For chief executives, operating officers and heads of technology
Design an AI programme that pays its way.
Most AI programmes stall in the same place: a portfolio of pilots nobody can rank, each with a sponsor and none with a number. We rank the portfolio by what it costs to find out, prove the first candidates inside the operating business, measured in the metrics it already tracks, and stop the rest early. The method we use on ourselves is published, so you can judge the advice by the practice.
A written reply within two working days. No mailing list, no price list, nothing gated.
If this is on your desk
- A list of AI pilots that grows every quarter, and a budget that does not
- A transformation plan with a vendor's roadmap where the business case should be
- Savings that have been announced but not stated as a number anyone will be held to
- A build-or-buy call on models, tooling and data that has to look sensible in two years
- A board asking what AI changes in the unit economics, not what it demonstrates
How the programme runs
The programme has four stages, and each ends with a decision rather than a presentation. Its first product is a ranked portfolio; its last is a result inside the business that your team is running.
| Stage | What it does | What it produces |
|---|---|---|
| Diagnose | Where AI changes the unit economics of your operation, and where it stays a demonstration. What data and integration each candidate needs before it works. | A map of candidates against the metrics the business already tracks |
| Rank | Every candidate scored by value at risk and by what it costs to find out, with a kill gate agreed before anything starts. A modest prize that can be settled in six weeks outranks a large one that takes nine months. | A ranked portfolio, the blocking rules and the first two candidates |
| Prove | The first candidates run inside the operating business, not beside it, with the saving or the revenue stated as a number someone will be held to. | A result, or a kill, in the business's own metrics |
| Scale or stop | What paid is scaled with your team running it. What did not is stopped and written up, so the lesson is kept and the budget is not. What must stay human is written down. | A programme that pays for its next stage |
How the first step works
- 01
Tell us what is on your desk
A paragraph is enough. The more specific the question, the more useful the reply.
- 02
A written reply within two working days
How we would approach it, what it would take and what it would cost. If the right answer is to change nothing, the reply will say so.
- 03
A one-day teardown, if it is worth one
A 90-minute working session, a written answer in two working days, and the model behind it, yours to keep.
Work you can check before you write
Everything we publish is free, opens without an email address, and shows the source behind every figure.
Tell us where the programme stands
Say what has been piloted, what is on the list, and what the business is being asked to show for it. You will have a written reply within two working days.
Questions people ask first
Do you implement, or only advise?
Both. The programme runs to a first result inside the business with your team running it; our five-stage method ends in an initial offering, not a report.
Which models and vendors do you recommend?
None by default. The build-or-buy call is made on your data and your constraints, and the reasoning is written down so it can be revisited when the market moves.
How do we know it is paying?
In the metrics the business already tracks, stated as a number someone will be held to, with the kill gate agreed before the pilot starts.
What stays human?
The decisions AI accelerates but does not make: which sources count, which assumptions hold, what the evidence means, and who signs the recommendation. We publish how we apply that rule to ourselves.
Why is there no price on this page?
Pricing is set per engagement and written into the first reply, so you see it before you commit to anything.