case study · arce tenders
Arce, in a diagram you can move around.
Four views of the system: the full circuit from the portal to the dashboard, the real n8n workflow, how every tender document is analyzed with local AI and how the team uses the dashboard and corrects the AI. Every box can be dragged; every view has a step-by-step walkthrough.
From the public procurement portal to the sales team: n8n orchestrates, a scraper extracts, PostgreSQL stores, a script analyzes with a local model and the dashboard shows only what matters. Everything in containers, except the AI, which runs on the host.
Two-minute script
- Arce monitors Uruguay’s public procurement portal and leaves the sales team only the tenders that matter for the trade. It used to be two hours a day of manual checking.
- Every 15 minutes n8n reads the portal’s RSS; when a new tender appears, a Puppeteer scraper opens the page and extracts details, attachments, clarifications and items.
- Everything is stored structured and deduplicated in PostgreSQL. Then a script takes the pending documents, extracts the text from any format and sends it to an AI model running locally.
- The AI returns a JSON: whether it is relevant, how confident, why, and useful data such as the site visit with date, place and contact. That is stored next to the tender.
- The team opens the dashboard, sees the opportunities already filtered and leaves feedback: whether the AI got it right. That feedback is stored next to the verdict to keep tuning the prompt.
Click a box to see what it is and what it connects to. Drag to arrange; “Rearrange” restores the original layout.