Does Your AI Support Vendor Bill the Attempt or the Outcome?

Everyone writing about AI support pricing explains the difference between per-conversation and per-resolution billing. Almost nobody says which vendors do which, which is the only part you can act on.

The five results ranking for this question when we checked were all published by companies selling one of the tools. So here is the version with names in it.

Why the meter matters more than the rate

A per-attempt meter charges whether or not the AI helped. If it fails and a human picks up, you pay twice — once for the attempt, once for the salary. The worse the agent performs, the more you owe.

A per-outcome meter inverts that: you pay for the share that worked. It also creates an incentive worth naming out loud, because the vendor is now the one deciding what counts as a success.

Neither is dishonest. But they behave in opposite directions as quality changes, and a rate card cannot tell you which you are looking at.

The two that moved this year

HubSpot repriced Breeze in April 2026, from $1.00 per conversation to $0.50 per resolved conversation. The headline rate halved and, more importantly, the meter moved from attempts to outcomes.

Kustomer went the other way, and its own historical pricing page is explicit about what triggers the charge. An engaged conversation at $0.60 is counted when an AI automation attempted to generate a response, regardless of whether it was ultimately sent — and also when an AI team viewed or opened the conversation, even without sending a response.

Read those two sentences next to each other. One vendor decided to charge only when it worked. The other charges when its software looked at the ticket.

What happens when the customer walks away

The sharpest test of a meter is abandonment, because it is the case where nobody was helped and somebody still gets billed. We read fourteen platforms' terms on exactly this question.

The largest group is the one that will not say. That is the finding: on the single question most likely to inflate a bill, half the vendors we could read published nothing definite.

Fin is the clearest of those that do answer, and it answers in the buyer's disfavour. Its pricing page defines a resolution as occurring when no further help is requested after Fin's last answer. A customer who gives up and closes the tab has not requested further help, so that is billable. A handoff to your team without an outcome is not charged.

So on the same platform, an unhappy customer who abandons costs you money and an unhappy customer who reaches a human does not.

The seat floor underneath the meter

The other thing a usage rate hides is what you pay before any usage happens.

This is where a cheap-looking per-resolution rate stops being cheap. A rate of $0.20 on top of a $99 seat is a different product from $0.99 with no seat, and the crossover depends entirely on your volume.

The three questions that settle it

What event triggers the charge — an attempt, a send, or a confirmed outcome? What happens when a customer abandons mid-conversation? And what do I pay in a month where the AI resolves nothing?

A vendor that answers all three in writing has designed its unit. A vendor that answers the third with a shrug is selling you a meter it has not finished thinking about, and you will find the edges in your first invoice.

Billing terms read on each vendor's own pricing or documentation pages and dated in our records. Where a vendor's terms do not address abandonment, we record that rather than inferring an answer.

WRITTEN BY AR · UPDATED 2026-08-17

I read the fine print. Vendor pricing pages, billing definitions, terms, funding filings and acquisition notices — then I do the arithmetic nobody publishes: what a platform actually costs at your volume, what its headline metric is really counting, and who owns it now. I do not run benchmarks, and no page here pretends otherwise.

Editorial policy