Skip to main content

How to discover customer problems

The biggest risk is not building the product badly — it is building the wrong thing well. Discovery is how you avoid that.

Talk to operators, not to the market​

Interview the people who do the work every day: the salon owner between clients, the site foreman, the shop owner cashing up at night. You are looking for recurring, expensive, annoying moments — not opinions about AI.

Good questions:

  • "Walk me through the last time this went wrong."
  • "Where does information arrive, and what do you do with it?"
  • "What do you re-type or copy between systems?"
  • "What did you have to chase, and who complained?"
  • "What did that cost you — time, money, a lost customer?"

Avoid leading questions ("Would you use an AI assistant that…?"). People are polite; politeness is not demand.

Look for the four signatures​

A problem is worth investigating when it shows at least one of these:

  1. Repetition — the same manual step, many times a day.
  2. Delay — work waits for a human who is busy elsewhere.
  3. Leakage — enquiries, revenue or evidence that quietly disappears.
  4. Re-keying — the same data typed into two or three places.

These are exactly the moments where controlled automation pays for itself.

Write the problem down before the solution​

For each candidate, capture one paragraph:

Today, <role> does <task> using <tools>. It breaks when <trigger>, which causes <consequence>. It happens about <frequency> and costs roughly <effort or money>.

If you cannot fill those blanks from a real conversation, you do not yet have a problem — you have a guess.

Next​

Once you have three or four well-described problems, move to validating an idea.