The end of predictable Microsoft spend

TL;DR

The problem: Microsoft E7 introduces consumption-based pricing for AI tools that most organizations have no reliable way to meter, forecast, or govern, making budget predictability a thing of the past.

Key insights: 

• E7’s variable AI credit model means a negotiated discount no longer tells you what you’ll actually spend over a three-year term

• AI agents create a growing identity problem. Organizations managing 25,000 users today could be tracking 40,000+ identities within a few years

• Without visibility into agent activity and consumption telemetry before renewal, you are negotiating blind

Your next step: Audit your current AI estate, including licensed tools, agent deployments, and API integrations, before E7 becomes your default contract.

When Microsoft spend was still predictable

For years, forecasting Microsoft costs was hard work, but it was manageable. You knew the per-user price of E3. You knew the per-user price of E5. You multiplied by headcount, applied the negotiated discount, and arrived at a number you could defend in a budget meeting. Not glamorous. But predictable.

Microsoft E7 is not just a new SKU. It is the beginning of a different relationship between enterprises and their largest software vendor.

What E7 actually changes

E7 is structured as a frontier package: higher-tier, designed for the most advanced users in an organization. What makes it structurally different from E3 and E5 is not the features it bundles, but how two of them are priced.

Security Copilot comes with a base allocation of units. Go over it, and you pay for the overage. You cannot opt out. Security tools do not work that way.

Agent 365, the control plane for AI agents, has a fixed monthly fee. The agents themselves, however, consume AI credits on a pay-as-you-go basis. Every workflow an agent executes has a variable cost. That cost compounds across the organization invisibly.

So now the Microsoft estate runs on two pricing models at once. Per-user licensing, which SAM teams know how to handle, sitting alongside consumption-based credits that almost no organization has a reliable way to meter, forecast, or govern.

Why procurement is losing its footing

When licensing was per-user, an informed enterprise could negotiate from a position of strength. You knew your entitlements. You knew your usage. You knew what you needed at renewal and what you could hand back. That information was the foundation of every serious negotiation.

Consumption metrics dismantle that foundation. If AI credit spend is unpredictable, a 5% discount on credits tells you almost nothing about what you will actually spend over the next three years. The discount exists. The power behind it does not.

CFOs who could forecast Microsoft spend with reasonable precision are now working without the same anchor. Many are accepting this quietly, treating AI costs as a line item that will land wherever it lands. That is not a governance approach. It is the absence of one.

“When you’re informed, you can negotiate. When you’re not informed, Microsoft knows that. And it will be easier to sell you anything.” — Alexander Golev, SAMexpert, speaking at a Xensam webinar

The organizations that stay in control will not be the ones who accept that framing. They will be the ones who build a new basis for negotiation before the next renewal arrives.

The identity problem no one is planning for

E7 also accelerates the deployment of AI agents, and agents introduce a structural SAM problem most organizations have not yet confronted.

An agent is, in operational terms, an employee. It accesses systems, executes tasks, and generates audit events. When properly configured, agents run under their own Entra ID accounts, subject to the same access controls and policies as human users. That is the right approach. It gives security and IT teams real governance over what agents do and what they cost.

But it also means the number of identities in your directory is about to grow significantly. An organization managing 25,000 user identities today could be managing 40,000 or more within a few years once agents are counted properly. Microsoft has been clear about its intent to grow average revenue by expanding beyond per-user into per-agent and consumption-based models. Identities are central to that picture.

Most traditional discovery tools were built to count humans and devices. They were not built to count agents. That gap is already opening.

What this means for SAM programs

Traditional SAM handles the per-user side of the Microsoft estate cleanly. License entitlements, usage reconciliation, renewal modeling. Those processes still apply to E3 and E5, and they continue to apply to the per-user components of E7.

The consumption side is a different problem. It requires continuous visibility into which agents are running, what they are consuming, and what credit spend looks like before reconciliation, not after. It requires connecting discovery data, identity data, and consumption telemetry in a way that produces usable analysis before the renewal window closes.

That is the architecture of third-generation SAM. Not a replacement for traditional license management, but an extension of it into the consumption layer that Microsoft is expanding across its entire product line.

Xensam is built for both. Complete visibility across SaaS, on-premises, and cloud. Entitlement intelligence that maps contracts to actual usage. An AI-powered analytical layer that makes sense of consumption metrics traditional tools cannot read. And because Xensam is built in Sweden and subject exclusively to EU law, European enterprises can manage their Microsoft estate without adding US-jurisdiction risk to their data governance exposure. No CLOUD Act. No ambiguity.

What to do before E7 becomes the default

Map your current AI estate. Not just what is licensed centrally, but what is actually running. Self-subscribed tools, team plans, API integrations, agent deployments. The Xensam Insights: 2025 SaaS Report found that AI tools are not consolidating. They are growing across every category. The tail is longer than most IT teams realize.

Understand your agent exposure. How many agents are deployed in your environment today? Who created them? What systems do they access, and what do they cost? If you cannot answer those questions with data rather than estimates, you do not have visibility into an increasingly significant share of software spend.

Do not buy E7 without a plan. For organizations with heavy E5 and Copilot deployments that already understand their consumption patterns, the package offers real value. For everyone else, the headline discount is smaller than it appears, and the consumption exposure is larger.

The enterprises that handle the next wave of Microsoft pricing well will not be the ones who negotiated the best discount. They will be the ones who built the capability to understand what they are actually consuming.

The forecast is gone. What replaces it is visibility.