October 3, 2026
GitHub's changelog entry from August 28, 2026, titled "Upcoming changes to GitHub Copilot policies and billing," laid out a quiet but consequential shift: every Copilot Business and Copilot Enterprise seat paid by credit card or PayPal now requires upfront payment at the start of each billing cycle, for every seat assigned -- regardless of whether the person holding that seat ever opens Copilot. New signups moved onto this model on September 1, 2026. Existing customers followed on October 1, 2026. GitHub says list prices for Business and Enterprise aren't changing. But the mechanics around the edges did: seat revocation no longer triggers a prorated refund, while seats added mid-cycle still get prorated from the assignment date through the end of the cycle.
Walk through what this means for an actual bill. Say an engineering lead assigns Copilot Business to a 40-person team at the start of a renewal cycle, because rolling it out to "everyone, just in case" is easier than figuring out who really needs it. Three weeks in, it's clear eight people haven't touched it. Under the old model, revoking those eight seats mid-cycle would claw back a prorated credit for the unused time. Under the policy that took effect October 1, that credit is gone. You already paid for the full cycle upfront, and freeing up the seat doesn't get you a dime back -- it just stops the meter for next cycle. Meanwhile, if you need to add a seat for a new hire halfway through, you still pay a prorated amount for the remaining days. The asymmetry is the whole story: adding is metered fairly, removing isn't. Over-provisioning now costs strictly more than it used to, with no offsetting upside.
It's tempting to read this as a GitHub-specific gripe, but the underlying dynamic is structural to per-seat SaaS pricing generally. The vendor's revenue is tied to seat count, not usage, so every billing policy will tend to drift toward making it cheaper to assign seats and more expensive to un-assign them. Nobody designs that on purpose as a punishment -- it just falls out of optimizing for predictable, upfront revenue. The fix vendors offer is almost never "buy fewer seats." It's a seat-reclamation dashboard, a usage report, a quarterly license audit. Entire product categories exist to help you manage the sprawl that per-seat pricing creates in the first place. That's not a conspiracy; it's just where the incentives point. If you want the fuller version of this argument, including where it does and doesn't apply, we laid it out in Buy vs. Build vs. ViibeStack.
To be clear, this isn't an argument against Copilot, or against per-seat software in general. A deeply integrated developer tool that sits inside your IDE and your CI pipeline earns its subscription -- rebuilding that in-house would be a waste of engineering time, and GitHub's list prices genuinely aren't going up. The place this billing change should actually change behavior is narrower: the internal tools and one-off workflows a team reaches for a SaaS seat to solve almost by default. A shared intake tracker. An internal approvals queue. A lightweight helpdesk for twelve people. A dashboard three managers glance at once a week. These are the tools most likely to get over-provisioned precisely because nobody wants to be the person who has to decide, seat by seat, who really needs access -- so everyone gets a license, and the bill reflects that decision whether or not it was the right one.
When you build that kind of tool instead of buying it per seat -- whether by hand or with an AI app builder -- there's no seat count to over-provision in the first place. Nobody is charged upfront for a hypothetical future user who might log in next quarter. There's no proration clock to watch when someone joins or leaves the team, because access to something you own isn't metered by a vendor's billing cycle at all. That's the practical case for tools like internal tools and admin builds or a purpose-built helpdesk you own outright: the entire category of billing-policy risk that GitHub's changelog just demonstrated simply doesn't exist when the software is yours. If your team is staring at a Copilot renewal and wondering whether you really need all 40 seats, that's a useful prompt -- but it's an even more useful prompt for the five other SaaS tools on your internal stack where the only reason you're paying per seat is that nobody has stopped to ask whether you should be.
Sources