AI consulting for a small business comes in four shapes, and the shape decides the cost more than the consultant does. Hourly advice is the cheapest way in and the hardest to budget. A fixed fee assessment runs two to four weeks and ends in a written document. With us that is €9,500 to €14,000 for two weeks. A fractional CTO is a monthly retainer, €3,500 to €6,000 with us. A build partner charges for working software, from about €10,000 for one feature. Whichever you buy, you should leave with a written answer you could hand to another supplier.
This guide is for companies with fewer than 250 people, the EU threshold for a small or medium-sized enterprise. In our experience the large consultancies rarely price for that size. Solo advisors often do, but they usually cannot build what they recommend. Both gaps matter when the budget is your own money.
What AI consulting costs a small business, by engagement shape
The table shows how the market is shaped, with our own fixed fees where we offer that shape. We have left out other firms' hourly rates on purpose. The next section explains why.
| Engagement shape | Typical cost | What you get | Best for |
|---|---|---|---|
| Hourly advice | Billed per hour or per session, open ended. Rates vary widely by seniority and type of firm. | Answers to the questions you bring. Rarely a written document unless you ask for one. | A second opinion on a vendor quote, or one narrow question |
| Five day review | €4,500 to €6,500 fixed, with us | Written findings on one existing AI feature or stalled pilot, and what to fix first | A pilot that demoed well and then stopped moving |
| Fixed fee assessment | €9,500 to €14,000 for two weeks, with us | Data inventory, use case shortlist with expected return, build or buy calls, GDPR and EU AI Act position, one costed first build | Deciding whether to build anything, and what to build first |
| Fractional CTO | €3,500 to €6,000 per month, minimum three months, with us | A senior person in your technical decisions every month: roadmap, supplier choices, architecture, review of what gets delivered | Companies without a CTO running several software or AI initiatives |
| Build partner | From about €10,000 for one feature. Most agents land between €25,000 and €80,000. | Working software in production, with an eval suite, monitoring and a runbook | A use case that is already chosen, with data that has been checked |
Discovery week sits between the last two rows. It takes one week at a fixed fee. It ends with a written scope and one price for a build you have already decided on. If you know what you want, start there rather than with an assessment.
AI consultant hourly rates, and why they are the wrong thing to compare
People search for an AI consultant hourly rate because it feels like the honest unit. It is not. An hourly rate tells you what the seller's time costs. It tells you nothing about how many hours the job needs, or what you will hold at the end.
AI consulting rates fall into three rough tiers. Independent advisors bill by the hour or the day and usually stop at advice. Boutique firms mix advice and engineering, and price either way. Large consultancies sit at the top of the market and are built around enterprise budgets. Rate guides for all three are easy to find online. Almost all of the ones we read were published by firms selling the service. Their ranges also disagreed with each other by a wide margin. We would rather not repeat numbers we cannot check.
So compare totals and deliverables instead. Ask any hourly advisor for a capped estimate and a list of what you will have in writing at the end. Then put that next to a fixed fee. For reference, our two week assessment works out at roughly €950 to €1,400 per working day, with the document included. We do not sell AI work by the hour at all, whether it is consulting or development.
Our view is that hourly billing on AI work is a trap for both sides. The outcome is not linear with hours. A senior engineer who has seen a failure before spends two hours where someone learning on your budget spends twenty. A fixed fee moves that risk to the person who can actually control it.
What you should get for the money
Whatever the shape, small business AI consulting should end in something you can act on without the consultant in the room. For an assessment, that means a written document with at least these parts:
- A data inventory checked against reality. What you hold, where it lives and which of it is usable. Judged from the data itself, not from its documentation.
- A shortlist of use cases with expected return. Usually two survive. The reasons the others were dropped are often the more useful half.
- Build or buy, per use case. If a product already does the job for a few hundred euros a month, the document should say so plainly.
- Your GDPR and EU AI Act position. Where data would be processed, under what agreement, and which risk tier each use case falls into.
- Running costs as well as build costs. Model usage, hosting and maintenance per month, so someone can put the number in a budget.
