Agent behavior

Prepare approved knowledge and fallback rules

What to give Havio before launch so the AI receptionist answers from approved facts and routes uncertainty correctly.

6 minUpdated 2026-06-30

Applies to

Knowledge setupSafe answersHuman fallback

Setup steps

Collect approved facts

Start with services, locations, hours, pricing boundaries, booking rules, cancellation rules, emergency rules, and frequently asked questions.

Mark restricted topics

List topics the AI should not answer directly, such as legal advice, medical advice, pricing exceptions, refunds, safety emergencies, or account disputes.

Define escalation language

Write what the agent should say when it is unsure, when the caller is upset, or when the call requires a person.

Review transcripts after launch

Use real call transcripts to improve the approved knowledge base instead of allowing the agent to guess.

FAQ

Can the agent answer pricing questions?

Yes, if pricing is approved and simple enough to explain safely. Exceptions, discounts, or complex quotes should route to a person.

What happens if the AI is unsure?

It should transfer, create a callback task, or say that a person will follow up. It should not invent an answer.

Approved knowledge base Qualification scripts Security

Still checking fit?

Map this article to your real call flow.

Support answers are useful only when they match your phone stack, caller types, fallback contacts, integrations, and compliance needs.

Map this workflow