How Google Labs Decides What AI Products to Build

Google Labs product director Jaclyn Konzelmann talks with Peter Yang about product judgment and AI-assisted work.

Start with the decision

Konzelmann favors a short strategy document, a presentation and a prototype over beginning with a long product requirements document.
She looks at both emerging model capabilities and changes in user behavior, then asks whether a proposed product should exist.
Her team works with small businesses and trusted testers; she distinguishes understanding the underlying problem from simply implementing every requested feature.
A writer should be able to defend the ideas and words in an AI-assisted document. Her own one-pagers often run about a page and a half.

Labs teams decide case by case whether to grow an experiment, incorporate its lessons elsewhere or wind it down.
Her principle is to retain the lessons when a prototype ends; teams share findings and sometimes reuse infrastructure.
As agents perform more tasks, she asks which interactions people still value, such as creativity, learning and entertainment.
She describes using an agent to repair a GPS track while still enjoying the social feed around a hike.

Konzelmann describes making a scary-story podcast with her daughter, who chooses characters, voices and revisions.
She also values outdoor time, boredom and learning to handle friction. These are her parenting choices, not evidence of developmental outcomes.

“Keep the lessons, not the prototype.”