Pay only when the AI actually does the job. That's the promise behind outcome-based pricing for AI agents, and it's catching on. Zendesk charges $1.50 per automated resolution on committed volume and $2.00 on pay-as-you-go, a model it introduced in August 2024. Intercom charges $0.99 for each conversation its Fin agent fully resolves. Sierra has built its business around outcome-based pricing. At Dreamforce last month, the recurring theme from enterprise buyers was that they want pricing tied to resolved issues rather than seats.
On paper, this is the fairest deal buyers have been offered in years. In practice, it moves the hard question from "what does it cost" to "what counts as done," and most organisations haven't answered that yet.
Why Seat-Based Pricing Stopped Making Sense
Seat licences made sense when software helped people do work. They make much less sense when the software does the work itself. An AI agent resolving customer queries doesn't need a seat, and charging per human user creates an odd incentive: the more work the agent absorbs, the fewer seats the customer needs, and the less the vendor earns. Outcome pricing lines up what the vendor earns with what the buyer actually wants.
The Definition Is the Contract
Here's the issue. A "resolution" sounds precise until you try to measure one. Zendesk's own framing is that an automated resolution happens when AI fully resolves an issue without human intervention, and several independent guides report it's commonly measured with a 72-hour quiet period during which the customer doesn't come back. That's a reasonable definition. It's also a vendor's definition.
Consider what it rewards. A customer who gives up and walks away counts as resolved. A customer who gets a wrong answer, doesn't notice for four days, then calls back angry may already have been billed as a success. Neither of those is what a customer service leader means by resolved, but both can fit a quiet-period definition.
What to Negotiate Before You Sign
Before agreeing to outcome pricing, buyers should know exactly how the vendor defines and measures the outcome, whether they can independently check that measurement against their own data, what happens to billing when a "resolved" case reopens outside the window, and whether there are quality measures, like satisfaction or accuracy sampling, that sit alongside the volume count. An outcome definition you can't audit is a pricing model you can't govern.
Outcome Pricing Changes the Business Case, Too
There's an upside that's easy to miss. Outcome pricing forces organisations to decide, before implementation, what success actually looks like, which is exactly the discipline many AI business cases skip. The organisations getting the most from this model use the vendor's outcome definition as a starting point for their own measurement framework, not a replacement for it. They track cost per genuine resolution, not just cost per billed resolution, and the gap between those two numbers tells them a great deal.
What This Means for Your Organisation
What we see across implementation engagements is that pricing gets negotiated by procurement, while outcome definitions get left to whoever configures the agent. Those two things are now the same decision. Organisations moving to outcome-based AI pricing should put the operational owner of the process in the room when the contract is negotiated, because the definition of success written into that contract will shape how the agent is built, measured and paid for.
Key Takeaways
- AI agent vendors are moving from seat-based to outcome-based pricing, with published per-resolution prices from Zendesk and Intercom and buyers at Dreamforce 2026 pushing for pricing tied to resolved issues.
- Outcome pricing lines up vendor revenue with buyer value, but it makes the definition of a successful outcome the most important term in the contract.
- Common definitions, such as a quiet period after a conversation closes, can count abandoned or incorrectly handled cases as successes unless quality measures sit alongside them.
- Buyers should negotiate auditable outcome definitions, reopening rules and quality checks, and track cost per genuine resolution against cost per billed resolution.
How Trusenta Can Help
AI Strategy Enterprise defines what success actually means for an AI agent before pricing is negotiated, so the contract reflects business outcomes rather than a vendor's default metric.
AI Agents and Automation builds and configures agents against measurable outcome definitions, with the reporting needed to check vendor billing against real results.
AI Integration Services connects agent activity data with existing CRM and service systems so outcomes can be independently verified.
Conclusion
Outcome-based pricing is a real improvement on paying for seats that no longer do the work. But it only protects the buyer if the outcome is defined carefully, measured independently and backed by quality checks. The organisations that treat the definition of success as a strategic decision, not a line in the vendor's terms, are the ones who'll actually get what they pay for.
