AI strategy consulting: how to choose a partner

Definition: AI strategy consulting is advice that helps a business decide where AI can add value, get the foundations ready (data, processes, systems and people), and put AI into live use with the right controls. The people who provide it are often called AI consultants.

In short: Choose an AI strategy consulting partner by how they work, not what they know about AI. Put them through the month-six test: will the same people still be working alongside your team, and will your people have helped write the plan? The technology is the easy part. Getting it into live use is where projects fail.

Stop asking about the technology. I've spent ten years in AI and consulting, and if I were the one buying, this is what I'd ask.

[Katie: add a moment from your own career that shaped how you'd choose a consultant, from either side of the table.]

Why isn't AI expertise the right thing to ask about?

Every AI consultant will tell you they know AI. It's the one thing you can assume, so it tells you nothing.

Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027, citing rising costs, unclear business value and weak risk controls. None of those is a model problem. Projects fail because the data was scattered, the process was never standardized, or the team didn't want the change and nobody asked them. Those are people and operations problems, and that's where a consultant earns their fee.

What is the month-six test?

It's the test I'd put every shortlisted consultant through. Three questions:

1. Who's in the room in month six? Are the people who pitched to you still working with your team, or has the work moved to someone new?
2. Are they working alongside us, or have they handed over? Is the consultant still helping to get it live, or has a report been delivered and the team left?
3. Did my people help write the plan? Or will they meet it for the first time when it lands?

Anyone can look good in month one. Month six is when you find out.

The shortlist scorecard

Put these five questions to every consultant on your shortlist. Score each answer 0 for weak, 1 for partly and 2 for strong.

1. Who is actually doing the work?

• Weak: "We have a strong team of experts." The pitch team and the delivery team are different people.
• Strong: Names, and how many days a week they'll spend with your people.

2. How will my people be involved?

• Weak: "We'll run a workshop at the start."
• Strong: Your end users help shape the approach from the first week, and are in the room for design decisions.

3. What happens the week after the plan?

• Weak: A report and a handover, or delivery by a separate team at extra cost.
• Strong: They work alongside your team until it's live, and stay as long as you need. A recommendation nobody acts on is an expensive opinion.

4. Will you tell me what not to do?

• Weak: Agreement with your first idea.
• Strong: Pushback. "That isn't the right first step, and here's why." If they're part of a software company, they tell you how they separate advice from sales, and admit when the answer isn't their own product. I work for a software company, so ask us the same question.

5. How will we know it worked?

• Weak: Vague benefits and no measures.
• Strong: A business case you help write, covering what the change is worth and what standing still costs.

Reading the score. Our rule of thumb: 8 to 10, shortlist them. 5 to 7, press for specifics. Under 5, keep looking. [Katie: confirm or change these thresholds.]

[Katie: add an example, anonymised, of a strong or weak answer you've heard or given.]

What should your risk and legal teams check?

Three questions belong outside the pitch.

• How will our data be handled? Where does it go, who sees it, and how is it kept separate?
• Who owns what's produced? Plans, designs and anything built with your people.
• What happens if we stop? Can you take the work and continue without them?

What does it look like when it works?

When we ran our own Team Readiness engagement at Enate, we put an AI Champion in every team and built the guidance around what people told us. The top worry, having to double-check AI output, fell to 5%.

[Katie: say in your own words what you think made the difference.]

Key takeaways

• AI expertise is a given. How a consultant works is what tells you something.
• Apply the month-six test: same people, working alongside your team, with your people's input.
• Score every shortlisted consultant on the same five questions, and check data and exit terms before you sign.

FAQs

What is AI strategy consulting?

AI strategy consulting is advice that helps a business decide where AI can add value, get its foundations ready, and put AI into live use with the right controls.

What does an AI consultant do?

An AI consultant helps a business decide where AI can add value, get its foundations ready (data, processes, systems and people), and put AI into live use with the right controls.

How do I compare AI consultancies?

Ask each one the same five questions and score the answers side by side: who does the work, how your people are involved, what happens after the plan, whether they'll tell you what not to do, and how you'll know it worked.

What is the month-six test?

It's a check on any consultant: will the same people still be working alongside your team in month six, and will your people have helped write the plan.

Do I need an AI consultant or a software vendor?

A vendor sells a product. A consultant advises on what you need. If a consultancy is part of a software company, ask how they separate advice from sales.

What should I prepare before speaking to an AI consultant?

Be ready to describe your goals, where AI is already being used, the main systems involved and who makes decisions. You don't need a finished plan.

Do I need an AI strategy before I hire a consultant?

No. A good consultant helps you work out where to start. If you're not sure, try our short diagnostic, which points you to the problem worth tackling first.

Sources: Gartner, June 2025, on the cancellation of agentic AI projects. [Add link.]