AI consulting services: what a good engagement should actually give you
Read this if you want to know what you should expect when hiring an AI consultant. And before you talk to anyone, you can check where your own business stands today.
Key takeaways
- A good AI consulting engagement hands you four things before anything is built: a map of how the work runs, the causes behind the slow parts, a price on each problem and a plan in order.
- The map has to come from the people who do the work, because the hours leak in the hand-offs between them, and AI can only follow steps someone has drawn out.
- A price on each problem lets you aim AI at the most expensive one first. Inside one 13-person branding agency, our consultation found $165,828 a year of waste, from hunting for information to extra briefing meetings.
- Whatever gets built should be yours, from the tools to the data and the accounts. The decision on what to build stays with you.
- If your work has never been mapped, start there. Our free AI readiness quiz shows which part of your business to fix first.
What AI consulting services cover, from advice to custom AI solutions
AI consulting services, sometimes sold as artificial intelligence consulting or simply an AI consultancy, do two jobs. They work out where AI belongs in your business, and then they build or set up whatever goes there.
Some firms only advise, and you're left with a strategy and nobody to build it. Others, often listed as AI software consulting services, mostly build what you ask for. We think the useful ones do both, in that order, so whatever gets built is aimed at a problem somebody has already counted.
What gets built depends on the problem. A custom AI solution for a business might be an AI agent that does a task from start to finish, or an automation that moves information between your tools without anyone chasing it. It might be a dashboard where every number finally agrees, or software shaped around how you already work.
And when a tool on the market already does the job, a good consultant helps you put it in place instead of building one from scratch. Either way, the tools, the data and the build should end up yours, on your own accounts.
Why AI projects stall before they pay off
We think most stalled AI projects share the same story. Somebody bought a tool before anyone mapped the work it was meant to change, so nobody could say which problem it was for or what fixing it was worth.
The founders we talk to have often tried AI already, and they still can't say where it fits in their own work. An AI tool with no map is a bit like a new hire with a login and no job description. It will work hard, and it won't know what to do first.
As of October 2026, much of what ranks for AI consulting services is written for large companies, with strategy and governance pitched at the boardroom. A founder-led business needs plainer answers, starting with where the hours go and which problem is worth fixing first.
What a good AI consulting engagement should hand you
Our own consultation report comes in six parts: the business as it runs today, the problems found and costed, the solutions, the investment, the returns and the game plan. Whoever you hire, those come down to four things you should be able to hold in your hands at the end, before anyone builds anything.
Each one below comes with a real finding from our own work. The names are left out, and the numbers are the ones we found.
1. A map of how the work really runs
The first thing you should get is a map of how work moves through your business today. It shows who does each step, in which tool, and where the work sits waiting for someone.
It has to come from the people who do the work, because they know where it gets stuck. A map drawn from the org chart or the list of tools you pay for misses the hand-offs between people, and the hand-offs are where the hours leak. It's also what any AI you add will work from, since AI can only follow steps that someone has drawn out.
At a 13-person branding agency, the map showed client details spread across five platforms. Finding one basic detail could pass from a team member to a project manager, then up to a partner and on to the head of operations. Each search took 10 to 20 minutes, and it happened dozens of times a day.
2. The causes behind the slow parts
Next, a good engagement should name what's causing each problem. A symptom like slow reporting or numbers nobody trusts usually has one cause underneath, and AI pointed at the symptom only makes the symptom faster.
The cause might be a step nobody owns, or two tools that disagree about the same number. An AI tool will read whichever one it's handed as the truth.
For an owner-run e-commerce brand, we found in 2026 that its ad platform was claiming 1.78 times the sales its own web analytics credited to the ads on last click. Each tool was counting the same sales its own way, so any budget set from one report alone started from a number the other report disagreed with. Read the e-commerce case
3. A price on every problem
Once each problem has a cause, it needs a price, in hours and in money, with the working shown so you can check it. That price lets you rank the problems and aim AI at the most expensive one first.
Without it, every problem feels equally urgent, and the AI budget tends to go to whichever one somebody complained about most recently.
