Nearly 1 in 4 senior leaders in large UK firms say they pretend to understand AI better than they do. I understand why: when I didn't know something, I asked, and learning AI in today's noise was still hard. That is why it pays to listen to how other people see AI, treat it and use it in practice.
This piece is about how to learn AI as a manager without faking it: a way that starts from the work you already run and the gaps you already have.
How many leaders pretend to understand AI?
More than you would guess from the meetings. In a September 2026 survey of UK leaders in companies of 1,000+ people, 23% of senior leaders said they pretend to understand AI better than they do, and 22% said they feel out of their depth (Emergn, via IT Brief UK).
Further down, most managers are still at the start. The Chartered Management Institute found in June 2026 that 68% of managers are still in the early stages of experimenting with AI or building pilots, and 70% go to generative AI for advice rather than to their own manager.
These are UK numbers and mostly larger firms. I would not stretch them to every company of 100 people. But the pattern matches what I hear from people who run operations, finance or marketing: the tools moved fast, and few people got the time to catch up.
Why do leaders pretend instead of asking?
Because asking feels expensive. If you run a team, "I don't know how this works" sounds like a weakness, and AI is the topic where everyone is supposed to have an opinion already.
There is a second reason, and I think it matters more. Managers often have nobody to ask. Gallup's 2026 report found that 23% of managers felt lonely for much of the previous day, and only 22% of managers are engaged at work. Your team looks to you for answers. Your boss expects results. The people who would understand your problem work somewhere else.
So the gap stays private, and pretending is the cheapest way to get through the next meeting.
What does pretending cost?
My guess, not a study: decisions get made on vibes. A tool gets bought because a competitor has it. A pilot runs for 3 months and nobody can say what changed. The team copies the leader, so if the leader fakes it, the team fakes it a little too.
You don't have to become an engineer to fix it. You have to make the gap small and specific enough to close.
How to learn AI as a manager, step by step
This is what worked for me. I run marketing, not engineering, and I learned by doing it on my own work. Something similar could help you; treat it as an example, not a recipe.
- Pick 1 process you own. Not "AI strategy". One thing that repeats: the weekly report, the supplier check, the onboarding of a new hire.
- Write down how it really works today. Including the workarounds, the spreadsheet only one person understands, the step everyone skips. This map is useful even if you never touch AI.
- Find the 1 step that eats the most time or breaks most often. That is where you try a tool first: small and reversible, with a manual fallback and a clear idea of what "better" means.
- Write down what broke. When I started running our marketing with AI agents, the first thing that broke was me: every draft came back to me and the whole system moved at the speed of my reading. I only fixed it because I wrote it down. (I described it here.)
- Show it to someone in the same seat. Alongside your own team, show it to someone who runs a similar process elsewhere and will tell you honestly that your step 3 is the wrong step.
Step 5 is the one most people skip, and it is the one that stops the pretending. It is very hard to fake understanding in front of someone who does the same job.
Why learn with peers, not alone?
Because the gap is easier to admit to people who have the same one. Courses give you structure. Peers tell you what happened when they tried.
A small example from our own work. When we first put our community page online, someone from exactly the group we wanted to reach told me they liked the idea, but the page read like the first step of a sales process: as if applying meant signing up to a company's funnel, not joining a group of people. What mattered to them was a human reason: there are not many people you can get honest, normal feedback from when so much of your work now runs through AI.
After that conversation I went through the page element by element and kept changing whatever made it look like a funnel. That is the whole point of a peer group: 1 honest comment that changes what you do next.
What are we doing about it?
Building software used to belong to engineers. That has changed: a finance lead, a logistics manager or a head of marketing can now build first versions of the tools that help them run their process. Production software still needs engineers. We think this is a structural shift, and we care about the people living through it, not only the tools.
So we show the path the only honest way we know: we work like this ourselves and show it as it is, including what breaks. Our team reserves time for a small group of people who run processes:
- You work on your own process, not on an example from a slide.
- The founding group helps decide how the group works and how often it meets.
- You share only what you want; general words are enough, no names or numbers.
- Nothing is sold in the group, and applying does not put you on a sales list.
- It's free for the founding group, which starts on 1 December 2026; applications are open until 27 November.
In my view, the people who will shape how AI is used in operations are the ones who start on a real process now, say out loud what they don't know yet, and learn with others instead of alone. A year from now, they are likely to be the ones others in their industry ask how they did it. That is the group we are building.
If you would rather learn this with people who run processes like yours, apply with your process. Applications close on 27 November.
FAQ
How do I start learning AI as a manager if I am not technical?
Start from 1 process you already run. Write down how it works today, pick the step that costs the most time, and try 1 tool on that step only. You learn faster on your own work than on a generic course, because you can judge the result.
Is it normal to feel behind on AI as a leader?
Yes, and you are far from alone. In 2026 UK surveys, 68% of managers said they were still in the early stages of experimenting with AI (CMI), and 23% of senior leaders in large firms said they pretend to understand AI better than they do (Emergn).
What is an AI peer group for leaders?
A small group of people in similar roles who work on their own real processes and give each other honest feedback. It is not a course and not a sales event: you pick 1 process you know well, and the others tell you what they would try.
How much time does it take to learn AI on the job?
For a first try, set aside about 30 minutes to write down how 1 process works today. After that, 1 step at a time is enough. How fast you go depends on the process and on you.
What is the 1 process you would map first?
