Intercom's pay-per-resolution AI play
Intercom at $343M/year
Intercoms move into AI was really a move to rebuild its business model before the old one eroded. Customer support teams were shrinking, which threatened per seat SaaS revenue, so Intercom pushed Fin as a product that gets paid when a ticket is actually solved, then tied that agent to its inbox, help center, customer data, and human handoff tools so AI automation would pull customers deeper into the platform instead of replacing it.
-
The core idea was not just adding a chatbot. Intercom had already learned from Resolution Bot that scripted and heavily trained bots were too much work for most customers. Fin used LLMs to answer from docs with much less setup, which made automation broad enough to matter to mainstream support teams.
-
This also meant accepting short term cannibalization. Intercom said AI could replace some paid seats, but believed pay per resolution would more than offset that over time, because a bot that solves large volumes of tickets can produce more revenue than one human seat while still saving the customer labor cost.
-
The competitive wedge was the full stack. AI native players like Sierra and Decagon were built to automate support from the outside in, while Zendesk initially leaned more into agent assist and workflow tooling. Intercoms bet was that the winning product would connect AI, the ticket inbox, customer records, docs, reporting, and human escalation in one system.
Where this heads next is toward a customer service market that is priced on outcomes, not headcount. If Intercom keeps improving resolution rates and expands Fin across more channels and third party stacks, it can turn AI from a defensive response into the product that defines the new system of record for internet support.