AI Training for Leaders & Boards UK

Every leadership team I sit with has a decision that has been open since spring. Ask what is actually stopping it and you get more data.

It is rarely the information. Most boards I meet already have more than they can read. The dashboard, the market report, two advisers who disagree, and a decision that has not moved.

What is missing is a method for interrogating a decision before it gets made. Nobody teaches one. You get taught to gather, to present, and then to decide on instinct in the last ten minutes of a meeting that overran.

The Army teaches one. It is called the Combat Estimate, seven questions in a fixed order, and I spent 23 years using it where getting it wrong cost more than money.

That is the day.

A method, not a briefing

Most leadership AI training is a briefing. Somebody stands at the front, explains what the models can do, warns you about the regulation, and leaves at four. Your team is better informed by Friday and no better at deciding anything.

This day is built the other way round. You bring a live decision, one that is real, unresolved, and costing you while it stays open. We run it through the estimate in the room. Then your team runs the next one while I watch.

The instrument is two things fused. The Combat Estimate supplies the questioning spine: seven questions, in a fixed order, that between them leave a decision nowhere to hide. British Army doctrine, introduced in 2001, and still how a plan gets built before anybody moves. CRIT supplies the discipline for working the same problem with AI: context before role, role before task, and the answer stress tested before you believe it. That one is Geoff Woods, from The AI-Driven Leader. He gives it away free and he sits on my advisory board.

Neither is enough alone. Seven questions asked coldly is an interrogation. CRIT without the spine is a good conversation that misses the resourcing question and the stop condition almost every time.

Who is in the room

Up to 8. Directors, the board, the senior team. The people who own the decisions rather than the ones who carry them out.

It works best when the room can actually decide something on the day. A leadership team that has to take everything away and check with somebody else gets less out of it.

Right for you if

✓You have a decision that has been open longer than a month

✓Your leadership team has less AI literacy than your most engaged junior staff

✓You are being asked to sign off AI work you cannot yet interrogate

✓Growth planning keeps stalling on a number nobody will commit to

Not right yet if

You want your team competent in the tools. That is the Team Training Day

You need an AI policy and the enforcement behind it. That is Govern

You want AI built into your systems rather than into your decisions. That is Build

How the day runs

01The Combat Estimate, on a decision you brought

Before the day, each of you writes down the decision you are stuck on. Handwritten, one side. Not a paper for the board, the actual thing, in the state it is in.

We take the biggest one and run the seven questions on it, live, on the wall. What is happening and what it is costing while nothing changes. What you are actually trying to achieve, who set that, and what is truly fixed as opposed to assumed fixed. What has to be different afterwards, in one line.

Those first three build the understanding, and the order carries weight. Question four answered before questions one to three gives you a plan aimed at the wrong thing, executed well. It is the most common failure in commercial decision making I see, and it is why the spine exists.

Question 1

What is the situation, and how does it affect me?

What is happening in facts rather than interpretation, and what it costs per week that this stays unresolved.

Question 2

What have I been told to do, and why?

The goal above the decision. Who set it. What is fixed, and what is only being assumed to be fixed.

Question 3

What effects do I want, and what direction must I give?

What has to be different afterwards, in one line. An effect that takes a paragraph is two effects.

Question 4

Where can I best accomplish each effect?

The decisive point. Which client, which channel, which part of the business. The obvious place, and the non-obvious one.

Question 5

What resources do I need?

Money, people, political capital, and your own hours. The last one gets left out of every plan I am ever shown.

Question 6

When and where, in relation to each other?

Sequencing and dependencies. What has to happen before what, and where it collides with the diary.

Question 7

What control measures do I need?

How you will know it is working, what will tell you it is not, and the condition under which you stop. A decision without a stop condition is a wish.

02Deciding when you do not have enough

Questions four to seven build the course of action, and they are where the room usually finds out how thin the plan was. Then I attack it.

That part is not optional and it is not theatre. Whatever you have all agreed on by three o'clock is the best you get by agreeing with each other, and agreeing with each other is not a test. So I change role and stress test it properly: the assumptions, the second order consequences, the downside nobody in the room is looking at.

Your team then does the same to the next decision. Most of them have never been asked to argue against a plan they helped build. It is a learnable skill, and it is the one that separates a board that governs from a board that approves.

