You will need
- A shared support inbox (Gmail or Help Scout)
- An agent with email access (Claude, or an n8n flow)
- A short list of your common ticket types
You will end up with
- Every new ticket tagged by type and urgency
- A drafted reply waiting on each one
- A morning queue you approve in minutes
Connect the inbox with read-only access
Give the agent permission to read messages and create drafts, but not to send. You can widen access later once you trust the drafts.
Write down your tags and rules
List the ticket types you see most (billing, bug, how-to, refund) and what counts as urgent. Keep it to one page: this becomes the agent's instructions.
Give the agent its triage prompt
Paste this into your agent's instructions and fill in the brackets.
Prompt You triage the support inbox for [COMPANY]. For each new message: 1. Tag it with one type: [billing | bug | how-to | refund | other]. 2. Mark it urgent if [your urgency rule]. 3. Draft a reply in our voice using [link to help docs]. 4. Never send. Save the reply as a draft. If you are unsure, tag it "needs-human" and explain why.
Run it on a schedule
Have the agent check the inbox every 15 minutes, or trigger it on each new email. Start with a small batch of tickets and read every draft.
Approve before anything goes out
A person reviews each draft, edits if needed and sends it. Don't give the agent permission to send until its drafts need almost no edits for a few weeks.
Human checkpoint. A person approves this step before the agent goes further.Tighten the rules each week
Look at the drafts you rewrote the most and add a line to the rules for each pattern. The prompt from step 3 is a living document.