Stop Counting Copilot Seats: GitHub Enterprise Now Includes Copilot, Billed by Usage

In one sentence: stop counting seats, start managing a budget

On October 28, 2026, GitHub announced at its Universe conference that enterprises working with a GitHub or Microsoft account team can now get Copilot included with GitHub Enterprise for every licensed user, without buying separate Copilot seats. Companies used to manage two things: how many Copilot seats they paid for, and how much AI usage those users consumed. Now the first one is gone. What is left is a single thing to manage: AI usage against the enterprise budget.

That sounds like a billing detail, but it changes how the cost is managed. Under the old model, cost followed headcount: more people meant a bigger bill, even if some of them only used completions twice a month. Under the new model, cost follows usage: it is metered from the first request against the enterprise budget, so heavy users cost more and light users cost almost nothing. GitHub's changelog puts it plainly: with Copilot included in the platform, usage against the enterprise budget becomes the primary thing to manage.

For people who write code every day, nothing changes on the day of the announcement. Completions do not get faster and chat does not get smarter. The change matters to the people who sit in front of the budget spreadsheet: procurement, platform teams, and engineering leaders. This article answers four questions: what actually changed, what happens when the budget runs out, who can use it now, and the five things a responsible leader should do next.

What changed: three contrasts

First, access changed. Copilot used to be an add-on subscription on top of Enterprise, with a seat bought for each person. Now it comes with the platform: every licensed user in the enterprise account gets it, with no per-person seat purchase.

Second, the billing unit changed. Under the old model, each seat came with a monthly AI allowance, and usage beyond it was billed extra. Under the new model, there is no monthly allowance per seat to track or top up. Usage is metered from the first request against the enterprise budget. That matters for forecasting, because you can no longer estimate spend from "how much each person gets per month." You have to look at actual usage.

Third, the management tools did not change. GitHub says usage is managed with the policies, alerts, and attribution the enterprise already has. Teams that already built usage dashboards do not need new tools. They need to read the same dashboards with a different question in mind: not "how many seats," but "where did the usage go."

What happens when the budget runs out

This is the most misunderstood part of the new model, so it deserves its own section. GitHub's changelog lists four points, in plain language:

Code completion and next edit suggestions keep working even with no budget available. The most common daily feature, the one that suggests the next line as you type, does not stop when the budget hits zero. GitHub deliberately kept this floor: however bad the overspend gets, the basic coding experience continues.

Copilot Chat, Copilot code review, and Copilot in the cloud require available budget. When the budget is empty, these three stop until budget is restored. These are also the most agent-heavy features: chat that rewrites whole files, review that reads every line, and cloud sessions that run tasks on their own. They burn usage fastest, and they are the first to be cut off.

An exhausted budget never removes access. GitHub's own words. Your seat stays, and the features that do not need budget stay. This is not a service shutoff. Your account is not locked and completions do not stop. Only the heavy work pauses.

People in the enterprise account who do not hold a GitHub Enterprise license get the same Copilot access through Copilot Standalone, at $10 per user per month, with the same features and models as Copilot Enterprise but no monthly AI credits per seat. This covers contractors, interns, and partners who work inside the enterprise account without a formal license. The price and the rules are stated outright, so there is nothing to guess.

Who can use it now, and who has to wait

There are two kinds of enterprise here, and it matters which one you are. Enterprises working with a GitHub or Microsoft account team can use the new model now, for GitHub Enterprise Cloud. To set up a new enterprise account on this model or move an existing one over, talk to the sales team.

Enterprises that manage their own subscription, the kind that buys and renews on the website, cannot use it yet. GitHub says self-serve availability is coming in stages, with dates to be shared in the coming weeks, and current self-serve subscriptions are unaffected. If that is you, the useful work right now is preparing usage dashboards and budget alerts, not changing the purchasing process.

The old plans are not being retired. Copilot Business and Copilot Enterprise remain available to buy, for new and existing enterprises. This is not a forced migration. It is a third road alongside the existing two. Teams that are happy with seat-based billing do not need to move. The new model fits best the companies whose usage is uneven: some people run agents all day, others barely touch them, and seat-based billing never fit them well.

For large companies, the real change is managerial

Replacing seats with a budget looks like one less invoice, but it moves cost management from procurement to engineering management. Procurement used to ask "how many people need Copilot," and the answer was a number. Now the question is "which teams and which tasks are burning usage," and the answer is a report. That shift has three practical consequences.

Attribution matters more than before. Without a per-seat allowance, looking only at the enterprise total tells you nothing. Is one team running cloud agents for a migration in bulk, or is one project repeatedly sending huge changes through review? Only attribution by team, repository, and feature can answer that. GitHub says the attribution tooling is already there, so it should actually be used, split by chat, review, and cloud, not just watched as a total.

