Playbook: Entrepreneurs and Founders

Founders have always been short of three things: time, money and honest feedback. AI helps with all three — it can play a sceptical investor at midnight, draft a landing page in minutes, and suggest twenty experiments before breakfast.

It can also help you avoid reality for months. Infinite analysis feels like progress. It isn't. Customers are the only source of truth.

---

1. The pre-mortem: find the ways it fails

Optimism is necessary to start. Blind spots are what kill companies. Ask AI to attack the idea before the market does.

Act as a sceptical early-stage investor who has seen 1,000 pitches
in India.

My idea: [2-3 sentences: who, problem, solution, how it makes money]

1. Give 10 specific reasons this fails within 12 months — market,
   customer behaviour, distribution, unit economics, competition,
   regulation, execution
2. Rank them by likelihood × impact
3. For the top 3: what's the cheapest test I could run this month
   to find out if it's true?

No encouragement. Be specific to my idea, not generic startup
advice.

Then ask, for each top risk, "How would I fix or mitigate this?" — and turn the answers into a failure-prevention checklist you revisit monthly.

2. Validate with people, not with AI

AI can't tell you whether customers will pay. It can help you find out faster:

1. Interview script: "Write 8 open questions to learn how [customer] handles [problem] today, what it costs them, and what they've tried. No leading questions, no pitching."
2. Talk to 15–20 real potential customers. Record (with permission) and transcribe.
3. Synthesise:

Here are transcripts of [N] customer interviews.
- What problems come up repeatedly, and how many people raised each?
- Quote exact words people used about cost or frustration
- Where did people contradict my assumptions?
- What did people say they'd pay for, vs what they currently pay for?
Only use what's in the transcripts.

1. Test demand cheaply: a landing page with a pre-order or waitlist, a manual "concierge" version of the service, or a small paid pilot.

3. Define the MVP

An MVP is not the smallest product you can build. It's the smallest thing that tests your core value proposition with real users and produces learning.

Core user: [specific]
Their single most painful problem: [from interviews]
My proposed solution: [description]

1. What is the ONE outcome the user must get for this to be valuable?
2. List every feature I've imagined, then sort into: essential to
   that outcome / nice to have / not now
3. Could the first version be manual, no-code, or a spreadsheet?
4. What would I measure in the first 4 weeks to know it's working?

Many good MVPs today are no-code: a form, an automation, a shared spreadsheet and a WhatsApp group — exactly what you built in this course.

4. Growth experiments, one a week

Current state: [users/customers, main acquisition channels, what
has worked, budget]

Suggest 10 growth experiments that are low effort and could have
high impact. For each: hypothesis, how to run it in under a week,
cost, and the single metric that decides success.

Track them in a simple sheet: Experiment | Hypothesis | Effort | Expected impact | Result | Decision. Run one per week for four weeks. Keep what works, kill what doesn't, write down why.

5. The founder's AI stack

Job	How AI helps
Research	Market scans, competitor summaries (verify sources)
Customer discovery	Interview scripts, transcript synthesis
Build	No-code apps and automations, AI coding assistants for prototypes
Marketing	Landing page drafts, ad variations, content pipeline
Operations	Support replies, SOPs, invoicing and follow-up automations
Finance	Model scenarios with code, investor update drafts
Legal/compliance	First-pass understanding — then a lawyer and CA for anything that matters

6. Time-box the research

AI will research forever if you let it. Set limits:

• 2–4 hours of AI-assisted research on a question, then decide.
• A written decision with what would change your mind.
• Ship something to real users every two weeks.

You'll learn more from ten weeks of real users than from ten months of analysis.

7. Gut feeling still matters

AI sees patterns in what's been written. You see context it can't: a relationship with a distributor, a regulatory change you heard about last week, the look on a customer's face. When AI analysis and strong first-hand evidence disagree, investigate — don't automatically defer to the model.

---

⚠️ Common mistakes

• AI over-analysis instead of talking to customers.
• Treating AI's opinion as market validation.
• Building a full product before testing the core value.
• Ignoring your own context and gut in favour of generic analysis.
• Pasting confidential investor or customer data into unapproved tools.

---

What's next: you've worked through the industry playbooks. Pick the one closest to your work, adapt its prompts into your prompt library, and bring your first result to office hours.