Outcome-Based AI Pricing Sounds Fair. Read The Contract.

Salesforce just paid $3.6 billion for Fin — an AI customer service platform whose whole pitch is you only pay when it resolves a ticket^[1]. HubSpot moved two of its Breeze agents to the same model in April^[2]. Zendesk launched an "Autonomous Service Workforce" priced only on verified outcomes at Relate 2026^[3]. Sierra, the $10B agentic AI darling, prices every contract by outcomes it defines, and starting enterprise deals reportedly hit $200K–$350K in year one^[4].

Every vendor is racing to the same headline: you only pay when it works.

If you're a $2M–$20M business shopping AI agents, that sounds like the first honest software pricing model in 20 years. Read the contract before you agree.

The pitch, in one sentence

Traditional SaaS charges per seat whether the tool works or not. Outcome-based pricing flips it: the vendor charges only when the AI agent completes an "outcome" — a resolved ticket, a qualified lead, a saved cancellation, a completed workflow.

HubSpot's Breeze Customer Agent and Breeze Prospecting Agent moved to this model on April 14, 2026^[2]. Intercom's Fin charges $0.99 per resolution with a 50-outcome monthly minimum^[5]. Zendesk sits at $1.50–$2.00 per resolution on top of Suite seats plus a $50 Copilot add-on^[6]. Sierra doesn't publish a rate card — every deal is quoted by their sales team^[7].

The framing sells itself: no ROI, no bill. Aligned incentives. The vendor eats the risk of a bad model.

Except that's not what's actually happening.

The vendor writes the definition of "outcome"

Here's the mechanic nobody explains at the demo.

An "outcome" is whatever the vendor's product team decides it is. Read Intercom's own help doc: if Fin gives an answer and the customer doesn't explicitly ask for a human or push back, the conversation is "resolved" — and you pay $0.99^[8]. The customer doesn't have to confirm anything worked. They just have to not complain loudly enough within the session.

That's not a resolution. That's a shrug.

A third-party test of Fin across 500 tickets found the actual resolution rate landed at 38%, not the 50%+ marketed^[9]. Intercom's own site claims 67% across 40M+ conversations^[10]. Both are true — they're measuring different things. The 67% is Fin's internal "outcome" definition. The 38% is what the human tester counted as an actual answered question that didn't need follow-up.

You're not paying per problem solved. You're paying per Fin-defined event that hit its own success threshold.

Where the money actually leaks

Three places, every time.

1. Ghost outcomes. A customer asks a two-part question. Fin answers part one, the customer never comes back. Resolved. $0.99. They just gave up. This shows up in support tickets a week later as churn, and now your CS team is chasing refunds while your AI bill sits there looking clean.

2. Escalation lag. Watch the routing rules. If the platform bills for an outcome the moment the agent gives a plausible answer — before the customer has actually confirmed it worked — you're paying for AI-generated hope. Analysts call this the "premature resolution" problem, and it's the single biggest lever in every outcome contract^[11].

3. Seat + outcome stacking. Every one of these vendors keeps a seat license underneath the outcome fee. Intercom Fin: $29–$139 per human agent per month plus $0.99 per resolution^[5]. Zendesk: Suite seats + $50 Copilot + $1.50–$2.00 per resolution^[6]. Sierra: no seats, but $150K–$350K annual floors^[4]. The outcome charge isn't replacing the base bill. It's on top.

For a business doing 3,000 support tickets a month, $0.99 per resolution at even a real 40% resolution rate is $1,188/mo in outcome fees before you've paid a single seat. That's $14K/year for the tickets your humans still have to review anyway.

What to negotiate before you sign

Every one of these platforms has room to move on the contract. Sierra is 100% custom-quoted. Fin will negotiate committed volume down from $0.99 to the $0.60–$0.75 range at real scale. Zendesk drops to $1.50 on committed tiers^[11]. Nobody publishes it, and everybody's willing to talk.

Four things I'd push on before signing anything:

1. Get the outcome definition in writing. Not "resolution" — the exact event that fires the charge. If it says "the AI agent provides a response and the conversation ends without escalation," you're paying for silence. Rewrite it to require a customer-side confirmation or a 24-hour re-open window that voids the charge.

1. Cap monthly outcomes. Nobody offers this by default. Everybody agrees when asked. Pick a number based on your last 12 months of ticket volume × your target resolution rate, and negotiate a hard ceiling. This alone kills the surprise-bill risk.

