Microsoft Plans to Charge a License for Every AI Agent You Hire
A few weeks ago, Rajesh Jha, an executive at Microsoft, made a statement at a conference that perfectly encapsulates the direction of the company's revenue model: "All of those embodied agents are seat opportunities". Translated without euphemisms: every artificial intelligence agent a company deploys in its operations could become another line on Microsoft’s monthly bill.
What's brewing is tentatively named Microsoft 365 E7—with a price already circulating in the analyses of the firm Directions on Microsoft: $99 per month per agent, combining the E5 layer with Copilot and a new management tool called Agent 365. To put this into perspective, the current E5 tier costs $57 monthly and will increase to $60 starting July 1, 2026. Copilot adds another $30. E7 would bundle all this plus governance controls for autonomous agents.
This move is not an accidental product engineering feat. It is a very deliberate business thesis on how to protect billions in recurring revenue when automation threatens to reduce the number of humans—and consequently licenses—within companies.
The Arithmetic Behind the Move
Rajesh Jha illustrated this with a concrete example: a company with 20 employees currently buys 20 licenses. In the scenario that Microsoft is betting on, that same company could operate with 10 humans and 40 AI agents—four per person—raising the seat count from 20 to 50. This represents a 150% increase in revenue from the same client, without that client hiring a single additional employee.
The financial logic is impeccable from Microsoft’s perspective. The risk the company wanted to avoid was the opposite: that automation would reduce the number of human users, thereby collapsing the subscription base on which their productivity business rests. Lane Shelton, director of advisory services at Directions on Microsoft, described it as positioning itself as the "business control plane for the emerging digital workforce". In other words, being the layer that manages and bills everything an agent needs to operate: identity, email, access to Teams, security policies, document analysis in OneDrive.
This last detail is not insignificant. Agent 365 requires a Microsoft 365 Copilot license to function. There is no ambiguity here: this is an architectural decision that turns dependence into a monetization model. If you want to govern your agents within the Microsoft ecosystem, you need Copilot. If you need Copilot, you pay an additional $30 per agent. If you want the integrated package, you pay $99. The chain is deliberately designed.
What’s Still Unverified and the Silence That Betrays It
Microsoft has made no official announcement regarding E7. The company acknowledged inquiries from specialized media but did not provide further details. Mary Jo Foley, an analyst at Directions on Microsoft, reported on the plans based on market signals and industry conversations in March 2026. This means what we have today is a product hypothesis with a tentative price but no confirmed launch date or defined contractual terms.
From my perspective, this is precisely where the highest risk for Microsoft lies. Not because the idea is wrong—the strategic direction is coherent—but because the price of $99 per agent is a figure floating in the air without public validation of willingness to pay. Nenad Milicevic, a partner at AlixPartners, represents the opposing stance with equal logic: if a single human can supervise 40 agents, the number of users interacting with corporate software decreases, which eventually pressures all license vendors, not just Microsoft. His argument is that charging machine access over machine access generates enough friction to push clients toward more open platforms.
Both visions are plausible. And the fact that both can be plausible simultaneously is precisely the signal that no one is certain about how much the market is willing to pay for a license for an autonomous agent. Microsoft is building a product tier without having seen clear signals of acceptance at that price. The conference statements and the subsequent official silence are the corporate equivalent of launching a product behind closed doors and waiting for the market to validate it afterward.
The Precedent That Turns This into an Industry Experiment
What Microsoft is doing with E7 will have consequences that extend far beyond its own billing. Salesforce, Workday, and any other enterprise software provider with a seat-based model are watching this move with calculated interest. If Microsoft manages to get companies to accept paying $99 monthly for each deployed agent, it will have set the benchmark price for the entire industry. If there is massive resistance, the licensing model per agent could take years longer to mature or shift towards consumption or outcome-based frameworks.
The closest historical precedent is Microsoft's own initial rollout of Copilot. Microsoft launched Copilot as a $30 add-on at a time when actual adoption within companies was limited and uneven. It took time, product iterations, and commercial pressure for the add-on to normalize. E7 faces the same adoption process but with greater complexity: it requires buyers to understand and manage autonomous agents, which is still not standard capability among most corporate IT teams.
Moreover, the Agent 365 category is currently in preview. This means Microsoft is considering setting the price of a premium tier based on a tool that is not yet in widespread production. Selling the promise before the product is stable is not unusual in the tech sector, but it does increase the risk of fixing the price before the product demonstrates the value that justifies that number.
The Seat Model Will Not Disappear, But It Needs Real-World Validation
The strongest argument that Microsoft has is also the simplest: agents consume resources, generate infrastructure costs, require managed identities, and security policies. Someone has to pay for that, and it makes sense that the entity generating the cost—whether human or digital—has an associated billing unit. The problem isn't the logic of the model, but the speed at which it’s being formalized before the market has operational clarity about what an agent is, what it does, and what measurable value it delivers.
Companies currently beginning to deploy pilot agents in specific departments do not yet have their own data on how much productivity these agents generate per month. Without that number, they cannot make the calculation that justifies $99 per unit. Microsoft needs that learning cycle to be completed in the market so that the price can encounter minimal resistance. In the meantime, the period between now and July 2026—when E5 and Copilot prices increase—will be the window where contractual negotiation signals will reveal how much actual resistance exists.
The business leadership that is a Microsoft client currently has one relevant lever: accurately document the operational value that their existing agents are generating. That evidence is the only basis upon which a new pricing tier can be negotiated, adopted, or rejected. Ignoring this exercise and allowing the contract to auto-renew equates to validating a price without having tested whether the value proposition holds up. The market that concedes without data cedes to Microsoft the authority to set the price of automation for the next decade.
