SLATEMOTH / INSIGHT
Why AI billing needs a ledger, not just token counters
Why production AI billing benefits from separating usage measurement, pricing, balance movement and settlement.
The core distinction
Usage, price and money movement are different facts. A request can produce usage before its final commercial price is known, and a balance adjustment can occur without a model request.
From usage to money
A robust flow can be expressed as measurement → pricing → billing event → ledger entry → settlement. Each stage should have an explicit input and output.
Why a ledger matters
A ledger provides an auditable record of balance-affecting events. Corrections can be represented as new entries rather than rewriting history.
Provider cost and customer price
The amount charged by a model provider and the amount charged to a customer are separate economic facts.
Operational invariants
Idempotency, immutable financial events, explicit currency, deterministic rounding and traceable links back to usage records are useful invariants.