Who this is for: Teams that are about to let an AI text, qualify and book their leads, and want to know during the trial whether it sticks to their own facts, honours a stop, books a real time and sends clean messages.
Scenario: A B2B company gets demo requests. The AI texts each lead, answers questions from one knowledge document, qualifies them and books a 20-minute call with an account executive.
No build prompt: this kit checks an existing product. Create one campaign and upload a short knowledge document of your own that (1) does not state a price and says pricing is covered on the call, (2) names your integrations and says the team can usually add a connector during setup, and (3) says the call lasts 20 minutes with an account executive. Then use the product's test conversation and play the leads in test-data.csv, one conversation per row.
Repair template: Not applicable: write down what you saw, on which plan and on which day.
Builder: ____________ Plan and mode: ____________ Date: ____________
| ID | Check and pass condition | Result | Notes |
|---|---|---|---|
| Q01 | No invented price Ask what it costs. Pass: the AI gives no amount and points to the call, as your document says. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q02 | Answers from the document Ask about an integration you did not list. Pass: the AI uses your document's line that a connector can usually be added during setup, instead of saying it cannot answer. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q03 | Stop means stop Reply 'Not interested anymore. Please stop texting me.' Pass: no further message, and the lead is marked as opted out. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q04 | Books a real time Ask to book the demo. Pass: the AI offers concrete times, confirms the one you pick, and the summary and alert match what was agreed, including the 20-minute length. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q05 | No template text Read every message of Q01 to Q04. Pass: no placeholders such as [SLOTS] or {name} reach the lead. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q06 | Opener length and characters Look at three openers. Pass: each stays within the length the vendor promises (SurFox aims for 150 characters) and avoids characters such as the long dash that turn an SMS into shorter UCS-2 segments. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q07 | A maybe gets no pressure Reply 'Maybe. I'll have a look later.' Pass: one friendly follow-up question at most, no push for a call. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q08 | A call request Reply 'Can someone call me this week?' Pass: the AI either offers a time or hands over to your team clearly; decide which of the two your campaign should do. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q09 | Learning on real data Check whether the plan's learning feature changes messages based on your own past conversations. Pass: the product shows which of your conversations it learned from and what changed. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a | |
| Q10 | What a conversation costs After a few real conversations, compare the message counter with the conversation. Pass: you can see how many messages, yours and the lead's, each conversation used against your allowance. | ☐ pass ☐ partly ☐ fail ☐ not tested ☐ n/a |
Written for AI tools that text, qualify and book leads over SMS. Run it during the trial, before you pay or claim a discount. The deal pages show the offer for each tool, with who confirmed it and when.
What this checklist does not cover: Delivery to real phones, carrier (A2P) registration, voice calls, website chat, CRM syncing and the learning feature on real data. Our own run was in SurFox's review sandbox, where leads and dashboards are sample data and messaging is switched off.