How Copilot usage-based billing works
Every Copilot seat comes with AI credits. Credits are pooled across the enterprise — then cost centers and budgets decide who can use what.
Seats become one shared pool of AI credits
Each seat's price converts 1:1 into AI credits, and every seat's credits land in the same pool.
50 seats
Copilot Business · $19 / seat / month
Each seat adds
Enterprise pool · this month
Resets on the 1st of each month · no rollover
One pool for everyone. Heavy users can draw more; lighter users balance them out. No per-person buckets.
Budgets layer on top of the pool
Every Copilot request passes through these checks, in order. Fail one and the request is blocked.
A user sends a Copilot request chat · CLI · agent - Check 1
User-level budget
ALWAYS ON · HARD STOPCaps each person's usage — from the pool and beyond. Most specific one wins:
- Individual e.g. @mona · $100
- Cost center e.g. Eng · $30/user
- Universal everyone · $10/user
Over budget → blocked
- Check 2
Shared pool
INCLUDED IN SEATSCredits left in the pool? Served at no extra cost.
Cost-center caps (if on) are checked here too.
Credits left → served
Pool empty → metered
- Check 3
Metered usage
$0.01 PER CREDITCharged to the first budget that matches:
- Cost center budget if in one
- Organization budget license org
- Enterprise budget otherwise
The enterprise budget still counts it (unless excluded). Tightest budget blocks first.
“Stop usage” at limit → blocked
Paid usage only happens if the “AI credits paid usage” policy is on (default). Off = blocked as soon as the pool is empty.
YOUR MAXIMUM MONTHLY BILL
The enterprise budget isn't your total bill — it only limits spend once the pool runs out.
Good to know
- Code completions and next edit suggestions don’t consume AI credits.
- Paid usage is on by default via the “AI credits paid usage” policy.
- User-level budgets are always a hard stop.
Updated September 22, 2026