AI Value Leaders

Field guide · October 2026

Six moves for thinking with AI on a hard problem

The gap between AI activity and AI value usually shows up in the quality of the question. This guide sets out six moves a leadership team can make on a real problem, with a prompt for each. It works in any enterprise AI assistant.

Start with your problem

The problem we haven't solvedThe outcome we want, ideally with a measure

Type them once and every prompt below fills in. Nothing you type leaves this page.

Four ground rules

01
Use a real problem, not a test one.
02
Give it the context only your people hold.
03
Read the output as a sceptic.
04
Keep what worked.

Each move is shown once on the same invented example: a freight business where quote turnaround has slipped and win rates are falling. Output excerpts are shortened and illustrative.

Frame

What it's for

Getting from a vague worry to a problem worth solving, before anyone proposes a solution.

Prompt

Here is a problem we haven't solved: [the problem]. The outcome we want is [the outcome]. Don't propose a solution yet. Ask me one question at a time to sharpen the problem. Think deeply between questions and let each answer shape the next. After about ten questions, restate the problem in two sentences and tell me what you think we've been assuming without noticing.

What good looks like

The restated problem is narrower and sharper than the one you started with, and at least one assumption surprises you.

How it fails

You answer quickly to get to the solution. The value is in the answers you have to stop and think about.

Illustrative

What the AI came back with (excerpt)

"Restated problem: quotes for non-standard freight take two days because pricing sits with three specialists who also handle escalations. An assumption you may be making: that speed is why you're losing. You haven't yet checked whether the lost quotes were also priced higher."

What the leader did next

Pulled last quarter's lost quotes to compare price as well as speed before going any further.

Widen

What it's for

Seeing the problem from positions your team doesn't hold.

Prompt

Look at this problem from five positions: our most demanding customer, a frontline employee who does this work every day, our CFO, a competitor who wants us to fail, and someone from another industry that solved something similar. For each, say what they'd see that we're missing. Then tell me which other people or functions in our organisation should be in the room, and why. The problem: [the problem].

What good looks like

At least one perspective changes who you invite to the next conversation.

How it fails

The perspectives all sound alike. Push back and ask it to make each one disagree with the others.

Stress-test

What it's for

Finding the weaknesses in a plan before reality does.

Prompt

Act as two reviewers. The first builds the strongest case for this proposal. The second is a sceptical board member looking for a reason to say no. Let them argue for three rounds. Then list the objections we'd struggle to answer and the evidence that would settle each. The proposal: [paste the proposal]. The problem it addresses: [the problem].

What good looks like

You find an objection you'd rather hear now than in the boardroom.

How it fails

The critic is polite. Tell it to be harder and to name the specific line it doesn't believe.

Decide

What it's for

Choosing between options on criteria you can defend later.

Prompt

Here are our options: [list them]. Before comparing them, help me agree the decision criteria and their weighting, given that the outcome we want is [the outcome]. Then score each option, show your reasoning, and tell me what would have to be true for the lowest-scoring option to be the right one. Finish with what would make us reverse this decision, and when we should check.

What good looks like

A decision with named reversal conditions and a date to look again.

How it fails

The scores get treated as the answer. They're a prompt for the team's judgement, not a replacement for it.

Hand off

What it's for

Turning a framed problem into work an AI agent or a colleague can run without you.

Prompt

From our conversation, write a one-page brief another person or AI agent could act on without asking me anything. Include the problem, the outcome and how we'll measure it, the context they need, the constraints, what's out of scope, the steps you'd suggest, and the points where a person must review the work before it goes further. Flag anything you've had to assume.

What good looks like

Someone who wasn't in the conversation could start work from it today.

How it fails

The brief leaves out the context that lives in people's heads. Read it as a newcomer would and add what's missing.

Reuse

What it's for

Making sure the next person starts further ahead. This is where individual use becomes organisational learning.

Prompt

Look back at how we worked on this problem. Write up the approach as a one-page guide for a colleague facing a similar problem: the questions that mattered most, the prompts that worked, the mistakes to avoid, and how to tell when the output is good enough. Leave out anything specific to this one case.

What good looks like

A colleague uses the guide and gets to a better result faster than you did.

How it fails

The guide sits in one person's files. Put it where the team will find it, and review it after the next use.

This is how learning velocity is built: each use makes the next one faster and better. See the learning system.

07Human in the lead

What stays with people

Every move above leaves three things with the leadership team: deciding which problem matters, judging whether the output is good, and owning the decision. AI widens what a team can consider in the time it has. The accountability doesn't move.

Prompts are starting points. Use them within your organisation's AI policy, and don't enter information that policy doesn't allow.