Guide
How to Track AI Spend Across Teams
To track AI spend across teams, connect a spend source (corporate cards via Ramp, plus vendor invoices) and a usage source (GitHub, Cursor, OpenAI, Anthropic, Okta). Spendassay maps every dollar to a team, then surfaces and prices waste as findings. Your admins act on the step-by-step instructions Spendassay generates.
Why is AI spend so hard to track across teams?
AI spend hides across seat subscriptions, API tokens, and infra, bought on different cards and vendor portals. Shadow AI adds tools no one approved. Vendor dashboards only show their own usage, so no single view ties spend to the team that incurred it or to what that spend produced.
A mid-market engineering org typically runs Copilot, Cursor, ChatGPT, and Claude seats alongside OpenAI and Anthropic API tokens. Each vendor reports its own numbers in its own console.
Seats get provisioned and forgotten. Tokens spike without an owner. Two tools often cover the same job. None of this is visible from any one vendor's dashboard.
Spendassay is the independent, cross-vendor audit that consolidates these sources and attributes every dollar to a team, producing a CFO-credible AI cost report.
What sources do you connect to track team AI spend?
Connect one spend source and one usage source. Spend comes from Ramp corporate cards and vendor invoices. Usage comes from GitHub, Cursor, OpenAI, Anthropic, Okta, and Google Workspace. CSV import works with zero live integrations, so you can start from an export.
Spend sources price what you pay: card charges and vendor subscription invoices. Usage sources show who actually used what, at the team level.
Connectors available: GitHub, Anthropic, OpenAI, Cursor, Ramp, Okta, Google Workspace, plus CSV import. All access is read-only and metadata-only.
Spendassay never reads source code, prompts, or message content, and the audit never writes to or throttles any vendor system. Okta and Google Workspace grants also let it detect shadow AI from SSO and OAuth activity.
How does Spendassay attribute spend to the right team?
Spendassay joins spend records to usage records, then rolls both up by team. It reports at the team level only, never as individual league tables. Every finding names the team, the dollar amount, and the proof level behind the number.
Seat spend maps to the team holding the seats. Token spend maps by the account or project generating the calls. Card charges for shadow tools map by the corporate-card owner and SSO grant.
Findings are dollar-priced: unused seats, underused seats, shadow AI, tool overlap, renewal opportunities, and quality-trend regressions.
Every number carries a proof level, ranked Counted, then Compared, then Estimated, and exposes its underlying formula so finance can trace it.
How do you turn tracked spend into recovered dollars?
Spendassay sets a recoverable target, typically 10 to 15 percent of seat and subscription spend within 30 days, then attaches a set of step-by-step instructions to each finding. Your own admins execute the step-by-step instructions. Recovered savings post to a banked "Recovered" statement so finance sees realized dollars, not projections.
A finding like unused Copilot seats on one team is priced at its annual subscription cost, then comes with a set of step-by-step instructions your admin follows in the vendor console.
For renewals, Spendassay produces a negotiation package with a cited public-list benchmark and drafted asks for the vendor account manager.
Spendassay surfaces, prices, and tracks the recovery. It does not act on your systems. Recoveries verify when the finding closes on a later sync, then post to the Recovered statement.
Step by step
- 1
Connect a spend source
Connect Ramp for corporate-card charges and add vendor invoices for seat subscriptions. No live integration? Upload a CSV export. This tells Spendassay what each team pays across Copilot, Cursor, ChatGPT, Claude, OpenAI, and Anthropic.
- 2
Connect a usage source
Connect GitHub, Cursor, OpenAI, Anthropic, Okta, or Google Workspace so Spendassay sees who actually uses each tool. Access is read-only and metadata-only. Okta and Google Workspace grants also surface shadow AI from SSO and OAuth activity.
- 3
Review findings by team
Open the dashboard to see dollar-priced findings rolled up per team: unused seats, underused seats, shadow AI, tool overlap, renewal opportunities, and quality regressions. Check each finding's proof level and open its formula to verify the number.
- 4
Act on the step-by-step instructions
For each finding, hand the attached step-by-step instructions to the responsible admin, who executes it in the vendor console. For renewals, use the negotiation package with its cited benchmark. Spendassay generates the step-by-step instructions; your admins take the action.
- 5
Track recovery on the Recovered statement
On the next sync, closed findings verify as banked dollars and post to the Recovered statement. Use it to report realized savings against the 10 to 15 percent recoverable target to your VP Engineering and CFO.
Common questions
Does Spendassay reduce or throttle our AI spend automatically?
No. Spendassay's audit is read-only and never writes to, routes, or throttles vendor systems. It surfaces and prices the waste, sets a recoverable target, and tracks recovery. Your own admins execute the step-by-step instructions and renewal negotiations that realize the savings.
Can we track AI spend without connecting live integrations?
Yes. CSV import works with zero live integrations. Upload spend and usage exports and Spendassay attributes them to teams and produces the same dollar-priced findings. You can add GitHub, Ramp, Okta, and vendor connectors later for continuous, automatic tracking.
Does tracking usage mean Spendassay reads our code or prompts?
No. Access is metadata-only. Spendassay never reads source code, prompts, or message content. It uses seat counts, usage signals, token metadata, and card and invoice records to attribute spend and price waste at the team level.
How much of our AI spend can we expect to recover?
Spendassay surfaces 10 to 15 percent of seat and subscription spend as recoverable within 30 days. The exact figure depends on your findings, such as unused and underused seats, tool overlap, and renewal opportunities. Realized dollars post to the Recovered statement as your admins act.
Run the audit on your own AI spend
Connect read-only sources and get a CFO-credible AI cost report. Free to start, about 10 minutes to connect.