12 Questions to Ask Before Buying an AI Assistant

AI assistant demos run smoothly: prepared questions, prepared answers. The real difference hides in the details that surface after the contract. These twelve questions are the shortest, highest-value preparation you can do before buying.

Data and ownership

• 1. Where are customer conversations stored, and in which country?
• 2. If the contract ends, in what format and how quickly can I take my data?
• 3. Are my conversations used for model training? Can that be turned off?
• 4. What is the process when a customer requests deletion of personal data?

If these four do not have clear answers, stop there. Data ownership is the hardest thing to fix later.

Capability and limits

• 5. Which channels are supported, and do WhatsApp, Instagram and the website share one knowledge base?
• 6. Can I update the knowledge base myself, or does every change require a support ticket?
• 7. How do I define the handover rule; can a customer reach a human instantly on request?
• 8. Can it talk to my calendar, order or stock system?

Question six is critical: if you cannot update the knowledge base yourself, every price change makes you dependent on the vendor.

Cost items

9. Can I see a sample monthly invoice covering all five items? This single question surfaces most surprises early.

Continuity and exit

• 10. What happens during an outage: are messages lost or queued?
• 11. When and through which channel is support available, and is there a response time commitment?
• 12. If I leave, do I keep my number, knowledge base and conversation history?

Run the demo with your own questions

Instead of the vendor's scenario, use your hardest cases: your three most common customer questions, one complaint, and a deliberately vague message ("are you free tomorrow maybe"). Watching how it behaves with incomplete information teaches far more than a scripted demo.

Final check before deciding

Before you decide, request three things in writing: the data retention and deletion policy, a sample monthly invoice, and the data handover timeline on exit. A vendor who struggles to provide these will struggle with bigger things later.