Contents · 8 min read
01 / 06 · Wrong start
Why is a tool list the wrong place to start?
In short
A tool list tells a partner what to build. It does not state why the work matters. Start with the business choice that your leaders must make.
Consider a recruitment firm whose fee rests on access to talent. AI now makes the basic transaction work open to all. A tool-first brief may still buy a sound system. Yet the leaders' main choice, what clients will pay for, can remain open while the fee model stays at risk.
The CEO sees the shift. She is ready to ask AI firms for bids. But her team has not yet agreed on the result it needs.
An IT-style brief feels neat. It names a system to install. It sets a date and asks for a bid.
That way of buying work has value. The brief is clear. The system may work just as planned.
But AI can change what the firm can do. That makes it a business choice for the leaders. It is not just a system for IT to install.
For the recruitment firm, a sound system could still miss the point. It may work while the fee model stays at risk.
The first line must therefore name the choice the leaders will make.
02 / 06 · The decision
What decision must your leaders make?
In short
State the choice that must be made when the work ends. Then state the result your company will own. That gives each partner a fixed end to work toward.
For the recruitment firm, the key choice is clear. Its leaders must decide how talent access will be priced.
The brief should say that in plain words. It should also say what success looks like.
The result is a clear choice on what clients will pay for. The firm must have a working way to act on it.
This keeps the brief firm without guessing the answer. A partner may find that the first idea is wrong. That is useful, and it is part of the work.
The leaders still own the call. The partner brings proof and helps test each path. The firm decides what it will do.
For the recruitment firm, this result is now clear enough to buy. It is also broad enough to let the partner learn.
That result now tells the firm what it must learn to do.
03 / 06 · The capability
What capability must the partner leave behind?
In short
Require a working skill that your leaders can use in real work. The partner may prepare the case and the options. Your leaders must make each call and know how the work runs.
For the recruitment firm, the new skill sits with its leaders. They must be able to test a fee model against the facts. They must also be able to make the next choice without the partner.
That calls for more than a deck. The firm needs a working setup that its leaders use.
The partner may do much of the early work. That is useful, and it is part of the job. But the work must move into the firm as it goes.
A good brief says who takes part from the firm. It says which calls stay with them. It also asks who will do the work on the partner side.
AI can help gather proof and draft options. It can show where facts clash. But the call stays with a person.
The same rule applies to outside help. The partner can prepare the work. Supplier judgment and knowledge transfer must stay with the company.
For the recruitment firm, the test is simple. Can its leaders run the work when the partner leaves?
The next test is whether the firm owns what makes that work.
04 / 06 · The boundaries
What must the company own and control?
In short
Set the data rules before any work starts. State that the company owns what gets built. Keep each key risk call with a named person inside the firm.
For the recruitment firm, the work will touch company data. The brief must set the lines for that use.
Name where data may be processed. Require the partner to name each subprocessor. Set how long data may be kept.
State whether company data may train a model. Require the partner to follow that choice.
Set the steps for an incident. Set the rules for any data transfer. A hosting region alone does not settle who may gain access.
Security calls also need an owner. The partner may flag a risk and suggest a fix. The firm must decide what risk it will accept.
The same goes for the work itself. The brief should say the firm owns each build. It should own the business rules used in it. It should also be free to use that work with someone else.
For the recruitment firm, those rights make the skill real. Without them, the firm still needs the old partner to keep moving.
With those rights set, the firm can decide when the work may stop.
05 / 06 · Stopping point
Where should the work stop?
In short
Set a clean stop after the first useful result. Make each next step a new choice. Keep the goal fixed while the partner finds the right tools.
For the recruitment firm, the first stage may end with one clear choice. It should also leave a working way to test that choice.
The firm can then stop. It can go on with the same partner. It can also take the owned work elsewhere.
This protects the firm from lock-in. It also protects the partner from a false plan set too soon.
Tools still matter. The partner must find ones that fit the work. But the CEO does not need to pick them before that work is known.
A strong brief leaves those means open. The goal stays fixed. The data lines stay fixed too.
The proposal should state what the first stage will prove. It should state what the firm will own at that point. It should not assume that a later stage has been sold.
