How to triage support emails and draft replies with n8n

Customer supportn8n

The problem

A shared support inbox mixes urgent problems with routine questions. Reading, sorting and drafting an answer for every message eats hours, but letting a bot reply on its own is risky.

Tools needed

• n8n
• Gmail
• OpenAI-compatible LLM (gpt-4o-mini in the shared workflow)
• Google Sheets
• Slack

Step-by-step recipe

1. Add a Gmail trigger in n8n that polls the support inbox for new messages.
2. Pass each email's sender, subject and body to an LLM node and ask for structured JSON: category, urgency, a one-sentence summary and a short draft reply.
3. Parse the JSON and append one row per email to a Google Sheet so you have a log.
4. Post the summary, urgency and draft to a Slack channel for the support team.
5. Keep sending manual: a person checks the draft, edits it and replies from Gmail.

Copyable prompt

You are a support triage assistant for [COMPANY]. Read the email below and reply with JSON only, using these keys:
- category: one of [billing, bug, how-to, account, other]
- urgency: low | medium | high
- summary: one sentence
- draft_reply: 3-5 friendly sentences in [LANGUAGE]; do not promise refunds, discounts or deadlines
If the message is spam or not a support request, use category "other" and an empty draft_reply.

From: {{from}}
Subject: {{subject}}
Body: {{body}}

What the builder reported

The builder says the five-node workflow took about 20 minutes to wire up and that the drafts were good. No volume or time-saved figures were shared, and community reviewers found bugs in the posted version.

These are the source author's own claims. AgentRecipes has not tested or verified them.

Watch out for

• The shared Code node only processed the first email of each poll; loop over every item instead.
• Read the sender's text field rather than the raw sender object.
• Add JSON-parsing error handling and a spam filter before using it in production.

Source

More customer support recipes