OpenAI

Setting usage limits for custom roles in ChatGPT Enterprise

Usage limits (spend controls) let you control how your workspace’s credit pool is used across different teams and roles, preventing unexpected over-spend while still giving your AI champions enough room to experiment and drive adoption.

Updated: 18 hours ago

Overview

Usage limits are part of Role-Based Access Controls (RBAC) in ChatGPT Enterprise, applying on a per-user, per-week basis. They let workspace owners:

  • Set a default usage limit for the whole workspace, and

  • Set a specific limit for each custom role (for example, “Power users – Engineering” vs “Light users – Contractors”)

This helps you:

  • Avoid unexpected spikes in advanced model spend

  • Prevent blocking heavy users who create most of the value

  • Use your shared credit pool efficiently across teams and user groups

See RBAC for more information on custom roles and groups.

Who can configure usage limits?

  • Available plans: ChatGPT Enterprise/Edu (and any plan with RBAC and custom roles).

  • Who can change settings: Only workspace owners are allowed to configure limits. Admins and members cannot configure limits.

Configure usage limits by access Workspace settings Permissions & roles and locating the Usage limits section. By default, we set the usage limit to None. Use the drop down to select recommended Low, Standard or High limits, or select a custom limit for a given role.

Types of limits

When selecting a custom limit for a role, you can set:

  • Admin alert

  • Hard cap

  • Recommended practice

  • Use Admin alerts as the default during early and mid AI adoption.

  • Use Hard caps only in exceptional cases, because strict caps can reduce early productivity gains and experimentation.

How we calculate recommended limits

  • Data sources: We use anonymized, aggregated usage data from many Enterprise/Edu workspaces. Workspaces are grouped by similar monthly credit tiers, size (users), and overall adoption patterns.

  • What we look at: We analyze total monthly credit consumption for similar workspaces, focusing on how usage is distributed across power users (~10% highest-usage), median, and light users. We also track how these patterns change over time as AI adoption grows.

  • How recommendations are derived: For each similar workspace segment, we measure typical per-user monthly usage, paying close attention to power user consumption. We then compute a weekly per-user baseline for each credit tier that accommodates power users, assumes most users will stay below the limit, and aims to keep total usage near the purchased monthly credits in typical scenarios.

  • Recommended weekly limits are typically 4–5× higher as this gives power users enough headroom, while still keeping the expected total usage within your monthly credit pool under normal patterns.

Example: why a higher weekly limit can still be safe

  • Workspace

  • Observed usage pattern

User segmentShare of usersWeekly usage per user (credits)Weekly total (credits)
Power users10% (10 users)4004,000
Typical users40% (40 users)1204,800
Light users50% (50 users)201000
Total100 users9,800

Outcome

  • Weekly usage ≈ 9,800 credits

  • Monthly usage ≈ 39,200 credits — still within the 40,000-credit pool

Even with a 400-credit weekly limit for power users (4× the naive 100-credit limit), the workspace stays within the overall pool because most users consume much less than their limit

FAQs

Can members see their limits and change them?

No, members cannot view or edit limits. They only see a generic message if they hit a hard cap. Note that admins cannot change limits - only workspace owners can.


When do usage limits reset?

The weekly spend limits reset on a rolling basis. This weekly window starts the moment a member sends their first message that consumes credits.


What happens if a user changes roles mid-week?

The new role’s limits apply immediately after the role change. Any usage already consumed in the week still counts toward the new role’s limit.


What exactly does a user see when they hit a hard cap?

Users see a generic block message indicating they’ve reached a usage limit. The message does not reveal credit amounts, thresholds, or role details. They may continue using features that don’t draw from the advanced credit pool (subject to plan limits).


Do admin alerts notify users when they’re close to a limit?

No. Admin Alerts are admin-only and delivered via weekly email digest. Users are not proactively warned before hitting a hard cap.


What if a user has multiple custom roles assigned?

If a user is assigned multiple custom roles, the most permissive (least restrictive) limits across those roles apply. For example, if Role A has an admin alert at 500 credits and Role B has a hard cap at 600 credits, the user will be subject to only the admin alert at 500 credits, with no enforced hard cap. If any assigned role has spend control set to None (no limit), that role becomes the most permissive and the user will effectively have no enforced weekly credit limit. Make sure that even experimental or feature-test roles include an appropriate limit if you want the user to remain capped.


How do I check a user's usage limits?

Only admins and owners can check a user's usage limits. Navigate to Workspace settings > Members, and click on the member name. In the resulting sidebar, check the Usage limits section.

Was this article helpful?