For the branding agency above, the waste came to $165,828 a year. It covered hunting for information, typing the same data into more than one tool, extra briefing meetings and the back-and-forth over who had time for what. We found and mapped that number in the consultation.
4. A plan in the right order, and the decision left to you
The plan should tell you which problems to fix, what each fix costs to build and to run, what it gives back in hours or money, and the order to build in. For each fix, it should also say whether you need AI at all. Sometimes an automation, a cleanup of your data or a tool you can already buy will do the job.
Then the decision is yours. You can go ahead with all of it, part of it or none of it, and a good consultant leaves that call with you.
For an owner-run contractor, the first build went straight at the number the business ran on. Its 3,030 scattered Drive items added up to a €3.37M sales pipeline. In 2026 we rebuilt it in one place, and the real number was €2.76M, with €612K of the old total in duplicates. Can you imagine making hiring and investment decisions on a number that's off by over half a million? Read the contractor's case
How to check an AI consultant before you hire one
The simplest test is to ask what you'll be holding at the end, before anything is built. If the answer is a demo of a tool, the map is missing. These are the questions we'd ask in your place:
- What will we have in our hands before anything gets built? Listen for the map, the causes, a price on each problem and a plan.
- Who will you talk to, and which tools and files will you look at? The people who do the work every day should be on that list, alongside the founder.
- How will you put a number on each problem, and will we see the working?
- If a tool on the market already fits, will you tell us, even when it means building less?
- Who owns what gets built? The tools, the data and the accounts should all be yours.
- How much of our team's time will this take, and when?
If you've been burned by an AI build before, the third and fifth questions matter most. A plan with numbers you can check and a build that sits on your own accounts let you change course without starting over.
When you don't need an AI consultant yet
Sometimes you don't, and it's worth knowing that before you spend anything.
If what you want is help with one person's writing, summaries or research, a subscription to Claude or ChatGPT will take you a long way on its own. Use it on real work first, since it's the cheapest way to learn what AI can do in your business.
Maybe you already know the problem, roughly what it costs you each month and which tool fits it. Then you may only need someone to set that tool up, which is a much smaller job than consulting.
And if your work isn't written down anywhere and two reports rarely give the same number, the first job is the map, whoever ends up doing it. Our free AI readiness quiz shows which of those gaps is biggest for you, before you talk to anyone.
Frequently asked questions
We think AI will take over much of the reading, counting and drafting that consulting used to be paid for, and that's good news if you're the one hiring. We run our own agency on the same kind of systems we build for clients, which is a big part of why a study costs us so little to run. AI can't sit with your team, hear how the work really runs and earn the trust to change it, so the interviews and the judgement stay with people.
We think the big general models are close enough that what you point them at matters more than which one you pick. Claude and ChatGPT both read long documents, summarize calls and draft well. In our AI readiness score, AI experience carries only 10 of the 100 points, because choosing a tool and learning it is the easiest gap to close. Data and processes carry 60.
Most of the job is listening and counting. A good AI consultant interviews the people who do the work, follows how information moves between your tools, counts the hours each step takes and prices what that costs. Then they design the fixes, work out what each would cost to build and run, and put them in order. The building comes after, and only for the fixes you choose.
Use it yourselves first, because it's the cheapest way to learn what AI can do in your business. Outside help earns its place when the problem sits between people and tools, in reports that disagree, steps nobody wrote down or the same data typed into two systems. ChatGPT can make each person faster at their own part, and it still won't map the hand-offs between them.
Hire for what you need at the end. A consultant who only advises hands you a plan, and you still have to find someone to build it. An agency that only builds will build what you ask for, which works when you already know the problem and what it costs. If you don't know that yet, look for a firm that does both and starts with the map. That's how we work, with a study, then a plan, then your decision. So weigh this answer knowing we're one of them.
Your next step
Take the free AI readiness quiz to see where your business stands on the five things AI needs, from your data and processes to your direction. It shows you which one to fix first.
And if your score comes back strong, we'd love to talk about a free consultation, where we map your business this way and hand you the whole plan, completely free.