FIN is probably the most mature product in the AI support agent category, but its "up to 50% autonomous resolution" promise only holds under very specific conditions.
How does Intercom count a FIN resolution?
The official definition is narrower than most people assume: a resolution counts when the customer confirms Fin resolved the issue, or doesn't request more help after an answer, and only if Fin actually answered. An exchange of pleasantries isn't one.
It remains an operational definition, not a quality one: a customer giving up because the answer was off leaves exactly the same trace as a satisfied one. Across audited accounts, reopen rates on Fin-resolved conversations sit between 12 and 27% at day 7.
Check this before comparing any series: Intercom revised how its metrics are computed, dropping "Fin Constrained" conversations from the denominator. The displayed rate went up without the product changing.
What do you need in place before deploying FIN Agent?
- An up-to-date knowledge base. FIN cites. If your articles are 18 months old, your customers get stale answers.
- Intentional routing. Letting FIN handle billing or advanced technical requests is counter-productive, the cost of a wrong answer dwarfs the tier-1 saving.
- A quality loop. Without a weekly audit of 30 to 50 FIN conversations, you'll miss semantic regressions when the KB evolves.
What does FIN Agent really cost?
The official price is $0.99 per outcome, and it's only a fraction of the total. Add seats, Copilot, KB maintenance (0.5 to 1 FTE depending on size), source curation, and the reopen cost (CSAT, churn). Actual spend commonly runs 60 to 80% above seat cost. Net, we observe a break-even from around 4,000 tier-1 conversations per month. The line-by-line breakdown is in our analysis of outcome-based billing.
When should you avoid FIN today?
- High-value B2B support where one wrong answer loses a contract.
- Heavily regulated verticals (health, banking) without legal sign-off on each output.
- Products in fast pivot: the KB can't keep up, FIN hallucinates with confidence.
FIN is an excellent productivity lever once the foundations are there. Without them, it's a paid-per-conversation NPS killer.