1. Track your own resolution rate. Instrument this before you start. Every ticket flagged "resolved by AI" gets a 72-hour follow-up flag. Compare your internal number to the vendor's dashboard weekly. When (not if) they diverge, that's your renewal ammunition.

1. Force a shadow-mode trial. Run the agent for 30 days answering questions your team also answers, and count real resolution rate against the vendor's counter. Don't sign anything longer than 6 months on the first contract. This is a market where prices are dropping every quarter — Fin's rate is already down from its 2025 launch pricing.

Why this pricing shift is happening now

Outcome pricing isn't a gift to buyers. It's a moat.

The vendors moving hardest — Salesforce, HubSpot, Zendesk, Intercom, Sierra — all have deep CRM/support data already inside their walls. They can price on outcomes because their models have the customer context to look competent^[12]. A standalone AI agent startup can't offer $0.99 per resolution without hemorrhaging money, because it doesn't have the ticket history to know when it's about to be wrong.

Which means outcome-based pricing is doing two things at once. It's giving buyers cover to spend more on AI (because "you only pay when it works" is easier to sign off on than a $50K/year seat contract). And it's letting the platform incumbents fence off the AI-agent market before any startup can undercut them on price.

That's the shift. Not "software is finally accountable." It's "the platform incumbents figured out how to charge more per interaction while making it sound like less."

The one-line summary

Outcome-based AI pricing is a pricing model, not a promise. The vendor still writes the definition, still keeps the seat license, and still bills you when the AI generates something plausible-sounding. Read the outcome definition, cap the volume, and instrument your own truth. Otherwise you're not paying for results. You're paying for whatever the vendor's dashboard says.

If you're evaluating one of these platforms right now and want a second read on the contract before you sign, that's what the audit call is for. 30 minutes, no pitch — bring the pricing PDF and I'll tell you where the traps are.

Sources 12 references

1. 1
   Salesforce acquires AI customer service platform Fin for $3.6B

   TechCrunchnews

   Salesforce is acquiring Fin (formerly Intercom's AI unit) for $3.6B

   ↩
2. 2
   HubSpot Shifts Breeze AI Agents to Pay-per-Result Pricing

   CMSWirenews

   HubSpot moved Breeze Customer Agent and Breeze Prospecting Agent to outcome-based pricing effective April 2026

   ↩
3. 3
   Zendesk Introduces the Autonomous Service Workforce

   Zendeskprimary

   Zendesk launched Autonomous Service Workforce with outcome-based pricing where customers pay only for confirmably resolved interactions, at Relate 2026

   ↩
4. 4
   Sierra AI Pricing 2026

   Fin.aianalysis

   Sierra enterprise deals reportedly start at $150K–$350K annual

   ↩
5. 5
   AI Agent Pricing Benchmark 2026: 18 Vendors Compared

   Aissist.ioreport

   Intercom Fin charges $0.99 per billable outcome on top of $29–$139/seat with a 50-outcome monthly minimum

   ↩
6. 6
   Zendesk AI Pricing in 2026, Explained

   Coworker AIanalysis

   Zendesk AI resolutions priced $1.50–$2.00 on top of Suite seats plus $50 Copilot

   ↩
7. 7
   Sierra AI: Guide to Features, Pricing & Limitations (2026)

   MyAskAIanalysis

   Sierra contracts are custom-quoted enterprise agreements, no public rate card

   ↩
8. 8
   Fin AI Agent outcomes

   Intercomdocs

   Intercom's own help doc: a conversation counts as resolved (and billable) if the customer doesn't explicitly ask for human help after Fin's answer

   ↩
9. 9
   Intercom Fin AI Review: We Tested It on 500 Tickets (2026)

   Builts.aianalysis

   Third-party test found real Fin resolution rate at 38% across 500 tickets, not the marketed 50%+

   ↩
10. 10
    Intercom Fin AI Guide: Features, Pricing & Limitations (2026)

    MyAskAIanalysis

    Intercom claims 67% resolution rate across 40M+ conversations at $0.99 per resolution

    ↩
11. 11
    Outcome-based Pricing in Practice: What Outcomes Are AI Companies Really Charging For?

    The Pricing Conundrumanalysis

    Zendesk resolutions ~$2 pay-as-you-go, ~$1.50 committed; premature-resolution is the biggest structural lever

    ↩
12. 12
    HubSpot's Customer Agent and Prospecting Agent: Now you pay when the task is complete

    HubSpotprimary

    HubSpot argues its embedded CRM context is what lets it price on outcomes reliably

    ↩