Triage a support ticket
Summarize, categorize and draft a grounded reply for each incoming ticket.
Support ticket triage decides whether a customer waits four minutes or four hours. This workflow reads each incoming ticket, categorizes and prioritizes it, drafts a reply grounded in your actual documentation, and routes the ones that need a human to the right human, with context attached.
What does this workflow do?
For every incoming ticket the agent produces four things: a summary, a category (bug, billing, how-to, feature request), a priority based on severity and customer context, and a drafted reply.
The reply is the part that has to be right. The agent answers from your knowledge base, help docs and past resolved tickets, and cites where each claim comes from. When the answer isn't documented anywhere, it doesn't improvise. It asks the person who'd know, an engineer for a technical question, finance for a billing edge case, and the answer flows back into the draft and into the knowledge base. The second customer who asks gets an instant answer.
How does it work?
- The workflow sweeps for new tickets. On a frequent schedule, every few minutes to hourly, it picks up new messages in your shared Gmail inbox or Slack support channel.
- The agent classifies it. Category, severity and priority. It recognizes VIP customers and active deals from your CRM if connected.
- It drafts a grounded reply. Answered from your docs and past tickets, with source links a support agent can verify in seconds.
- It escalates the unanswerable. Undocumented questions become information requests to the right teammate. Suspected bugs become Linear issues with the ticket linked, deduplicated against existing ones.
- A human approves or the reply sends. You choose per category. Many teams auto-send how-to replies and keep approval on billing and bugs.
Why ground triage in a knowledge base?
An AI reply is only as good as what it's grounded in. Generic AI support tools answer from whatever text they can find, which is how customers get confidently wrong answers about your refund policy. This workflow answers from your maintained, permission-aware knowledge base, flags contradictions between docs for a human to settle, and closes documentation gaps by asking people, so the knowledge compounds with every ticket.
Triage also improves with feedback. When an agent recategorizes a ticket or edits a reply, the correction persists. Priorities tune themselves to how your team actually works the queue.
Works with
Slack, Gmail, Notion, Linear, GitHub. Runs on a frequent schedule, sweeping for new tickets each time.
Frequently asked questions
Does the reply go out automatically?
Your choice, per category. Common setup: how-to answers with strong sources auto-send, billing and bug replies wait for approval. Every draft carries source links so review is fast.
What happens with tickets the agent can't answer?
It says so, routes an information request to the teammate who'd know, and holds a draft for the customer. The teammate's answer completes the reply and gets saved, so the gap closes permanently.
Can it file bugs from tickets?
Yes. Suspected bugs become Linear issues with the ticket summary, reproduction hints and a link back to the customer thread, deduplicated against existing issues.
How does it prioritize?
Severity signals in the ticket, customer context from your CRM (plan, deal stage), and your own rules. Corrections from the team adjust prioritization over time.
Which helpdesks does it work with?
Gmail shared inboxes and Slack channels today. If your helpdesk forwards ticket notifications to either, those tickets are covered too.