Alerts must sit on the budget, not on seats. Under seat-based billing, running out of seats was a natural brake: someone had to request more seats. The new model has no such brake. Usage can keep climbing until the budget is gone and the heavy features stop. So budget alerts need to be set early, for example at around 70% to warn team leads and around 90% to warn the platform team. Alerts should go to people who can make decisions, not only to a finance mailbox.

Cost discussions move from "whether to buy" to "how to use." The old argument was whether a team deserved seats. The new argument will be whether a task is worth an agent running all night. That is a healthier argument, because it points at the real question: which tasks genuinely save time with an agent, and which ones just burn money for reassurance.

For individuals and small teams: do not misread this news

This news is easy to misread as "Copilot is free now" or "Copilot got cheaper." Neither is true. It applies only to GitHub Enterprise customers, and right now only to the ones working with an account team. Individual users, free users, and small teams buying their own subscriptions see no change.

It is also not a price cut. Usage is metered from the first request with no included allowance to cushion it, so a heavy user may pay more than under seat-based billing. Small teams working with an enterprise customer should note the Copilot Standalone route: $10 per user per month for the same features and models, but usage billed separately. Do not budget for it as "$10 all-in," because the $10 is only the entry fee.

One reassurance applies to everyone under the new model: completions keep working when the budget runs out. Even if the first year's budget estimate is wrong, the team does not grind to a halt. That floor makes the new model much safer to try.

Worked examples: three fictional teams

Rules alone stay abstract, so here are three fictional teams. The examples are invented, and the numbers only illustrate the mechanism, not any real company's bill.

The first team uses agents heavily across the board: twenty people running cloud tasks and large reviews every day. Under seat-based billing they paid per seat plus overages. Under usage billing their bill follows usage directly and will probably be higher, because heavy work is genuinely expensive. Their gain is not savings but clarity: they can finally see which project and which kind of task burns the most, and where to cut.

The second team has very uneven usage: thirty people, but only five use agents heavily and the rest mostly use completions. Under seat-based billing they paid for thirty seats no matter what. Under usage billing the light users cost almost nothing, so the total may actually fall. This team is the clearest winner from the change, and the one GitHub most wants to win over.

The third team uses agents occasionally: a ten-person company that only leans on agents during crunch weeks. Under seat-based billing they had to buy seats for everyone or go without. Under usage billing everyone has access and the bill follows the crunch weeks. For them, the new model lowers the barrier to starting, not the unit price.

Across all three, the conclusion is the same: whether the new model saves money depends on the shape of usage, not on headcount. Many people does not necessarily mean expensive. Heavy usage does. Before switching, pull the last three months of usage by person and by team and look at the distribution.

Questions people get wrong

Do we stop managing seats entirely? No. Licenses are still per person, and Copilot access follows the license. What changed is that the Copilot part is no longer billed per seat. It is billed by usage against the budget.

Do completions really keep working with an empty budget? According to GitHub's changelog, yes. Completions and next edit suggestions do not depend on budget. Chat, review, and cloud features do.

Can a self-serve enterprise switch right now? No. GitHub says self-serve availability is coming in stages, with dates in the coming weeks, and current subscriptions are unaffected. Prepare dashboards and alerts first.

Are Copilot Business and Copilot Enterprise being discontinued? According to GitHub, no. Both remain available for new and existing enterprises. The new model adds a road; it does not close the old ones.

What about collaborators without a license? They go through Copilot Standalone at $10 per user per month, with the same features and models, but no monthly included allowance. Usage is billed separately.

How this differs from the other Copilot news this week

GitHub's announcements came in a cluster, and readers mix them up. Here is the boundary. Copilot Fusion is about model selection: the system picks models and orchestrates them step by step. Copilot Projects is about working across repositories: one goal, many repos, parallel agents. Copilot code review Max and approvals are about review quality: multi-agent review plus approvals that count toward required approvals. This article is about billing: seats replaced by a budget. Four announcements from the same conference, four different questions.

If you only care about writing code, Fusion and Projects affect your day more. If you manage budgets, teams, or procurement, this is the one to read first. Keeping the four angles apart avoids the feeling of "I think I already read this."

The judgment: a long-overdue correction

In the end, this change corrects a mismatch that has been obvious for a while: the cost of AI coding tools was never really per person. It was per task. One person running an agent overnight on a migration and another person using completions twice a day cost wildly different amounts, yet once paid the same seat fee. Usage billing makes the bill honest about that for the first time.

Honesty has a price, though: it requires actually watching usage reports, setting alerts, and agreeing on rules. Seat-based billing was imprecise but simple. Budget-based billing is precise but demands management. If your team has no usage dashboard and no budget alerts yet, now is the time to build them, not because the new model forces you, but because agent usage will only grow.

One last reminder: the new model is currently open only to customers working with an account team, and self-serve follows in the coming weeks. Teams that cannot switch yet can use the wait to prepare reports and rules. Teams that can switch should run a pilot with one uneven-usage team first. Either way, do not read "Copilot included with Enterprise" as "Copilot costs nothing." It costs money. It just costs it differently.