Catch churn signals in support chats and loop in the CSM with full context

Step by step

1. 1

   Create a Flow that runs every time a conversation closes

   Flow

   In Customerly, go to AI Studio → Flows → New Flow. Name it Churn signal watcher. Set the trigger to Conversation Closed. This way the Flow runs in the background on every conversation, without any extra work for the support team.

   Conversation Closed is the right trigger here (not "every message") because we want the full context of the resolved interaction, including how the customer felt by the end of it.

   textFlow name

   Copy

   Churn signal watcher
   

2. 2

   Add an AI Mission that classifies the conversation for churn risk

   AI Mission

   Add an AI Mission step. This Mission does not write to the customer. It reads the full conversation history and outputs a structured churn risk classification, which the next step branches on.

   Paste the prompt below into the Mission's instructions. Aura will return a single JSON object you can route on.

   promptAI Mission prompt

   Copy

   You are a senior Customer Success analyst reviewing a closed support conversation for renewal risk signals.
   
   Read the full conversation. Decide whether any of these churn risk signals are present:
   
   1. The customer says they are "evaluating alternatives", "looking at other tools", or names a competitor
   2. The customer asks about cancellation, downgrade, or "pausing" their account
   3. The customer expresses sustained frustration across multiple messages (not a one off complaint)
   4. The customer says the product "is not working for our team" or "doesn't fit our use case anymore"
   5. The customer mentions a budget cut, layoff, or "we have to consolidate tools"
   6. The customer's tone shifts from positive to flatly transactional across the conversation
   
   If two or more signals are present, severity is "high". If exactly one strong signal is present, severity is "medium". If none, severity is "none".
   
   Output exactly this JSON and nothing else:
   
   {
     "severity": "none | medium | high",
     "signals_found": ["<short label of each signal you saw>"],
     "summary": "<two sentence summary of why this customer is at risk, or 'No risk detected.'>",
     "suggested_action": "<one sentence: what the CSM should do this week>"
   }
   

3. 3

   Branch the Flow on the severity score

   Flow

   Add a Conditional step right after the Mission. Read the Mission's severity output and:

   • If severity == "none" → end the Flow. Nothing to do.
   • If severity == "medium" || severity == "high" → continue to the next steps.

   This is the rule that keeps your team from getting paged on every conversation. Only the ones that actually matter fire the rest of the Flow.

   textConditional rule

   Copy

   {{mission.output.severity}} != "none"
   

4. 4

   Send a Slack alert to the account's CSM with full context

   REST API block

   Add a REST API step. Post to a Slack incoming webhook (one webhook per CSM team channel, or use Slack's routing to direct messages by account owner). The message should include the severity, the AI summary, the suggested action, and a direct link back to the Customerly conversation so the CSM can read it in context.

   You'll need:

   • A Slack incoming webhook URL for the CS channel (Slack → Apps → Incoming Webhooks → Add to channel)
   httpHTTP method + endpoint

   Copy

   POST https://hooks.slack.com/services/T0000/B0000/XXXXXXXXX
   

   jsonRequest body

   Copy

   {
     "text": "Churn signal detected on {{contact.email}}",
     "blocks": [
       {
         "type": "header",
         "text": { "type": "plain_text", "text": "Churn risk: {{mission.output.severity | upper}}" }
       },
       {
         "type": "section",
         "fields": [
           { "type": "mrkdwn", "text": "*Customer*\n{{contact.name}} ({{contact.email}})" },
           { "type": "mrkdwn", "text": "*Company*\n{{company.name}}" }
         ]
       },
       {
         "type": "section",
         "text": { "type": "mrkdwn", "text": "*What we saw*\n{{mission.output.summary}}\n\n*Suggested next step*\n{{mission.output.suggested_action}}" }
       },
       {
         "type": "actions",
         "elements": [
           { "type": "button", "text": { "type": "plain_text", "text": "Open conversation" }, "url": "{{conversation._url}}" }
         ]
       }
     ]
   }
   

5. 5

   Open a "Renewal at risk" task in HubSpot on the company record

   REST API block

   Add a second REST API step. Create a HubSpot task on the customer's company record so the risk shows up inside the CRM the CSM uses daily, not just in Slack.

   You'll need:

   • A HubSpot private app token with crm.objects.tasks.write scope
   • The HubSpot owner ID of the CSM (or use a Mission that looks up the right owner by company)
   httpHTTP method + endpoint

   Copy

   POST https://api.hubapi.com/crm/v3/objects/tasks
   

   httpRequest headers

   Copy

   Authorization: Bearer pat-eu1-XXXXXXXXX
   Content-Type: application/json
   

   jsonRequest body

   Copy

   {
     "properties": {
       "hs_task_subject": "Renewal at risk: {{company.name}}",
       "hs_task_body": "{{mission.output.summary}}\n\nSuggested next step: {{mission.output.suggested_action}}\n\nSource conversation: {{conversation._url}}",
       "hs_task_priority": "HIGH",
       "hs_task_status": "NOT_STARTED",
       "hs_timestamp": "{{now}}",
       "hubspot_owner_id": "12345678"
     },
     "associations": [
       {
         "to": { "id": "{{company.hubspot_company_id}}" },
         "types": [{ "associationCategory": "HUBSPOT_DEFINED", "associationTypeId": 192 }]
       }
     ]
   }
   

6. 6

   Tag the contact so future conversations auto route to the CSM

   Flow

   Add a final Set Attribute step (or Tag step) on the Customerly contact. Setting a risk_status attribute means every future conversation from this customer can be auto routed to the CSM via a separate routing Flow, not the support queue. The CSM sees every word the customer says from this point on.

   textContact attribute to set

   Copy

   risk_status = "at_risk"
   

More AI Recipes