How to hire an AI consultant: 7 questions to ask

To hire an AI consultant is a decision worth several thousand euros for most companies, and it still often gets made after one sales call and a good-looking deck. Eurostat found that 70.9 per cent of EU companies that considered AI and did not start named missing expertise as the reason, ahead of unclear legal consequences at 52.5 per cent.¹

Missing expertise is what an AI consultant sells, which is exactly why the choice of partner matters. The guide below covers what the work includes, seven questions that separate a good consultant from an expensive one, and what the EU AI Act changes about the scope. It ends with an honest answer on when you are better off alone. How satori. structures the same work, with fixed prices, is on our AI consultant page for Europe.

What does an AI consultant do?

An AI consultant helps a company pick AI tools, set them up safely and train the staff to use them, and the work splits into three parts that each leave something behind.

Strategy means choosing the tool and the first tasks, for example Claude for a finance team that reads long contracts or ChatGPT for a marketing team that needs images every week. The output is one page with the chosen tool, three first tasks and the hours each one takes today.

Setup means configured accounts, single sign-on, data settings and a first AI policy, meaning the written rules on what staff may paste into the tool. The difference between a good and an average consultant shows most clearly here, because a wrong data setting survives unnoticed until someone audits it.

Training means your team practises on its own documents instead of generic examples. Licences nobody uses cost the same as licences everybody uses, and training is the only part of the work that moves that number.

Why do companies hire an AI consultant?

Most companies can start with AI on their own, since creating an account takes five minutes and the vendors publish decent guides. What goes wrong comes later, in three patterns we see again and again.

Tool sprawl comes first, with half the team on ChatGPT, three people on Claude, one developer on a coding tool paid by private card and no shared rules. The security gap comes second, since most people know not to paste customer data into a chat, although few can say exactly which data counts. The missing measurement comes third, with licences paid every month and nobody able to say whether anyone saves an hour.

A consultant who has run the same rollout before closes all three with a plan, written rules and a number that shows whether the licences pay back.

How to hire an AI consultant: seven questions to ask

1. Which AI tools do you run in production?

A consultant who knows only one tool will recommend that tool for every problem. Ask which tools they have running at customers today, and which ones they advise against for your case, since the second answer tells you more.

2. Do you train our staff, or only set things up?

Setup without training leaves you with configured accounts nobody opens. Training is also a legal point now, because the AI literacy duty in the EU AI Act has applied since 2 February 2025. It requires staff who use AI to have a level of understanding that fits their role.²

3. Where will our data be stored, and under which terms?

Ask for the region, the data processing terms and whether the vendor trains on your data. Anthropic stores Claude Team and Enterprise data in the US by default and does not train on commercial data by default. A company that needs processing inside Europe can run Claude on Amazon Bedrock through its EU regions.³⁴⁵ A consultant who answers that it is probably fine has not read the terms.

4. Which parts of the EU AI Act apply to what you build?

The transparency rules have applied since August 2026, meaning a customer who talks to an AI system must be told so, and the rules for high-risk areas such as recruitment apply from 2 December 2027.² A consultant who builds a CV screening flow without mentioning that date is building something you may have to rebuild.

5. Can you show a measured result?

Ask for one customer case with a number before and after, for example hours per month-end close or minutes per quote. A consultant who has never measured anything cannot tell you whether their own work paid back, and that is the answer to take home.

6. What is included after go-live?

The first month after launch is when questions pile up and habits either stick or fade, so ask what support, follow-up and reporting are included. A quarterly usage report showing which seats nobody opens pays for itself the first time it cancels three of them.

7. Is the price fixed, and what exactly does it cover?

Hourly billing puts the risk of a slow week on you, while a fixed price puts it on the consultant. Take a consultant who quotes 40 hours at €150 an hour, where the bill reaches €6,000 as planned. If the work runs 50 per cent over, which happens often on a first AI project, the bill lands at €9,000.

What does an AI consultant cost?

Most European AI consultants bill by the hour or by the project, which makes quotes hard to compare. Ask each candidate for the total for your team size and what happens if the work runs over. satori. publishes fixed prices per user instead, and every number below is excluding VAT.

A launch for ten people costs €5,600 as a one-time fee. The sum pays back within five months when those ten people save about three hours each per month at a loaded cost of €40 an hour. That arithmetic holds only if the team actually uses the tool, which is why the training and the follow-up weigh more than the setup. The full scope of the launch is on get started, and training on its own is on Claude training.

Small company or large organisation

A company with ten to fifty staff usually needs a fast start: one tool, trained people, an AI policy and three tasks that show results in the first month. A workshop plus a month of follow-up often covers it, and the whole budget stays under €6,000.

A large organisation needs governance, integration with existing systems and a rollout in phases, with GDPR and the AI Act handled in writing for every flow. The cost is higher, although so is the risk, since a wrong data setting multiplied by five hundred users becomes an incident report.

When you are better off without a consultant

A team with one confident admin, one tool and no personal data in its prompts can get far alone with the vendor's own guides and a free afternoon. Hire help when a second tool appears, when personal data enters the picture or when licences have run for three months with nobody able to say what they saved.

Open a blank document and write the three tasks that take your team the most hours each week, with the hours next to each one. In fifteen minutes you have a list of three numbers, and that list is the brief any AI consultant should ask you for first.

---

Sources

Footnotes

1. Eurostat (2026). Use of artificial intelligence in enterprises. Statistics Explained. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises ↩

2. European Commission (2026). AI Act: regulatory framework for AI. Shaping Europe's digital future. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai ↩ ↩²

3. Anthropic (2026). Where are your servers located? Do you host your models on EU servers? Anthropic Privacy Center. https://privacy.claude.com/en/articles/7996890-where-are-your-servers-located-do-you-host-your-models-on-eu-servers ↩

4. Anthropic (2026). Is my data used for model training? Anthropic Privacy Center. https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training ↩

5. Amazon Web Services (2026). Regional availability of models in Amazon Bedrock. https://docs.aws.amazon.com/bedrock/latest/userguide/models-region-compatibility.html ↩