Skip to main content
Next Step PM
Back to the library

Define · Add-on

Check your data is ready for AI

AI is only as good as the data behind it. Check you have the right data, of good enough quality, and that you are allowed to use it.

About 2-5 daysEvergreenLast reviewed 5 November 2026

Who this applies to

Skills it builds
Data and measurementTechnical and AI fluency
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. List the data the AI will need, for training, examples or retrieval.
  2. Find where it lives and who owns it.
  3. Check quality: is it complete, accurate, up to date and representative of all your users?
  4. Check you have the legal right and consent to use it this way.
  5. Plan how you will fill any gaps.

Intermediate (1 to 4 years)

  1. Inventory data sources and owners.
  2. Assess quality and bias.
  3. Confirm legal basis and plan gaps.

Experienced (5+ years)

  1. Design a data strategy that improves as the product is used.
  2. Plan data governance and retention for AI use.
  3. Assess the risk of sensitive data leaking through outputs.

Who can do this

Owner: Product manager

Can be delegated to

  • Data Scientist / ML Engineer or Data Analyst: Assess data quality and fitness
  • Policy / Legal / Compliance: Confirm legal basis and consent

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

Tool examples

Looker
Profile the data to see gaps and distributions.
Power BI
Chart data completeness by source and user group.

Definition of done

  • Data sources and owners listed
  • Quality and bias assessed
  • Legal basis confirmed

Common pitfalls

  • Assuming the data is clean
  • Data that does not represent all users
  • Using data without the right to

If you lead the team

What good looks like: Teams check data quality, coverage and permissions before committing to an AI feature.

Ask your PM

  • Do we have the right data, of good enough quality, and permission to use it?
  • What happens if the data isn't ready?

Warning signs

  • Data problems are discovered after the build starts.
  • No one has checked the legal basis for using the data.

Do these first

Sources