Glossary
AI cost & FinOps glossary
The vocabulary of controlling AI spend. Each definition stands on its own.
- AI FinOps
- AI FinOps is the practice of measuring and attributing AI spending across seat subscriptions, API tokens, and infrastructure, so finance and engineering share one view of cost and value. It extends cloud FinOps to AI's mix of per-seat and metered pricing. Spendassay runs the audit side, surfacing and pricing recoverable waste.
- Shadow AI
- Shadow AI is AI tooling employees adopt without IT or finance approval — personal ChatGPT accounts, unsanctioned Copilot seats, API keys on corporate cards. It creates untracked spend, security exposure, and duplicate tools. Spendassay detects it from SSO and OAuth grants plus corporate-card spend, then prices the overlap and unsanctioned seats as findings.
- Token cost
- Token cost is the price of the input and output tokens a model processes, billed per million tokens at rates that differ by model and vendor. It makes API spend variable and hard to forecast. Spendassay reads token usage as metadata, never message content, and prices token anomalies as findings.
- Cost attribution
- Cost attribution assigns AI spend to the team, project, or cost center that incurred it, turning one vendor invoice into per-owner accountability. Without it, a single Anthropic or OpenAI bill hides who spent what. Spendassay attributes spend at the team level, never individual league tables, using connector metadata and corporate-card data.
- Chargeback vs showback
- Chargeback bills each team's AI spend back to its own budget; showback reports the same per-team cost without moving money. Chargeback drives accountability but adds friction; showback informs without enforcing. Most teams start with showback. Spendassay produces the team-level cost breakdown either model needs, with the proof level behind every number.
- Unit economics of AI
- Unit economics of AI expresses AI cost against a unit of output, such as cost per merged PR, per ticket resolved, or per active user, so spend is judged by what it produces rather than its raw dollar total. It reframes AI as an investment, not overhead. Spendassay builds these ratios into a CFO-credible AI cost report.
- Seat utilization
- Seat utilization is the share of paid tool licenses actually used in a period — active seats divided by purchased seats. Low utilization means you pay for access no one uses. Spendassay measures it from connector activity and flags unused seats (no activity) and underused seats (trailing-30-day low use) as dollar-priced findings.
- AI spend audit
- An AI spend audit is an independent, cross-vendor review of what an organization pays for AI and what that spend returns. It inventories seats, tokens, and infrastructure, then prices the waste. Spendassay is buyer-side and grades no vendor's own homework; it targets 10 to 15% of seat and subscription spend as recoverable within 30 days.
- Cost per merged PR
- Cost per merged PR divides AI tooling spend by the number of pull requests merged in a period, giving one unit-economics view of whether AI coding tools pay for themselves. It is directional, not a productivity verdict. Spendassay can express this kind of ratio in its team-level AI cost report, with the proof level and formula shown.
- Model and tool sprawl
- Model and tool sprawl is the unmanaged accumulation of overlapping AI products, like three coding assistants or two chat tools, each with its own bill and contract. It inflates cost and fragments usage data. Spendassay detects overlapping tools across connectors and prices the redundancy as a tool-overlap finding your admins can consolidate.
- AI cost report
- An AI cost report is a profit-and-loss view of AI: total spend on one side, the output and recoverable waste it maps to on the other, built to be credible to a CFO. It replaces scattered invoices with one accountable picture. Spendassay produces this cross-vendor AI cost report, with every number carrying its proof level.
- Recoverable spend
- Recoverable spend is the portion of AI cost you can reclaim without losing capability — unused seats, underused licenses, tool overlap, and renewal overpayments. It is the target an audit sets and then tracks to banked savings. Spendassay surfaces 10 to 15% of seat and subscription spend as recoverable within 30 days and tracks the recovery.
- Renewal opportunity
- A renewal opportunity is an upcoming contract renewal where usage data supports cutting seats to the size you need or negotiating a better rate before you re-sign. Timing matters: negotiating power peaks about 60 days out. Spendassay surfaces renewals early with a negotiation package: current spend, a cited public benchmark, and the seat count your own usage justifies.
- Tool overlap
- Tool overlap is paying for two or more AI products that do the same job — for example Copilot and Cursor for the same engineers. It signals consolidation savings without capability loss. Spendassay detects overlap across connectors, prices the cheaper redundant seats of the low-usage pair, and hands admins consolidation steps.
- Metered pricing trap
- A metered pricing trap is usage-based billing, charged per token or per request, where cost scales silently with adoption and no fixed cap warns you. A small pilot can become a large bill. Spendassay tracks metered spend as metadata and flags token anomalies and price drift as findings.
- Proof level
- A proof level rates how strongly a finding is proven: Counted (from observed activity) beats Compared (versus a peer benchmark) beats Estimated (from assumptions). It tells a CFO which numbers are hard facts and which are projections. Spendassay tags every finding with its tier and exposes the formula behind it.
Put the terms to work
Spendassay is the independent audit that turns these into a CFO-credible AI cost report. Free to start.