A common AI project begins like this:

Client: “We want AI in our application.”
Developer: “Which model should we use?”

The developer’s question comes too early.

The first questions should be:

What business decision needs improvement?
What information is available before that decision?
What should the system predict or generate?
What action will be taken from the output?
How will success be measured?

Consider this vague requirement:

“Use AI to improve policy renewals.”

It does not specify:

  • What should be predicted
  • When the prediction should occur
  • Which policy is being evaluated
  • Who will use the result
  • What action will follow
  • How improvement will be measured

A technically accurate model can still create no business value if nobody acts on its predictions.

Today’s core lesson:

Start with the decision and measurable outcome—not with the algorithm.

PreviousNext