Customer Email Automation with AI Draft + Human Approval
Johnny Z
Customer Email Automation
min read
The problem
Feel’s customer service team was overwhelmed by incoming support emails ranging from delivery issues to return requests. Manually sorting and drafting replies slowed response times and led to inconsistent communication with customers.
What we built
Inbox audit — Defined categories for incoming messages: General Queries, Delivery Issues, Return/Exchange, Account/Billing, Other.
Auto-categorize — Built an n8n workflow to scan incoming Gmail messages, apply keyword + AI classification, and route them to the right queue.
AI draft reply — Used OpenAI 4o mini to generate draft replies tailored to the customer’s message and aligned with Feel’s tone of voice.
Human-in-the-loop — Drafts were saved in Gmail with a “Review” label. Agents could approve, edit, or reject before sending.
Tune & track — Weekly adjustments to prompts and categories based on real support emails, ensuring ongoing accuracy.
Results
10 hours per week saved on triage and first-draft writing.
First response time decreased by 25%, with tickets reaching the right queue faster.
Return rate decreased by 5% due to clearer, more consistent customer communication.
84% of AI-generated drafts were sent with only minor edits, showing strong alignment with the brand’s tone of voice.
Stack
Gmail + n8n + OpenAI 4o mini + Slack
Save time and increase your profits with these flows.
