The problem
Many follow-ups depended on someone remembering to check: expiring raw-material documents, certificates and trademarks, trips missing paperwork, or attendance scans. Things slipped and were noticed late.
What I built
Made n8n the company’s automation hub: it pulls from SQL Server, Google Sheets and agency websites on a schedule, or receives webhooks from my own systems (TimeStamp, X-ray Sentinel, the production dashboard), then formats the message and sends it to the right people.
Key features
- An attendance alert on Telegram for every fingerprint scan, in the employee’s language (Thai, Myanmar, Khmer, Lao, English)
- A Telegram sign-up bot that checks the employee ID, links the chat ID automatically and sends a welcome in the right language
- Missing-scan alerts before each payroll cut-off, and a monthly Excel summary of new and leaving staff
- Email alerts for expired or expiring Test Report and HALAL documents with a summary table and file links, plus customer-complaint follow-up every two hours
- FoodLexAI: 16 workflows that scan food-law agencies in 16 countries and bodies (e.g. Thai FDA, Department of Fisheries, ACFS, WTO SPS), de-duplicate, and feed FoodLex
- Alerts when a production line reaches its plan, packaging idle for 180+ days, trademarks nearing expiry, and trips missing documents
Under the hood
- My systems post webhooks to a single n8n entry point, so channels and wording change without touching the source code
- Separate “Test” and “Production” send nodes in each workflow, so messages are checked before they reach real users
- Send results are written back to SQL Server (e.g. a trip-reminder log, a complaint “sent” flag) to avoid duplicates and share cooldowns with the source systems
- Each FoodLexAI workflow has 30–40 nodes: it pulls several sources per country, parses them in Code nodes, checks what it has seen before, and stores only new items