Stop 13 · PM Specializations & Career TracksTopics · 6/8

Data & AI PM

In short

Recommendations, fraud alerts and AI assistants all depend on data that quietly flows through pipelines. Data and AI PMs own that plumbing and what comes out of the tap.

Owns data infrastructure, ML/AI-powered features, or internal data products — requires comfort with data pipelines, model evaluation, and the AI fundamentals from Phase 12.

The purple blob turns a big valve on a filter tank next to a glass of clean water.

Think of it like this

The water treatment plant

Clean drinking water starts as muddy river water. It's filtered, tested and treated before it reaches your tap. If any step fails, people get sick, and nobody blames the tap. Data products work the same: bad data in means bad decisions out.

Watch it happen

From raw data to an AI feature

Collect

Clean

Model

Evaluate

Feature

1/5A "you might also like" feature needs purchase and browsing events. The PM checks what's tracked and what's missing.

The big ideas

What you need to know

Data quality is product quality

Missing or wrong data produces wrong outputs. The PM cares where data comes from.

Define success clearly

Pick model metrics and real user metrics, like click rate and repeat purchase, not just accuracy.

Watch for drift

User behaviour changes, so a model that worked last year can quietly get worse.

Privacy and fairness

Data PMs handle consent, access rules and bias checks as everyday work.

Save yourself the pain

Common mistakes

Oops

Judging a model only by offline accuracy.

Do this instead

Confirm it improves real user behaviour in a live test before full launch.

Oops

Ignoring where training data came from.

Do this instead

Check consent, coverage of different user groups, and freshness.

Your mission

Trace a data feature

~25 min

Quick check

A recommendation model performed well at launch but got worse over a year. What's a likely cause?

Go deeper

1Revisit Phase 12 (AI & Modern PM) as core preparation for this track.
2Read a Data/AI PM job description and note the statistics and ML-literacy expectations.
3Consider this track if Phase 4 (Data & Metrics) was your favorite part of the roadmap.