Build a discovery question bank for one persona, with cost-of-inaction probes

What this does

You describe a persona and your offer. The AI builds a question bank in groups (situation, problem, impact, decision, timing) and adds follow-up probes that help a buyer put a real cost on leaving the problem alone.

Who it’s for

AEs and SDRs who want better first-call questions; managers building a shared question list for a team. Not a script to read in order; pick 5-8 per call.

What you need

• Persona (title, industry, company size)
• What you sell and the problem it solves, in your words
• What a qualified buyer looks like to you
• Topics you must learn on a first call
• Tool: Any AI chat tool (for example ChatGPT, Claude, Gemini, or Grok), A notes doc
• Cost: $0 with a free AI chat tier (limits change)
• Setup: 3 minutes (estimate)

Steps

1. Fill in the persona and what you sell. Add the three things you most need to learn on a first call.
2. Run the prompt.
3. Cross out any question that sounds like a quiz or that you’d never say out loud.
4. Pick 5-8 for your next call; save the rest in your notes.
5. After a few calls, note which questions got real answers and delete the rest from your bank.

The prompt

You are a sales coach. Build a discovery question bank for one buyer persona. Use only the information I provide about my product; don't invent features, customers, or statistics.

NO-ADDITIONS RULE: Do not add facts, tips, statistics, scarcity or urgency, promises, references or testimonials, or the gender of any person that I did not supply.

Persona: [TITLE, INDUSTRY, COMPANY SIZE]
What I sell and the problem it solves: [DESCRIPTION]
A qualified buyer looks like: [CRITERIA]
Three things I must learn on a first call: [1, 2, 3]

Output, with 3-5 questions per group, all open-ended and short enough to say out loud:
A. Situation (how they do it today)
B. Problem (what isn't working)
C. Impact (what it costs in time, money, risk, or stress; ask them for numbers, never supply numbers yourself)
D. Decision (who else is involved, how they've bought before)
E. Timing and alternatives (why now, what else they're considering, including doing nothing)

Then add a section "COST-OF-INACTION PROBES": 6 follow-up questions that help the buyer say, in their own words, what happens if nothing changes. Don't be pushy or leading.

Finally, list 3 questions NOT to ask on a first call and why.

Sample input (fake data)

All names and companies are fictional. This is the input for run 4 (the output below), on the current prompt. Runs 5 and 6 used different inputs; see the test log.

Persona: Compliance Manager at a regional credit union (assets under $1B)
What I sell: Software that tracks policy attestations and training deadlines for staff
Qualified buyer: 100+ employees; today uses spreadsheets and email reminders
Must learn: How they track attestations today, what happens when someone misses a deadline, who approves tools

Sample output

Real output from run 4 on 2026-10-14, using the current prompt, from an unspecified AI assistant (Grok Bot agent session; model name not shown). Copied as produced. Nothing was fixed after the fact; the problems we found in it are listed under “How it went”.

**A. Situation**
1. Walk me through how you track staff attestations today.
2. Who sends reminders, and how?
3. Where is the record kept?
4. How do you prepare when an examiner asks for it?

**B. Problem**
1. What part of tracking takes the most effort?
2. When does a deadline get missed?
3. How do you find out someone hasn't completed training?
4. What do employees say about the process?

**C. Impact**
1. How many hours a month does tracking take your team?
2. How many people are usually late?
3. What happens when someone misses a deadline?
4. What would an incomplete record mean for you?

**D. Decision**
1. Who else weighs in on a tool like this?
2. How did you choose the last system you adopted?
3. Who would need to approve security or vendor reviews?
4. What would the approval steps look like, and about how long do they take?

**E. Timing and alternatives**
1. Why look at this now?
2. What else are you considering?
3. What happens if nothing changes this year?
4. What would have to be true for you to switch?

**COST-OF-INACTION PROBES**
1. If the process stays as it is, what do you expect next year?
2. What would that mean for your team's time?
3. What would it mean for your next exam?
4. What have you already tried?
5. What would you do with the time if tracking were easier?
6. Who else would feel the effect if nothing changes?

**QUESTIONS NOT TO ASK**
1. "Are you worried about failing your next exam?" Leading and alarming; ask what they expect instead.
2. "What's your budget?" Too early; ask about impact first.
3. "Are you the decision maker?" Sounds like a screen; ask how decisions get made.

Check before you send

• No question supplies a statistic; they should ask the buyer for it.
• Questions don’t assume facts about the buyer.
• You’d say each one in your own voice.
• Impact probes don’t pressure the buyer.

