Skip to content
AI Customer Research & ValidationLesson 2 of 13

2 · What AI can and can't do in customer research

AI is a powerful research assistant and a terrible research substitute. Knowing exactly which is which is the whole skill. Get it wrong and you'll mistake a rehearsal for the real thing.

What AI is genuinely good for (the "prep and digest" jobs):

  • Drafting better questions. Turn a leading question like "do you like my idea?" into open, neutral ones about what people actually do today (Lesson 4).
  • Designing a first-draft survey that you then send to real people.
  • Summarizing and clustering anonymized answers into themes after you've collected them — a real time-saver.
  • Mapping the landscape — who might have this problem, what alternatives exist, what to go investigate. The SBA calls this secondary research: fast, general, and a starting point, not an answer (SBA, n.d.).

What AI can NEVER do (the "be the customer" jobs):

  • Be a real customer. A model can imagine a buyer; it cannot be one. Its made-up answer is not market evidence.
  • Tell you if people will pay. Only a real person reaching for their wallet — or refusing to — answers that.
  • Replace the conversation. The most valuable thing in a young business is a real exchange with a real person. Don't automate it away.

The line that keeps you honest: AI handles the work before the conversation (prep) and after it (digest). The conversation itself — the only part that produces facts — is yours, with real humans. This is exactly the difference between secondary research (existing sources) and direct research (surveys, interviews, focus groups with your actual customers) that the SBA draws (SBA, n.d.). You need both, and AI only helps with one side of it.

Trust DNA: AI prepares you for reality and helps you digest it. It is never a stand-in for reality.

Check yourself. Name one research job you'd happily hand AI and one you must never let it do — and say why the second one can't be faked.

Sources

Sign in to track your progress