How AI changed the product manager and product designer dynamic in 2026

Gabriela Naumnik of the Product Management Society argues that AI brings product managers and designers together earlier, but each must keep its own sign-off, on scope and on usability.

Gabriela Naumnik is a staff product manager on AI platforms for large companies and the founder of the Product Management Society, a community of product managers. She sets out how product managers and designers should review research and test AI-drafted prototypes together in 2026, instead of handing work from one to the other. She warns that a convincing AI summary can make an assumption look like a fact, and that drafting something does not give anyone the right to approve it.

• The team writes one shared discovery brief with five fields: user problem, evidence, assumptions, business outcome and open questions. Each open question gets a named owner and a next step to test it.
• Themes that AI finds in interviews, tickets and analytics are treated as hypotheses and traced back to the source records. The calculations are checked too, such as whether monthly and annual payments were put on the same basis.
• The product manager can draft an early prototype from the brief with AI tools. The designer reviews its usability, accessibility and fit with the design system before it goes to users.
• Product managers approve scope and designers approve the experience, separately. Pricing changes, legal disclosures and account deletion stay behind strict human approval, even in experiments.

Product and design teams that prototype with AI can agree in advance who approves scope and who approves usability. A fast draft then does not become the plan by default. Naumnik writes from her own practice and recommendations, not from one team's account.

Derived by Working Surface from the article; more in the Playbook. Source line: Creating a draft doesn’t give someone final authority.

23 October 2026: teams get dependable work from AI agents by writing the rules first and keeping the checks with people and ordinary code that the agents cannot override.