Scaling AI responsibly at Microsoft: Lessons from our Customer Zero journey

Microsoft's internal teams govern thousands of AI agents by giving each one a narrow job, reviewing it before launch and requiring explanations that people can check.

Microsoft's Inside Track blog reports what the company learns by using its own AI products before customers do. In this piece by David Hirning, teams in supply chain, legal and customer support describe how they deploy AI agents without losing control of them. Alberto Carrozzo, who leads responsible AI reviews for Europe, the Middle East and Africa, says his team has reviewed more than 2,000 internal tools.

• Jeevan Pathuri, vice president of software engineering for the cloud supply chain group, has more than 100 AI agents in production. The group tests each one to confirm it stays within its intended job.
• The legal department's AI agent works only on a curated set of documents. It gives a written reason for each relevance decision, so that attorneys can check it against the source.
• Customer support teams serve customers in nearly 200 countries. They add new security, privacy and governance requirements to their AI tools as those requirements change, rather than approving a tool once.
• Carrozzo's team uses patterns from past reviews to direct reviewers to the higher-risk cases.

A team deploying several AI agents can define each agent's job before launch and test that it declines work outside that job. This is Microsoft's account of its own practice, and the piece promotes Microsoft's tools.

Derived by Working Surface from the article; more in the Playbook. Source line: A narrow scope is a safety feature.

24 October 2026: teams keep the time AI saves only when they decide in advance what an AI agent may do alone, what waits for a person and who maintains what it builds.