BRO Barbershop is a chain of 4 barbershops. At each location, the administrator simultaneously serves walk-in clients and answers inquiries on Telegram/email — roughly 400 inquiries/month across the network, 65% of which are routine (booking, pricing, rescheduling, cancellation). Over six months NPS dropped from 56 to 38, and last month the business lost ~30 bookings due to slow responses. Budget for any software solution: up to $100/month, ruling out enterprise-tier tools (Zendesk/Intercom).
Before: Gmail and Telegram, answered by hand by the administrator at each location; no log, no priorities, no reports.
— All inquiry types (booking, pricing, rescheduling, cancellation, complaints) arrived in a single channel with no queue or prioritization
— Routine inquiries (65% of volume) were handled manually the same way as complex cases that genuinely need a human
— No automated way to tell promotional/spam emails apart from real customer inquiries
— The smallest mistake on a sensitive case (an injury or allergic reaction, a threat, a payment/refund question) could turn into a legal or reputational risk if a bot replied instead of a person
— No logging or reporting: the owner had no visibility into inquiry volume, quality, or how it was trending
I built a 3-workflow system in n8n that covers the full email-handling cycle:
— Email Triage — a Claude-based AI classifier on a Gmail Trigger determines priority (P1–P4), category (booking / rescheduling / pricing / complaint / other), customer tone, and whether human escalation is required; each email is simultaneously logged to Google Sheets, tagged with a category label, and routed through priority-specific logic — P1/P2 go to the administrator with no auto-reply, P3 gets a 30-second-delayed auto-reply, P4 is only labeled
— Analytics — Period Report — on request, calculates inquiry volume, % negative, and % requiring human escalation for a given period, with date validation
— Analytics — Period Comparison + AI Agent — calls the first analytics workflow twice and shows the change in metrics with a conclusion (increase/decrease, improvement/decline); a chat-interface AI agent understands natural-language requests (including arbitrary date formats) and decides on its own which of the two analytics workflows to call