Italiano 
Why questions matter more than the demo
A well-made demonstration always impresses. But the problems of an AI project appear later: which data is needed, how it connects to existing systems, who checks the answers, who maintains it over time. These are things a demo rarely shows.
A good supplier does not tell you straight away that everything is possible. They first ask what you want to achieve and how you will check it.
The questions below are organised by theme. You can bring them to a first meeting and use them to compare proposals.
Questions about the goal
- What problem are we solving and how do we measure it today?
- What is the first perimeter, the smallest part to start from?
- What does «it worked» mean and when do we assess it?
- What stays outside this first project?
If you have not yet chosen the task to start from, you will find a method in Where to start with AI in business: how to choose the first process to improve.
Questions about data and security
- Which data is needed and where does it come from?
- Where and how is it processed?
- Can our content be used to improve models or services?
- Who can access the data, inside the company and in the supplier's team?
- How is data deleted at the end of the project?
For privacy and compliance matters, talk to an adviser: the supplier's answers should be read together with them. On company data we wrote in Company data and AI: what you can share with an assistant and what you cannot.
Questions about integration
- Which systems must it work with: CRM, email, management software?
- What can it read and what can it change?
- What happens if one of these systems changes?
- Who handles the connection and who maintains it?
Connecting to systems is the topic of the page on AI integration and workflows (Italian page); a similar case, for events, is in Integrare la segreteria organizzativa con i sistemi aziendali: come farlo bene (Italian).
Questions about controlling the answers
- Which sources are the answers based on?
- How are errors reduced and detected?
- When and how does a case pass to a person?
- Who reviews the conversations and how often?
- Is there a record of what the AI has done?
An example on the customer support side is in AI for customer care: faster answers without losing the company's tone.
Questions about testing, timing and maintenance
- How is it tested before release and with which real cases?
- Who maintains the system after release and what is included?
- Who owns configurations, content and data at the end of the project?
- How can we change supplier or leave without losing the work done?
Warning signs
- Promises of results without asking how you work today.
- Vague answers about where data ends up.
- No perimeter: a project that wants to change everything at once.
- No handover to a person in doubtful cases.
- Only demos with sample data, never with your real cases.
For examples of AI applied to a company's processes, also outside events, see Segreteria eventi 4.0: come usare l'IA (Italian) and IA negli eventi: come rivoluzionare l'organizzazione (Italian).
What to bring to the first meeting
Prepare these six points
The problem in one sentence, with today's starting point measured.
Data and systems involved, with who manages them.
Who in the company follows the project and can say what is correct.
Criteria to assess a first test, with a date.
The rules on data, already agreed with your adviser.
The questions in this guide, chosen for your case.
Frequently asked questions
Having an order of magnitude helps, but the quote depends on the case. It is normal for a serious supplier to first ask to understand the process and the data involved. Often yes. A small perimeter lets you check the result before extending the work. Whoever knows the process and whoever manages the data and systems involved, plus whoever decides on the budget. Yes. Ask to understand what was done and how it was checked, not only the declared results.Do I need a defined budget before talking to a supplier?
Is it better to start with a pilot project?
Who should attend the meetings?
Can I ask for examples of similar projects?
Are you evaluating an AI project?
Let's talk about your case: we will tell you honestly whether it makes sense to start and from which perimeter.
Let's talk about your case
