AI Product Management: Start with the problem

Free preview | Frame the problem

What makes an AI feature different

Software that behaves differently each time

Most software does the same thing every time you give it the same input. A calculator, a booking form and a tax rule all work that way. You can write down the rule, test it once and trust it.

Many AI features do not work like that. A language model can answer the same question in two different ways. A recommendation can be right for one person and useless for the next. That is not a bug to remove. It is how the technology behaves, and it changes how you plan the product.

Three questions to ask first

  • Does the answer need to be exactly the same every time? If yes, a plain rule is usually better than a model.
  • Is a good answer easy to describe but hard to program? Summarising a long email is like that. This is where AI helps most.
  • What happens when it is wrong? A wrong movie suggestion costs nothing. A wrong refund amount costs money and trust.

What this means for you

Before you think about models, prompts or tools, you need a clear problem and a clear way to tell whether you solved it. The next four lessons show how.

Like what you see?

The full course has more lessons and a deliverable for every module. Enrolment opens soon.

See the full course