Bulk export conversation content via UI

Hi ​@AlexCaravitis 

Jennifer here from Intercom

Thanks for this feedback, I will pass it on to our Product Team for you 🚀

Really in need of this as a regulated business entity

Any update on this suggestion in the road map? It would be an extremely useful feature 🙏

Any updates?

Any updates on this? Really need this done via workspace UI for ease of access. Extremely useful feature for our business. 

​@Ken A ​@Ashley Foreman ​@Ross Findlay ​@AlexCaravitis - I’ve used a workaround with Claude. Inside Claude (whether in the chat or cowork/code), you can connect to Intercom via the MCP (like a native AI API). I’ve pulled conversations in bulk, analyzed responses, and then directly made edits or ‘pushed’ new public articles directly into Intercom.

Here are the steps that I used: (The set up has a learning curve, and then it becomes super easy and powerful!)

You'll need a Claude account (Pro or Team plan) and admin access to your Intercom workspace. In Claude, go to Settings > Connectors and add the Intercom MCP connector (MCP is like a native AI API). You'll be prompted to authenticate through Intercom, which grants Claude read access to conversations, contacts, and articles. Once connected, the tools are available in any new conversation automatically. Before your first pull, grab the tag IDs you'll be filtering by (audience tags, topic tags, etc.) You can also use Postman to get the IDs  (I had Claude ‘explain the process like I’m a non-technical 5 year old)

You typically want to be specific in what you pull (think of the report filters you’d want if Intercom had a bulk export feature).

• Filters: tag IDs, date range, CSAT score, message count
• A cap: how many conversations to search, and how many full threads to actually read
• A goal: what you want out of it (cluster themes, find gaps, validate a specific topic, etc.)

Here’s an example prompt for finding knowledge base gaps once you’ve set everything up:

"Search conversations from the last 90 days tagged [tag ID & tag ID] with a CSAT score of 4 or 5 and at least 6 message parts. Return up to 100 conversations, making sure the sample is spread across days of the week and times of day rather than clustered in one window. Then pull the full threads and cluster them into themes.
For each theme, give me:

• Theme name (short, descriptive)
• Volume (how many conversations fall under it)
• What users are asking (the underlying question or problem, in plain language)
• How support resolved it (the typical answer or fix)
• Resolution consistency - flag if the support response varied significantly across conversations (e.g. different answers, different workarounds, or escalation paths). This usually signals the product behavior is unclear or support doesn't have a canonical answer yet, meaning the article needs product input before it can be written.
• Representative conversation IDs (2-3 examples I can open in Intercom)
• KB coverage status - one of:
  • ✅ Covered (link the existing article)
  • ⚠️ Partial (article exists but is missing key info, note what's missing)
  • ❌ Gap (no article exists)

At the end, give me a prioritized list of the top 10 gaps or partial articles to address, ranked by volume and how clearly the support response could be turned into self-serve content. Call out separately any high-volume themes where resolution consistency is low, since those need product alignment before KB work."

This would also be super helpful for us, especially when it comes to data evaluation