GitHub Copilot Business is the organizational tier of the AI pair programmer, billed per user per month with policy controls, data handling commitments, and admin management on top of the individual product. The capability is real and the developer enthusiasm is genuine, which is exactly why the seats get rolled out broadly and fast. The bill is a function of assigned seats; the value is a function of active seats; and the gap between the two opens within the first quarter of any deployment. Seats land on developers who try it twice and drift, on contractors who leave, and on teams that never adopt the workflow. Copilot Business is where the per seat AI premium is the easiest line to grow and the hardest to govern, and where active usage is the only honest measure of what to buy.
Copilot Business is a per user subscription that adds organizational controls to the AI coding assistant. Above it sits Copilot Enterprise, which layers deeper repository awareness and additional capability at a higher rate per seat.
Over the individual product, Business adds centralized seat management, organization wide policy controls, the commitment that prompts and suggestions are not used to train the foundation models, and the administrative surface a regulated enterprise needs. It is the floor for any serious corporate deployment, and the controls are the reason to buy it rather than the individual plan.
Copilot Enterprise sits above Business and adds repository aware chat, knowledge base grounding, and tighter integration with the codebase, at a higher per seat rate. It is justified for organizations whose developers genuinely work the deeper features. Defaulting the whole population to Enterprise because it is the top tier repeats the over editioning pattern that inflates every developer line.
Copilot produces three recurring exposures. The first is the broad rollout to a population whose active usage never matches the seat count. The second is the Enterprise tier assigned without evidence. The third is the seat that survives the developer who stopped using it.
Copilot gets deployed to the whole engineering org on the strength of early enthusiasm. Within a quarter a meaningful share of seats show little or no active usage as developers try it, drift, or never build the assistant into their workflow. The bill holds at the assigned count while the value collapses to the active count, and the gap is the first thing nobody is measuring. The enthusiasm that drove the rollout is exactly what makes the line hard to question later, because cutting seats reads as cutting a popular tool rather than removing waste.
The Enterprise tier gets selected because it is the most capable, and the repository aware features that justify the premium go unused across most of the base. Paying the Enterprise rate for developers who only use the inline completion that Business already provides is the same over editioning that inflates the rest of the developer estate, now on a per seat AI line.
Copilot seats assigned to contractors and movers keep billing after the person leaves the team or the company. Because the monthly per seat charge is modest, the stranded seats escape scrutiny individually. Across a large org the unreclaimed seats compound into a standing monthly line that grows quietly with every onboarding and never shrinks with the departures. Because the assistant is provisioned at hire and rarely deprovisioned at exit, the count only ratchets upward over the term.
Copilot responds to three levers. The activity reconciliation right sizes the seat count to active developers. The tier review moves the base to Business where Enterprise is unused. The reclaim discipline catches the stranded seats at the leaver and mover events.
The admin usage reporting shows which seats are active and which have gone quiet. The seat count is rebased on developers who genuinely use the assistant, the quiet seats are reclaimed or held for redeployment, and the contract reflects active adoption rather than the optimistic rollout. This is the single largest correction on a Copilot line in its first year.
The active count then feeds the broader Microsoft and GitHub position negotiated at the EA renewal.
The Enterprise seats are tested against use of the repository aware features, and the developers who only use inline completion move to Business at the lower rate without losing a capability they exercise.
A reclaim process at the leaver and mover events keeps the seats mapped to active developers, so the monthly line tracks real adoption instead of accumulating every seat the org has ever assigned.
Copilot increasingly negotiates inside the broader Microsoft and GitHub agreement, where the active count resets the baseline and the developer volume across Copilot, GitHub, and the IDE seats combines into a single position.
A renewal that carries the original optimistic rollout forward locks the over assignment into the term. The reset happens against the usage reporting: the active seat count established, the tier mix corrected, and the baseline rebased before the rates are set. A buyer who arrives with that evidence negotiates the Copilot line against real adoption rather than the headline deployment number.
Copilot, GitHub Enterprise, and the Visual Studio seats describe the same developer population. A buyer who negotiates them as one combined position carries more volume and more leverage than treating each as a standalone renewal. The AI line is set inside the full developer tooling spend, where the overlap and the total commitment shape the rate rather than the Copilot seat count in isolation.
The engagement is an adoption and tier diagnostic, an active seat and reclaim model, and the integration of the clean baseline into the broader developer tooling negotiation. The output is a Copilot line priced at active use.
We pull the admin usage reporting, separate the active seats from the quiet ones, test the Enterprise tier against use of the repository aware features, and surface the contractors and movers still holding seats. The output is a defensible picture of real Copilot adoption and the seat count the organization genuinely supports.
We rebase the seat count on active adoption, move the over editioned seats to Business, reclaim the stranded ones, and fold the clean count into the combined developer tooling position across Copilot, GitHub, and Visual Studio. We secure the rates and lock protection on the AI line. The output is a Copilot position priced at active use and defensible through the term.
The Copilot diagnostic reconciles assigned seats against active usage, corrects the tier mix, reclaims the stranded seats, and folds the clean count into the combined developer tooling negotiation across Copilot, GitHub, and Visual Studio. The result is an AI line priced at real adoption, not the rollout number that launched the deployment.