Cost per Agent Cycle
Definition
Measures the cost of one agent reasoning cycle focusing on agentic workloads where multi-step reasoning drives cost enabling Engineering, Product, and FinOps Practitioner Personas a metric for tracking spending across complex agent workloads. This KPI excludes simple single-call workloads.
Formula, as published
Cost per Agent Cycle = Total AI Spend / Agent Span Count
- Where:
- A “span” is a granular monitoring technique that breaks down an agent’s execution into specific internal segments that can be used to track the multiple-reasoning steps used by an agent. It is the building block of a trace.
- Data sources:
- Cloud Cost Billing Data Usage Telemetry (eg. Azure Cognitive Services metrics) LLM Observability/Application Telemetry
Calculate it
Example values to start with; replace them with yours.
From the FinOps Foundation
More Budgeting KPIs
Definition and formula: Cost per Agent Cycle, FinOps KPI library, FinOps Foundation, licensed CC BY 4.0, read 16 Oct 2026. Changed: HTML removed, dashes and British spellings changed, formula split from its data sources and examples. The calculator and the notes are Finitizer’s.
Measured once, a KPI is a number. Tracked daily, it is a practice.
Finitizer computes the KPIs your billing data supports every day, by team and product, across AWS and Google Cloud.
See how Finitizer works