Microsoft uses many AI agents. We’re starting to think of them like human employees to help us manage them.
CEO Satya Nadella said the software giant is considering what tools and policies it needs to oversee all the agents it creates. This includes giving agents specific permissions about what they can and cannot access within the company and how to audit their work, he said.
“You need to give them an identity, give them a sandbox, and set policies to govern them,” Nadella told Reid Hoffman on an episode of “The Possible Podcast” posted Friday.
During the discussion, Hoffman also said he plans to leave Microsoft’s board after 10 years and return to what he calls “founder mode.”
Businesses are spending billions on implementing AI, but many still don’t understand how AI agents work with human employees. Understanding how to manage AI agents is a particularly difficult problem.
It’s a challenge Nadella himself has dealt with, he told Hoffman. The Microsoft CEO said he often runs 100 AI coding agents at once, and guiding each one through a chat interface is difficult. “The cognitive load on me managing this is very high,” he said.
Microsoft has developed Entra, its digital identity and network access product, and Agent 365, a suite of tools that includes Purview, which the company uses to label data created by its AI agents, he said.
“I think security, containment, manageability and observability are how we can have confidence in these agents,” Nadella said.
Any tips? Contact information for these reporters is as follows: dreuter@businessinsider.com and abitter@businessinsider.com. Use a personal email address, non-work WiFi network, and non-work device. this is our A guide to sharing information securely.