Gong documents per-user and workspace controls for AI credit usage

Gong exposes a clearer operating model for AI credits

Gong's credits documentation, updated September 27, describes credits as usage units for selected AI-powered features that perform heavier processing or run autonomously. The page lists question-based AI Trackers, APIs for AI Ask Anything and AI Briefer, the MCP server and deep research among the features that can consume credits. Gong says credits are shared at the company level rather than assigned only to individual users.

The documentation also describes administrative controls around that shared pool. Admins can monitor balance and usage trends, receive notifications at defined usage thresholds, set monthly limits for user-initiated credit actions and allocate portions of the pool to workspaces. These are vendor-described controls; they do not establish that one credit corresponds to a standard unit of business value across features or companies.

Sources: Gong: About Gong credits

User limits and background processing are different boundaries

Gong says a monthly user limit caps credits a team member can consume through credit-based actions that person initiates, such as Gong Enrich. The same page notes that company-wide background features are not counted against individual monthly limits. A user can also be blocked when the shared company pool is exhausted even if that person's monthly limit still has room.

That distinction matters for RevOps because a per-user control is not a complete system-level budget. Interactive research, enrichment or MCP requests may be attributable to a person while trackers or other background processing continue on a different cadence. Teams should map recurring consumers separately from ad hoc usage and assign a named owner to each automated feature that can draw from shared capacity.

Running out of credits can become a data-freshness event

The documentation says that when all credits are consumed, data processing stops for credit-based features while existing data remains available. It also says question-based AI Trackers stop processing new conversations, APIs and MCP requests requiring credits return errors, automated briefs stop generating or updating, and dependent features may continue displaying existing data that becomes outdated over time.

A revenue workflow can therefore remain visually intact after its evidence stops refreshing. RevOps should define a freshness rule for any decision that depends on credit-based processing. The operator needs to know when the underlying evidence was last processed, whether the feature is currently active and what fallback applies if the result has crossed the approved age threshold.

The unit is useful for capacity, not a universal productivity benchmark

Gong publishes current consumption examples for different data types, including email, call and web-search processing. Those rules help admins understand how the product consumes its own credits. They should not be compared directly with model tokens, another vendor's credits or a generic agent action, because the underlying processing, commercial entitlement and business consequence differ.

A better local measure is usage per verified business outcome inside one named workflow. A team can track the credits consumed by a tracker or research process and separately count the reviewed signals, accepted briefs or verified actions it supported. That keeps capacity planning connected to business evidence without inventing an external benchmark that the vendor documentation does not provide.

Workspace allocation adds another governance layer

Gong also documents workspace allocation for distributing the shared pool. Credits assigned to a workspace create a ceiling for that workspace, while unallocated credits remain in the company pool and are unavailable until assigned. Future credit additions can follow percentage-based allocation rules. The feature matters for organizations where business units or teams should not compete invisibly for the same shared resource.

RevOps should pair allocation with workflow ownership. A workspace ceiling is meaningful only when the organization knows which recurring processes live there, which teams depend on them and which outputs become stale when capacity is exhausted. Review the allocation after material workflow changes rather than treating the first split as a permanent forecast.

What to verify before making credits part of a production workflow

Inventory each credit-consuming feature that supports a recurring revenue process and classify it as user-triggered or background. Record the business owner, current limit or allocation, alert threshold, expected workflow output and fallback. For evidence used in forecast, prioritization or customer work, add a last-processed or checked-at field that an operator can inspect before acting.

Then test the low-capacity path. Confirm what the user sees, which API or MCP requests fail, whether background processing pauses automatically and what manual step is required to resume it. Gong has made the product behavior more explicit in its current documentation. The RevOps task is to translate those controls into a workflow contract that stays understandable when capacity becomes constrained.

Gong currently says each paid core seat includes 2,000 annual credits in the shared pool; separately purchased credits expire at the end of the contract term. That packaging is subject to change and is not a promised capacity for every customer. An operator should check the actual entitlement, scheduled processing load and workspace allocation before setting a local alert. Gong lists notifications at 60, 80, 90, 95 and 100 percent consumption, but those notices are useful only when the organization knows which dependent workflow to pause or inspect at each threshold.