Auto-escalate angry support tickets to Slack

01/The build

How to build it

1. 01

   Decide what 'angry' actually means — on paper first

   Needs a human

   Write the rule before you build it, because 'angry' is a policy choice, not a feature. Three buckets work well: sentiment (the message reads as negative or hostile), high-stakes wording (specific phrases that mean money or reputation is on the line), and signals (a repeat contact within 24h, a VIP/high-LTV customer, an all-caps shout). A ticket escalates when it trips a high-stakes phrase outright, or when negative sentiment stacks with at least one signal. Mildly annoyed is not angry — escalating everything is the same as escalating nothing.

2. 02

   Trigger on every inbound customer message in Gorgias

   Runs itself

   Build a Gorgias Rule that fires on ticket created and on new customer message (not on agent replies — you don't want your own team's words scored). Gorgias Rules can run actions directly, or hand off to glue (n8n / Zapier) for the heavier logic in the next steps via an HTTP request action. Trigger on each customer message, not just ticket creation, so a calm thread that turns hostile on the third reply still escalates.

3. 03

   Score the sentiment of the message body

   Runs itself

   Use Gorgias's built-in sentiment/intent detection where your plan includes it — it tags incoming messages as negative or offensive — or pass the message text through a sentiment classifier in your glue step. Treat the score as one input, not the verdict: sentiment alone is noisy (sarcasm, blunt-but-fine customers, non-native phrasing), which is exactly why it gets combined with the wording and signal passes rather than firing on its own.

4. 04

   Run the high-stakes keyword and signal pass

   Runs itself

   Match the message against a curated list of phrases that mean real risk regardless of tone: 'chargeback', 'dispute', 'lawyer' / 'attorney', 'fraud', 'BBB', 'reporting you', 'never again', 'cancel everything'. These escalate on their own — a flat, polite 'I will be filing a chargeback' is more dangerous than an exclamation-heavy rant. Layer in structural signals: a high ALL-CAPS ratio, this customer's second or third ticket in 24 hours, or a high order count / lifetime spend that Gorgias already surfaces from your Shopify data. Keep the phrase list in one place so it's editable without touching the build.

5. 05

   De-dupe so one ticket pings exactly once

   Runs itself

   Before posting, check for an 'escalated-slack' tag on the ticket; if it's there, stop. If it isn't, post the alert and add the tag in the same run. Without this, a customer who fires off three angry follow-ups in a row triggers three identical Slack pings, and the team learns to tune the channel out. One ticket, one alert — re-escalate only on a deliberate condition, like a brand-new high-stakes phrase appearing later in the thread.

6. 06

   Post a Slack alert with the context to act on

   Runs itself

   Send to a dedicated channel (e.g. #support-escalations) via a Slack Incoming Webhook for the simple version, or chat.postMessage with Block Kit for a richer card. Include what a person needs to triage without opening anything: customer name, the reason it escalated (the matched phrase or sentiment), the order count and total spend, and the actual message snippet. Add an 'Open in Gorgias' button linking straight to the ticket (https://YOURDOMAIN.gorgias.com/app/ticket/{ticket_id}). The goal is a glance-and-go, not a notification that makes someone go digging.

7. 07

   Route the real emergencies harder than the rest

   Runs itself

   Not every escalation is equal. A high-stakes phrase or a VIP gets an @here and lands in the on-call channel; ordinary negative-plus-signal tickets post quietly to #support-escalations. Branch on business hours too: during the day, the support channel is enough; after hours, a chargeback-language ticket can route to a separate channel or an on-call DM so it isn't waiting until morning. Routing is automated — who picks it up is still a person.

8. 08

   Review the misses weekly and tune the list

   Needs a human

   Log every escalation to a sheet or a thread with its trigger reason, then spend ten minutes a week on two columns: what got escalated that shouldn't have (tighten the phrase list or raise the sentiment threshold) and what blew up that never pinged (add the phrase you missed). An escalation rule that nobody tunes drifts into noise within a month — the weekly pass is what keeps the channel something the team actually trusts.

02/The stack

What it’s built on

Gorgias

The helpdesk where tickets land. Its Rules engine fires on inbound messages, its sentiment detection supplies one of the escalation inputs, and it already pulls each customer's Shopify order history so the alert can show real spend. Rules can post to Slack directly for the simple build, or hand off to n8n / Zapier for the scoring, de-dupe, and routing logic.Built-in AI sentiment detection sits on Gorgias's higher / add-on tiers. You don't need it to ship this — the high-stakes keyword and signal pass runs on the Rules engine available on any plan, and you can add sentiment later.

Slack

Where the alert lands and the team acts on it. A dedicated escalations channel keeps the fires out of general chatter; Block Kit gives you the context card and the deep-link button.Incoming Webhooks and chat.postMessage are free on every Slack plan, including the free tier — there's no upcharge to receive these alerts.

03/Questions

Before you build