Outcome-based support pricing: how the shift rolls out by 2027

Under outcome-based pricing, the invoice is only as reliable as the vendor's definition of a resolved ticket.

Support software is moving to outcome-based pricing, and the unit it bills on is the resolution: a customer issue the AI agent closes without a human stepping in. Buyers pay when the problem is solved, not per seat or per message. That shift is real. It is also uneven, because the rule deciding what counts as solved changes from one vendor to the next.

In 2025, Pragmatic Institute reported that Intercom's Fin had quadrupled its revenue year over year, crediting A/B tests built to lift the resolution rate. The same article flagged Zendesk's trouble with ambiguity over what exactly qualifies as a resolution. Growth and dispute arrived together.

The cost side explains why. A 2026 pricing analysis showed an AI agent earning the same flat fee for a 30-second password reset that cost $0.05 in compute and for a stubborn billing dispute that cost $1.20, more than the fee itself.

I read that spread as a warning, not a reason to walk away. A vendor losing money on hard tickets can tighten its definition or add a minimum commitment, and both moves land in the buyer's contract. So the question that decides adoption is narrower than price. Who gets to say a ticket is closed?

Is support software actually moving to outcome-based pricing?

Yes, but slowly: per-resolution fees from vendors such as Intercom and Zendesk are arriving as add-ons inside hybrid contracts, while fully outcome-based pricing remains rare among major software vendors.

Before you weigh any forecast, place your own contract on the map with three questions:

1. Does your current plan bill for seats, for usage such as conversations or credits, or only for resolved outcomes?
2. If the AI agent resolves nothing next month, does your invoice fall, stay flat, or rise?
3. Does the per-resolution fee replace any seats, or does it sit on top of the seat bill?

Most support teams will answer with some version of seats plus a meter. That matches the market. AlixPartners found that only 4 of 65 major SaaS and AI-native companies have fully adopted outcome-based pricing, and more than half of the remaining 61 still rely primarily on per-seat models. The same July 2026 analysis found that 72% use a hybrid approach layering AI credits or usage metering onto the base price, and that for 82% of companies the customer bears the risk when the software does not deliver.

One detail matters more than the headline count for support buyers. All four outcome-based adopters operate in customer service or customer-facing agent workflows, where a ticket either closes or it does not. They cluster below $1 billion in annual revenue, while seven of the nine companies above $5 billion keep the risk with their customers.

The common assumption is that seats are disappearing. They are shrinking, not vanishing. In 2025, Growth Unhinged's State of B2B Monetization report, as summarized by Pragmatic Institute, found that seat-based pricing fell from 21% to 15% of companies in 12 months. Around half of the 65-company cohort prices an anchor product per seat and layers a consumption model on top, so if you are shortlisting help desk platforms with live chat, expect that two-line structure on many quotes.

Buyer sentiment runs further ahead than contracts do. A billing vendor's 2025 guide claimed that 77% of business leaders say customers are pushing for outcome-based pricing, while only 32% of businesses define usage that way. An analyst firm's 2026 press release went further and called outcome pricing a market standard, although its own survey showed buyers preferring consumption models ahead of outcome-based ones.

Read together, the 5 sources behind this baseline point one way. Appetite for paying on results is loud. Signed outcome contracts are scarce. Zazachat's reviews are written and edited by our editorial team, led by Daniel Calloway, and the forecast that follows starts from this baseline rather than from the enthusiasm. In practice, your next renewal will most likely add a meter rather than remove seats, so budget for both lines and read the meter's trigger closely, because the word resolution does not describe the same event from one vendor to the next.

What counts as a resolution at each support vendor?

A resolution is whatever each vendor's billing rule says it is, and those rules differ: some count 72 hours of silence, some need customer confirmation, one requires a satisfaction lift.

CX Foundation's August 2026 comparison puts the published definitions next to each other, and The SaaS CFO's June 2026 pricing guide adds the fine print on exclusions and credits. Combined, the picture looks like this:

Vendor and product	Published rate	What triggers the charge	Waiting window
Zendesk AI Agents for Service	$1.50 per successful resolution	No repeat contact, with an LLM check that the request was satisfied	72 hours
Salesforce Help Agent	$2.00 per successful resolution	Customer says the case was resolved	None stated
Crescendo	$1.50, typically	Ticket marked closed and satisfaction rose during the conversation	None stated
Intercom Fin AI Agent	$0.99 per resolution	Customer confirms satisfaction or leaves without escalating; configured procedures also count	24 hours
Help Scout AI Answers	$0.75 per resolution, billed in arrears	Full answer given and no further help requested; greetings and requests for a person excluded	Once per chat session
HubSpot Breeze Customer Agent	$0.50 per resolved conversation, or 50 credits	Resolved conversation	None stated
Gorgias	Not stated in the sources	Customer does not need a person; can add an AI fee on top of a helpdesk ticket fee	72 hours
Sierra, Decagon, Omilia, Retell AI	No public rate card	Negotiated per enterprise contract	Negotiated

The spread matters more than the rates. Zendesk and Crescendo both list $1.50, yet one bills when a customer stays quiet for three days and the other only when the ticket closes with satisfaction rising mid-conversation. Those are two different purchases at the same price. Fin's window is the shortest on the list, and the same guide notes that an outcome there can also mean executing a configured procedure, which widens what counts.

