Outcome pricing

Assumed resolutions, and why verified outcomes matter.

Some AI vendors bill for outcomes that may not have happened. What an assumed resolution is, why it matters, and what to ask any outcome-priced provider.

Outcome pricing is only as honest as the definition of an outcome. In customer service AI, some vendors count a conversation as resolved when the customer confirms it, or when they leave without asking for a human. That second case, an assumed resolution, can include people who gave up. If you are buying AI on outcomes, the first question to ask is what exactly is being counted.

How assumed resolutions work

Intercom was one of the first to price an AI agent on outcomes, charging per successful resolution. As Foundation Capital described it, Intercom counts a resolution either when the customer confirms the answer was satisfactory or when the customer exits the conversation without escalating to a human.

The logic is understandable. Most customers who get what they need do not bother to say so. But silence has more than one cause. A customer who could not get a straight answer and closed the window looks the same in the data as one who was helped.

Others have moved the other way. According to a 2026 pricing benchmark, Zendesk moved to a three-tier model in May 2026 where only verified resolutions are billed.

Why it matters outside customer service

The same problem shows up anywhere AI is priced on results. A document is "processed" when it was read, even if the data extracted was wrong. An invoice is "matched" when a match was proposed, even if someone had to fix it. A renewal is "prepared" when a file was created, even if half of it was empty.

Each of those counts looks like an outcome and is really activity. The buyer pays for work that still has to be done again.

What a verified outcome needs

A verified outcome has three things:

  1. A written definition agreed before work starts: what counts, what does not, and how long it must hold.
  2. Evidence in the record: the inputs, the checks run, where the result landed and any human approval.
  3. A holding period where the result can still be reversed, such as an invoice not reversed within 30 days.

If any of the three is missing, the outcome should not be on the invoice.

Questions to ask an outcome-priced provider

  • What exactly counts as one billable outcome? Can I see the definition?
  • Do you bill anything based on silence, timeouts or assumed success?
  • Can I click from any invoice line to the evidence behind it?
  • What happens to billing if an outcome is later reversed?
  • Who can dispute an outcome, and how long do we have?

Good providers answer these quickly, in writing. Vague answers usually mean the definition favors the provider.

We bill verified outcomes only. Every invoice line links to its record in the Evidence File. See how to write an outcome definition for a template you can use with any provider.

Common questions.

What is an assumed resolution?

An assumed resolution is a support conversation counted as resolved because the customer left without asking for more help, rather than because they confirmed the problem was solved. It can include customers who simply gave up.

What is a verified outcome?

A verified outcome meets a written definition and has evidence behind it: the inputs, the checks run, the result in the system of record and any human approvals. Only verified outcomes should be billed.

Sources

Want this run for you.

Agentic MSP runs back-office work with governed AI agents and bills only for verified outcomes.