Self-taught business owner builds free local invoice reader using Ollama and n8n

A small business owner with basic coding skills automated monthly invoice data entry by building a fully local AI pipeline using Ollama and n8n, eliminating the need for cloud APIs or paid subscriptions. The seven-node workflow extracts key invoice fields — supplier, date, concept, and total — using a locally run language model (qwen2.5-coder:7b) on a standard laptop at roughly 12 seconds per invoice. To prevent the model from fabricating data on blurry or poorly scanned invoices, the creator added a confidence scoring mechanism that flags uncertain results for manual review instead of silently passing them through. The setup runs entirely on-device, keeping sensitive supplier and financial data private, with no per-call costs. The creator has offered to share the full workflow and is seeking community feedback on handling flagged invoices and production trigger options.