delivered. Activity is not the same as impact.
AI gets pointed at whatever the business currently does, and the result is the same broken process running faster. We say it often in our workshops:
Don't just automate what you have today, automate to get the best outcome.
Real value comes from stepping back and asking what the desired end state is, then working out where AI changes how you reach it.
Many leaders, under pressure to make AI happen, jump straight to the technology question. Which platform, which licences, which provider. The foundations of successful AI use sit underneath all of that and skipping them produces a predictable kind of failure where no tool, however good, can deliver what the demo promised.
You do not need to become a technical expert. You do not need to know how a transformer works or how the latest models are trained. What you do need is enough understanding to ask the right questions, sense-check the answers you are given, and recognise the difference between a strategy that fits your business and a tool being sold to you.
Two principles do most of the work here:
The first is to accept openly that you cannot know everything, and to build a small group of trusted advisors around you who can fill the gaps. That might be an internal AI lead, an external partner who works with this technology every day, or a peer leader who is a year or two further along. All of them are worth more than another whitepaper.
The second is to remember that AI is not one thing. The discussions about chatbots, agents, machine learning, image generation, and embedded SaaS features are not the same conversation. Different applications need different approaches, and any strategy that treats AI as a single homogeneous decision is going to fall over the moment it meets the reality of your business.
In our workshops, we work through what we call the four foundations of successful AI use. They are the bedrock for any business that wants AI to deliver commercial impact rather than activity. Whenever a proposal lands on your desk, internal or external, the foundations are your test.
Data. AI is only as good as the context you give it. If your data is fragmented, out of date, or sitting across systems that do not talk to each other, no AI initiative built on top of it will perform reliably. We have a page on data hygiene for AI readiness you might find interesting.
Systems and Technology. What matters is not which tool you pick, but whether it integrates with the systems you already run. Your AI needs a connected view of the business to do its job, and a best-in-class tool that sits in isolation will deliver less than a good-enough tool that talks to everything else.
Processes. AI delivers value when it is mapped onto a process that has been properly understood and, where necessary, redesigned. It does not fix processes that are broken to begin with.
People. Adoption is where most AI projects quietly fall over. The technology is rarely the problem.
If a proposed AI initiative has solid answers across all four, it is the start of a strategy. If it only has answers on the technology side, it is a sales pitch in better clothing. The detail of how to assess each foundation, and how to build a roadmap from where you are to where you want to be, is the bulk of what we work through in the AI Workshop for Leaders.
There is a useful distinction worth holding on to here. Some AI projects are painkillers, and some are vitamins. A painkiller solves a problem the business is already feeling, where the cost of doing nothing is obvious. A vitamin is a nice-to-have that might help over time but is easy to deprioritise when something more pressing comes up. The strongest AI projects are painkillers. The ones that fail tend to be vitamins bolted on the side.
The projects that deliver are pointed at where the business is losing today:
That last one matters more than most leaders realise. Making your most expensive people more efficient delivers a much bigger return than making your cheapest people more efficient by the same margin.
If you are trying to work out which problem to point AI at first, that is the place to look. The pain, not the tool.
The leaders who make real progress with AI are not the ones with the biggest budgets or the most sophisticated technology stacks. They are the ones who have quietly accepted they cannot know everything, built a small group of trusted advisors around them, and gone back to the fundamentals. They start with the pain rather than the tool, and they treat AI as a question about how the business runs rather than a question about which platform to buy.
Your competitor using AI more powerfully than you do is a bigger threat to the business than AI itself. Closing the understanding gap is the difference between being in that position and being on the other side of it.
If any of the patterns described above feel familiar, the most useful next step is structured, facilitated time with people who do this work every day.
That is what our AI Workshop for Leaders is built for. A full day with me and the team, working through:
It is a focused, practical session built around your real situation, not a generic overview of AI capabilities.
Reserve your place at the next AI Workshop for Leaders.