AI Limit Approvals

The problem

AI usage was growing in every department, and everyone had the same monthly Claude limit. Hitting it stopped work mid-task. Department heads knew whether the work justified more, but they had no view of what their team used or who was close to the cap.

The solution

A self-serve limit approval tool. Anyone who hits their cap types one command in the team chat with the new limit and a reason, then picks when it expires: end of month, permanent, or a custom date. Their department head sees every teammate's limit, spend, utilization, 30-day trend, and models used in one portal, and approves in a click. I designed it with the Director of IT, who wanted increases to expire by default.

How it works

1. Per-user Claude cost syncs daily from the provider's admin API into a snapshot that feeds the portal.
2. Each department head approves only for their own team, from an explicit approver list rather than job titles.
3. Requests start in the team chat: a slash command with the new total and a reason, then a pop-up to pick an expiry.
4. Increases are time-boxed by default, so a one-off need doesn't become a permanent limit.
5. Every change is logged with who asked, why, how much, and until when.

The results

• Every department is live on real spend data.
• Raising a limit went from a blocked task to a chat command and a one-click approval.
• Department heads see their own team's AI usage in one place.
• Every override has an owner, a reason, and an expiry.
• Part of the program that took AI tooling to 97%+ adoption.