Thai government tender tracker
Monitoring Thai government procurement manually across 8 sector categories was consuming 15+ hours a week, with poor coverage and frequent missed opportunities. We built a system that now surfaces every relevant tender in Thailand within hours of publication — scored for relevance, summarised in plain language, and ready for a bid decision.
The problem
The Thai government publishes thousands of procurement notices every week across dozens of departments and state-owned enterprises. For a mid-sized supplier, scanning even a handful of these sources manually is impossible — and missing a tender means losing a bid window entirely. The client was spending 15+ hours per week on what amounted to patchy, reactive monitoring, and still missing most of the opportunities relevant to their business.
What we built
A self-contained tender intelligence pipeline, running on the client's own infrastructure. No SaaS lock-in, no per-seat pricing, no foreign data residency.
The system runs on a scheduled n8n workflow: scrapers pull new notices from the national e-GP system and targeted state-enterprise portals; Claude reads each tender, scores its relevance against the client's sector profile, and generates a plain-language brief covering scope, budget, timeline, and probable incumbent. Results land in a tagged Google Sheet and surface in a dark-themed dashboard filtered by sector.
The AI scoring is the critical piece. Thai tender notices are often poorly structured PDFs with inconsistent terminology. Keyword matching on its own produces too many false positives to be useful. Claude, prompted with the client's real bid history as context, produces a relevance score that actually correlates with what the BD team wants to see.
Architecture
The whole system is self-hosted on a single Hostinger VPS, fronted by Traefik for routing and TLS. n8n orchestrates scheduled scrapes and the Claude scoring steps. SerpAPI provides supplementary search when the primary scrapers time out. A lightweight dashboard reads from the Google Sheet backend.
Total running cost: under 2,000 THB per month of infrastructure plus Claude API usage. No enterprise SaaS bill, no vendor lock-in, and the client can take the whole system in-house whenever they want.
What we'd do differently
The SerpAPI fallback was an operational pain point — intermittent hangs from the Hostinger egress caused missed polling cycles. In hindsight we'd have built our own scraping layer from day one rather than reaching for SerpAPI for edge cases. We've since moved most sources to direct scraping with retry logic, and the SerpAPI layer is being phased out.
Why it matters for outbound
This system was built for one supplier, but the pattern generalises. If you sell into government or large-enterprise procurement, the quality of your tender intelligence directly determines your BD team's hit rate. Most Thai industrial suppliers are flying blind on 80% of the notices they could bid on. The same pipeline — adapted for your sectors and your bid history — is exactly the kind of system we build as part of a full outbound engagement.