glossary · term
AI Operating Ledger
My framework for measuring what enterprise AI returns. It accounts for the full path from request to trusted result: inference, context preparation, review, rework, escaped defects, evaluation, governance, and institutional memory. Its metric is value per verified outcome, which counts work only once it has been checked against the intent.
in practice
How I use the term
The model invoice is the visible line, and many dashboards stop there. A high-token session that produces a verified architectural migration may be cheap. A low-token session that creates a subtle security flaw may be expensive.
Verified means the output matched the intent, the constraints held, and a competent human or automated gate checked the result against something more durable than vibes.
You can't measure verified outcomes without a verifier. So the ledger leads to gates: checks that run every time, with people writing the standard and deciding whether it's good enough.
in the essays · 5
Where it shows up
- The AI Operating Ledger
The framework: eight ledger lines and one metric.
- The Subsidy, Itemized
A baseline from 26 days of my own Claude Code usage.
- Trust the Gate, Not the Actor
Why the ledger needs a verifier, from one I built.
- The Instrument and the Ledger
The hand-built ledger becomes an afternoon of tooling.
- Where the Saved Hour Goes
The time side of the same ledger.