How to choose an artificial intelligence agency. Nine questions to ask before you sign.
This guide is written by an agency. The questions apply to us too, and our answers are at the end.
You choose badly when you compare the prices of different things. Two quotes for “an AI system” can describe a subscription to an off-the-shelf product and a system built around your processes. The questions make offers comparable before you look at the price.
Nine questions matter, and they cover four things: who keeps what gets built, how the result is measured, what the machine does on its own, and what the AI Act and the GDPR require.
Each question has an expected answer, written down before you sign. An answer that only comes verbally, or only after the contract, leaves that item uncovered.
Why the choice is hard for an SME
Because judging an artificial intelligence proposal takes skills that businesses often don't have in-house. According to Istat, Imprese e ICT, 2025, a lack of the right skills is holding back AI adoption in almost 60% of the businesses that considered investing in AI but then didn't.
Meanwhile the gap is widening. In 2025, 15.7% of small and medium-sized businesses with at least 10 employees used at least one artificial intelligence technology, against 53.1% of large ones; a year earlier the figures were 7.7% and 32.5%. Anyone arriving now is buying in a fast-moving market, where the right questions protect you better than technical knowledge.
The nine questions, and the answer you should get
Nine questions, each with the answer you should get. You don't need to know the technology to judge them: the answer just has to be written, precise and verifiable. A vague answer on a point is an uncovered point, and it's worth asking again before you sign.
| Question | Why it matters | The answer you should get |
|---|---|---|
| Who owns the data, the system and the accounts at the end?ownership | A system built around your processes is only worth it if it stays yours. If accounts and data are in the supplier's name, changing supplier means starting again from scratch. |
A written clause: the data stays yours, what's built for you is yours once paid for, and the accounts are in the business's name. |
| What happens if we stop paying?the exit | The answer tells you whether you're buying a system or renting it. The fee covers maintenance and services, and it's right that it exists: what matters is what's left when it ends. |
What keeps working, what stops, how many days until you receive your data and exports, and in what format. |
| Which number shows it worked, and what is it today?the measurement | Without a number measured beforehand, the after can't be checked and every result becomes a matter of opinion. You can work out the starting figure yourself, as with the cost of answering enquiries by hand. |
A single indicator, for example average time to first reply, measured before starting, and the date when it will be measured again. |
| Who builds it, and who steps in when it stops?the people | Knowing who works on the system tells you who'll be able to fix it. A name and a response time are worth more than a description of the team. |
The name of the person building it, the response time for a fault, and who takes over if that person isn't available. |
| What does the system do on its own, and what goes through a person?oversight | A wrong message to a customer costs more than the time saved. The line between automatic and approved is decided beforehand, in writing, and it can be moved. |
The list of automatic actions and those that wait for approval, and who can change where that line sits. |
| Does the system talk to your customers?AI Act, Article 50 | From 2 August 2026, a system that interacts directly with people must let them know it's artificial intelligence, as required by Article 50 of Regulation (EU) 2024/1689. |
Where and how the system says so, with the exact text of the message. |
| Who processes personal data, and where is it kept?GDPR, Article 28 | If the supplier, or the models it uses, process your customers' data on your behalf, you need the agreement required by Article 28 of the GDPR. |
The data processing agreement appointing them as processor, the list of suppliers used, AI models included, and the country where the data is kept. |
| Who pays when the model changes price?dependency | AI systems rest on third-party models, which change versions and price lists. If usage costs go up, someone pays, and it's worth knowing who in advance. |
Which models are used, who covers the usage costs, and what happens to the fee if those costs rise. |
| What does the price cover, and when does the fee start?the scope | Two quotes can only be compared if they describe the same work: what gets built, what's left out, by when. Otherwise you're comparing the prices of different things. |
A written scope with a fixed price, and the date the fee starts, tied to the system working. |
1. Who owns the data, the system and the accounts at the end of the work? 2. If we stop paying, what keeps working, and in what format do we get our data? 3. Which number measures the result, what is it today, and when will it be measured again? 4. Who builds it, and how quickly do they step in if the system stops? 5. What does the system do on its own, and what waits for approval? 6. If it talks to customers, where and how does it disclose that it's an AI? 7. Can you send us the agreement required by Article 28 of the GDPR and the list of suppliers, models included? 8. Who pays if the model's usage cost goes up? 9. What's the scope, what's the price, and when does the fee start?
Three signs a proposal needs a second look
A percentage result with no starting number. A promise like “70% less time” only means something if someone has measured the time it takes today. If that measurement is missing, the percentage stays a promise, and you should ask how it will be checked.
