AI FinOps Glossary
Definitions for AI spend management terms: token maxing, context tax, cost per outcome, AI spend attribution, and more.
Clear definitions for terms used in AI spend management, LLM cost attribution, and AI FinOps. Some of these terms are still new — we define them here as clearly and neutrally as possible.
Terms
- Token maxing — using an expensive frontier model for a task a cheaper model handles equally well.
- Context tax — the recurring cost of sending a large, mostly-static system prompt with every API call.
- Cost per outcome — measuring AI spend against the business result it produced, not just raw token usage.
- AI spend attribution — tracing AI API costs back to the feature, team, or department that generated them.
- Agentic cost simulation — projecting what an AI agent workflow will cost at scale before rolling it out company-wide.
- AI Efficiency Score — cost per business outcome delivered, tracked month over month to show whether AI spend is getting more or less efficient.
- OpenAI proxy — a server between your app and OpenAI's API that adds cost attribution and budget enforcement OpenAI's own API doesn't have.
- LLM cost tracking — measuring what your app spends on AI providers, broken down by feature or team, not just one invoice total.
See also: Waste Detection · Attribution Headers · Outcomes & ROI
Related articles
What is Token Maxing?
Token maxing is the practice of using expensive frontier AI models for tasks that cheaper models handle equally well.
What is Context Tax?
Context tax is the recurring cost of sending a large, mostly-static system prompt with every API call.
What is Cost Per Outcome?
Cost per outcome measures what it costs to achieve a specific business result with AI, not just raw API spend.