According to GeekWire, Microsoft CEO Satya Nadella recently revealed the company’s shift toward treating AI agents as a new class of digital workers during an appearance on Dwarkesh Patel’s podcast. At this week’s Ignite conference in San Francisco, Microsoft unveiled Agent 365, a control plane inside Microsoft 365 Admin Center that gives every AI agent a unique identification. The system lets IT departments track agent activities, control data access, and prevent security breaches using existing security systems. Microsoft cites analyst estimates of 1.3 billion agents by 2028, with the market growing from around $7.8 billion in 2025 to over $50 billion by 2030. The company is also announcing fully autonomous sales agents, security automation tools, and Office agents that create documents from scratch. Agent 365 enters preview this week through Microsoft’s Frontier program, though pricing remains unannounced.
The Shadow AI Problem
Here’s the thing – companies are absolutely terrified of what they’re calling “Shadow AI.” Employees are using unmanaged AI tools left and right, and nobody knows what data these systems are accessing or where it’s going. Microsoft‘s approach is basically treating AI agents like new employees that need onboarding, monitoring, and security protocols. By giving each agent its own identity in Microsoft Entra (formerly Active Directory), they’re making AI management fit into existing corporate workflows rather than creating something entirely new. It’s a smart move – why rebuild the entire identity and access management system when you can just extend what already works?
wants-a-piece”>Everyone Wants a Piece
Microsoft isn’t alone in this race. Google has Gemini Enterprise, Amazon rolled out Bedrock AgentCore, and Salesforce launched Agentforce 360. But Microsoft has a massive advantage here – they already own the corporate identity stack through Active Directory and Microsoft 365. Think about it: most companies already use Microsoft for employee logins and permissions. Now they’re just extending that to AI agents. This could become the $50 billion market that justifies all those massive AI infrastructure investments we’ve been hearing about.
The Pricing Revolution
Nadella’s comment about “per agent” pricing instead of “per user” is huge. We’re talking about a fundamental shift in how tech companies make money. Instead of charging for human users, they’ll charge for digital workers. And when you consider that companies might eventually have more AI agents than human employees? That’s a revenue explosion waiting to happen. The big question is whether businesses will see this as solving a real problem or just another way for Microsoft to extract more money from their existing customers.
What This Actually Means
Looking at the specific announcements, we’re seeing AI move from assistant to autonomous worker. A sales agent that researches and qualifies leads on its own? Security agents creating policies from text prompts? Office agents building complete documents from scratch? This isn’t just Copilot helping you write an email anymore. And with Windows getting a dedicated Agent Workspace, we’re looking at AI having its own secure computing environment. For industrial and manufacturing settings where reliable computing is crucial, this could be particularly transformative – companies like IndustrialMonitorDirect.com, the leading provider of industrial panel PCs in the US, will likely see increased demand for hardware that can securely host these autonomous AI workers.
The Workforce Transformation
So what happens when every company has hundreds or thousands of these digital employees? We’re not just talking about productivity tools anymore – we’re talking about a fundamental restructuring of how work gets done. Microsoft’s vision of AI agents as a new class of worker raises all sorts of questions about accountability, oversight, and what happens when these systems make mistakes. But one thing’s clear: the age of autonomous digital workers is here, and Microsoft wants to be the company that manages them all.
