5 Revenue Leaks in Usage-Based Billing for SaaS and AI

API calls, AI tokens, inference workloads, agent actions, transactions, and other consumption events originate outside the billing system.

Before they can become charges, those events have to be captured, validated, associated with the right customer and service, and prepared for rating.

Events can be dropped during ingestion. The same event can arrive more than once and be processed again. Late events can miss a billing period. An event can also arrive without the customer, product, or usage attributes needed to price it correctly.

AI makes this particularly important. A single user request can trigger multiple model calls, tool invocations, or agent actions. What looks like one customer interaction may produce many underlying consumption events.

Billing cannot charge for an event it never receives.

Capturing usage correctly does not guarantee revenue accuracy.

An AI event might be priced according to the model used, token type, customer contract, volume tier, or commitment. An API transaction may depend on the endpoint, customer tier, included allowance, and overage rate.

That means two customers can consume exactly the same quantity and legitimately owe different amounts because their contracts, commitments, discounts, or pricing tiers are different.

Revenue leaks when the usage event is correct but the wrong commercial rule is applied. An expired rate may remain active. The wrong volume tier may trigger. A discount may persist beyond its agreed term. Usage may map to the wrong product or customer-specific rate.

This becomes more important as SaaS and AI companies move beyond simple subscriptions into combinations of recurring, usage, credit, commitment, and outcome-based pricing.

Hybrid models are increasingly useful for SaaS and GenAI because they combine predictable committed revenue with flexible consumption.

A commercial model might look like:

That may be simple for the customer to understand, but behind the scenes the contract, entitlement, usage, credit balance, and billing all have to remain synchronized.

A customer can consume the service correctly while the wrong credit pool is depleted. An overage may fail to trigger. Credits may expire incorrectly. Usage may continue beyond an agreed limit. Or a customer may be charged for consumption that should have been covered by a commitment.

The challenge is not simply counting usage. It is keeping usage, entitlements, commitments, credits, and billing aligned.

This is one of the most overlooked revenue leaks in B2B SaaS.

A customer expands. A discount expires. Included usage changes. A new AI service becomes billable. An amendment introduces a different rate, credit allocation, or commitment.

The commercial agreement changed. Did the revenue system change with it?

If quoting, contracts, provisioning, usage, and billing operate independently, an amendment can be correct in CRM and still never reach the rating engine.

For example, a customer could move from a platform fee with 5 million included tokens to a higher platform fee with 10 million included tokens and a discounted inference overage. The quote can be correct, the contract can be correct, and the service can be provisioned correctly, while billing continues to apply the old overage rate.

This creates contract drift: what the customer agreed to and what the revenue system actually applies begin to diverge.

Usage-based monetization is therefore not only a billing problem. It is a commercial-to-revenue continuity problem.

Billing answers what should we charge? Revenue recognition answers when and how should that revenue be recognized?

Those are not always the same thing.

A SaaS or AI contract may combine a recurring platform fee, prepaid commitment, usage charges, credits, and professional services. Usage may be billed in arrears while another portion of the agreement was prepaid.

As monetization models become more flexible, the connection between consumption, billing, and revenue recognition becomes increasingly important. A clean invoice does not automatically mean a clean financial close.