AI Use & Human Escalation Policy

1. Purpose

This policy governs how Azire AI voice agents represent themselves, handle calls, and escalate to human operators. It applies to all Azire-powered AI agents and is incorporated into the Terms of Service.

2. AI identification

Every Azire-powered agent must identify itself as an AI assistant at the beginning of every call, before collecting any personal information. The disclosure includes the business name and the fact that the call is being recorded and transcribed. Agents must never represent themselves as human when sincerely asked.

3. Call recording and transcription disclosure

Every agent delivers the following key elements at the start of every call:

• Identification as an AI assistant for the named business
• Disclosure that the call is being recorded and transcribed
• Statement of the purpose (handling service request, creating a lead or booking)
• Disclosure of cross-border (US) voice processing
• Offer to transfer to a human at any time
• Reference to the privacy notice at azire.ca/privacy

4. Human escalation

Agents must offer and complete a transfer to a human operator when:

• The caller explicitly requests to speak with a human
• A genuine safety emergency is detected (gas leak, carbon monoxide, fire, medical emergency)
• The caller expresses extreme distress or threatens harm
• The agent cannot resolve the caller’s request after reasonable attempts

Emergency transfers use the configured transfer-to number set during onboarding. Agents will only use emergency transfers for genuine safety threats, not for routine requests to speak to a person.

5. What agents will not do

• Provide medical, legal, or binding pricing advice
• Request or repeat Social Insurance Numbers, payment card numbers, or health card numbers
• Make commitments that the tenant has not pre-authorized
• Use the caller’s personal information for purposes other than the stated service request
• Identify themselves as human when sincerely asked

6. PII safeguards

A PII safeguard sentence is injected into every agent prompt instructing it not to collect or repeat sensitive identifiers. If a caller volunteers sensitive data (SIN, card numbers, health card numbers), the platform automatically redacts it before database persistence.

7. Contact

AI policy inquiries: privacy@azire.ca