The three questions you will always be asked
Eight minutes and one page. The founder has read half of it and has already decided they disagree.
How do you know, what does it cost, what if you are wrong
Pressure-test your recommendation
- Recommendation: The call you want
- Evidence: How do you know?
- Economics: What does it cost?
- Reversal: What if it fails?
Almost every pushback is a version of those three. Have them answered before you walk in.
“I do not know” is a complete answer, followed by “here is how I would find out, and by when”. Bluffing is the only unrecoverable mistake. Once you are caught inventing one number, every number you have ever given becomes retroactively suspect.
Review questions are a chance to locate the weak parts of your recommendation before committing resources. A strong answer can be “we do not know yet” when it is followed by a useful next step.
Use your chosen course project throughout. The additional examples below are fictional practice cases; transfer the method to your own evidence.
Prepare the evidence trail
For each major claim, keep a source, an assumption label, and a limitation. Reviewers may ask how often a problem occurs, why this segment matters, or what happens if the estimate is wrong. You do not need an answer to every question; you need to avoid confusing a guess with a fact.
Answer the actual objection
Listen for whether the concern is about evidence, feasibility, scope, or priority. Restate it briefly before responding. A question about operational cost is not answered by repeating user enthusiasm. If the question changes the decision, update the proposal rather than defending the original wording at all costs.
Turn uncertainty into a plan
Name the smallest check, an owner, and a date. Explain which possible result would change your recommendation. This prevents “we will investigate” from becoming an indefinite delay. Record open questions after the meeting and distinguish blockers from details that can be resolved during implementation.
Make a hard question useful
- Concern: What is at risk?
- Evidence: What do we know?
- Unknown: What is missing?
- Check: Owner and date
How Superhuman handled it
Superhuman
Rahul Vohra needed to prove to his own team, and to investors, that Superhuman had product market fit. A claim usually made on vibes.
He ran Sean Ellis's question, how would you feel if you could no longer use this product, and made the 40 percent very disappointed threshold the entire argument. Then he segmented the somewhat disappointed group and built the roadmap from what would move them up.
The most defensible arguments compress to one number and one sentence. If you need five slides to make the point, the point is not sharp yet.
Doing it with AI, and where it breaks
Have a model role play the specific person you are presenting to, their priorities and known objections, and run the eight minutes three times.
It will invent a much nicer executive than you will face. Tell it to interrupt within the first thirty seconds and to reject any answer that does not include a number.
Your AI workbench
Start with your own notes or clearly labelled practice data. Remove private details before sharing. Replace the placeholders, run the prompt in your chosen AI tool, and keep the output beside its source.
Challenge this proposal [paste] using only supplied facts. Ask six questions across evidence, feasibility, economics, scope, and measurement. For each, identify what a supported answer would require. After I answer, flag overclaims and missing evidence. Do not invent hostile stakeholder personalities or new research findings.
Before you use the output
- Answers address the stated concern
- Unknowns remain explicitly unknown
- Follow-up checks could change the decision
Stuck? Try this next
Practise a thirty-second answer with AI, then ask it to remove every unsupported assertion. Keep the source table open while rehearsing so fluency does not replace evidence.
Keep a brief AI log: input used, useful output, what you checked, and what you rejected. The decision remains yours.
Build it
Written answers to the three hardest questions, and a memo revised where they exposed a hole.
Checkpoint
If you did the build, these take two minutes. If you cannot answer one of them, that is the part to go back to.
Moving on marks this lesson complete. Finish the build first, it is the part that counts.