Intercom: turn AI topics into a customer improvement list

Your customer gets quick support replies but asks the same questions every month. Before the next account meeting, you need to see which problems repeat and decide what to change. Intercom's customer portal can now show AI-grouped topics for the customer's company, with links back to the conversations behind the numbers.

Intercom introduced the new Insights tab on October 29. Here we use it for a joint review between a support lead and the customer's contact person. The result is an improvement list with evidence, an owner, and a follow-up date.

Source: Intercom: Customer Portal Improvements.

Check access before the customer meeting

This routine suits teams already using Intercom's customer portal. It is part of Intercom Help Center and Inbox, not a feature included with Fin on platforms such as Salesforce. AI topics and subtopics come from Topics Explorer and require a Pro plan, according to the documentation. Without it, the rest of Insights is available, but the topic breakdown is not.

The portal requires a Help Center custom domain, Cookie Forwarding, and Identity Verification. The customer needs an email address, an association with a company, and at least one ticket associated with that company. Use a computer: Intercom says the customer portal is not yet supported on mobile devices.

Sources: Intercom: how the customer portal works. Intercom: setup and limitations.

An administrator carries out these steps in Intercom. The menu names follow the English documentation.

1. Open Settings > Inbox > Tickets > Customer portal and select the right Help Center.
2. Under Insights visibility, turn on Show Insights. Use Add visibility conditions if only selected contacts should see the report. Without conditions, everyone with portal access sees the tab.
3. Select Show topics in Insights to display AI topics. Use the preview to check which users are included.
4. Separately check which tickets and conversations the customer can see. Conditions under Ticket visibility restrict matching users to their own tickets and conversations; they do not grant extra access. The same rules apply to both types.

Source: Intercom: configure Insights and visibility.

Workflow: agree on one improvement with the customer

This is our suggested meeting routine, not a built-in automatic analysis or a prompt. AI groups the conversations. You decide what the group means and what needs changing.

• Open the customer's Insights and note the filters you use. Read the conversation volume and topics. Choose a recurring topic that the customer finds time-consuming, not automatically the largest one.
• Open underlying tickets or conversations from the report. Read a few actual examples together. Do they concern the same problem, or has AI grouped different questions under one heading? In particular, check whether Swedish conversations have been grouped sensibly.
• Write a concrete observation with links to the examples. Separate what the conversations show from your explanation. Lots of invoice questions do not, by themselves, establish that the invoices are wrong.
• Agree on a change you can follow up. If the examples show that the customer cannot find invoices, the change might be clearer instructions for that step. Name the owner and decide when to check whether the same question returns.

Intercom documents that customers can filter Insights and open the ticket or conversation behind a row. Topics are AI-generated labels, not established causes of the problems.

Source for report features: Intercom: customer portal Insights.

Copy this improvement list into your meeting notes

Complete one entry for each agreed change.

• Customer and selection: company, review date, and filters used.
• Recurring question: topic name and what you found in the actual conversations.
• Evidence: links to relevant tickets or conversations that participants can access.
• Agreed change: what you will change, or which question you need to investigate first.
• Owner and date: who does the work and when you will follow up.
• Next check: whether the same question returns in a comparable selection after the change.

Save the list where you already keep the customer's meeting notes. At the next review, compare the same type of selection and read new examples. Fewer questions may be a good sign, but changes in the customer's activity can also affect volume. Judge the change from the conversations, not just the chart.