AI Expert

Guides/Working with ML teams

Working with ML and data science teams

AI product work runs through a different partner than typical engineering collaboration — model iteration is empirical, timelines are probabilistic, and “done” is a moving target. Shared vocabulary, early alignment on evaluation, and explicit ownership handoffs keep the partnership from breaking down.

Ownership across the build lifecycle

Roles vary by org size — a startup may compress several of these into one person, but the phases and the question of who signs off on each one don't go away.

Frame problemPM
Assess data & feasibilityDS + PM
Build & iterateML
Evaluate & sign offML + PM
Ship & integrateEng + MLOps
Monitor & retrainShared
Ownership shifts at every arrow — the handoffs, not the phases, are where AI projects usually stall.

Where the partnership breaks down

Friction pointWhy it happensFix
Success metric defined after the model is builtML optimizes a proxy metric with no product input up frontDefine the eval rubric and business metric together before training starts
Model updates treated like feature releasesModel iteration is empirical and unpredictable, not fixed scopePlan in confidence intervals and experiment budgets, not committed dates
No shared vocabulary for “good enough”ML thinks in offline metrics (F1, perplexity); PM thinks in user outcomesTranslate offline metrics into a shared, product-level acceptance threshold
Monitoring ownership unclear post-launchData science moves to the next model; nobody owns driftAssign an explicit monitoring owner as part of the launch checklist