Inspect usage and spend

Inspect usage

1. Open Gateway → Usage for the intended organization. Captured reporting requires Cortex Atlas and the visibility granted to your account.
2. Select 7D, 30D or 90D. Check the reporting window shown beside the chart before comparing totals.
3. Select Filters. Under Group results by, choose Model, Provider, Team or Member.
4. Choose a Time bucket and any Team, Member, Model or Provider filters. Apply the filters and confirm the selected chips above the results.
5. Read Estimated spend, Captured activities, Input tokens and Output tokens for the same filtered population. Changing the grouping reorganizes that population; changing a filter changes which activity is included.
6. Compare a high-spend group with another using the same range, filters and pricing coverage. Check AI spend before treating a captured estimate as an organization total.

An empty result means no matching captured activity was returned for this view. An unavailable result is a read failure; retry rather than treating it as zero. Sample views use a fixed sample population and disable the live filters.

Understand estimated spend

Estimated spend prices captured usage using available model rates. The published model catalog contains base token rates; it does not establish the exact enabled models or prices for every organization.

Confirm the organization's pricing configuration, its effective time, and the treatment of input, output, and cache usage. A missing usage field or price can leave activity unpriced. Keep unpriced activity visible rather than treating its cost as zero.

Comparing provider charges

A configured contracted rate does not guarantee that Cortex's estimated total matches a provider invoice. Complete provider reconciliation requires the same usage coverage, pricing version, currency, time window, adjustments, and treatment of retries or incomplete requests.

Captured spend is not the complete organization bill or an authoritative admission balance. See AI spend.