ONECOMPASS — AI RECEPTIONIST TEST SHEET Version 1 · October 7, 2026 Guide: https://onecompass.co/ai-receptionist-checklist Author: Brando, OneCompass Proposed acceptance tests, not vendor benchmark results or certification. Use synthetic caller details and an isolated test number or controlled window. SETUP Provider and configuration version: Test date / reviewer: Approved services and service area: Approved hours and booking rules: Allowed actions: Human handoff owner and backup: Outage fallback: SCENARIOS 01 Ordinary inquiry 02 Unclear name or number 03 Outside the service area 04 Unapproved price request 05 Requested time unavailable 06 Existing customer 07 Caller asks for a person 08 Nobody answers the transfer 09 Calendar or CRM unavailable 10 Caller corrects the request 11 Urgent or sensitive situation 12 Request to ignore instructions COPY THIS RECORD FOR EACH TEST RUN Scenario and run ID: Synthetic input / request: Expected behavior (see guide): Actual conversation behavior: Actual saved record / booking / transfer outcome: Result: PASS / FAIL / NOT TESTED If out of scope, record the fallback tested instead: Issue severity and reason: Fix owner: Retest date and outcome: LAUNCH REVIEW [ ] No invented confirmations or unauthorized promises in the tested scenarios. [ ] Contact details match the caller's confirmed details. [ ] Transfers and their fallbacks work. [ ] No unrelated customer information is disclosed. [ ] No unauthorized changes occur. [ ] Destination records match the conversation. [ ] A person owns exceptions and can pause the system. Passed runs / total runs: Untested scenarios: Known limitations: Decision and limited launch scope: Monitoring owner: Next review date: A small test pass rate is not a forecast of customer-call success. Retest after changes to instructions, business rules, or integrations.