AI Pulse · Business Models & Winners

Intercom’s AI agent, Fin, is closing in on $100 million in annual recurring revenue. That’s roughly half of Intercom’s total revenue, and it’s running on a pricing model most SaaS companies avoid: pure outcome-based charging at $0.99 per resolved customer conversation.

No seats. No tiers. You pay when the agent closes the loop.

According to the original report, Intercom’s CFO confirmed that Fin’s resolution rate has climbed from around 25% at launch to between 65% and 70% today. That means the product got better while customers paid more only because it delivered more. The incentive alignment is tight. If Fin doesn’t resolve the issue, Intercom doesn’t get paid for that interaction.

This matters because outcome-based pricing has been a niche experiment for years, mostly talked about but rarely scaled. Intercom is proving it can outpace traditional per-seat revenue when the AI actually works. The model rewards improvement. As resolution rates rise, so does revenue per customer, without renegotiating contracts or upselling features.

For operators, the lesson is straightforward. If you’re deploying an AI agent, whether it’s handling support, lead qualification, or internal workflows, you need to track resolution or completion rates from day one. That metric tells you whether the agent is worth expanding or needs tuning. It’s also the number that justifies cost, internally or to a vendor. When you’re building or buying AI that touches customers, outcome tracking isn’t optional anymore. It’s the business model. Tools like the Omni Command Centre let you monitor these kinds of agent performance metrics across your stack in one place, so you’re not flying blind on what’s actually closing versus what’s just logging activity.

Intercom didn’t invent AI agents, but they’ve shown a repeatable way to charge for them. That’s the part worth copying.