Contents · 7 min read
01 / 06 · The mandate
What must you decide before you research firms?
In short
Decide if you face an open business question or a set build. The first needs judgment. The second needs a team that can build what you chose.
A mid-market CEO is about to research AI firms. Yet the team has not agreed what those firms must solve. Some leaders want a new system built. Others first need to know what AI changes about the business.
Some leaders want a new system built. Others want to know what AI changes about the business. Those are two different mandates.
Vendor lists can help, and they do. They give the company names to research. But they cannot decide which type of help it needs.
The CEO must make that call first. If the business question is open, the firm needs help with the choice. It should not use a build to answer a strategy question.
That is why AI consulting starts with what changes in the business. It does not start with a tool.
For the mid-market company, the first task is one written line. It states what the outside firm must solve. That line tells the company which skill to seek.
02 / 06 · The skill
Do you need judgment or build skill?
In short
An open business question needs sound judgment. A set system needs the skill to build it. Do not score both as if they were the same.
For the mid-market company, the open question comes first. The team does not yet know what AI should change. It needs a firm that starts with the business model.
A set build starts later. The company knows what it wants built. It now needs the right team to do the work.
A strategy boutique may fit the first mandate. A build firm may fit the second. The company may keep the work inside if its team fits.
A Big Four firm also has a clear place. It may stay when scale or a regulatory program demands it. It may also stay when the board asks for that type of brand.
These routes have different strengths. The comparison of AI consulting options shows where each one fits.
For the mid-market company, the mandate decides which strength counts. Once that is clear, it can cut the list before it weighs brands.
03 / 06 · The cut
Which firms should leave the list?
In short
Remove each firm that does not fit the mandate. Then apply two more pass-fail gates. Only firms that pass should reach the final choice.
For the mid-market company, there are two gates before the final choice.
Test mandate fit
Does the firm start with the business question or the tools? Keep it only if that start fits the mandate.
Test management and ownership
Will management stay in the work? Will the company own what gets built? A no on either removes the firm.
Compare the firms left
Only now should the company compare skill and proof. Brand and claimed breadth come after the gates.
This order keeps the test clear. A broad firm can still be wrong for the work. A known brand can still fail on ownership.
The same tests guide how to choose an AI consultancy. They sort the field before a long sales process starts.
For the mid-market company, the first list may be long. The real shortlist is what remains after the cuts. The first gate is whether management stays in the work.
04 / 06 · Management
Will senior management do the work?
In short
Treat management involvement as pass-fail. Your own leaders must stay in the work. The firm must also name the senior people who will do it.
For the mid-market company, the mandate concerns the business. It cannot be handed to a tech team and checked at the end.
The CEO need not run each task. But the management team must make the key calls. A proposal with no role for them is a warning sign.
The outside firm must answer one more point. Who will do the work after the sale?
Senior names in a pitch are not enough. The company needs to know who will work on its mandate. It should also know what those people will decide.
A large team can help, and may be needed. That does not replace senior judgment on an open business question.
Established firms often bring trusted ties. New firms can shape their work around AI from the start. Each type has a different test to pass.
For the mid-market company, both sides need a clear role. The next gate tests what the company keeps.
05 / 06 · Ownership
Who owns what gets built?
In short
The company should own the agents or workflows built for it. If the outside firm would keep them, remove it. Ownership is a gate rather than a score.
For the mid-market company, this point must be clear before the final choice. The sales pitch may still sound strong. That does not cure a no on ownership.
Ask who owns each agent or workflow at the end. Put the answer in the terms. A proposal that denies client ownership is a warning sign.
The company should also be free to stop after the first step. It should not need a long deal to keep using what it paid to build.
This tests the firm's way of working. GRAIL states that clients own the agents and workflows built for them. It also states that each step can stand alone.
For the mid-market company, ownership keeps its next choice open. With both gates passed, it can make a fair comparison.
06 / 06 · The choice
How should you compare the firms that remain?
In short
Compare only the firms that pass every gate. Judge them on the skill your mandate needs. Then choose one that commits to a clear result and end point.
For the mid-market company, this is now a small field. Each firm fits the mandate. Each keeps management in the work. Each gives the company ownership.
The CEO and management team can now compare like with like.
Have the CEO rank the key skill
For an open question, test judgment about the business. For a set build, test the firm's ability to deliver it.
Have each firm show its own use of AI
Ask how AI changes the work it does. A method slide is not proof of a working system.
Have the management team set the result
Ask each firm for proof of likely P&L impact. A lack of proof is a warning sign.
Have the buyer set the end point
State what the company will own when the work ends. A fixed price for a result is clearer than hours on a meter.
Do not weigh strategy fit against build skill. Choose which one the mandate calls for. Then compare the firms on that skill.
For the mid-market company, the result is a list it can explain. Each firm earned its place against the same mandate. With one firm picked, the next step is to test the deal before signing.
Test the final fit
Before signing, the CEO and management team need to test each surviving firm against the mandate, then confirm that the proposed deal preserves management involvement and company ownership. A 30-minute conversation with GRAIL lets the team apply that test to GRAIL and talk through the specific choice it is preparing to make. GRAIL will say honestly whether it can help, giving the company a direct answer on one candidate before it decides.
Choose a 30-minute slot. If we are not the right fit, we say so and point you elsewhere.
Questions CEOs ask about this
How should we shortlist AI consultancies without relying on self-ranked vendor lists?
Build a serious AI vendor shortlist by setting the mandate first. The CEO must choose between an open business question and a set build, or the company may use a build to answer strategy. Apply pass-fail gates for fit and management involvement. Treat client ownership the same way. Compare survivors on the skill the mandate needs. Choose the firm with a clear result and end point.
What should be decided before researching firms?
Write down the mandate in one line. Decide if you need help with an open business question or a system whose scope is set.
How many different provider types should make the shortlist?
Use no type quota. Add only types that fit the mandate. Keep a large firm when scale or regulatory work requires it, or when the board calls for a known name.
Which criteria should be pass-fail rather than scored?
Mandate fit is pass-fail. So are management involvement and client ownership. Use the result and end point to compare the firms that remain.
How should strategic fit be weighted against implementation capacity?
Do not weigh them against each other. First decide which one the mandate needs. Then compare firms on that skill.
When should a vendor be removed before the first meeting?
Remove it when its written terms already fail a gate. That includes denied client ownership or no role for senior management. If public material does not answer, use the first meeting to test it.
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
- How to choose an AI consultancy: the five questions that sort the field before a company signs. (accessed October 8, 2026)
- What is AI consulting: why a business question should be settled before a system is built. (accessed October 8, 2026)
- GRAIL vs alternatives: where different types of firms fit. (accessed October 8, 2026)
- What is GRAIL: client ownership and work that can end after each step. (accessed October 8, 2026)
- Your Consulting Firm Hired a Country of Geniuses: the strengths of established firms and new firms. (accessed October 8, 2026)
Next scheduled review: January 28, 2027