Support teams face rising ticket volume and the cost of scaling human agents linearly with growth. Intercom's bet was that an AI agent could resolve a large share of conversations autonomously — and that customers would only pay when it actually worked, rather than per seat.
The challenge
What was deployed
Intercom built Fin, an AI customer-service agent, and pivoted to outcome-based pricing in early 2023 — charging per successful resolution rather than per seat. Fin handles chat and email end-to-end, escalating to humans when needed, and reports its resolution rate as the share of conversations closed without human involvement.
The results
Intercom reports Fin averages a 76% resolution rate across 12,000+ customers, with many above 85% and a 51% out-of-the-box rate. One customer (Clay) saw Fin participate in 90% of conversations and resolve up to 50%, and Intercom cites customers absorbing volume increases up to 690% without proportional headcount. In June 2026, Salesforce agreed to acquire Fin for $3.6B.
Twarx analysis
Original interpretationIntercom's bet wasn't really the AI — it was the pricing. Charging $0.99 per resolution instead of per seat aligns vendor incentives with outcomes, and that model (not the 76% number) is what the $3.6B Salesforce deal actually validated.
Most readers anchor on the resolution rate (~76% reported). The more transferable insight is the business model: Fin charges per successful resolution, not per seat. That single decision re-aligns a vendor's incentives — they only get paid when the agent actually works — and it is why outcome-based AI pricing is spreading. The June 2026 $3.6B Salesforce acquisition is the market putting a price on that thesis, not merely on the technology.
The honesty caveat matters more than usual here: the resolution figures are vendor self-reported, and "resolution" means closed-without-a-human, not verified-correct. For buyers, the practical move is to negotiate on the same outcome basis Intercom prices on — pay for resolved tickets, define "resolved" tightly, and measure CSAT on AI-handled conversations independently.
Illustrative Twarx model
Illustrative monthly support cost: all-human vs Fin blended
10k tickets — all human
60 $/mo (k)
10k tickets — Fin blended
25 $/mo (k)
50k tickets — all human
300 $/mo (k)
50k tickets — Fin blended
125 $/mo (k)
Method & assumptions: Compares a ~$6 fully-loaded cost per human-handled contact against Fin at $0.99/resolution for the 70% it resolves, plus $6 each for the 30% still escalated to humans (blended). A Twarx illustration only — not Intercom's or any customer's actual economics; excludes setup and oversight.
Read the numbers honestly
Most numbers here are vendor self-reported — read them as such.
- Resolution-rate figures (76%, 85%+, 51%) come from Intercom / fin.ai marketing, not independent audits.
- Resolution rate counts conversations closed without a human, which is not the same as verified-correct or satisfying answers.
- The 90% involvement / 50% resolution and 690% volume figures are for specific customers, partly via secondary sources.
- Outcome-based $0.99/resolution pricing is from a third-party listing and may change.
- The $3.6B Salesforce acquisition (June 2026) is reputable press, but reflects market valuation, not operational ROI.
Timeline
Early 2023
Intercom pivots to AI; launches Fin with outcome-based pricing
2024
Fin AI Agent: 51% out-of-box resolution; customers scale volume sharply
Jun 2026
Salesforce agrees to acquire Fin (formerly Intercom) for $3.6B
Frequently asked
How well does Intercom Fin actually resolve tickets?
Intercom/Fin reports an average resolution rate around 76% across 12,000+ customers, with many exceeding 85% and a 51% out-of-the-box rate. These are vendor self-reported figures; resolution rate measures conversations closed without a human, not verified customer satisfaction.
How is Fin priced?
Fin uses outcome-based pricing — listed at $0.99 per successful resolution for chat and email, so you pay when it resolves a query rather than per seat. Voice pricing is quoted separately. Pricing is per third-party listings and can change.
Are these independent results?
Mostly no. The headline performance numbers come from Intercom's own marketing and blog, and the 90%/50% figures are for a specific customer via a secondary source. The $3.6B Salesforce acquisition (June 2026) is from reputable press. Treat the performance figures as vendor-reported.
AI AgentsCustomer ServiceSaaSOutcome-Based PricingROI
TR
Analysis by
Twarx Research Team · Applied AI Research
Twarx researches and deploys enterprise AI agents with a measurement-first method: every metric traced to a primary source, projections labelled as estimates, and limitations stated up front.
AI agentsAgentic AIOutcome-based pricingCustomer service automationSLMs
