Let Me Teach You AI

Part 1 of 5

Choose the idea

Start with a real problem, a specific person, and a simple paid outcome.

Your outcome for this part

A one-sentence business concept with a target customer, problem, offer, and first test.

Start with pain, not AI

Do not begin with ‘What can I build with AI?’ Begin with ‘What annoying, expensive, slow, confusing, or repetitive problem already exists?’ AI is useful when it makes the solution faster, cheaper, clearer, or easier to deliver. The business is the solved problem, not the model behind it.

Narrow the customer

A broad audience makes every decision harder. Pick one group you can describe plainly: local contractors who miss leads, creators who hate clipping videos, small shops that need product descriptions, or busy professionals who need research summarized. You can expand later after one group responds.

Define one outcome

The first offer should produce one visible result. Examples: ‘turn every form submission into an approved follow-up,’ ‘turn one long video into ten short clips,’ or ‘turn a messy idea into a launch-ready landing page.’ Avoid selling vague access to AI. Sell the result someone can judge.

Run a cheap reality test

Before building a full product, show the offer to real people. Ask whether they would use it, what they do now, what they dislike about the current process, and what would make the result worth paying for. Five useful conversations beat fifty imagined features.

Worked example

From vague idea to something testable.

Suppose the starting idea is: “I want to build an AI tool for contractors.” That is too broad to build intelligently.

What we observe

Small home-service contractors often lose leads because inquiries arrive while they are driving, on a job, or after hours. Following up consistently is a real operational problem.

Narrowed concept

“We help small home-service contractors respond to new website leads faster by drafting a ready-to-approve follow-up and logging what happened.”

Customer

Small home-service contractors

Problem

Slow or inconsistent lead follow-up

Test

Interview five contractors before building

Interactive worksheet

Score three ideas before choosing one.

Give each idea a 1–5 score. Do not use the total as proof of demand — use it to decide which idea deserves your first real-world test.

IdeaUrgencyFrequencyWill payEasy to reachAI leverageTotal
5/25
5/25
5/25

1–12: weak candidate; learn more before building.

13–18: worth validating with real people.

19–25: strong hypothesis — still not proof until customers confirm it.

Customer validation

Ask these questions. Do not pitch yet.

Your job is to learn how the problem works in real life. Avoid leading questions like “Would you use my AI app?” People are polite. Past behavior is better evidence than compliments.

  1. 1Walk me through the last time this problem happened.
  2. 2What do you do today to solve or work around it?
  3. 3What is the most frustrating or expensive part of the current process?
  4. 4How often does this happen in a normal month?
  5. 5Have you ever paid for a tool, service, or person to help with it?
  6. 6If a solution reliably produced the result we described, what would make it worth trying?

Your assignment

Leave the lesson with evidence, not excitement.

01 · Score

Score three candidate ideas above and pick one to investigate.

02 · Interview

Use the six questions with five people who actually fit the customer profile.

03 · Decide

Write: “We help [person] get [result] without [pain]” and list what remains unproven.

Do this before moving on

  • Write down three problems you personally understand.
  • Choose one narrow customer group for each problem.
  • Turn the strongest option into: We help [person] get [result] without [pain].
  • Ask five real people how they solve that problem today.
  • Keep the idea only if the problem is real enough to deserve another week of work.
Use this AI prompt

Act as a practical business coach. I will give you three problems I understand. Help me score each one for urgency, frequency, willingness to pay, ease of reaching customers, and how much AI can improve delivery. Then recommend one tiny offer I can test this week. Do not invent demand; tell me what I still need to validate.

Open AI Helper
Completion checkpoint

Do not move to branding until these are true.

I can name one specific customer — not ‘everyone.’
I have evidence the problem actually happens, not just a guess.
I can describe the result in one sentence without mentioning AI.
I have spoken to at least five relevant people or have five equivalent pieces of real customer evidence.
I know what I still need to validate before spending serious time or money.

If you can check all five, you have enough clarity to give the idea a name, a promise, and a home on the internet.

Continue to Part 2: Build the brand and site