Microsoft CEO: AI agents need to be ‘treated like employees’

“You need to give them an identity, have their own ‘playground’, then set management policies like a real employee,” Nadella told host Reid Hoffman in the episode. Possible Podcast last weekend. “We’re looking at tools and policies that can be applied to these ’employees’, such as giving them specific permissions about what can and cannot be accessed, as well as ways to check for safe work.”

According to him, similar to other businesses in the AI ​​trend, Microsoft also spends huge amounts of money on artificial intelligence, while also finding ways for AI agents to work effectively when combined with humans.

However, figuring out how to manage them is “exceptionally difficult”. The Microsoft CEO said that he can operate 100 AI programming agents at the same time, but instructing each agent through a chat interface is not easy to do. “The cognitive burden for me in managing this is very high,” Nadella said.

He also said the company currently operates Agent 36, a toolkit that includes Entra – a digital identity and network access product, and Purview – a data labeling system generated by AI agents. “I think security capabilities, control issues, management and monitoring will be factors that we can trust in artificial intelligence agents when operating businesses,” Microsoft CEO added.

Microsoft introduced an AI agent named Scout at the Build event on June 2. The tool builds on the power and flexibility of OpenClaw, is cloud-based, and compatible with desktop and web browsers. Users can connect to their inbox, calendar, and other systems on their computer, as well as into the Microsoft 365 ecosystem, using it to manage their calendar, create meeting agendas, and more. However, according to the documents 404Media gathered, Microsoft planned to make users “addicted” to this agent.

Nadella’s statement comes as many large companies such as Nvidia, Microsoft and Google are racing to develop new technology that allows computers to use AI agents to replace humans in complex tasks. In particular, Nvidia and Microsoft are improving Windows and incorporating the RTX Spark chip to achieve their goals, while Google is about to release Googlebooks with the ability to suggest actions when the user hovers over a location on the screen.

Analyst Neil Shah, co-founder of market research company Counterpoint Research, predicts that Nvidia’s direction in developing AI agent computers that process data themselves will revolutionize computing in the next 10 years.

However, according to Business Insiderthe rapid development of autonomous AI also raises concerns. On June 4, Anthropic called on leading AI developers to establish a coordination mechanism to slow or stop development if AI begins to improve itself at a rate that exceeds society’s ability to manage risks. This startup believes that AI with the ability to self-improve will be a major development in the history of technology, but also increases the risk of humans losing control of artificial intelligence.