What does an AI strategy consultant actually do?

An AI strategy consultant works out which few jobs AI should do in your business, in what order, and what each one will return in your own numbers, before anything gets built. Flairr's AI Assessment takes about three weeks and ends in a costed plan that also says what not to build.

Order matters as much as choice: at an immigration firm we built a knowledge hub first and reply drafting on top of it, and that second build returned 150% in its first year and paid for itself in roughly five months.

Most of the job is counting: how often something happens, how long it takes today, and who it waits on. The AI comes last.

What does an AI strategy consultant do in the first few weeks?

In the first few weeks, an AI strategy consultant learns how your business actually runs from the people who run it, and writes down where the time and the money go. Ours runs in four steps over about three weeks, with a few hours of your team's time.

1. Kickoff. We agree the goals, who we talk to, and the numbers we need from you.
2. Interviews. Short conversations with leadership and with the people doing the work. Each person gets 20 to 25 questions written for their role, and we don't send them ahead, so the conversation stays natural.
3. Findings. Where the time and money go, and what's worth fixing first.
4. The blueprint. The design of your AI operating system and a phased, costed plan to build it.

The interviews matter most, and they're the easiest part to get wrong. If we suggest a number in an interview and the client only agrees with it, we record it as our figure, and we don't cost anything on it until the client gives their own. We ask "how long does it take?" open, and wait for their number.

How does an AI consultant decide what is worth building?

A good AI consultant decides from two numbers your own people give: how often a task happens, and how long it takes each time. Multiply them and you have the hours.

We put an hourly rate on those hours and write the rate down, so you can argue with it. The headline saving is the sum of the lines, never rounded up. If someone says a task takes 30 to 60 minutes, the range stays a range all the way to the total.

Then every candidate gets two scores.

• Impact. High means 150 or more hours a year back, or removing a mistake that costs revenue or breaks a compliance rule, like a mispriced quote or a missed deadline. Medium is 40 to 150 hours a year. Under 40 is Low.
• Difficulty. Low means one system, with the information already in usable shape. High means four or more systems, an old system with no way to connect to it, information trapped in documents and images, or a process people will have to change.

The scores decide the order. Quick wins, anything of low or medium difficulty, go in the first 0 to 3 months, because they pay back first and buy trust for the rest. Foundational work goes in months 3 to 9, and the big, hard pieces in months 9 to 18. The first phase is never allowed to be empty. If a plan has nothing you could have running within three months, send it back.

We also never fill a gap with an industry benchmark. A benchmark can explain a price, but it can't tell you how long your team spends on a task.

What should you get at the end of an AI strategy engagement?

You should get a plan you could hand to any builder: what to build, in what order, what each phase costs, and what it returns in your own numbers. Flairr's AI Assessment ends in four things.

• The blueprint for your AI operating system: what it brings together, what leadership will be able to ask it, and what it will do.
• A phased plan with costs: what gets built first, what follows, and what each phase costs.
• The return, traced to your numbers: every saving is your volume times the time it takes today, both from your own team, so you can check the maths.
• What not to build: the automations worth doing and the ones that aren't. Sometimes the honest answer is "don't build this", and we say so.

Then you decide: build with us, build some of it, or take the plan and do nothing yet.

My test for any deliverable is whether you could check one line of it against your own week. A list of use cases with no hours against them is a brochure, however many slides it runs to.

Why not skip the strategy and just build something?

Because most companies have little to show from AI so far, and the ones that do had the groundwork in place first.

In BCG's survey of 1,250 senior executives, published in September 2025, 60% of companies reported minimal revenue and cost gains from AI (BCG). In PwC's 2026 Global CEO Survey of 4,454 CEOs, only one in eight said AI had delivered both cost and revenue benefits, and CEOs whose companies had strong AI foundations were three times more likely to report meaningful financial returns (PwC). PwC's release puts it this way: "Foundations matter as much as scale."

My reading is that a smaller business goes wrong the same way, for a simpler reason: the AI goes in before anyone has counted what it is for, or what has to be in place under it.

What does getting the order right look like in a real business?

Getting the order right means building the less exciting piece first when the second one depends on it. An immigration firm we built for is the clearest case we have.

Their rules lived in scattered files, old email threads and the memory of whoever had been there longest. Meanwhile, writing replies to a steady stream of enquiries took roughly two hours of someone's day, every day.

We built the knowledge hub first: 32 documents in one place, with every answer pointing to its source (case study). Then we built the reply drafting on top of it. An enquiry comes in, the system pulls the relevant rules from the hub, and a draft reply lands in the right employee's Outlook. They skim, adjust and send. Nothing goes out without a person.

The result is 44 hours of enquiry writing gone every month, a 150% return in the first year, and a build that paid for itself in roughly five months, measured against the staff hours it gave back (case study). You can check the 44 yourself: two hours a day over about 22 working days.

The drafts are written from the rules in the hub, so the reply drafting could only come second. These were two builds rather than an assessment, but that dependency is exactly the kind of call a good plan settles before anything is built.

We have builds with bigger headline numbers. I lead with this one because five months is a number an owner can plan around.

When is an AI strategy consultant not worth it?

An AI strategy consultant isn't worth it when you already know the one job to fix, when nobody on your side will give the numbers, or when the plan has nowhere to go afterwards.

• You already know the job. An adviser told me about a firm that had been offered a full AI system. They said that was nice to have, but one process that had annoyed them for years was their biggest thing. The right step there is a fit check on that one workflow.
• Nobody will give the time or the numbers. It takes a few hours of your team's time. If leadership won't sit for an interview and nobody will say how long a task takes, the plan rests on our guesses instead of your numbers.
• The consultant sells one tool. If every recommendation lands on the same product, you've paid for a sales call.
• Nobody builds it. A plan with no one to build it stays a document. Ours can go to anyone, and if you build with us, the people who designed it are the people who build it.
• You want a guarantee. A strategy finds the savings and puts a price on them. It doesn't promise them, and nobody honest will.

Frequently asked questions

How much does AI strategy consulting cost?

Our AI Assessment is a fixed fee, agreed after a free intro call, so you know the full price before we start. The build is quoted from the plan.

How long does an AI strategy engagement take?

Ours takes about three weeks and a few hours of your team's time: a kickoff, a short interview with each person, and a findings session. Most systems are live 3 to 8 weeks after that.

Do you have to build with the consultant afterwards?

No. The plan is yours to act on however you choose: with us, with someone else, or not yet. A good plan is written so that any builder could pick it up.

Is AI strategy consulting the same as an AI readiness assessment?

They overlap. A readiness assessment asks whether your information and your team are ready for AI. Strategy asks what is worth building, in what order and for what return. Ours covers both, because how ready your information is shows up in the difficulty score of every item.

If you want to see how we work, book a free intro call.