Use casesTriage

Catch the support conversation that is about to go wrong

Mitra checks Intercom every hour for conversations going the wrong way: replies stacking up, a frustrated tone, a named account waiting. Each one goes to Slack with the thread summarized and linked. Anything with a real product defect behind it also opens a Linear issue.

Runs: Every hour
Used by: Customer Support

The prompt

Every hour, find the Intercom conversations going badly and post each one to #support-escalations with a summary and link.

Intercom

Linear

Slack

Build this agent

200 credits to start · no credit card

What the agent actually does.

Sweep the open conversations
Every hour the agent reads the open Intercom threads, with who replied last, how often and in what tone.
Score what is heating up
Repeated replies, frustration in the wording and a name from your account list all count against a thread.
Post it with context
Each flagged thread goes to Slack with the customer, how long they have waited and a link.
File the real defects
When a product problem sits behind the thread, the agent files it in Linear and links the conversation.
Skip what it already raised
Threads flagged on an earlier run stay quiet unless they get worse, which keeps the channel worth reading.

Make it yours.

For support teams with no escalation manager, who find out an account was angry when it asks for a call. Change any of this by saying so — there is nothing to rewire.

Watch only accounts above a certain plan, and check every fifteen minutes.
Skip Linear and message the assignee and their lead directly.
Add your Zendesk tickets to the same sweep, so one agent watches both.

Questions, answered.

How does Mitra know a support conversation is escalating?

Mitra reads the thread rather than counting fields: how often the customer came back, how the last message is worded, how long they waited, and whether the account is one you named. Any of those can raise a conversation. Tell Mitra what else matters and the next sweep includes it.

Will Mitra flag the same conversation every hour?

No. Mitra records what the agent raised on earlier runs and stays quiet on a thread unless it gets worse, so the escalation channel keeps its weight. A conversation that picks up more replies or a harder tone is raised again, with what changed.