1. Turn the Idea Into Something You Can Test
A vague idea like 'I want an AI app for freelancers' becomes a testable hypothesis: target user, problem hypothesis, current alternative, value hypothesis, risks, and a test. AI clarifies the hypothesis, but AI is not evidence the problem matters. The founder still has to run the test and talk to real users.
AI can help
- ✓Turn idea notes into a structured hypothesis
- ✓Surface risks and assumptions to test
- ✓Draft a simple test plan
Human / specialist keeps
- ●Decide which risk matters most
- ●Talk to real users
- ●Own the go-or-no-go decision
Note
AI clarifies the hypothesis. AI is not evidence the problem matters. Customers validate the business.
2. Use AI to Prepare Customer Interviews—Not Replace Them
Durable YC guidance emphasizes talking to users, launching, and iterating. Current Stripe startup guidance similarly emphasizes talking to target customers, building the simplest useful version, testing demand, and observing real behavior. AI can draft neutral questions, identify leading questions, summarize transcripts, cluster pain, compare interviews, identify contradictions, and prepare follow-ups. Do not use synthetic personas instead of real customers.
AI can help
- ✓Draft neutral, past-behavior interview questions
- ✓Critique the question set for leading questions
- ✓Summarize and cluster interview transcripts
Human / specialist keeps
- ●Run the real interview
- ●Hear objections and customer language directly
- ●Do not substitute synthetic personas for real users
Warning
A synthetic customer cannot tell you whether a real customer will pay. AI prepares the interview; it does not replace it.
3. Research Faster Without Turning Search Results Into a Business Thesis
AI can find competitors, summarize pricing, compare positioning, extract review themes, map categories, organize market reports, and surface regulatory questions. Use a source hierarchy: customer and first-party signal, competitor primary sources, government or industry primary data, high-quality secondary, community or reviews, and AI summary only as a discovery aid. Verify market size, competitor pricing, funding, and product capability before relying on them.
AI can help
- ✓Find competitors and summarize positioning
- ✓Extract review themes and map categories
- ✓Surface regulatory questions to investigate
Human / specialist keeps
- ●Verify market size, pricing, and capability claims
- ●Do not build a thesis from a single AI summary
- ●Own the research-to-strategy decision
4. Build the Cheapest Credible Demand Test
A test can be a landing page, a manual service, a prototype, a waitlist, a preorder, or a demo. The landing page can include the problem, value proposition, audience, proof or demo, and a CTA like join waitlist, request demo, book call, start trial, preorder, or buy. AI can draft, design, create FAQs, and generate variants. Specialist builders may be stronger depending on whether the test is a website or a software product.
AI can help
- ✓Draft landing-page copy, FAQs, and variants
- ✓Generate page structure from offer facts
- ✓Suggest CTA and proof options
Human / specialist keeps
- ●Verify facts, proof, and the conversion path
- ●Choose the cheapest test for the biggest risk
- ●Own the offer and the promise
5. Build the Smallest Product That Tests the Biggest Risk
Before building, ask what the biggest uncertainty is: will they sign up, will they pay, can the technical system work, will they return, can we acquire them, will they trust it. Build the smallest test for that risk — a manual concierge service, a spreadsheet, a form, an automation, a landing page, a prototype, a vibe-coded app, or a simple SaaS MVP. Do not build advanced admin, analytics, mobile apps, affiliate systems, and complex permissions before the core risk requires them.
AI can help
- ✓Build a prototype or MVP faster
- ✓Draft the simplest version of the core workflow
- ✓Generate supporting assets like FAQs and onboarding
Human / specialist keeps
- ●Identify the biggest risk and test that first
- ●Do not overbuild before the core risk is tested
- ●Own scope and prioritization
Tip
Do not build advanced admin, analytics, mobile apps, affiliate systems, and complex permissions before the core risk requires them.
6. Use AI to Prepare Sales—Do the First Sales Conversations Yourself
AI can do account research, meeting prep, CRM summary, outreach draft, follow-up, call summary, and objection clustering. The founder should still talk to users, hear objections, learn customer language, understand who decides, and learn why they buy. Do not outsource all early sales to an autonomous agent. Your first customers are a research system.
AI can help
- ✓Research accounts and prepare meeting packs
- ✓Draft outreach and follow-up
- ✓Cluster objections from early calls
Human / specialist keeps
- ●Run the first sales conversations personally
- ●Hear objections and customer language
- ●Own pricing and commitments
7. Use AI to Model Pricing—Test It With Real Buyers
AI can structure packages, compare competitor plans, model scenarios, and identify assumptions. It cannot know what customers will actually pay. Test with real offers, proposals, checkout, and conversations. Do not use 'AI says $49 is optimal' as evidence.
AI can help
- ✓Structure packages and compare competitor plans
- ✓Model pricing scenarios and surface assumptions
- ✓Draft proposal language
Human / specialist keeps
- ●Test pricing with real buyers
- ●Do not treat an AI model as market evidence
- ●Own the pricing decision
8. Turn One Launch Idea Into a Complete Launch System
Inputs include the customer problem, positioning, product facts, offer, proof, CTA, and channels. AI can draft the landing page, email, social, demo script, FAQ, sales deck, and video outline. The human verifies claims, pricing, availability, screenshots, customer stories, technical capability, and compliance. Do not publish unverified claims just because the launch feels urgent.
AI can help
- ✓Draft launch assets across channels
- ✓Generate a demo script and FAQ
- ✓Repurpose one launch idea into channel-native content
Human / specialist keeps
- ●Verify claims, pricing, and availability
- ●Approve customer stories and screenshots
- ●Own the launch decision
9. Use AI to Learn From Feedback
Sources include support, sales calls, interviews, reviews, cancellations, product notes, and surveys. AI can cluster themes, summarize pain, identify contradictions, group requests, and draft hypotheses. The founder and product team decide severity, customer value, strategic fit, effort, frequency, and retention impact. Feature-request frequency alone is not product strategy.
AI can help
- ✓Cluster feedback themes across sources
- ✓Summarize pain and identify contradictions
- ✓Draft hypotheses for the product team
Human / specialist keeps
- ●Decide severity, value, and strategic fit
- ●Do not let request frequency alone drive strategy
- ●Own the roadmap
10. Automate the Business After the Process Becomes Real
Good automation includes lead capture, CRM, booking, invoices, payment reminders, onboarding, support triage, review requests, email nurture, report summaries, and content distribution. Bad automation is automating a process you have not manually understood. Automate the work you understand — not the uncertainty you still need to learn from.
AI can help
- ✓Automate proven repetitive operations
- ✓Draft onboarding and nurture sequences
- ✓Summarize reports and support themes
Human / specialist keeps
- ●Understand the process manually before automating
- ●Own exceptions and judgment
- ●Do not automate uncertainty
Note
Automate the work you understand — not the uncertainty you still need to learn from.