I built a no-code workflow that reads my Gmail and texts me only the emails that actually need a reply

I wired up a no-code workflow for the worst part of running sales and support out of one Gmail inbox: figuring out which of this morning's emails actually need me. Here is the part I am happy with. Most AI email triage is basically keyword matching, so a newsletter with URGENT in the subject outranks a quiet note from a real customer. This workflow looks up the sender's prior Gmail history first, so the classifier has relationship context before it decides anything. That is what lets it tell an existing customer's blocking issue apart from a brand-new cold inquiry. The mechanics are straightforward. It sweeps the last 24 hours of Gmail, up to 50 messages, and ignores anything I sent myself. For each email it reads the full body, then pulls the sender's history to see whether this is a long-running thread or a first contact. It classifies intent into existing customer, new inquiry, marketing, or automated, and scores urgency from the content and that history. Only the high and medium urgency emails that need a reply survive; marketing blasts, no-reply notifications, and one-way confirmations get dropped. What I get back is one Telegram digest, grouped by urgency, with existing customers separated from new inquiries, so I know which fire to put out first. A concrete example: on a morning with a few dozen unread, it handed me a short list, the customers with open threads first and the new inquiries right after, instead of a screen of unread badges. End to end it saves me about 25 minutes a day. It runs on demand, but you can also schedule it or trigger it after a call or form, and you can launch it from Hermes, OpenClaw, Codex, Claude Code, or ChatGPT with no local setup. Happy to share the exact tool chain if it is useful. The workflow link is in the comments, and I would love to hear how it changes your inbox routine.