Intercom Conversation Topics: What They Are and How to Turn Them On

Conversation Topics are an Intercom reporting feature that automatically tags every conversation — including ones already closed — based on keyphrases you define, so you can filter and report on volume, response time, and CSAT by subject. They're off by default: nothing gets categorized until you create at least one topic under Reports. Setup takes a few minutes per topic, and one workspace can run as many as it needs.

What are Conversation Topics?

A Conversation Topic is a saved rule: a list of keyphrases, and optionally exclusion keywords and conditions, that Intercom checks every conversation against. Any conversation that matches gets the topic applied automatically, including conversations that happened before the topic existed — Intercom applies topics retroactively, which is the main thing that separates them from tags.

Matching runs on "or" logic across your keyphrases, not "and" — a conversation matches if it contains any one of them. A few rules worth knowing before you write your first list:

• Words in a multi-word phrase match up to three words apart, so "tracking time" also matches "tracking my work time."
• Root words match their variations — "planning" also catches "plan" and "planner."
• Wrap a phrase in quotation marks to require the exact sequence instead of loose proximity matching.

Once you've created a handful, Intercom's Topic Suggestions (Reports → Conversation Topics → Suggestions) starts proposing new ones once a workspace passes roughly 5,000 conversations, generated nightly from short, English-language, inbound Messenger conversations that didn't touch a workflow or Custom Answer. A suggestion isn't live until you save it — nothing gets categorized off a suggestion alone.

Topics vs. Fin's AI-generated topics

This is the distinction that trips people up, because Intercom uses "topics" for two unrelated features. The Conversation Topics report — the one this article is about — is keyphrase-based and something you configure. Fin's Topics Explorer generates its own topics and subtopics from conversation content using AI, with no keyphrase setup at all. Intercom is explicit that these two features are not related and work independently.

In practice: use the Topics Explorer to see what Fin thinks your customers are asking about, unprompted. Use Conversation Topics when you need a defined category you control — one that shows up as a filterable property in every report, survives a schema you set once, and doesn't shift if Fin's model changes how it clusters things.

Topics vs. attributes vs. tags

Intercom gives you three ways to categorize conversations, and they're built for different jobs. Intercom's own guidance is to use a combination of all three rather than picking one.

A rough way to choose: if you need a number in a report (how many conversations were about billing this month), reach for a topic. If you need to route or filter live in the inbox based on a structured value, use an attribute. If you need to mark one exact reply as noteworthy inside an otherwise unremarkable conversation, use a tag.

How to turn on a Conversation Topic

Topics aren't a single switch — each one is its own saved rule, and a workspace has none until you create the first.

1. Go to Reports → Conversation Topics → Topics, then click New Topic.
2. Enter the keyphrases that should trigger this topic, one per line.
3. Add exclusion keywords for phrases that would otherwise cause false matches.
4. Choose whether matching looks at customer language, teammate language, or both.
5. Add conditions if you want to scope the topic further — by product area or another conversation attribute, for example.
6. Check the live preview of matching conversations before saving, and adjust the keyphrase list if it's pulling in conversations that don't belong.
7. Name the topic and save it.

Repeat per topic. There's no bulk import for keyphrase lists, so most teams start with 5–10 topics covering their highest-volume intents, then add more once Topic Suggestions has enough data to propose them. If you're not sure which intents are worth a topic first, the Fin Topics Explorer is a reasonable place to check what customers are actually asking before you write the keyphrase list.

Where topics show up once they're on

Once saved, a topic appears under Conversation attributes in the inbox for every matching conversation, and teammates can add or remove it from an individual thread with the + or × icon — doing so doesn't change the topic's keyphrase definition, just that one conversation's tagging.

Everywhere else, a topic behaves as a filterable property:

• Filter any standard report by topic, or build a custom report segmented by topic.
• Compare volume, median first response time, median time to close, and CSAT per topic, with trend against the previous period.
• Pull topic values through conversation exports and the API.

Two limits worth knowing going in: topics can't be manually assigned to standalone tickets, and converting a conversation to a ticket removes it from the Topics report. If your workspace converts a lot of conversations to tickets, treat topic volume as a floor on ticket-eligible intents, not a full count. For the rest of what a topic's numbers do and don't tell you, see every Intercom metric explained and seven Intercom reporting gaps.

The rest of what we publish on running Fin — the audit framework, measurement, and handoff quality — is indexed in the Fin library. If you're segmenting Fin's numbers by topic, pair it with what Fin's resolution states actually count.

Frequently asked questions