Boosting product manager productivity at Microsoft with AI

Microsoft product managers built AI agents to draft leadership reports and clean planning data, and now spend their time reviewing and approving the agents' work.

Product managers in Infrastructure Engineering Services, part of Microsoft's internal IT organisation, spent weeks each month turning Azure DevOps planning data into leadership reports. They built one AI agent to draft the reports and another to flag missing or inconsistent data. Reports now arrive on the first day of the following month, and the team describes its role as reviewing and approving agent output.

• A monthly report used to arrive near the end of the following month; July's report arrived at the end of August.
• A second AI agent checks planning items against written rules, flags gaps and tells each person how to fix them. The team added it because poor data had produced misleading AI summaries.
• The team keeps reusable agent instructions in one shared repository so product managers adapt existing ones instead of building duplicates.
• Nevedita Mallick, a principal product manager, says prototyping that took a day or two now takes under an hour.

Product teams that hand reporting to an AI agent may need to fix their source data and name the reviewer before the agent is useful. This is an inference and is not the author's claim. The piece is Microsoft's account of its own products used inside Microsoft.

Derived by Working Surface from the article; more in the Playbook. Source line: At the end of the day, the AI can only be as accurate as the data it has in front of it

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.