Trabalho de transcrição de documentos

I built a small Windows tool that turns invoice PDFs and phone photos into one clean Excel file - with AI extraction and, more importantly, human verification.The problem it solves: 80 invoices a month is about 3 h 30 min of typing, and every typo becomes a reconciliation problem at the end of the month.What it does:1. Put the invoices in a folder and press one button.2. The Claude API reads each document and returns the fields with a strict schema.3. Anything it is not sure about is flagged with a confidence value instead of being quietly filled in - and the raw text of every field is kept next to the extracted value.4. Save, and you get one .xlsx with 12 columns per invoice: provider, date, number, description, taxable base, VAT, total, currency, category, confidence and the original text.3 h 30 min becomes about 30 seconds. The AI step costs roughly 0.02 USD per invoice (paid with your own API key), and from the second month you run it yourself - the program is yours, no subscription, no account.The part I care about most is #3: a tool that silently invents a date is worse than no tool at all. That is why every value keeps its source text, so you check the doubtful lines in two minutes instead of the whole file.Full case study with the screenshots: https://contra.com/p/PQPWRbOR-facturas-ia-ai-invoice-extraction-to-clean-excelI am looking for people who do this work for a living - bookkeeping, admin, finance ops - and would rather have clean data than another afternoon of copy and paste.What is your most painful data-entry task? I am collecting the ugly ones.