Guide
How to attribute AI costs to teams (chargeback and showback)
Attribute AI costs to teams by connecting a spend source (Ramp or CSV) and a usage or identity source (GitHub, Okta, Google Workspace, Anthropic, OpenAI). Spendassay maps seat subscriptions and token spend to teams, builds a team-level AI cost report, and prices the waste. Your finance team then runs showback or chargeback from that allocation.
How do you attribute AI costs to teams?
Connect one spend source and one usage or identity source. Spendassay allocates seat subscriptions and API token spend to teams using your directory and GitHub team membership. It reports each team's AI cost, usage, and dollar-priced waste at the team level. Showback reports the allocation; chargeback bills it back to team budgets in your finance system.
Attribution starts with two inputs. The spend source tells Spendassay what you paid. The usage or identity source tells it who used what and which team they belong to.
Spendassay is READ-ONLY and METADATA-ONLY. It reads seat rosters, SSO and OAuth grants, corporate-card line items, and token usage totals. The audit never reads source code, prompts, or message content, and it never writes to vendor systems.
Every allocated dollar carries a proof level: counted, compared, or estimated. Measured attribution comes from real usage and roster data. Modeled attribution is labeled as modeled and exposes its formula, so a CFO can trust the split.
The output is a team-level AI cost report: cost per team, usage per team, and the recoverable waste inside each team's line.
What is the difference between chargeback and showback?
Showback reports each team's AI cost without moving money; it drives awareness and behavior. Chargeback goes further and bills the allocated cost back to each team's budget. Spendassay produces the attribution both models need. Showback uses the report as-is. Chargeback feeds the same allocation into your finance system, which posts the internal charge.
The difference is whether money actually moves between internal budgets.
| Dimension | Showback | Chargeback |
| --- | --- | --- |
| Money moves | No | Yes, to team budgets |
| Main goal | Visibility and behavior change | Cost accountability and budgeting |
| Rollout friction | Low | Higher, needs finance buy-in |
| What Spendassay supplies | Team-level AI cost report | Same allocation, exported for billing |
Most teams start with showback to build trust in the numbers, then graduate to chargeback once the allocation is stable. Spendassay does not post internal charges. It produces the audited allocation; your finance team runs the billing step in its own system.
What data does Spendassay need to attribute AI spend to teams?
Spendassay needs one spend source and one usage or identity source. Spend sources: Ramp corporate cards, seat invoices, or CSV import. Usage and identity sources: GitHub, Okta, Google Workspace, Anthropic, and OpenAI. CSV import works with zero live integrations, so you can attribute costs before wiring up any connector.
Spend sources capture the dollars: Copilot, Cursor, ChatGPT, and Claude seat subscriptions; OpenAI and Anthropic API token spend; and shadow AI charges detected from corporate-card line items.
Usage and identity sources capture who used what and their team. GitHub supplies team membership and Copilot activity. Okta and Google Workspace supply SSO grants and directory groups. Anthropic and OpenAI supply token usage by account.
You do not need every connector. One spend source plus one identity source is enough to produce a team allocation. Each connector you add raises the proof level from modeled toward measured.
CSV import covers vendors with no live integration, so a finance-only team can still attribute spend on day one.
How does Spendassay map people and spend to the right team?
Spendassay resolves each person from SSO and directory data (Okta, Google Workspace) and GitHub team membership, then attaches their seat cost and token spend to that team. All reporting is team-level. Spendassay never builds individual league tables. Unattributed or shadow AI spend surfaces as its own finding instead of being silently spread across teams.
Identity resolution joins seat rosters, SSO grants, and directory groups so one person maps to one team even when they appear across several vendors.
Seat costs attach to the seat holder's team. API token spend attaches by account. Shadow AI, detected from SSO or OAuth grants plus card spend, is flagged rather than absorbed into an existing team's number.
Reporting stops at the team. Spendassay is designed for VP Engineering, CTO, and CFO buyers who need team accountability, not individual surveillance.
When a cost cannot be attributed with confidence, it is shown as unattributed with its proof level, so the AI cost report never hides a guess as a fact.
How do you turn team attribution into recovered dollars?
