# Auto-triage new Kustomer conversations before an agent replies > Every new support conversation gets a severity, topic tags, an internal summary and a Slack alert routed by urgency, before anyone picks it up. - Workflow type: agent - Services: Kustomer, Slack Bot - Categories: Customer Support - Published: 2026-08-07 ## What it does - Reads every new Kustomer conversation the moment it opens, including the customer's own words and who they are. - Checks whether that same customer has contacted you recently about the same thing, so repeat issues stand out instead of starting from scratch. - Assigns a priority: P1 for outages, failed payments, legal issues or cancellation threats, P2 for anything blocking the customer from working, P3 for general questions. - Tags the conversation, leaves a short internal note with the reasoning, and posts a Slack alert carrying the customer name, a one-line summary, the priority and whether this is a repeat contact. ## What you'll need - A Kustomer account with permission to update conversations and add internal notes. - Kustomer set up to notify this workflow whenever a new conversation is created. - A Slack workspace, ideally with two channels: one for urgent escalations and one for everyday triage. ## Prompt I want every new Kustomer conversation triaged automatically before a human agent picks it up. Trigger this with a webhook: Kustomer supports outbound webhooks that subscribe to Conversation Create events, so a new conversation being created in Kustomer should start the workflow. The webhook payload gives you the conversation id to work from. First, gather the full picture. Use Kustomer Get Conversation by ID to pull the conversation itself, then Get Messages by Conversation to read the customer's actual words rather than just the subject line. Use Get Customer by ID to load the requester so you know who they are. Then call Get Conversations by Customer to look at their recent history and work out whether this person has already contacted us recently about the same thing. Treat it as a repeat contact when a recent conversation covers substantially the same issue, not merely because they have written in before. Based on everything you read, assign a severity. P1 is for outages, service being down, payment or billing failures, legal threats, and any signal the customer is about to churn or cancel. P2 is for problems that block the customer from doing what they are trying to do, but where the product is otherwise working. P3 is for questions, how-to requests, and general feedback. If a case sits between two levels, choose the higher one and say why in the note. Also pick a small number of topic tags describing what the conversation is actually about, for example billing, outage, integration, onboarding, or cancellation. Write the result back into Kustomer. Use Update Conversation to set the severity and the topic tags on the conversation so the support team can filter and sort by them. Then use Create Note by Conversation to post a short internal triage summary: a one-line statement of the issue in plain language, the severity you assigned and the reason for it, whether this is a repeat contact and what the earlier conversation was about, and anything the agent should know before replying. Keep the note to a few lines. It is a running start for the agent, not a transcript. Finally, alert the team in Slack using Send a Message. Route by severity: P1 goes to the escalations channel, and P2 or P3 go to the general triage channel. Every alert should carry the customer name, the one-line issue summary, the severity, and whether this is a repeat contact. Make the repeat contact flag visually obvious, since a customer writing in twice about the same unresolved problem is the thing most worth catching. Include a link back to the conversation in Kustomer so an agent can jump straight to it. Keep alerts to a few lines and use Slack formatting, where bold is written with single asterisks. One hard rule: this workflow must never send anything to the customer. It writes internal notes and Slack alerts only. Do not create or reply to customer-facing messages in Kustomer, and do not use any operation that would add a message to the conversation. Per Kustomer's own documentation, creating a message records an instance on the conversation timeline, so stay strictly on notes plus Slack. A human always sends the first reply. If anything fails, for example the customer cannot be loaded or the history lookup returns nothing, still post the Slack alert with whatever you do know and say plainly in the note which part was unavailable, rather than silently skipping the triage. ## How to customize - Redefine what counts as urgent. Move billing questions up to P1, or add categories of your own such as shipping delays or onboarding blockers. - Point the alerts wherever you want. Route P1 to an on-call channel, or send everything to a single channel if your team is small. - Change how far back the repeat contact check looks, and how much detail the internal note includes. ## FAQ ### Will this ever send anything to my customer? No. It only writes internal notes inside Kustomer and posts alerts in Slack. Nothing it produces is delivered to the customer, so a human always sends the first real reply. ### How does it know whether someone has contacted us before? Before deciding anything, it looks at that customer's recent conversations. If a previous one covers the same issue, the alert and the internal note both flag it as a repeat contact so you can spot problems that were never really resolved. ### What happens if it gets the priority wrong? The priority is written onto the conversation as an ordinary field your team can edit, and the internal note explains the reasoning behind it. Anyone can override it in seconds, and nothing is blocked waiting on the decision. ### Do I need two separate Slack channels? It works best with two, so urgent issues are not buried among routine questions. If you would rather keep everything in one place, point both routes at the same channel. ### How fast does the triage happen? It runs as soon as the conversation is created in Kustomer, so in most cases the severity, tags and summary are already there by the time an agent opens the ticket. Use this prompt in General Input: https://www.generalinput.com/prompts/auto-triage-new-kustomer-conversations-before-an-agent-replies