Satya Nadella, Chairman and Chief Executive Officer of Microsoft, has called on business leaders to rethink how they manage artificial intelligence at work, arguing that AI agents are no longer just tools but entities that require the same governance structures applied to human employees.
Speaking on an episode of Reid Hoffman’s Possible podcast recorded fresh off Microsoft Build 2026, Nadella said organisations must extend formal structures to their AI systems if they hope to operate them responsibly at scale.
Give Them an Identity
“You need to give them identities, you need to give them sandboxes, then you need to set policies to govern them,” Nadella told Hoffman, outlining a framework that mirrors how companies onboard and manage human staff.
He added that the challenge is not simply technical but organisational. Nadella revealed that he personally runs as many as 100 AI coding agents simultaneously, and that guiding each through a standard chat interface is far from straightforward. “The cognitive load on me managing this is so high,” he said, pointing to the need for smarter interfaces built specifically for agent management.
A New Kind of Workforce
Nadella’s broader vision is of a workforce in which human capital and what he calls “token capital” are deeply intertwined. He described a future where every knowledge worker manages a growing network of AI agents on their behalf, with a new layer of infrastructure required to handle the exceptions, notifications, and instructions flowing between humans and machines. The knowledge worker, in his framing, does not get replaced. They become responsible for directing, governing, and extracting accountable output from an entirely new class of colleague.
Microsoft’s enterprise AI roadmap reflects this vision directly. The company’s Copilot platform is being positioned as the primary interface through which this agent-led model of work operates, with Build 2026 serving as the clearest signal yet that Microsoft is betting its next decade of enterprise growth on this shift.
Security and Accountability at the Core
Nadella was equally direct on governance, emphasising that confidence in AI agents will only come through security, containment, manageability, and the ability to observe what agents are actually doing inside enterprise systems. Without those guardrails in place, he argued, organisations cannot operate AI at the scale he is envisioning. Microsoft has been building out its enterprise tooling in this direction, with identity management and data monitoring capabilities forming a growing part of its AI infrastructure offer.
What This Means for Leaders
For business leaders, the shift Nadella is describing is significant. The question is no longer simply whether to adopt AI but how to build the organisational muscle to manage it with the same rigour applied to human teams: clear roles, defined boundaries, audit trails, and accountability. Organisations that treat agent deployment as a technology decision alone, rather than a workforce and governance decision, risk building capability without control.
The Possible episode also addressed Reid Hoffman’s departure from Microsoft’s board after a decade to focus on his AI drug discovery startup Manas, reflecting a broader moment of transition for the industry’s most prominent voices as AI moves from experimentation into enterprise infrastructure at scale.
Sources: Satya Nadella, Possible Podcast with Reid Hoffman, Business Insider, Windows Central, Firstpost
TagTechnology
