Product Management for AI

Preface

The increasing push in many industries to develop products that incorporate AI puts the intersection of AI and product management into sharp focus. It’s easy to say, “So what’s changed? Product management is product management.” But that’s wrong. AI brings many challenges to traditional product management. Nondeterministic outcomes, uncertainty (schedule, accuracy, relevance), opacity (models can be difficult to understand and explain), fairness (are results biased against a particular group?), and other factors make AI a difficult sell to decision makers and upper management. Product management for AI is different. The differences aren’t unsurmountable, but they’re real, and they cause many projects to fail before they’re deployed.

In this report, we lay out, in detail and with helpful supporting examples, what makes AI different, how to address those differences, and how product managers can better align their efforts in pragmatic support of business goals. If you’re aware of the difficulties you will be facing, you’ll be able to anticipate and solve the problems you face as you develop an AI project and shepherd it into production.

Acknowledgments

We would like to thank the many people who have contributed their expertise to the early drafts of the articles in this series, including: Emmanuel Ameisen, Chris Albon, Chris Butler, Ashton Chevalier, Hilary Mason, Monica Rogati, Danielle Thorp, and Matthew Wise.