Usable with a quick edit11 of 12 checks passed
The guide is thorough and actionable; a PM could use it with minor edits like marking priority questions for time management.
Output
Autochase discovery call guide
Interviewers: Mei and Tomas · 8 calls, 30 minutes each Decision: Should Paydeck build automatic, escalating overdue-invoice reminders next quarter, priced at $10/month?
1. What we need to learn
Test Sam’s hypothesis—not whether people like the idea:
- Problem: How often do late payments create meaningful cash-flow problems or chasing work?
- Current behavior: How do users chase invoices today, and why do some not use Paydeck’s reminder button?
- Fit and risk: Which reminders could safely be automated? Where might escalation damage client relationships?
- Value: Is solving this worth an additional $10/month, given existing workarounds and alternatives?
Our data establishes lateness, not demand: 34% of invoices are late, but only 12% of accounts used reminders recently. Support requests and frequent reminder users are useful leads, not representative evidence.
2. Who to recruit
Don’t book the first eight respondents from the top reminder users. That would disproportionately select enthusiastic, high-pain users.
Recruit invoicing-active accounts from these groups:
- 3: Recurring overdue invoices; frequent Paydeck reminder use.
- 3: Recurring overdue invoices; little or no Paydeck reminder use.
- 2: Few overdue invoices, or late payments that appear manageable.
Across the eight, include freelancers and small agencies, plus both Pro and free accounts. Aim for roughly four of each plan; this is purposeful sampling, not a representative survey. Include at most two customers who requested automatic reminders.
Speak to the person who actually manages invoices and chasing; establish whether they also approve software spending. Recruit within each group rather than simply taking the fastest replies. Use a neutral invitation: “Help us understand how you manage invoices and payment follow-up.” Offer the same incentive regardless of feedback.
3. Call questions and timings
0–2 minutes: Welcome and context
“We’re learning how people handle overdue invoices. We’re testing an idea, not selling anything; honest criticism is helpful. There are no right answers.”
Ask permission to record. Confirm their role, business size, and who handles invoices and software purchases.
2–12 minutes: Reconstruct a real example
“Tell me about the most recent invoice that went past its due date.”
Follow the sequence: - When was it due, and when did you notice? - What did you do next? Then what happened? - Who followed up, using what tools, messages, and timing? - How did you decide whether to remind the client again? - Was it paid? What do you think caused the delay?
If comfortable, ask them to show a redacted invoice or follow-up message. Don’t collect client-identifying information.
For low-lateness participants: “Tell me about the last late invoice—or how you usually prevent late payments.”
12–18 minutes: Frequency, consequences, alternatives
- “Over the last three months, how many invoices needed follow-up?”
- “Roughly how much time did you spend chasing them?”
- “What, if anything, did the delay affect?” Probe for concrete consequences, not just frustration.
- “What have you tried to make this easier? What did it cost?”
- “Have you used Paydeck’s reminder button? Walk me through why or why not.”
- “When do you deliberately avoid sending another reminder?”
18–26 minutes: Test the concept, then price
Only now introduce it:
“We’re considering Autochase: scheduled reminders for overdue invoices, with reminders becoming firmer over time. It isn’t built, and details aren’t decided.”
Ask: - “Thinking about that invoice, where would this fit—or not fit?” - “Which steps would you automate, and which would you keep manual?” - “What would you need to control or check before enabling it?” - “What could go wrong? Tell me about a client situation where it would.” - “What would you use instead?”
Then disclose: “We’re considering $10/month, additional to your current plan.”
“What would make that worth paying—or not? Which current effort or expense would it replace? Who would approve it?”
Don’t offer a $5/$10/$15 menu or treat “yes, useful” as purchase evidence.
26–30 minutes: Commitment and close
“If we offered a paid pilot at $10/month, what would you need before deciding?”
Ask whether they’d join a follow-up evaluation or discuss a pilot with the budget owner. Record the specific next step; don’t imply availability or collect payment.
“What important part of chasing payments haven’t we covered?”
4. Notes for first-time interviewers
Alternate moderator and note-taker roles; do one practice call together. Ask one question at a time, allow silence, and follow concrete examples. Avoid praise, pitching, defending, and questions that assume pain. Separate observed behavior, exact quotes, estimates, and your interpretations.
Debrief immediately: frequency, cost, workaround, automation boundaries, price reaction, commitment, and contradictory evidence.
5. Decision rules
Agree these directional gates before interviewing; eight calls cannot estimate market demand.
Go toward a scoped build if at least five describe recurring, consequential pain—including two non-reminder users—and at least three connect $10 to concrete value and take a specific pilot-evaluation step. There must also be a credible way to address relationship and control risks. Validate paid adoption before committing substantial engineering.
Stop or reshape if lateness rarely causes meaningful cost, chasing is easy, automation cannot address the underlying causes, or users prefer manual control and won’t pay.
Mixed findings mean targeted follow-up—not “eight customers validated it.”
Check by check
Got wrong · 1
- Fits the callTimings add up to 30 minutes but no must-ask questions are marked and no guidance on what to cut if time runs short.
Got right · 11
- Uses the supplied evidence correctlyAll factual claims about the current situation are directly from the supplied context.
- Addresses the actual decisionProvides clear go/stop conditions tied to specific call outcomes, framed for the team.
- Respects explicit constraintsIncludes all required sections, stays under 900 words, and is written for Mei and Tomas.
- Identifies material uncertaintyNames specific unknowns (cost of lateness, ease of chasing, automation fit, willingness to pay) and says mixed findings mean targeted follow-up.
- Avoids unsupported claimsInterpretations are clearly separated from facts; no unsupported causal claims are presented as established.
- Produces the required deliverableComplete call guide with learning goals, timed questions, coaching, recruiting plan, and decision rules, usable as-is.
- Talks to the right peopleExplicitly avoids only heavy reminder users, gives a balanced split with rationale and includes non-reminder users.
- Coaches first-time interviewersProvides specific, actionable instructions: alternate roles, practice call, avoid pitching, allow silence, separate observations from interpretations.
- Designed to falsifyEvery learning goal has disconfirming questions, e.g., 'What did you do next?', 'When do you avoid sending another reminder?'
- Asks what people did, not what they thinkCore questions ask for the most recent overdue invoice and specific actions taken, not opinions or predictions.
- Doesn't lead the witnessQuestions are neutral and open; Autochase is introduced only after 18 minutes of problem exploration.
Grades and run details
Decision model 96 · LLM judge 12 of 13 checks
Decision model checks
- passUses the supplied evidence correctly69%
- passAddresses the actual decision51%
- passRespects explicit constraints54%
- passIdentifies material uncertainty100%
- passAvoids unsupported claims76%
- passProduces the required deliverable93%
- passTalks to the right people100%
- passCoaches first-time interviewers87%
- passDesigned to falsify98%
- passAsks what people did, not what they think82%
- passDoesn't lead the witness80%
- partialFits the call56%
Run
- Run
- #1
- API response time
- 44 s
- Submitted
- 30 Sept 2026