Fin
Intercom ships Fin surface faster than most support teams can change their settings. This is everything we have written about running it: the audit framework, the numbers worth reporting, where handoffs go wrong, and what a resolution costs.
New here? Start with the audit framework, then check your own numbers against what resolution rate leaves out.
Start here
The audit framework, the numbers that tell you whether it worked, and the version we run for you.
How to QA Intercom Fin
A five-check audit: resolution accuracy, handoff quality, source citation, policy compliance, and CSAT on the threads Fin handled. Includes the weekly sampling workflow.
Is Fin actually resolving conversations?
Resolution rate, surveyed CSAT, handoff quality, and repeat contact, plus why a single Fin-versus-teammates number blends work that is not alike.
Fin audit
We read a slice of your Fin conversations and send back a scored report: where each thread ended up, what the answers missed, and the changes worth making first.
Audit and QA
Reviewing what Fin said, on a rubric, often enough that a drift in answer quality shows up before CSAT does.
QA tools for Intercom teams
Eight options compared, from native Scorecards and Monitors through to building your own, with the case for staying native.
Calibrating the AI judge
An AI reviewer scoring Fin is only as good as its agreement with your humans. How to measure that agreement and fix it.
Fin AI quality monitoring
Score the conversations Fin answered first and put the weak ones in front of a human, without exporting anything.
Reviewing AI support responses
The review loop itself: what gets sampled, who adjudicates, and where the decision lands.
Fin Audit Dashboard
Every Fin-first conversation with its outcome and rubric score in one view. Announced, not shipped.
Measuring Fin
Intercom reports a resolution rate. These pages cover what that number counts, what it leaves out, and what to report instead.
Seven Intercom reporting gaps
Survey coverage, plan-gated quality tools, and the escalation reporting Intercom does not do, with what to use for each.
Every Intercom metric, explained
What each number in Intercom Reports is actually counting, including the AI ones that look interchangeable and are not.
Building a CSAT report in Intercom
The native report, step by step, and the parts of the Fin picture it will not show you.
Intercom CSAT benchmarks
What a good CSAT score looks like for human agents and for AI support, and why the two need separate targets.
Comparing AI and human CSAT
Split ratings by who handled the conversation so an AI-heavy queue stops hiding inside one workspace average.
Deflection rate tracking
Track the conversations Fin closed without a human, and watch the ones that came back.
Resolution rate
The definition Intercom bills on, and the difference between a closed thread and a solved problem.
Deflection rate
What deflection counts, and why a high rate can mean customers gave up rather than got an answer.
Containment rate
The third number in this family, and the one most often quoted as if it were the other two.
Fin resolution states
Confirmed resolved, assumed resolved, escalated, abandoned: what triggers each state, which of them bill at $0.99, and why a blended resolution rate is mostly silence.
Intercom Conversation Topics
What the keyphrase-based Topics report is, how it differs from the AI topics Fin generates on its own, and the setup steps to turn one on.
Handoffs and escalation
Most Fin complaints are handoff complaints. Where the bot should stop, and what it owes the human it passes to.
AI's job stops at the human handoff
Two Live Chat guests from different corners of support tech landed on the same rule for where AI should stop.
Handoff quality
Score the moment Fin passes a conversation over: whether it happened late, and what context survived.
Fin escalation tracking
See which topics escalate, how often, and which of them should have been an article instead.
What Fin answers from
Fin can only answer from what someone wrote down. The source material, and the holes in it.
Structuring a help center Fin can use
One job per article, the answer in the first sentence, titles phrased the way people ask. A vague title produces a vague Fin answer.
Knowledge gap detection
The questions Fin could not answer from your Help Center, ranked by how often they came up.
What Fin costs
Fin bills per resolution. Both calculators run in your browser and send nothing anywhere.
Keeping up
Intercom ships Fin surface faster than most teams change their settings. What is worth reacting to.
What's new in Intercom
Ongoing release coverage, filtered down to the changes that should make you open Settings.
Every Intercom release in July 2026
The complete list for the month, with the Fin-facing entries called out.
Grok Bot with Intercom
Intercom's own MCP server, five jobs worth giving it, and the things it cannot do.
Get the next one
We publish a Fin piece every few weeks. One email when it lands, nothing else.