GitHub Copilot’s token billing shows why AI costs need guardrails

GitHub is turning Copilot from a flat subscription into token-based billing, and that change matters because it turns AI coding from a predictable software seat into a usage-management problem. Starting June 1, the cost of writing code with Copilot is no longer just a line item for access. It is becoming a variable expense that engineering leaders have to monitor, approve, and explain.

The sticker shock is already doing the rounds. Users on Reddit and X have pointed to examples where a monthly bill that once sat around $29 could climb to nearly $750, while another case jumped from roughly $50 to about $3,000. Whether those numbers reflect ordinary work or inefficient “vibe-coding” is part of the debate, but the larger signal is hard to miss: once AI use is metered, enthusiasm can turn into accounting.

That is the practical lesson for monday.com teams. Engineers using AI to speed up code generation will need to think not just about whether the tool saves time, but whether the token cost is lower than the labor it replaces. If a workflow burns through usage quickly, the productivity gain can disappear as soon as finance starts asking for a breakdown. For product managers, the shift is a reminder that buyers do not evaluate AI on model quality alone. They also want predictability, especially when a feature moves from novelty to habit and starts touching every sprint.

Sales teams will feel the same pressure from the other side of the table. Enterprise buyers are increasingly likely to ask where the guardrails are, whether there are quotas, and how pricing behaves when usage scales. A product can look inexpensive during early adoption and then become politically difficult once the bill lands in procurement. That is especially true for workplace software that encourages heavier automation across teams.

The bigger risk is trust. When vendors invite broad use and then change the pricing mechanics underneath customers, the reaction is not just annoyance. It is a recalculation of whether the tool is a helper or a liability. For monday.com, which is expanding AI across workflows, the message is straightforward: if the product encourages more automation, the commercial model has to stay legible. Otherwise, customers will start treating usage as something to control, not something to celebrate.