Your team is already using AI. No one has shown them how to use it well.

Somewhere in your organisation right now, someone has a chatbot open next to a piece of real work. You didn't sanction it. You don't know what they're doing with it, whether it's any good, or what it touches. Meanwhile a client has started asking what your AI answer is, the board wants one sentence they can repeat, and everything you read insists you're behind without ever saying where to start.

I sat in one of the free AI workshops councils run. The room was full, and the mood was unmistakable: everyone believed this mattered, nobody trusted their own judgement about the first step, and the fear of starting wrong hung over the whole afternoon. When the Department for Science, Innovation and Technology surveyed 536 employers about AI upskilling in 2026, it found the same thing at national scale — what holds organisations back isn't doubt, it's cost, capacity, unclear provision, and fear of failing.

The result, in most firms, is drift. Each person invents their own way of using AI, or doesn't. The output is inconsistent and invisible. The British Chambers of Commerce found 54% of UK SMEs now use AI in some form — and 97% report a skills gap. Adoption arrived before capability. That gap is where the risk lives: not in using AI, but in using it untrained.

This is in-person training, at your offices, working on tasks your team actually did last week — a report that took a day, correspondence that eats every morning, the document nobody wants to own. Before any session I run a discovery call to pull real examples from your own work, and the room works on those.

What your team learns is a method, not a bag of tricks: how to get a result you can rely on instead of a one-off lucky answer (the discipline is called context engineering — you'll leave knowing exactly what that means); how to teach an AI your way of working once, so it works that way every time after (skills, and stacking them); how to hand it a multi-step job and have it finish, not just start (agentic workflows); and how to connect it to the tools you already use so it can act, not just answer.

Three things this is not. It is not e-learning — no licence, no module, no completion rate to chase in six months. It is not a software pitch — I sell nothing but the training, and it works with the tools you already have. And it is not a prompt-engineering course: wordsmithing a single clever instruction is the previous era of this field, and courses still teaching it are teaching a method the frontier has already left behind.

Most training fails at the same point: the trainer leaves, the quarter gets busy, and six weeks later nothing is different. Whatever you buy in this category, that is the failure to price in.

I spent eighteen years coaching elite rowing crews — seventeen international medals, four of them gold. Coaching at that level is one long lesson in what makes a skill hold up under pressure when the coach isn't in the boat: you train on the real thing, not a simulation; you build the habit into the week, not the away-day; and you check what survived contact with reality, then fix that, not what a syllabus says comes next. That is exactly how these sessions are built — your tasks, your tools, follow-up calls after the Briefing and the Team Day, and return days in the Programme that start from what your team actually used.

Before coaching, I built and ran the UK arm of a global serviced-office brand — 750 affiliated centres worldwide — through the 2008 crash: lost half the client base when Lehman collapsed, rebuilt to multiple-six-figure turnover. I know what a stretched week in a mid-sized business looks like from the inside, and I don't design training that pretends you have spare capacity you don't.

And I work this way daily. My own business runs on the method I teach: a modular library of AI skills that write, research and check work to my standards, and an agentic knowledge system that manages my own operation. When I show your team how to build a skill, I'm showing them something I built and use, not something I read about.

Here is how the next year goes in a firm that does nothing deliberate. The quiet AI use spreads — inconsistently, invisibly, with nobody checking what's pasted where. A client asks the question directly, and the honest answer is "some of our people use it, unofficially," which lands worse than "no." Clients begin assuming an efficiency you haven't actually captured, and fee conversations get harder. And the person who was best at it leaves, taking the only version of the capability with them.

None of that arrives as a crisis. It arrives as drift — which is why it keeps not being anyone's job to stop. One trained team, one working quarter, ends it: the use becomes visible, consistent, and yours rather than each individual's.