Company

Outcome-based pricing for AI agents: how it works, with examples

Outcome-based pricing for AI agents means the customer pays only when the agent completes a defined result, such as a resolved support conversation, instead of paying per seat or per unit of raw usage. It has become the headline model for customer-service agents, but the price, the definition of "outcome" and the fine print differ a lot by vendor.

This page explains how the three main models compare, what Intercom, Zendesk, Salesforce and Sierra publish about their pricing, and how a buyer can check that an "outcome" is really an outcome. Prices are taken from the vendors' own pages as of October 2026 and can change.

Seat-based, usage-based and outcome-based pricing

Seat-based

The customer pays a fixed amount per user per month. Revenue is predictable and easy to forecast. The weakness for AI agents is that an agent has no natural "seat": a bot that handles 10,000 conversations costs the vendor far more to run than one that handles 100, yet the seat count may not change.

Usage-based

The customer pays for activity: tokens, API calls, actions, minutes or messages. Cost follows the vendor's compute bill closely, which protects margins. The buyer carries the risk, because an agent that loops, fails or gives a bad answer still generates a charge.

Outcome-based

The customer pays for a result. If the agent does not resolve the request, there is generally no charge. Risk moves from the buyer to the vendor, and the vendor is rewarded for making the agent better, not busier. The hard part is agreeing on what counts as success.

What vendors publish, as of October 2026

Intercom Fin

Intercom lists Fin at $0.99 per outcome. A resolution is a conversation where no further help is requested after Fin's last answer. The page also lists other billable outcomes: procedure handoffs, disqualifications and self-serve routings at $0.99 each, with "qualifications" priced at $9.99 each. A conversation that goes to a human team without an outcome is not charged. For Fin used with a non-Intercom helpdesk, the page lists a minimum of 50 outcomes per month.

Zendesk

Zendesk bills AI agents per Automated Resolution, defined as a customer request resolved by the AI agent without escalation to a human. As of October 2026 the listed price is $1.50 per resolution on a committed volume and $2.00 on pay-as-you-go. Support Team and Suite Team plans include 5 automated resolutions per agent per month, and Suite Professional includes 10. Enterprise pricing is handled by sales.

Salesforce Agentforce

Salesforce is the clearest example of a vendor offering several models side by side on its Agentforce pricing page:

• Flex Credits: $500 per 100,000 credits. A standard agent action uses 20 credits, about $0.10, and a voice action uses 30 credits, about $0.15. This is usage-based pricing.
• Conversations: $2 per conversation for customer-facing agents. It cannot be combined with Flex Credits in the same org.
• User licenses: $5 per user per month for the Agentforce User License (which requires Flex Credits) and $50 per user per month for the Coworker license, plus per-user add-ons for employees.

Salesforce's per-conversation price is a flat charge per interaction, not a charge for a verified resolution, so it sits between usage-based and outcome-based pricing.

Sierra

Sierra, the customer-experience agent company, describes its approach in a December 2024 company post. According to Sierra, customers pay when the agent achieves a specific outcome, such as a resolved conversation, a saved cancellation, an upsell or a cross-sell, and the criteria are agreed upfront. Escalated conversations are generally not charged. For routing or greeting, Sierra says it may propose a blended model based on conversation counts. Sierra does not publish a price list on that page, so the unit price is negotiated.

How the published prices compare

The per-unit numbers are not like for like. Intercom's $0.99 and Zendesk's $1.50 to $2.00 buy a resolved conversation, but each vendor defines resolution differently. Salesforce's $0.10 buys one action, and a single customer conversation can involve several actions. Comparing vendors fairly means estimating your own cost per resolved case, including the share of conversations the agent cannot finish.

How to define an outcome

Most disputes in outcome-based contracts come from vague definitions. Before signing, pin down:

1. The trigger. Is a resolution a customer saying thanks, a closed ticket, or silence for a set period? Intercom's definition relies on no further help being requested, which means a customer who simply gives up can be counted the same as a satisfied one.
2. The escalation rule. If the agent hands over to a human after doing half the work, is anything billed? Zendesk and Sierra both say escalated cases are generally not charged.
3. Reopens. If the customer returns within a few days about the same issue, can the charge be reversed?
4. Non-resolution outcomes. Routing, qualification and handoffs can be billable at different rates, as Intercom's page shows.
5. Minimums and commitments. A per-outcome price with a monthly floor behaves partly like a subscription.

How to audit outcomes

• Ask for conversation-level logs that show why each charge was counted.
• Sample a set of billed conversations each month and read them. A resolution rate that rises faster than customer satisfaction deserves a closer look.
• Track your total human-handled volume alongside billed outcomes. If outcomes rise but agent workload does not fall, the savings may be smaller than the invoice suggests.
• Compare against a baseline: your cost per human-resolved ticket before the agent.

Pros and cons

For buyers

• Pro: no payment for failed attempts, and cost tied to value received.
• Pro: easier to justify internally, because spend maps to resolved work.
• Con: bills are variable and hard to budget, particularly during a spike in tickets.
• Con: the definition of an outcome is set mostly by the vendor.

For vendors

• Pro: pricing that matches how buyers think about value, and an incentive to improve resolution rates.
• Con: revenue is lumpy and depends on customer ticket volume, which affects how investors read recurring revenue. Our guide to AI wrappers covers a related question: who captures the margin between model cost and customer price.
• Con: the vendor carries the compute cost of every failed attempt.

More coverage of how AI companies make money is in AI Business, and startup monetization stories are collected under Startups. New pricing announcements usually appear first under Launches.

Frequently asked questions

What is outcome-based pricing for AI?

It is a pricing model where the customer pays for a defined result delivered by an AI agent, such as a resolved ticket, rather than for seats or tokens. If the agent fails or escalates to a human, the charge is generally waived.

How much does Intercom Fin cost per resolution?

As of October 2026, Intercom's pricing page lists $0.99 per Fin outcome. Some setups carry a minimum, for example 50 outcomes per month for Fin on a non-Intercom helpdesk.

What is the difference between usage-based and outcome-based pricing?

Usage-based pricing charges for activity such as actions, tokens or messages, whether or not the task succeeded. Outcome-based pricing charges only when the agreed result is reached.

Does Sierra use outcome-based pricing?

Yes. In a December 2024 post, Sierra said customers pay when the agent achieves an agreed outcome and that escalated cases are generally not charged. It does not publish a standard price list.

Is seat-based pricing dead for AI agents?

No. Salesforce still sells per-user Agentforce licenses next to credit and conversation pricing. Seats remain common for tools used by employees, while customer-facing agents are where per-outcome pricing is spreading.