n8n
First replies in 20 seconds flat
Engrave Me Bukidnon is a six-person engraving shop in the Philippines. Every item is made to order, and most customers write after their own workday ends. A builder put a Facebook Messenger chatbot on the shop's real page. The bot answers first inquiries from one plain file written from the owner's own answers, in the Bisaya-English mix the shop already uses. It collects order details for the team's job order form. It hands the chat to the team when a customer types 'team' or the answer is not in the file. A code check blocks any price that is not in the file. The workflow runs on n8n with a small model through OpenRouter.
How it wentIn the first week, October 22 to 29, the bot was first to reply in 62 of 65 chats, in about 20 seconds. The team's median first reply was 70 minutes, and 10.4 hours after hours. The page response time improved from 3 hours 7 minutes to 38 minutes 54 seconds. The bot got three facts wrong in week one, and it cannot see the status of an existing order. The builder states he can show faster replies, but not an effect on orders.
Try it · based on the source's workflow
Answer first inquiries on my [Messenger or chat channel] from this facts file only: [paste hours, prices, products, and policies]. Reply in [language or tone]. Collect [name, contact, order details] for each new order. If a customer asks about an existing order, or types [handoff word], pass the chat to the team at once and send an alert to [phone or channel]. Never state a price that is not in the facts file.
A starting template based on the workflow the source describes. Review it before you use it; you are responsible for what it does.
First replies in 20 seconds flat — from The SmallOffice Brief, a daily brief of real AI cases for small businesses: https://smalloffice.space/case-first-replies-in-20-seconds-flat