Before buying an AI receptionist, test these seven calls
A practical scorecard for fees, service areas, booking, after-hours requests, existing jobs, complaints, and human handoffs.
Test the work after the conversation
Prepare a small test business with explicit fees, service areas, hours, availability rules, and escalation contacts. Use a designated test environment and numbers you control. Tell the vendor you are evaluating the workflow; do not create appointments in a real customer's calendar.
After each call, inspect the message, booking, and follow-up visible to the operator. A friendly answer followed by a missing or incorrect record is an incomplete result.
Seven calls to put on your scorecard
| Scenario | Ask the receptionist | Inspect afterward |
|---|---|---|
| Fees | “What will you charge just to come out?” | It used the supplied policy and recorded what was disclosed. |
| Service area | Give an address just outside the supported area. | It handled the boundary without promising unavailable service. |
| Availability | Request a time that is not available. | It offered only verified slots or clearly requested a callback. |
| After hours | Ask whether someone can come tonight. | It distinguished answering the call from dispatch and arrival. |
| Existing job | Ask about an unresolved visit or warranty concern. | It captured the job reference and assigned the appropriate next step. |
| Complaint | Explain that nobody arrived at an agreed time. | It acknowledged the concern, captured it accurately, and followed the escalation policy. |
| Human help | Ask to speak with a person and interrupt once. | It followed the configured transfer or callback path without pretending a transfer occurred. |
Use both English and Spanish if you serve callers in both languages. Repeat important calls with a corrected address or a changed preference. This checks whether the final record reflects the correction.
Use three outcomes, not a vague star rating
Mark each scenario pass, needs clarification, or fail. A pass means the spoken commitment and recorded result agree with the supplied policy. “Needs clarification” means you cannot yet verify an outcome. A fail means the system invented a fact, lost necessary information, or promised an action that did not happen.
Record the setup, expected behavior, observed behavior, and evidence for each result. Ask the provider to fix a failed scenario and rerun it with the same inputs. Do not average away a critical failure just because the other conversations sounded good.
Questions for the provider
- Who updates business policies, and when do changes take effect?
- What happens when the calendar, transfer destination, or notification service is unavailable?
- Can the operator inspect and correct a captured request?
- Which languages and workflows are supported in the configuration you would actually buy?
- How are duplicate requests and repeated calls handled?
Sources and scope
Selected customer accounts checked on September 27, 2026 described automated phone loops and after-hours systems that only took messages. Those accounts informed the scenarios; they are not a comparative product test. This original scorecard applies to AugmentDesk too. Our commercial relationship and evidence policy explain that distinction. Verify each capability in the actual configuration rather than inferring it from a vendor's demo.