How to use Claude with Intercom: a practical guide

Intercom is where most growth-stage SaaS companies handle customer support, onboarding conversations, and proactive messaging. Claude is useful at several points in the support workflow — particularly anywhere that involves writing, classification, or turning raw conversation data into structured content.

There is no native Claude connector inside Intercom (Intercom has its own AI features, Fin, built in). So the approaches here are either copy-paste workflows or Zapier/Make automations that move data between the two tools.

Writing better responses faster

The most immediate use case: you are in the middle of a support queue and need to write a clear, accurate, empathetic response to a customer who is frustrated or confused.

The workflow:

• Read the customer's message and understand what they need
• Paste the key context into Claude: what the customer is asking, what the actual answer is (or what you know about their situation), and any tone constraints
• Prompt: "A customer sent this message: [paste]. The actual situation is: [what you know]. Write a response that: answers their question directly, acknowledges their frustration if appropriate, and sets clear next steps. Tone: warm and direct. No corporate language."
• Edit the response for any account-specific details and send from Intercom

This is fastest when you are writing from knowledge — you know the answer, you just need to write it well quickly. It is less useful when you are still investigating the issue, because Claude cannot investigate for you.

Summarizing long conversations

Intercom conversations sometimes run for days or weeks, with multiple agents handling them at different points. When you pick up a conversation mid-stream, getting up to speed can take 5 to 10 minutes of reading.

How to do it:

• Copy the full conversation thread (Ctrl+A in the conversation, or export via Intercom's conversation export)
• Paste into Claude: "Summarize this customer support conversation. Include: what the customer originally needed, what has been tried so far, what is still unresolved, and what the next step should be. 5 bullets max."
• Use the summary to get current, then respond

This is especially useful for team handoffs and for managers reviewing escalations.

Building and improving your knowledge base

One of the highest-leverage things you can do with Intercom data is turn frequently asked questions into knowledge base articles. Most teams know they should do this but find the writing work too slow.

How to do it:

• Look at your Intercom conversations over the last 30 to 60 days and identify the questions that came up 3 or more times
• Copy 2 to 3 versions of the same question asked by different customers
• Prompt: "Customers keep asking about [topic]. Here are three examples of how they ask it: [paste]. Write a help center article that answers this question. Format: short direct answer first, then explanation, then step-by-step instructions if applicable, then related questions. Plain language — written for non-technical users."
• Publish to your Intercom knowledge base or help center

This converts existing support conversations into content that prevents future support volume. It also gives Intercom's AI features (Fin) better source material to answer from.

Classifying and routing incoming conversations

If your Intercom inbox receives high volume from different customer types — technical questions, billing issues, bug reports, general feedback — you may spend time manually triaging and routing before the actual support work starts.

How to do it with Zapier:

• Trigger: new conversation opened in Intercom
• Action: send the conversation opener to Claude with: "Classify this support message into one of: [billing, technical-bug, how-to question, feature request, general feedback]. Return only the category name."
• Action: use Zapier to apply the corresponding Intercom tag based on Claude's response

This is not perfect — Claude misclassifies ambiguous messages — but it handles 70 to 80 percent of straightforward volume correctly and reduces the time humans spend on routing.

Writing proactive outreach messages

Intercom is also used for proactive messaging — reaching out to users who hit a milestone, started a trial, or have not returned in a while. Writing these messages at scale is tedious.

How to do it:

• Define the trigger: user has been on the free plan for 14 days and has not used [key feature]
• Prompt: "Write a short in-app message for a user who has been on our free trial for 14 days but has not [done the key action]. We want to encourage them to try it. Max 3 sentences. Tone: helpful and direct, not pushy. No 'just checking in' phrasing."
• Use the result in Intercom's message composer

For onboarding sequences specifically, write each message for the specific moment — what the user has done, what they have not done, and what one thing would move them forward. Claude writes better messages when you give it behavioral context, not just "write a nurture email."

What does not work well

Claude cannot send messages through Intercom. All Claude-written content needs to be copied into Intercom manually or via automation. There is no live integration that writes back to Intercom.

High-sensitivity customer situations need human judgment. When a customer is threatening churn, reporting legal issues, or is genuinely distressed, the response needs human judgment about what to commit to and what not to. Claude can help you write the response once you know what you want to say, but the judgment call is yours.

Intercom Fin already handles common questions if you have a good knowledge base. If you have invested in your Intercom knowledge base, Fin (Intercom's built-in AI) covers most of the easy, common-answer questions. Claude's value in Intercom is highest for complex responses, knowledge base authoring, and conversation summarization — not for the routine Q&A that Fin can handle.

The setup that gives you the most leverage

For support teams without a developer: start with conversation summarization for handoffs and use Claude for drafting responses to complex or frustrated customers. Both of these are high-value, low-setup.

For teams with a developer: the Zapier classification + tagging flow is worth building once volume justifies it. More impactful long-term: a scheduled job that pulls Intercom conversations tagged as "common question" once a week and drafts knowledge base articles from them, which a human then reviews and publishes.

---

This guide is part of the Claude + Tool series — practical guides for using Claude alongside the tools your team already uses. 14 guides published.

Further reading

• Discover tools that work with Claude — the connectors directory