4d (edited) • General discussion
Near Zero Cost Invoice Processing Workflow
So, invoice processing as a first project is pretty simple. Everyone recommends tools like PDFVector or similar services that require credits in your account — and to be fair, they’re really good at what they do. But I set myself a very simple constraint: make the entire workflow as costless as possible.
I used Google Drive as the entry point. Whenever an invoice is uploaded to a specific folder, the automation starts automatically. The file type is checked first — if it’s an image, I run it through OCR (Tesseract) to extract the raw text. If it’s a PDF, I used LlamaParse to parse it first and then extract the text.
Once I have clean text, I pass it to an Information Extractor node powered by the Google Gemini Chat Model, using a very strict extraction schema. Instead of asking the model to “understand the invoice”, I only ask it to return specific fields like invoice number, date, seller, totals, tax, and currency. If something isn’t present, it returns null.
The extracted data is then logged into Google Sheets. I also added a basic duplicate check so the same invoice doesn’t get logged twice.
Finally, once everything is saved, the workflow sends a Slack message with only the most important information — invoice number, seller, and totals — so accounting can see it instantly without opening the invoice.
No paid APIs, no credits, completely free.
That said, there are some limitations:
- LlamaParse has a free limit of 1,000 pages parsed per day
- Gemini has a requests-per-day limit. I’m using Gemini 2.5 Flash, which is fast but comes with a lower free request quota compared to other models
Still rough around the edges, but it works — and that was the goal.
Curious how others here would approach this differently.
2
0 comments
Near Zero Cost Invoice Processing Workflow
powered by
skool.com/ai-first-client-formula-8589
From zero to first $1k/month with AI automation in 30 days. Get the exact formula + templates that landed 100+ their first client.
Suggested communities
Powered by