That is how the firm can seek results before it commits to scale.
For the recruitment firm, the partner now has room to learn. The CEO still knows what she is buying.
Those fixed ends and open means now fit into one matrix.
06 / 06 · The matrix
What should every partner have to prove?
In short
Issue the same matrix to every partner. In each row, state what the company needs. Then ask the proposal to prove how the partner will meet it.
For the recruitment firm, the matrix has ten rows:
Business decision
The firm names the choice its leaders must make. The partner shows how its work will help them make it.
Owned outcome
The firm states what must be true at the end. The partner shows what proof will support that result.
Leader role
The firm names who will take part. The partner shows where they must judge and decide.
Skill transfer
The firm states what its team must run alone. The partner shows how that skill will move into the team.
Data use
The firm sets where data may go. The partner names access, storage terms and each subprocessor.
Security calls
The firm names who may accept risk. The partner shows how it will flag harm and handle an incident.
Work ownership
The firm requires full use of what gets built. The partner confirms who owns the work and its business rules.
Buying terms
The firm asks for a fixed result. The partner states what is in scope instead of selling open hours.
Clean stop
The firm sets the first point where it may leave. The partner states what the firm will hold at that point.
Room to learn
The firm does not pick the tools in advance. The partner shows how it will test the means without making the goal vague.
Ask each partner to name who will do the work. Ask who owns each result. Ask if the firm can stop after the first step.
Those answers sort an AI partner from a vendor. They also make the bids easier to judge on the same terms.
For the recruitment firm, the matrix keeps the main choice in view. Each partner must now show how it will help reach that result.
The matrix is ready; now use it to choose.
Put the matrix to work
Your next task is to turn the matrix into a brief that keeps the business outcome fixed while giving each partner enough room to show how it would reach it. In a 30-minute conversation, GRAIL can talk through the decision your brief must protect and the point at which the first stage should stop. You will get a direct view on whether GRAIL can help before you ask for a proposal.
Choose a 30-minute slot. If we are not the right fit, we say so and point you elsewhere.
Questions CEOs ask about this
What requirements should we give an AI partner before asking for a proposal?
Name the outcome your company must own, not the tools. Your leaders must decide the business result before buying, or a sound system may leave the main choice and fee model at risk. Keep that result fixed and leave the means open. Set the skill and data terms. Set ownership and a clean stop. The matrix helps you judge which partner can deliver that result.
Which business outcomes should the matrix specify?
Name the choice your leaders must make. Then state what the company must be able to do when the work ends. Do not turn the outcome into a list of tools.
How should we define skill transfer?
State what your leaders must run without the partner. Name the calls that stay with them. Ask the partner to show how the working skill will pass into your team.
What should it say about data, rights and lock-in?
Set where data may be processed and who may gain access. Set how long it may stay and whether it may train a model. Confirm that you own the work and can leave after the first stage.
Which proof and references should partners provide?
Ask for proof tied to each row in the matrix. Make the partner name who will do the work and who owns the result. A reference call should not replace those answers.
How can we leave room to learn without getting vague bids?
Keep the outcome and the stop point fixed. Leave the tools open. Each partner must still state what its first stage will prove and what you will own.
Founder, GRAIL
Johan has spent fifteen years in management consulting and executive leadership, including as Chief Product Officer of a SaaS company. He founded GRAIL to help European leadership teams work out what AI changes at the core of their business. He is the author of The Execution Revolution and the host of the ThinkRoom podcast.
Related reading
Sources
- Recruitment in the Age of AI: why the price of talent access is a leadership question. (accessed October 8, 2026)
- What is AI consulting: why AI work starts with business change rather than a fixed system. (accessed October 8, 2026)
- What does an AI-augmented product organization look like?: why proof may be prepared by AI while the decision stays human. (accessed October 8, 2026)
- What does an AI-augmented engineering organization look like?: why supplier judgment and skill must stay with the company. (accessed October 8, 2026)
- AI in engineering consulting: why the company should see a result before it commits to scale. (accessed October 8, 2026)
- How to choose an AI consultancy: the tests for people, terms, ownership and a clean stopping point. (accessed October 8, 2026)
Next scheduled review: January 25, 2027