Free samples attract people who aren't your audience. A publisher's specimen book ends up with a student, a relative or a reseller; a doctor's sample goes to someone who just wants the free item. Each wasted sample costs the product, the courier and the follow-up. In one deployment, a CBSE exam-prep book publisher uses an AI call to verify that each person is a practising teacher before a specimen book ships. This post covers how to design that check.
Verify before you ship: Start free on Edesy — Rs 50 free credit, no demo needed.
Step 1: write down who is eligible
Before the script, decide the rule. For the publisher:
- a practising school teacher (not a tutor-only, student or parent);
- teaching a class and subject the book is for;
- one sample per teacher.
Be specific. "Teachers" is too loose; "currently teaching Class 9–12 science at a school" is a rule an AI call can check.
Step 2: verify with open questions
The agent asks questions a genuine person answers easily and in detail:
- "Aap kaunsi classes padhate hain?" (Which classes do you teach?)
- "Kaunsa subject?"
- "School ka naam kya hai?" (only if you need it)
Never offer the answer options. "Are you a TGT or a PGT?" invites anyone to pick one. More on phrasing in verification questions on AI calls that verify.
Step 3: listen for volunteered detail
The strongest signals are the ones the person gives without being asked. Real teachers often mention:
- a designation such as PRT, TGT or PGT;
- a qualification such as B.Ed or CTET;
- specific sections, board-exam pressure, or their timetable.
Vague answers ("haan, padhata hoon... sab kuch") or answers that echo your own wording are weaker. Detail volunteered unprompted is hard to fake.
Step 4: record a verdict, not just an outcome
Post-call extraction turns each call into fields your dispatch team can act on:
| Field | Values |
|---|---|
| Verdict | Eligible, not eligible, unclear |
| Reason | Not a teacher, wrong person, requesting for someone else, wrong class or subject, refused to answer |
| Class and subject | As stated |
| Designation or qualification volunteered | Yes / no + what |
Only eligible rows go to dispatch. Unclear rows go to a person for a quick check. Everything exports to CSV or Google Sheets, or to your CRM by webhook.
Step 5: be polite to people who don't qualify
The agent doesn't accuse anyone. "Thank you, this offer is only for teachers currently teaching these classes" is enough. People requesting for someone else can be asked to have that person call or reply, and a WhatsApp message can carry the details.
Watch out for
- Keyword checks that break in transcription. On Hindi-English calls, speech-to-text can write English acronyms like "TGT" in Devanagari, so a check for the English word misses it. See Hinglish STT, acronyms and keyword checks.
- Over-strict rules. A new teacher may not volunteer a designation. Use "unclear" rather than rejecting.
- Duplicate requests. One sample per person: match on phone number before dispatch.
Beyond books
The same verify-then-ship pattern works for pharma samples to doctors, trial packs to pharmacists, lenses to opticians and product samples to trade buyers. Only the eligibility rule and questions change.
Try it
Create a free account, start from your own eligibility rule, and test on 30 calls. See the full programme in how a publisher verifies teachers with AI calls.