Prepare a fair evaluation
Write a one-page offer brief with price, eligibility, exclusions, booking method and escalation rules. Use the same brief for each tool. Record the plan and configuration so a comparison can be repeated.
Build scenarios from recurring questions without exposing prospect identities. Use fictional details where needed and label the examples as simulations. Avoid feeding a real lead into several live automations at once.
Ten scenarios worth testing
| Scenario | What a good response does |
|---|---|
| Clear fit | Asks only missing questions and offers the approved next step. |
| Not a fit | Explains the relevant limitation without pushing a call. |
| Unknown offer detail | Acknowledges the gap and asks or hands over instead of inventing a fact. |
| Price objection | Uses the actual terms without inventing a discount or guaranteed outcome. |
| Long message | Preserves the main question and details already provided. |
| Voice note | Responds to the spoken content; requests clarification if it is unclear. |
| Image | Uses visible context without treating a screenshot as proof. |
| Stop request | Respects the request and does not keep selling. |
| Human requested | Pauses the automated sales exchange and supports takeover. |
| Booking | Uses the right event link and distinguishes sent links from confirmed appointments. |
Set acceptance criteria before the pilot
A fabricated price or ignored stop request should block launch even if most replies sound good. Decide which failures are critical for your offer before scoring. Keep a separate category for issues caused by unclear instructions so you can improve the setup and rerun the same test.
Do not turn an internal quality score into a claimed revenue uplift. Passing ten scenarios demonstrates behavior in those scenarios, not universal performance.
- Record the answer, expected behavior and reason for each score.
- Review failures with the person responsible for coaching sales.
- Retest changed instructions against earlier scenarios.
- Check delivery and takeover in the actual connected channel.
Run a limited, measured pilot
Start with a manageable set of inbound enquiries and an available human owner. Compare similar lead sources and time periods. Track unique eligible conversations, qualified bookings, attended calls, handovers and sales where available.
Record changes to ad spend, pricing or offers during the pilot. Those changes can affect results. Include tool cost, human review time and setup effort in the decision.