Customer service software used to be priced by the seat. You paid for the humans answering tickets. In 2026 the headline vendors price by the result instead: a resolved conversation, a saved cancellation, a qualified lead. The unit of billing is now the outcome. The catch is that each vendor defines the outcome, measures the outcome, and then invoices against its own definition.
The market, in three transactions
The clearest signal came in September, when Salesforce completed its acquisition of Fin, the company that renamed its entirety from Intercom to its AI agent's product name in May 2026. Salesforce signed the deal on June 15 for approximately $3.6 billion and closed it on September 10, 2026, folding Fin's agent and 30,000-plus customers into its AI Labs alongside Agentforce.
Sierra, the startup that made outcome-based pricing its brand, raised $950 million at a $15.8 billion post-money valuation in a round led by Tiger Global and GV, reported in May 2026. The company topped $150 million in annual recurring revenue in eight quarters and says it serves more than 40 percent of the Fortune 50. Co-founder Bret Taylor, who is also chairman of OpenAI, estimates $400 billion is spent annually on customer service and that most of it is migrating to agents.
Decagon, the third independent in the tier, closed a $250 million Series D in January 2026 at a $4.5 billion valuation, led by Coatue and Index Ventures. Sacra estimates it reached roughly $100 million in annualized revenue by mid-2026, up from $44 million at the end of 2025.
The published rate card
Three of the four major vendors publish numbers. They are not interchangeable.
Fin (Salesforce) charges $0.99 per outcome, charged at most once per conversation. An outcome is a resolution, a completed procedure including handoffs, a disqualification, or a self-serve routing. A qualified sales lead bills at $9.99. Running Fin standalone on another helpdesk carries a 50-outcome monthly minimum and no seat fees; on the Intercom platform it stacks on seats from $29 to $132 a month.
Zendesk charges $1.50 per verified resolution on committed volume and $2.00 pay-as-you-go, on top of seats. Every Suite and Support plan includes a small allowance, 5 automated resolutions per agent per month on Team plans and 10 on Suite Professional.
Salesforce Agentforce launched at $2.00 per conversation, whatever happened in it. In May 2025 the company added Flex Credits at $0.10 per agent action ($500 per 100,000 credits, 20 credits per action), and it has since layered on per-user licenses. The result is three simultaneous pricing models for one product.
Sierra publishes nothing. The rate is negotiated per contract. Third-party guides estimate roughly $1 to $2.50 per successful resolution, a figure Sierra has never confirmed. What is documented is the mechanic: charges trigger on agreed outcomes like resolved conversations and saved cancellations, and escalations to humans generally carry no outcome fee, in Taylor's own description.

The metric nobody can agree on
Per-outcome billing sounds like the perfect alignment of incentives: the vendor eats the cost of failure. The problem is that the denominator, and even the numerator, are not settled.
Ada is the only vendor that publishes both of its key metrics on the same base. Across more than 550 deployments, its containment rate, meaning conversations that never reached a human, averages about 72 percent. Its automated resolution rate, meaning problems actually fixed, averages about 52 percent on those same conversations. The 20-point gap is customers who gave up, went silent, or came back later through another channel. A headline number without a written definition is probably containment.
Billing definitions make the gap commercially interesting. Fin's own pricing FAQ counts an outcome when the customer confirms resolution or when "they don't ask for more help after Fin responds." Silence is billable. Zendesk at least attaches the word "verified" to its unit. Sierra's definition lives in the contract.
The claimed rates climb fast. Fin's cross-customer average went from 66 percent across 6,000 customers in October 2025 to 76 percent across 12,000 in 2026, and its own caveat, published in the launch post, notes that a 66 percent resolution rate is not 66 percent of the work. Decagon publishes case-study figures like 80 percent deflection for one customer while its own glossary benchmarks e-commerce at 55 to 75 percent and financial services at 25 to 45 percent. Against all of these, an Ada and NewtonX consumer study from March 2026 found that 24 percent of consumers said AI fully resolved their most recent service chat.
What buyers should do about it
The structure rewards measurement by the party being paid. That is the one design flaw per-outcome pricing could not engineer away. Practical counters, in order of leverage:
- Negotiate the definition of a billable resolution as a contract term with teeth: confirmed resolution, or resolution plus no repeat contact on the same issue across any channel within a defined window.
- Audit reopened tickets and channel-hop-to-voice volumes yourself, quarterly, before the invoice lands. Fin refunds outcomes on reopened conversations; whether your vendor's instrument catches reopenings is a diligence question, not a feature assumption.
- Convert every quote to effective cost per truly resolved case at your own volume before comparing vendors. A $0.99 rate on a generous definition can beat a $1.50 rate on a strict one, or the reverse, and the math is specific to your queue.
The seat era at least had an honest metric: you could count humans. The outcome era sells accountability and grades its own homework. The line item is only trustworthy if the buyer holds the pen on the definition.
Sources
- Salesforce Completes Acquisition of Fin
- Salesforce Is Acquiring Fin (Intercom): What It Means
- Intercom Fin pricing and Fin AI Agent outcomes
- Bret Taylor's Sierra raises nearly $1B at $15.8B valuation (CNBC)
- Sierra AI pricing mechanics and outcomes, incl. Taylor quote (Aissist)
- Sierra AI pricing estimates (Coworker AI)
- Decagon funding rounds and valuation (RevenueMemo), Sacra Decagon profile
- Zendesk AI agent pricing (kapa.ai roundup), Macha breakdown
- Salesforce Agentforce Flex Credits press release, SaaStr on Agentforce pricing history
- Ada: automated resolution guide (52% vs 72% containment)
- Lorikeet resolution-rate benchmarks 2026
- Fin AI resolution rate benchmarks
- Aissist industry benchmark page, incl. 24% consumer study and Decagon glossary ranges
