Outcome vs conversation billing: what should you count?

By Flynn Traeger · Sources checked November 5, 2026

A common worksheet for AI support billing. Separate the vendor charge, the customer result, and remaining human work.

The label does not define the work

An outcome, conversation, or resolution is a unit defined by a vendor's rules. Read those rules alongside the rate.

Fin includes configured Procedure handoffs among billable outcomes. Front charges each conversation at its highest automation level.

That means the presence of a human handoff alone cannot tell you whether the AI work is billable.

Billable event: Which contract rule created the charge?
Customer result: Did the customer get the answer or completed task they needed?
Human work: How much agent effort remained?

Use a row for each request. Keep these fields separate so a billing status does not become your only quality measure.

Imagine Product A bills 600 events at $1. Product B bills 900 events at $0.75. Their AI charges are $600 and $675.

The lower unit rate produces the higher usage bill in this invented example. It says nothing about either product's quality.

Next, add platform costs and human work. Product B might still be preferable if it meets the required result with less effort.

Our suggested measure is total support cost per request that passes your acceptance test. Define that test before comparing products.

Keep the vendor's billing count beside your own accepted-result count. Investigate gaps rather than assuming one replaces the other.

This is an editorial analysis of public sources. Examples use stated assumptions. We have not tested vendor billing or product performance.