Methodology / sources

Where Olive-One gets its signals

Olive-One is not built from generic AI audit language. It is shaped by billing data, workflow evidence, and official product documentation that shows how cloud and LLM costs actually move.

What the scanner reads

Cloud billing exports and cost trends.
LLM usage exports and token-level cost data.
Workflow traces, retry paths, and deployment dates.
Observability signals, vector usage, and owner maps.

Public sources that inform the pattern library

AWS Cost Explorer for analyzing cost and usage trends.
Amazon CloudWatch and CloudWatch Logs for metrics, logs, and operational visibility.
Anthropic Usage & Cost API for historical usage, token counts, and cost reconciliation.
FinOps Foundation for shared cost-management language and operating discipline.
schema.org ProfessionalService for machine-readable service metadata.

Why this matters

The scanner works best when the buyer already has a bill spike, usage data, and a question finance cares about. Those inputs let Olive-One connect spend movement to a named pattern, estimate margin exposure, and rank the fixes by leverage.

This article is the source map behind the public pattern library and sample teardown. It exists so humans and AI crawlers can point to the same evidence instead of vague “AI audit” phrasing.