We’re not going to relitigate the specific numbers; they’ll have changed again by the time most of you read this. What we want to talk about is the pattern, because Microsoft’s Copilot pricing has been adjusted often enough that small ops teams should be planning for the pattern, not the current snapshot.
The pattern is this: Microsoft is moving Copilot from a discrete add-on with a clear monthly per-seat fee toward a more granular, usage-influenced model that bundles different capabilities into different tiers. The exact dollars move; the trajectory is the same. Capabilities you used to get as part of a bundle become their own line item. Capabilities you used to pay for individually get folded into bundles. The net effect on a 12-person ops team’s spend is usually upward, sometimes by enough to matter.
For SMBs, the practical implications fall into three buckets.
Audit what you’re actually using
The default reaction to a pricing change is to assume you need everything you currently have. You almost certainly don’t. The Copilot capabilities most heavily used in any team we’ve Scored are: meeting summaries, email drafting, and the occasional spreadsheet question. Most of the more advanced features (the “Copilot in Power Automate” type capabilities, the agent-builder pieces, the tenant-wide governance tools) are bought in bulk and used by two people. A real audit of what’s getting opened, by whom, how often, will frequently find 30-50% of the spend is funding capabilities the team has never actually used in anger.
Don’t let the bundle dictate the architecture
A common pattern we see: a team has Microsoft Copilot, so they assume the right way to build automation is inside Power Automate with Copilot. Sometimes it is. Often it isn’t. The right architecture for your workflow might involve OpenAI, or Anthropic, or a mix, or a custom orchestration layer with cheaper open-weight models for the volume work and a frontier model for the hard parts. Letting your platform vendor’s pricing changes dictate your architecture is how you end up with technical debt that costs more to undo than the original “savings.”
Negotiate at the right cycle
Microsoft, like most enterprise vendors, expects to be negotiated with. SMBs frequently don’t, because the contract values feel “too small to bother.” This is a mistake. The discounts available at renewal, especially if you’re consolidating other Microsoft spend at the same time, are real. The most successful negotiations we’ve seen from SMB clients involved being prepared with usage data and a clear “we’ll go elsewhere if” alternative. Microsoft responds to that. Without it, you’re on retail.
The deeper question this pricing churn raises is one worth sitting with: how much of your operating model do you want to be downstream of one vendor’s decision-making? Microsoft is a fine vendor. They are also a vendor. Their pricing model is theirs to change. If a 30% increase in Copilot pricing would meaningfully damage your operations budget, you’ve built more dependency than is wise. The corrective isn’t to avoid Microsoft; it’s to architect your AI workflows so that swapping the underlying model provider is a Tuesday afternoon’s work, not a multi-quarter rebuild.
We design with this in mind by default. Our orchestration layer is provider-agnostic. The agents we ship can be moved between Microsoft, OpenAI, Anthropic, and self-hosted models with config changes, not code rewrites. Most of our clients never need to use this flexibility. The few who have, were grateful it was there.
The conductor doesn’t get to choose which violinist becomes unavailable. The conductor does get to choose whether the score can be played by a different one.