Understand estimated AI spend
Understand estimated AI spend by provider, team, or workstream and interpret changes with coverage in view.
Read the cost breakdown
Use the same reporting window, visibility and filters when comparing amounts. Open the provider, model, person or work item behind an outlier, then inspect its captured sessions.
What AI spend measures
Estimated model cost covers captured activity with available usage and pricing, not a provider invoice or the organization's complete AI bill. Unknown/unpriced usage is not zero. For example, $0.42 plus $0.18 gives $0.60 available cost; an unpriced session still counts as activity.
The screenshot compares $120 with $96 for its displayed periods, with 100% pricing coverage in each. Those estimates cover different sessions from the example above. Cost alone does not establish savings or ROI.
Time windows and work-unit rollups
Selected-range totals describe the selected captured activity. A work list filters units by latest activity, while detail uses a 30-day cost rollup. A short activity window does not necessarily recalculate a unit's cost for that window. Do not add overlapping groupings as distinct activity.
Use the controls
Date, provider, team and workstream filters change what is counted. A breakdown regroups that population. State the range and filters beside a figure you share.
Interpret a change
Inspect retries, context length and tool calls behind a spike. Follow work units or user profiles to the evidence and confirm coverage stayed comparable.
Partial period means the current-period estimate is incomplete; it does not itself mean pricing coverage is partial.