Zendesk will charge $1.50 per resolution. Outcome pricing has arrived.

For years, outcome-based pricing was the pricing conference topic that never survived contact with a contract. Everyone agreed it was the future: pay for results, not tools. Nobody could define "result" tightly enough to bill for it.

That era is ending in the customer support market. At Relate this month, Zendesk announced outcome pricing for its AI agent lineup: $1.50 per automated resolution on committed volume, $2.00 on pay-as-you-go, with a verification layer that confirms the customer's issue was actually resolved before the meter runs. Zendesk's CEO framed the logic simply: customers should pay for the value they realize, not the tools they deploy. Intercom got there earlier with its Fin agent at $0.99 per resolution. Salesforce sells Agentforce conversations at $2 each. The most competitive category in enterprise software now prices its AI on delivered work.

Why support went first

It is not an accident that outcome pricing became real in customer support before anywhere else. A support resolution is unusually well-behaved as a billing unit:

• It is definable. A ticket is opened, handled, and closed without human escalation. The boundaries are crisp.
• It is attributable. One agent handled the conversation. Nobody argues about whether the marketing team deserves credit for the resolution.
• It is verifiable. Did the customer come back about the same issue? Both sides can check. Zendesk's "verified resolution" mechanism exists precisely because early per-resolution billing generated disputes about what counted.
• It has a known human cost. Every support leader knows their cost per human-handled ticket. A $1.50 resolution against a $8-12 human ticket is arithmetic a CFO can do in the meeting.

Run your own product through that list and you will know whether outcome pricing is available to you or still aspirational. Most products fail on attribution: when your software contributed to a closed deal, a filed report, or a saved hour alongside six other tools and three humans, you cannot bill the outcome without a fight. This is why we advise most companies toward hybrid packaging as the intermediate step: bundle the baseline, meter the consumption, and move to outcomes only where attribution is airtight.

If you sell: the mechanics that decide everything

Define the outcome with your most skeptical customer in the room. The definition that survives is never the one product marketing writes. It is the one your angriest future disputant would accept. Write the edge cases into the contract: partial resolutions, reopened tickets, outcomes the customer's own systems caused to fail.

Price the risk transfer, not just the work. Outcome pricing moves performance risk from buyer to seller. That risk has a price, and it is why per-outcome rates should carry a premium over the equivalent usage math. If your outcome price is just your usage price divided by average outcomes, you gave away the risk transfer for free.

Expect your revenue to become a forecast problem. Outcome revenue arrives when outcomes do, not when contracts sign. Finance, sales comp, and investor reporting all need rework before the first deal, not after. The companies that stumbled here did not fail on pricing theory; they failed on operations.

If you buy: this is better for you, if you negotiate it properly

Buyers should generally welcome this shift. It converts vendor promises into vendor risk. But three clauses separate a good outcome deal from an expensive one:

• Verification rights. You need the ability to audit what was counted as an outcome, with a defined dispute window. Take the vendor's verification layer, but keep your own telemetry.
• Rate step-downs at volume. Per-outcome rates should fall as volume commits rise. The marginal cost of resolution number 100,001 to the vendor is close to zero; your rate card should acknowledge that somewhere.
• A quality backstop. Pure per-resolution incentives push vendors toward closing tickets, not solving problems. Tie a customer satisfaction floor or reopen-rate ceiling to the outcome billing, so the meter only runs on outcomes you would have paid a human for.

The bigger picture

Step back and the sequence is clear. Seats priced access. Usage priced activity. Outcomes price results. Each step moves the billing event closer to the value event, and each step is enabled by AI doing more of the work end to end, because software that completes work can be measured on the work it completes.

Support is first, not unique. Anywhere an AI agent owns a process with clean boundaries, collections, lease abstraction, invoice processing, first-pass contract review, the same pricing physics will apply. If you operate in one of those categories, the vendors are already modeling it, and the analyst frameworks, from Bessemer's AI pricing playbook on down, are already teaching your competitors the playbook.

The strategic question for 2026 is not whether outcome pricing is coming to your category. It is whether you will be the company that defines the outcome metric for your market, or the company that accepts a definition written by someone else.