Skip to main content
Next Step PM
Back to the library

Define · Add-on

Decide whether AI is the right solution

AI is powerful but costly, unpredictable and sometimes unnecessary. Check that AI is the best way to solve the user's problem before committing to it.

About 1-3 daysFast-movingLast reviewed 5 November 2026

Who this applies to

Skills it builds
Technical and AI fluencyDiscovery and research
PM types
AI
Product types
Data or AI product
Starting points
All
Ways of working
All
Experience
All
Size of organisation
All

Steps, by experience

Beginner (under 1 year)

  1. Write the user problem without mentioning AI.
  2. List simpler options: rules, better design, search or a manual process.
  3. Ask where AI would add real value: handling variety, scale or language that rules cannot.
  4. Ask what happens when the AI is wrong, and how bad that would be.
  5. Estimate the running cost per use.
  6. Decide and record why AI is, or is not, the right choice.

Intermediate (1 to 4 years)

  1. Compare AI with non-AI options on value, cost and risk.
  2. Assess the cost of errors and the need for human review.
  3. Record the decision and the evidence.

Experienced (5+ years)

  1. Check the AI approach creates lasting advantage, not just novelty.
  2. Assess build, buy or partner options.
  3. Agree quality and cost thresholds before committing.

Who can do this

Owner: Product manager

Can be delegated to

  • Data Scientist / ML Engineer or Tech Lead: Assess feasibility and likely quality
  • Policy / Legal / Compliance: Flag legal and ethical constraints

Reviewed by: Data Scientist / ML Engineer, Policy / Legal / Compliance

Tool examples

Claude
Prototype the AI part quickly with a handful of real examples to see if quality is close to good enough.
Google Docs
Write a one-page options comparison.

Definition of done

  • Problem stated without AI
  • Non-AI options compared
  • Cost of errors and running cost assessed
  • Decision recorded

Common pitfalls

  • Starting with 'we need AI'
  • Ignoring running costs
  • Underestimating the impact of wrong answers

If you lead the team

What good looks like: AI is used where it's clearly the best way to solve the problem, not because it's fashionable.

Ask your PM

  • Why is AI better here than a simpler approach?
  • What's the cost of it being wrong?

Warning signs

  • The starting point is 'we need an AI feature'.
  • No simpler alternative was considered.

Do these first

Sources