Email management is one of those sneaky ops problems that just steals hours every week—and it gets worse as volume climbs. In this build, I walk through the exact n8n workflow I made for the n8n Community Challenge: an AI email classifier + autoresponder that takes an incoming email (from/subject/body), routes it into the right bucket, and drafts (or sends) the right reply. I show every node, every prompt, and the decisions behind the architecture so you can deploy the same pattern in your own business today. I start with two triggers (a webhook for real inbound emails and an evaluation trigger for testing), then standardize fields with Edit Set nodes so webhook data and test-set data look identical. From there I add an IF guard for empty bodies, then use the Text Classifier node to label emails into categories like pricing, setup, security, HR, escalations, spam, and misdirected—using subject + body with clear category descriptions and tie-breakers. After classification, I selectively pull only the relevant Google Sheets docs for that category (instead of stuffing everything into the model), convert multi-row sheets into a single “document text” via small JavaScript code nodes, merge/concatenate where needed, and feed that into an agent to draft the reply with structured output. The big takeaway: make the classifier rock-solid first, keep context targeted, and set up evaluations so you measure accuracy before you let it run live.