The Agentic Shift

Google Labs product director Jaclyn Konzelmann argues that cheap building makes worth the main question, and that products now serve people and their AI agents.

Jaclyn Konzelmann, a product director at Google Labs, wrote up a talk she gave at The AI Conference in San Francisco. She sent her own personal agent to use the fitness app Strava while logged in as herself. She argues that every product now has two users, and that setting goals and judging quality stay with people.

• She says the interesting question stopped being whether something can be built and became whether it should be.
• She asks builders whether an AI agent can do everything a person can do in their product.
• She advises keeping the lessons from a prototype rather than the prototype itself.
• She recommends listing the guardrails built only to make earlier AI models behave, and removing those that newer models no longer need.

Product teams may need to test their product as an AI agent would use it, not only as a person would. This is an inference and is not the author's claim. Only the free opening of the post was read; the worksheet is for paying subscribers.

Derived by Working Surface from the article; more in the Playbook. Source line: An agent will happily build you something plausible.

9 October 2026: when AI agents take on routine work, teams write down in advance what a person still decides, put those rules where the agent reads them, and keep the people who approve able to understand the work.