Detect customer sentiment and flag churn risk in Intercom

Frustration shows up in the tone of a conversation long before it shows up in your churn dashboard. An Orkeste agent scores sentiment across Intercom threads and alerts your CS team on accounts that need a save play.

How the workflow runs

1. Stream Intercom conversations: The agent reads new messages as they arrive.
2. Score sentiment: Assigns a sentiment and frustration score per message and per thread.
3. Aggregate per account: Rolls up scores to account level, weighted by recency.
4. Human review on high-risk flags: A CS lead reviews flagged accounts before any save play is triggered.
5. Notify CS owner: Posts an alert in Slack with the account, the trigger thread, and a suggested next step.

Human-in-the-loop: Save plays and outbound saves are never triggered without a CS lead's explicit approval.

Works with your stack

• intercom
• slack
• hubspot

Related workflows

• Summarize Intercom support conversations automatically
• Alert on stalled deals in Salesforce
• Escalate high-priority tickets from Zendesk to Slack

Questions about this workflow

How accurate is sentiment scoring?

It's directional, not absolute. The agent surfaces candidates for human review, it doesn't make retention decisions on its own.

Does it handle multiple languages?

Yes. Sentiment scoring works across the major European languages out of the box.

Can we tune what counts as 'at risk'?

Yes. Thresholds, weighting and account segmentation are all configurable per workspace.