How I Automated 6 Hours of Invoice Work in a Weekend

A friend of mine runs a small construction business. Every Monday morning, he would sit down with a stack of PDF invoices and manually type the details into a spreadsheet. Vendor name, invoice number, date, amount, category. Over and over. It took him about 6 hours every week. Sometimes more.

He told me about this casually over lunch. I told him I could probably fix it over the weekend. He laughed. I built it anyway.

The Problem

His vendors sent invoices as PDFs. Some were scanned. Some were digital. All different formats. He had about 40 invoices a week and was copying each field by hand into Google Sheets. One typo meant his accountant would catch it weeks later and the whole reconciliation process would stall.

The real cost was not the 6 hours. It was the errors, the delays, and the fact that he was doing data entry instead of running his business.

What I Built

I used Python with PyPDF2 for digital PDFs and Tesseract OCR for scanned ones. The pipeline looked like this:

• Drop all PDFs into a Google Drive folder
• A script watches the folder every 10 minutes
• It extracts text from each PDF using OCR when needed
• Regex patterns pull out invoice number, date, vendor, and total amount
• Everything gets written to a Google Sheet automatically
• He gets a Slack message summarizing what was processed

The whole thing is about 200 lines of Python. Nothing fancy. No machine learning, no AI. Just solid scripting with good error handling.

How Long It Took

Saturday afternoon I started. Saturday night I had a working prototype. Sunday I added the OCR layer and the Slack notification. Sunday evening I deployed it on a $5/month DigitalOcean droplet.

Total time: about 10 hours including testing. Total cost: $5/month for the server plus a few dollars in API calls.

The Result

Those 6 hours of Monday morning work became 15 minutes of reviewing the spreadsheet for any edge cases. His error rate dropped to nearly zero. His accountant started getting clean data every week without follow ups.

The biggest surprise was how much mental bandwidth he got back. He told me he actually started looking forward to Mondays again. That sounded dramatic but I understood what he meant.

What I Learned

You do not need cutting edge tools to solve most business problems. A well written script that handles 90% of cases reliably is worth more than a perfect system that never ships. Start simple, deploy fast, and iterate based on real usage.

If you are sitting on a repetitive task that eats hours every week, there is almost certainly a straightforward automation that can handle it. The ROI is usually obvious within the first week.