- One costed first build, including the option of not doing it.
The last test is portability. The document should belong to you, and it should be specific enough that another supplier could quote from it. Ours is written that way on purpose. If a consultant's output only makes sense with the consultant attached, what you bought was a sales process.
How to tell an AI consultant who has shipped from one who has not
This matters more than price when you compare AI consulting companies for small business, and you can test it in one call. People who have put AI into production talk about different things than people who have only advised on it. Ask these questions and listen to how specific the answers get.
- 01What did you put into production that is still running, and what broke in the first month? Shipped people answer with a story. Everyone else answers with a list of capabilities.
- 02How would we know the system gives right answers? Look for a test set of real cases, scored automatically on every change. If the answer is a pilot and some user feedback, they have not run one in production.
- 03What will it cost per month to run? Someone who has operated a system thinks in cost per request and monthly usage, not only in the build price.
- 04What would you tell us not to build? A consultant who has never delivered a negative assessment is not assessing anything.
- 05Where would our data be processed, and under which agreement? Expect a specific answer naming the provider, the region and the data processing agreement.
- 06Who writes the code if we go ahead? Either answer can be fine. You just need to know whether the advice and the build come from the same people.
The best AI consulting for small businesses is rarely the firm with the biggest client logos. It is the one whose answers to those six questions are short, specific and a little uncomfortable.
What is different about AI consulting for a small business
The job of an AI advisor for small business clients is different from advising a bank. The budgets are smaller, the data is thinner, and nobody in house owns the result once the advisor leaves. Eurostat's 2025 survey shows the gap clearly. Among EU enterprises with 10 or more employees, 17% of small ones used AI, against 55% of large ones.
The same survey asked companies that had considered AI but not adopted it why they held back. Lack of relevant expertise came first, at 71%. Unclear legal consequences followed at 53%, and data protection concerns at 49%. Those three answers make a fair job description for whoever you hire.
Proportion matters as well. If the whole AI budget is under €25,000, spending half of it on an assessment is a poor split. Buy a few hours of advice instead, or pick a well reviewed tool and watch what your team actually uses. Below about €10,000 we point people to off-the-shelf tools rather than quote a build.
On regulation, most small business uses are lighter than people fear. Support assistants, internal search and document extraction are usually limited or minimal risk under the EU AI Act. Classification follows the use case, not the model or where it is hosted. Since 2 August 2026, an assistant that talks to customers must tell them it is an AI. Where heavier obligations do apply, SMEs get simplified documentation, and the 2026 Digital Omnibus extended that relief to small mid-caps. A good consultant tells you which tier you are in and does not sell you a compliance program you do not need.
Advice from people who also build it
Big consultancies advise and leave. That works for a strategy question and fails for an engineering one. The hard parts of AI only show up when somebody builds it: messy data, integration, evaluation and running cost. Advice written without that experience tends to be confident in the wrong places.
We write the assessment and can then build what it recommends. If we do, on a project above €50,000, the assessment fee comes off the first invoice. The credit is valid for six months from the document. If you take the document to another supplier or build it in house, that is a fair outcome too.
There is an obvious conflict in that model, and you should weigh it. A consultant who also builds has a reason to recommend building. The protection is in the document itself. Read the build or buy call for each use case, and the list of use cases that were dropped. Then check that the first build includes the option of not doing it. If every recommendation ends in a custom build, be suspicious, and that includes being suspicious of us.
Which shape to pick
Choosing an AI consultant for small business work mostly comes down to which of five situations you are in. Each has a sensible starting point.
- You have a vendor quote and want a sanity check. A few hours of advice is enough, ideally from someone who builds.
- You have an AI pilot that demoed well and then stalled. A five day review will usually find the reason.
- You think AI could help but cannot say where, and the initiative budget is above €25,000. A two week readiness assessment fits.
- You have no CTO and three suppliers who all want a decision. A fractional CTO for at least three months fits.
- You know the use case and the data is clean. Skip consulting and start with a discovery week before the build.
Our consulting overview describes each of these engagements. If you are unsure which one applies, describe the situation in a few lines and we reply within one business day.