A price with no scope. “All inclusive” is a convenient phrase for the seller. For the buyer, what matters is the list of what gets built and what's left out, because that's where extra costs come from once the work has started.
A proposal with no name on it. A document signed by “the team” leaves open who builds and who answers. A name and a direct contact are the simplest way to know who to call when something stops.
The same questions, put to Itria
Itria's answers are already written down in the Terms (in Italian), the principles and the AI transparency page, and they stand as commitments. Below are the relevant statements, question by question, and where to find them.
| Question | What it says |
|---|---|
| Ownershipquestions 1 and 2 | Once what's owed has been paid, the client keeps ownership of their data and of what was built for them; and every engagement ends with the client owning the system. Terms, point 8. |
| Measurementquestion 3 | No work is proposed until we've looked at where the business is losing time or revenue today. Terms, point 7. |
| Peoplequestion 4 | Itria is run by Mattia Esposito himself, who builds and answers. About us. |
| Oversightquestions 5 and 6 | “The machine prepares, a person always decides.” Ethics. Where a reply goes out on its own, the owner decides that channel by channel. AI transparency. |
| Data and modelsquestions 7 and 8 | “We never use a client's data to train models.” The third-party model is named, along with the agreement that rules out training. AI transparency. Who covers usage costs is written into the scope of each engagement. |
| Pricequestion 9 | The price is fixed and agreed in advance, and no fee starts before the system works. Terms, point 7. |
We take on only a few clients at a time, because every system is built by one person. That's why the questions on this page are worth putting to us too, before we start.
Questions and answers
How do you choose an artificial intelligence agency for an SME?
With nine questions, asked before signing and answered in writing. They cover four things: who keeps the data, the system and the accounts; which number measures the result and what it is today; what the system does on its own and what goes through a person; and what the AI Act and the GDPR require.
The questions make different offers comparable before you look at the price. A vague answer, or one that only comes after the contract, leaves that item uncovered.
Who owns a system built by an AI agency?
It depends on the contract, and it's the first thing to get in writing. The most common split separates what's yours, meaning the data, configurations and workflows built for you, from what stays with the agency, meaning the methods and generic tools it uses with everyone.
It also matters whose name the accounts for the services are in. If they're in the supplier's name, changing supplier means starting again from scratch.
Does a business's AI system have to comply with the AI Act?
Yes, for the parts that apply to it. From 2 August 2026, the transparency obligations in Article 50 of Regulation (EU) 2024/1689 apply: a system that interacts directly with people must let them know it's artificial intelligence.
Since 2 February 2025, Article 4 has also applied, as rewritten by Regulation (EU) 2026/1744: businesses using AI take measures to support the AI literacy of the people using it, as explained on the page about AI Act training requirements. If the system processes personal data on the business's behalf, the GDPR applies, with the Article 28 agreement.
Is it better to pay a monthly fee or a one-off price for an AI system?
It depends on what the fee covers. An AI system has costs that continue after delivery: models are paid for by usage, services have subscriptions, and someone has to step in when something changes. The fee pays for those.
Two things to get in writing: what it covers, item by item, and when it starts. A fee that begins before the system works is paying for work not yet delivered.
How do you compare two quotes for an artificial intelligence system?
By bringing them to the same scope: what gets built, what's left out, by when, and which number will be used to check the result. If a quote doesn't say, ask before comparing prices.
Then compare the exits: what stays yours if the relationship ends, and in what format you get your data. A lower price with a more expensive exit costs more over time.
Notes on sources
- Istat, Imprese e ICT, 2025 (in Italian), published on 15 December 2025. The source of the 16.4% of businesses with at least 10 employees using AI (8.2% in 2024), the 15.7% of SMEs against 53.1% of large businesses, and the statement on skills, given here as a faithful paraphrase. The survey covers only businesses with at least 10 employees: micro-businesses are left out.
- Regulation (EU) 2024/1689 on artificial intelligence, Article 50(1), and Article 4, as replaced by Regulation (EU) 2026/1744.
- Regulation (EU) 2016/679, the GDPR, Article 28 on processors.
- The statements in the section on Itria's answers come from the Terms (in Italian) (points 7 and 8), given as a faithful paraphrase, and are quoted from Ethics, AI transparency and About us. Istat and Itria's pages were opened and read on 25 September 2026; the texts of the regulations are those already checked for this site's AI Act pages.
Bring the nine questions to the first meeting. Ours included.
The first step with Itria is a fifteen-minute video call: we start from where the business is losing time or revenue today, and the questions on this page get asked there, before any quote. Drop us a line about what's slowing you down. We'll make the first move: we'll look at what a customer sees when they search for you, and tell you what we found. Even if we never end up working together.