Once spend is attributed, Spendassay prices the waste inside each team: unused seats, underused seats, tool overlap, shadow AI, and renewal opportunities. It sets a recoverable target and hands your admins a set of step-by-step instructions to act. Spendassay surfaces and tracks recovery; your admins reclaim seats and renegotiate. A banked Recovered statement records what landed.
Attribution is the setup. The payoff is dollar-priced findings tied to specific teams: idle seats, underused seats, overlapping tools doing the same job, and shadow AI.
The audit sets a recoverable target. The north-star benchmark is surfacing 10 to 15 percent of seat and subscription spend as recoverable within 30 days.
Each finding ships with a set of step-by-step instructions: the exact seats to reclaim, the renewal to renegotiate, or the tool to consolidate. Your own admins execute those steps in the vendor consoles. The audit never throttles usage or writes back.
As actions complete, recovered dollars are banked into a Recovered statement, so showback and chargeback conversations rest on realized savings, not projections.
How accurate is team-level AI cost attribution?
Accuracy depends on which sources you connect, and Spendassay is explicit about it. Every attributed number carries a proof level: Counted beats Compared beats Estimated. Each figure exposes its formula. Connecting more usage and identity sources moves attribution from Estimated toward Counted, so you can defend the split to finance and to each team lead.
Attribution is only as good as its inputs, so Spendassay grades every number instead of presenting one blended figure.
Measured: derived from real roster and usage data, the highest confidence. Cohort-delta: inferred by comparing similar teams. Modeled: estimated from a stated formula when direct data is missing.
Because the formula is always visible, a CFO can audit any team's allocation line by line. This is what makes the attribution CFO-credible rather than a black-box split.
To raise accuracy, connect an identity source alongside your spend source. GitHub, Okta, or Google Workspace turn a modeled allocation into a measured one.
Step by step
- 1
Connect a spend source
Connect Ramp for corporate-card AI charges, or import seat invoices and card statements by CSV. This tells Spendassay what you paid across Copilot, Cursor, ChatGPT, Claude, and API token spend. CSV import works with zero live integrations.
- 2
Connect a usage or identity source
Connect GitHub, Okta, Google Workspace, Anthropic, or OpenAI. These map each person to a team via directory groups and GitHub team membership, and supply usage totals so seat and token costs attach to the right team.
- 3
Review the team-level AI cost report and findings
Open the team AI cost report to see cost, usage, and dollar-priced waste per team: unused seats, underused seats, tool overlap, shadow AI, and renewal opportunities. Check each figure's proof level and formula before you act.
- 4
Choose showback or chargeback
For showback, share the team AI cost report to drive visibility and behavior. For chargeback, export the same allocation into your finance system and post the internal charge to each team's budget. The underlying attribution is identical.
- 5
Act on the waste via step-by-step instructions
For each priced finding, follow the step-by-step instructions. Your own admins reclaim idle seats, consolidate overlapping tools, or open a renewal negotiation using the provided package. The audit never writes to or throttles vendor systems.
- 6
Track recovery with the Recovered statement
As admins complete step-by-step instructions steps, banked savings post to the Recovered statement. Use it to close the loop on each team's allocation and to report realized recovery against the 10 to 15 percent target.
Common questions
Does Spendassay bill teams automatically for their AI costs?
No. Spendassay attributes and prices AI spend at the team level and produces the allocation. It does not move money. For chargeback, your finance team exports that allocation and posts the internal charge in its own system. Spendassay's audit is read-only and never writes to vendor or finance systems.
Can I attribute AI costs to teams without connecting any live integrations?
Yes. CSV import lets you attribute spend from invoices and card statements with zero live connectors. Adding an identity source such as GitHub, Okta, or Google Workspace raises the proof level from modeled toward measured, which makes the team allocation more defensible to finance.
Does Spendassay show individual employee AI usage?
No. All attribution and reporting is team-level. Spendassay never builds individual league tables. It is designed for VP Engineering, CTO, and CFO buyers who need team accountability, not per-person surveillance.
How do I know a team's attributed cost is accurate?
Every attributed number carries a proof level (counted, compared, or estimated) and exposes its formula. You can audit any team's allocation line by line. Connecting more usage and identity sources moves attribution from modeled toward measured.
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.