Now set the vendor windows against how support teams have long measured their own work. Many contact centers consider a case resolved only if the customer does not call back within 30 days. Every published vendor window above stops at 72 hours or less. A customer who returns the following week has already been billed as resolved under any of those rules.

My own view is that neither main design is neutral. Confirmation rules are the easiest to check but bill only when customers bother to answer; inactivity rules capture silent successes but also charge for customers who quietly gave up. Pick the bias you can live with, then price it.

Some meters are not measuring resolutions at all. A billing platform's 2025 write-up described Salesforce's Agentforce as charging $2 per conversation, counted when a customer sent at least one message or picked a menu option other than End Chat within 24 hours. That is an activity meter wearing outcome language. A former customer service software executive made a related point on a 2025 podcast: vendor and customer definitions of a resolution diverged so far that buyers were told they could not plan for the difference, and some ran a platform for a couple of weeks before agreeing to any price.

Before you compare headline rates in any set of AI customer support software rankings, line up the trigger first. If I were drafting the contract language, I would start from Help Scout's style of explicit exclusions, because exclusions are what make a count checkable. And for the four vendors that publish no rule at all, the order form, not the pricing page, is the document that sets your bill.

How will outcome-based support pricing roll out through 2027?

Expect three stages: resolution meters bolted onto seats, then contract fights over definitions, then a split where only verifiable workflows keep outcome pricing.

The stages below are my forecast, not a projection any vendor or analyst has published. They rest on one pattern. The consultants behind the 65-vendor analysis reduce it to one diagnostic: "Can you measure success in a way your customer will accept?"

Chargebee's analysis of Intercom shows the pattern at its best. Fin fit outcome pricing because of three conditions: it performs a single task with a focused outcome, its success signals are clear, and the outcome is measurable in-product. Fin's logs record every resolution and serve as a single source of truth for Intercom and its customers alike. The payoff was real. Intercom CEO Eoghan McCabe said Fin is now "a strong 8 figure ARR business that in Q1 grew at an annualized rate of 393%."

Bret Taylor of Sierra shows the edge of the pattern. Sierra charges a pre-negotiated rate when its agent resolves a case with no human intervention, and escalations to a person are free. Yet Taylor concedes "there's not a great way to do it for every type of agent right now," adding that Sierra can always fall back to usage-based pricing. Even the model's strongest advocate keeps a usage meter in reserve.

From those two signals, I expect the rollout to move like this:

1. Now into early 2027, meters on top of seats. Resolution fees spread as add-ons wherever an AI agent answers bounded questions with a clear success signal. Negotiate the trigger, the exclusions and a monthly cap before volume grows.
2. Through 2027, the definition becomes the contract. As AI bills grow, finance teams start disputing counts, and vendors whose logs a buyer can inspect gain an edge. Negotiate log export, a reconciliation window and a dispute process.
3. Late 2027, the split. Workflows with verifiable outcomes keep or deepen outcome pricing. Anything whose result lives in another system reverts to usage, credits or seats, and there you should negotiate transparent usage terms rather than accept an outcome label on activity.

Stage two has a mechanical cause. A metering vendor's 2025 explainer argued that an outcome's success signal can take days or weeks to settle and that LLM judgments of success will likely be debatable. Chargebee adds that business outcomes usually sit outside the vendor's product, which turns billing on results into a game of inference. Delay plus inference produces disputes.

A voice AI vendor that bills by the minute makes the opposing case in its own favor: when customers control the agent, usage pricing offers transparency, and when the vendor owns the result, outcome pricing creates alignment. It also points to an outsourcing precedent in which buyers said they preferred paying for outcomes while many contracts stayed priced per headcount. That precedent worries me more than any vendor pitch. Stated preference did not move contracts then. I doubt it will now without a count both sides can check.

If you run a small team comparing live chat tools for small businesses, stage one is where you will meet this shift, most likely as an optional AI add-on beside a familiar plan. The buyers who reach stage two with a reconciliation clause already signed will negotiate from their own records, while everyone else negotiates from the vendor's dashboard.

What should support leaders do before outcome pricing reaches their contract?

Treat the resolution definition as the real price. Get each vendor's counting rule in writing, test it against your own closed tickets, and set a spending cap before you sign.

Ask for that rule on the order form, not the pricing page: a June 2026 pricing guide notes that in enterprise deals success criteria "often cannot live on the pricing page" and become "a negotiated definition on the order form." Price is no longer the obstacle; list fees for an AI resolution already sit between $0.99 and $2. What still varies is the definition behind each fee.

Gabe Larsen, who sold per-conversation and per-resolution pricing in customer service software, described the buyer's side on The Transaction podcast in 2025. CFOs asked for standard pricing they could budget, and vendors answered with a price per conversation. His verdict on that answer was blunt: "A price per conversation that is not an outcome-based pricing."

He is right about the label. My bet for 2027 is that the billing trigger, not the marketing name, decides who carries the risk. A plan that charges when a chat opens keeps it with you. A plan that charges only after an auditable fix moves some of it to the vendor.

So run the test before renewal. Take a recent sample of tickets the AI closed, apply the vendor's written rule, and bring finance the gap between the vendor's count and yours.