We cover the harder version too, which is what to do when the honest answer is that you do not know. There is a move for that. It never ends the conversation and it never involves guessing.

03Data paralysis

A decision waiting on more data is usually a decision somebody has not framed. The dashboard is not the problem. The absence of a named effect is the problem, and more data will not supply one.

This is also where AI makes things worse before it makes them better. Bring it your five options and ask it to improve them and it will, politely, inside the boundary of what you already thought. Ask it for alternatives to your five instead. One word does more work than any tool switch here, and that word is non-obvious. Without it the model confirms your existing thinking and tells you what you want to hear, which is the failure mode nobody notices because the output looks excellent.

We do this live, on your decision, so the room can see the difference between the two answers side by side. The first output is around half of the answer. Your judgement is the rest, and knowing which half is which is most of the skill.

04Growth planning, and the enablement behind it

Growth plans fail on question five and question seven. Five, because the plan quietly assumes somebody has twenty spare hours a week and nobody says whose. Seven, because there is no signal that says this is working and no condition under which you stop.

So we build the growth decision the same way. Where the decisive point is. What it costs in all four currencies. What has to happen first, and the two dates in the diary where you review it candidly rather than optimistically.

Then the enablement. Every estimate leaves a one page record: the decision, the situation, the intent, the course of action, the control measures, what the challenger said, and what is still unknown. Six of those records is a syllabus. That is how the method stays in the building after I have gone, which is the whole point of buying a day rather than an adviser.

What leaves the room with you

✓

A decision, made

The one you brought in. Run through the estimate, red teamed, written down.

✓

The seven questions

A one page working aid your team uses on the next one without me in the room.

✓

Written records

Every decision from the day, to the same structure, including what is still unknown.

✓

A stop condition

For each one. What says it is working, what says it is not, and when you review it.

Skills that stay when we go.

Who runs the day

Chris Duffy. Founder and Chief AI Officer. Army veteran and ex Special Forces Communicator, 23 years. Certified Chief AI Officer. Regular expert for Forbes in AI and cybersecurity. Claude Partner Network member.

I was taught the Combat Estimate as doctrine and used it for two decades on decisions where the cost of getting it wrong was not financial. The business version is the same instrument with the stakes changed. I would rather teach your board that than talk them through a slide on the state of AI.

Two decades of getting the right information to the right people at the right time, for decisions that mattered. Same job. New tools.

Where governance sits

The day produces decisions and the records behind them, which is a good part of the evidence a board is asked for. It is not an AI policy and it is not an assurance framework. Where you need those, they live on Govern and they are a separate piece of work. I would rather tell you that here than sell you half of one.

See Govern

Investment

£5,995 for the day, exc VAT. See all pricing

One day, up to 8 senior leaders. A scoping call beforehand, the facilitated day, and the written records within five working days.

Priced by the day, not by the head.

Questions

A live decision each, handwritten, one side of paper. Something real and unresolved that is costing you while it stays open. Money, direction, a hire, a client, a commitment that is hard to unwind. If the room cannot name one between them, the day is not the right spend and I will say so on the scoping call rather than after it.

Both, in that order. The questioning method works with a whiteboard and no technology at all, and that is where the day starts. The second half runs the same decision with Claude in the room so you can see where it earns its place and where it simply agrees with you. Nobody leaves being asked to prompt for a living.

You do, and most of them turn out fine. The method is for the few that will not. The bar is a decision that drives most of the result, where being wrong is expensive and hard to reverse. Almost nothing clears it. Part of what your team learns on the day is telling those apart from the rest, because running the full estimate on a small decision wastes a morning.

Then you leave knowing exactly what is still open and what would have to be true to close it, written down. An estimate that produces no open questions was not an honest one. In practice most rooms decide, because what was stopping them was the framing rather than the information.

Leaders set the conditions, teams do the work. This day is about the decisions. The Team Training Day is about the daily work, on Claude, using your real tasks. Running both is how you end up with a business that decides well and can execute on it. Neither one replaces the other.

The day covers up to 8 senior leaders, a scoping call beforehand, the facilitated day itself, and the written records within five working days. Every price is published on our pricing page.

AI Training for Leaders and BoardsThe Leadership and AI Strategy Day