How it went

• Test date, tool, model: 2026-10-14. Unspecified AI assistant (Grok Bot agent session; the model name is not shown to the agent). One tool only.
• Runs logged: 6, on 6 different fake inputs (see the test log). Runs 1-3 used the original prompt. Runs 4-6 are a second round on new fake inputs after the prompt was fixed, as listed below. Every run was done by the same AI assistant session that wrote the fake input, and in round 2 the same session also edited the prompt, so these are not blind or independent runs. A script counted words and characters against the prompt’s limits. No second person re-ran it from the written steps.
• Why the label is “Tried in an AI tool,” not “tested”: we reserve “tested” for plays with at least 3 logged runs in at least 2 different AI tools, and we ran one tool. We also have no independent reproduction of the steps and no timed manual baseline. The runs are real, but this is one tool’s behavior, and 3 runs per round is too few to say a problem is gone.
• Time saved: Not measured. We did not time a manual baseline or an assisted run.
• What we changed after round 1: Added: no-additions rule only.
• Round 2 (current prompt, runs 4-6). What worked:
  • No statistic or benchmark appeared in any run, including the credit-union case (run 4) that is the regulated-industry setup that produced ‘Industry data says…’ in the first round. Impact questions all asked the buyer for numbers.
  • No ‘unlock’ or similar hype words; structure was right in all 3 runs (4 per group, 6 probes, 3 ‘do not ask’ items).
  • The ‘questions not to ask’ lists were specific (for example, not asking ‘Are you worried about failing your next exam?’).
• Round 2: what failed or needed fixing:
  • Some questions still assume a fact about the buyer (‘your next exam’, ‘unbilled hours’, ‘security or vendor reviews’). The rule covers statistics and facts supplied, not assumptions built into a question.
  • Decision and Timing groups are close to identical across personas (‘Why look at this now?’, ‘What else are you considering?’), so the bank is only partly persona-specific.
  • 29 items per run is still more than a rep needs for one call; picking 5-8 takes time.
• Checked against “Check before you send”: ‘No question supplies a statistic’ passed in all 3 runs; ‘questions don’t assume facts’ failed mildly in a few questions; none pressured the buyer.
• Edge cases: Regulated-industry persona (run 4) did not produce a statistic. Only 3 runs, so this is not proof the problem is gone.
• Round 1 (original prompt, runs 1-3): what failed, kept for reference:
  • Run 2, impact question 5: ‘Industry data says unplanned downtime costs hospitals an average of $X per hour.’ The prompt says never to supply numbers; this one implies a benchmark exists. It has to be deleted.
  • Run 2 probe 6 contains ‘unlock’, a word our own voice rules ban.
  • Some probes lead the witness: ‘What would you tell your director if records were incomplete?’, ‘How many jobs do you think you lose to slower quotes?’. The prompt says not to be leading, and the model did it anyway.
  • The Decision and Timing groups were nearly the same across all three personas (‘Why now?’, ‘What else have you looked at?’), so only about half the bank is persona-specific.
• Not yet observed: how a different model or tool handles the same prompt (a second-tool run is still to do).

Watch-outs

If you record or transcribe the call to use with this, get consent first (see Safe Use). Don’t use the cost-of-inaction probes to create pressure that isn’t real.

See the Safe Use page: AI drafts, you decide, you send.

Variations

• Manager version: run it for each persona on your team and compare to what reps actually ask on calls.
• Owner version: swap in ‘homeowner’ or ‘office manager’ and keep to 5 questions.
• Renewal version: replace ‘problem’ with ‘results so far’.

Sources

None. Original wording. (Add any public framework or post used, with link and checked-on date.)

Retest

Re-check by: 2026-12-13 (60 days after the 2026-10-14 runs), or sooner when run in a second tool.

Test log

Changelog

• 2026-10-14: Draft written. Not run yet.
• 2026-10-14: Ran 3 times in one AI assistant on fake data. Filled in sample input and output, how it went, and the test log. Front matter set to runs_logged 3, last_tested 2026-10-14, retest_by 2026-12-13. The 2-tool bar for the word “tested” is not met, so the play is not labeled tested. The prompt text was not changed after the runs.
• 2026-10-14: Prompt revised after round 1. Added: no-additions rule only. Re-ran 3 fresh runs (4-6) on new fake inputs in the same AI assistant. Updated the sample input and output to run 4, the “How it went” section, and the test log (original runs 1-3 kept, marked). Front matter set to runs_logged 6, last_tested 2026-10-14, retest_by 2026-12-13. The 2-tool bar for the word “tested” is not met, so the play is not labeled tested.
• 2026-10-15: Published with the evidence label “Tried in an AI tool” (6 runs, 1 tool, fictional inputs). Earlier wording that implied more evidence than one tool’s runs was replaced with the label; the runs, results, and limits above are unchanged.