What Exactly Does an AI Consultant Do?

The first phase is mostly watching

The opening move is a walk through your operation.

I want to see where work piles up. What gets rekeyed from one system into another. Which reports someone rebuilds by hand every Monday. Where things wait on one person. Which tasks make your best people say “I could do this in my sleep,” because that sentence is almost always pointing at something a machine should have taken already.

Nobody enjoys hearing that the first phase is looking. It is still the phase that decides whether everything after it works. Pick the wrong process and you can build something flawless that saves four minutes a week.

The time needed depends on the process, the systems involved and the information available. We agree on that scope before work begins.

Then comes the part everyone skips: order

Once you can see the whole operation, you have to decide what comes first.

This is where most AI projects quietly die, and rarely because the tools were bad. Teams automate the flashy customer-facing thing while the operational mess underneath it stays untouched. Then the flashy thing starts generating problems faster than anyone can clean them up.

Foundation first. Then operations. Then the exciting use case. That sequence is the entire thesis behind The Sequence Model, and it is the thing I get hired for more often than any specific piece of technology.

The decision at this stage is worth more than the build itself: this one, first, for this reason.

Building, in the tools you already pay for

The build is less exotic than people expect.

Most of what a small business needs already exists inside software it owns. The work is connecting it and shaping it around a real process, so the output is good enough that nobody quietly goes back to the old way. Sometimes that means an automation between two systems, sometimes a drafting workflow that turns an hour into ten minutes.

Buying a new platform is rarely the answer. If the first recommendation you hear is a purchase, ask what problem it solves that your current stack cannot. I wrote out how to think about that stack in What Is the Best AI for Small Business?.

Adoption: making the handoff work

I use a split I have written about before, the 10-20-70 rule: roughly ten percent of an AI project’s outcome comes from the model, twenty percent from the technology and data around it, and seventy percent from people and process. Where that breakdown comes from and why it holds is in What Is the 10-20-70 Rule for AI?.

For JBK, that means building for the people who will run the system. Setup and training happen around the client’s own authorized accounts. The handoff includes instructions for everyday use and recovery, and a representative task the client can complete independently. Any maintenance or ongoing support is agreed as part of the scope.

A tool nobody adopted is not a partial win. It is the same as never having built it, except you paid.

How this differs from the other people you could hire

A trainer teaches your team to use AI tools. Most businesses buy this too early. Training people on tools while the process underneath is still broken produces staff who are excellent at doing the old thing slightly faster.

A developer builds software to a spec. Hire one when you know exactly what you want built. Discovery at build rates is an expensive way to learn you did not need the build.

An AI implementation strategist is the closest neighbor, and the line is genuinely blurry. Both titles get used for the work described above. Where they differ in practice is emphasis: “consultant” is the broader word and covers everyone from advice-only to fully hands-on, while “implementation strategist” is often used to emphasize implementation and adoption support; the actual responsibilities still need to be agreed. I broke that role down separately in What Does an AI Implementation Strategist Do?.

The title is a weak filter. The next section has a better one.

What it costs, roughly

Independent consultants tend to run between $150 and $350 an hour. That range, what moves it, and how firms price against it is covered in How Much Does an AI Consultant Cost?. Plenty of good work gets scoped as a fixed-price project instead.

A first project should be small and fixed in scope, tied to a number you agreed on before it started.

How to tell in one conversation

Ask what they would do first, and why that.

A real answer names a process in your business and gives a reason it comes first. If the answer names a tool, keep looking. And “I would need to look at your operation before I could tell you” is a strong answer, as long as they say it plainly and do not pitch.

Then ask what the handoff includes: who runs the system, what training you receive, and what support is included after delivery. Those answers tell you more than a title.

Book a 30-minute strategy call. Let’s talk about what you need and how we could work together.

A tool nobody adopted is not a partial win. It is the same as never having built it, except you paid.

FAQ