Define one buyer and one painful moment
Write the idea as a sentence: “When [buyer] needs to [job], they currently use [workaround], which causes [cost or delay].” If you cannot name the workaround, you are still exploring the problem. “Small businesses need AI” is too wide. “An export-services founder loses client approvals across email and WhatsApp” is narrow enough to investigate.
Choose one buyer segment for the test. A finance manager at a distributor, a solo agency owner, and a plant operations lead have different budgets and buying paths. Mixing them makes ten interviews look like evidence when they are really ten different markets.
Day one: collect recent examples, not opinions
Ask five to ten potential buyers about the last time the workflow happened. Record the trigger, the people involved, the tools used, the time spent, the consequence of failure, and who decides whether to pay. Avoid leading with a product pitch.
- “Walk me through the last time this happened.”
- “What did you do before and after that step?”
- “What was the cost of getting it wrong or doing it late?”
- “What have you already tried?”
- “Who would approve a paid fix?”
A complaint is a lead. A recent incident with a workaround and a named owner is stronger evidence. Keep exact notes, but ask before publishing anything a buyer told you privately.
Day two: sell the outcome before building software
Offer a small manual version of the result: reconcile ten exceptions, prepare one proof packet, or triage one shared inbox. State the scope, turnaround, and price clearly. A paid pilot is the cleanest signal, but a scheduled follow-up with access to real examples is still useful. Praise and waitlist clicks are weaker signals.
Show a one-page mockup only after you understand the workflow. Ask the buyer to correct it: which field is missing, where would it enter their process, and who would actually use it? Do not count “I would use this” as a commitment unless the buyer agrees to a next action with a date.
Write a decision memo before you build
At the end of the second day, record four numbers: qualified conversations, buyers with a recent incident, buyers who showed their current workaround, and buyers who agreed to a paid or scheduled pilot. These are your test results, not market-wide conversion rates.
Continue if the pain repeats within one buyer segment and somebody will give you access, time, or money to solve it. Narrow the buyer if the same problem appears only in one subsegment. Stop if the alleged problem is hypothetical or the current workaround is good enough. For the next step, see our first-customer plan or open the live opportunity board for a specific idea to test.
