Product questions
Ask them to show not only a chat reply, but the full lifecycle of an AI employee.
- Who can create and change the employee?
- How is knowledge structured?
- Are there versions and publishing?
- Can it be tested before launch?
Questions about actions and security
The main risk appears where the AI does not only speak, but also changes an external system.
- Which permissions does the integration receive?
- Can confirmation be required?
- Is an action log kept?
- How is data from different projects isolated?
Questions about operations
After the pilot, the system needs to be updated, evaluated, and supported.
- How is cost calculated?
- Where are errors visible?
- How is the knowledge base updated?
- Who is responsible for quality after launch?
How to compare pilots
Give participants the same set of real conversations, success criteria, and forbidden actions. Compare not the impression of a demo, but the share of correctly completed scenarios, errors, cost, and the effort required for support.