You've made up your mind: it's time to bring artificial intelligence into your business. You open your browser, search for “AI consultant” and find a hundred websites all promising the same thing: revolution, efficiency, the future. How do you pick the right one for you? Help, curiously enough, comes from the workshops of the thirteenth century.
The master, the apprentice and the notary
In medieval Italian cities, from Genoa to Florence, the know-how of a trade wasn't bought: it was learned in the workshop, alongside a master. And entering a workshop was a serious matter. Very often it was set out in an apprenticeship contract drawn up by a notary, and notarial archives, starting with Genoa's, preserve a great many of them. Historians study them precisely because they show, in black and white, how learning a craft worked.
What did they say? Usually the duration, often a few years; the master's obligations, who undertook to teach the trade and often to provide board and lodging; and the apprentice's obligations, who had to work diligently and stay until the end of the term. The guilds, through their statutes, also set rules on who could take on apprentices and under what conditions.
You may ask: Professor, what does a thirteenth-century shop boy have to do with choosing a consultant? Think about it for a moment. Those contracts put three things in writing: learning happened on the workshop's real work, the commitments were clear and verifiable, and by the end the apprentice had to be able to practice the trade alone.
Your company today is that apprentice, and the consultant is the master. Those three criteria still apply.
8 criteria for choosing an AI consultant
1. They ask questions before showing you a demo
A good consultant wants to understand how you work: who does what, where time is lost, which decisions need human experience. If at the first meeting they only talk about tools, they're selling a product, not consulting.
2. They're independent from vendors, or they say so clearly
Many consultants have commercial agreements with software platforms. That's not a crime, but you need to know. Ask openly whether they earn commissions on the tools they recommend.
3. They also tell you what not to do
A consultant who advises you against automating something is saving you money. Not everything should be automated: some tasks are too variable, others too sensitive, others simply not worth the investment.
4. They know the rules: GDPR and the AI Act
An AI project that processes personal data without proper safeguards is a problem waiting to happen. The consultant must know where the data goes, which contracts are needed with providers and which uses count as high-risk under the AI Act. We covered this in What the AI Act means for SMEs.
5. They suggest starting small, with a measurable goal
Beware of huge projects right off the bat. A good start is a pilot on a single process, with a clear goal (“cut customer response times”, “stop re-typing orders”) and a concrete way to measure it.
6. Pricing is clear, by phase
Analysis, development, training, support: each phase with its own cost and deliverables. If the quote is a single figure with no breakdown, ask for one.
7. Your data and tools stay yours
Accounts, data, automations and documentation must belong to your company. Always ask: “If we stopped working together tomorrow, what would I keep and how would I carry on?” A serious consultant has an answer ready.
8. They leave you self-sufficient
A workshop contract had an end date: the best master was the one you eventually no longer needed. Training your team should be part of the project, not an extra. Besides, the AI Act already requires organizations using AI systems to ensure a sufficient level of AI literacy among their staff.
The red flags
- Guaranteed percentages. “You'll cut costs by 40%” said before seeing your processes is a promise nobody can keep. Honest estimates come after the analysis, and they're ranges, not certainties.
- “AI will replace your staff.” Anyone who starts there hasn't understood where your company's value lies: in its people.
- One-size-fits-all solutions. The same package for a law firm and a logistics company? Unlikely to suit both.
- Jargon as a shield. If after an hour you still don't understand what they're proposing, it's not your fault.
- No questions about privacy. If the consultant doesn't ask what data the tool will handle, you'll discover the problem later.
- Vague or unverifiable success stories. Ask for concrete examples. And appreciate those who honestly distinguish between real clients and demo scenarios.
Questions to ask at the first meeting
- Which process would you start with in my company, and why?
- How will we measure whether the project works?
- Where will our data be processed, and with what safeguards?
- How much of my team's time will it take, on top of yours?
- What happens to the data and automations if we stop working together?
- What training is included, and for whom?
What about Penzai?
We're consultants too, so the most honest thing we can do is invite you to apply these criteria to us as well. Our method, which we call Human Tech, starts from people and processes and arrives at the minimum technology needed. You can see how we work in our typical scenarios and on our consulting services page. If you run a professional practice, you'll find specific ideas on the page for professional firms.
What to do on Monday morning
- Write down the problem, not the solution. In half a page, describe the process that costs you the most time, without naming any tools. It'll be the basis for every conversation with a consultant.
- Prepare the six questions. Print them and bring them to the first meeting. Note each consultant's answers and compare them.
- Decide how you'll measure success. A number you already know today: hours spent, response times, errors. Without a starting point you'll never know whether the project worked.
Want to put us to the test? Book a free intro call and ask us all six questions. If AI isn't the right answer for you, we'll tell you.
The Professor at Penzai
