Don’t Drain the Planet to Read an Invoice
Would you print an email just to scan it again? Of course not. Yet that is exactly how much of the market processes e-invoices today.
Since January 2025, every invoice arriving in a German business is already pure data. Supplier, amount, tax rate, IBAN. Machine readable, by law.
And what happens next? That clean data gets rendered into a PDF. The PDF goes through an AI-powered OCR engine. A language model then guesses back the exact fields that were already there. And your AP team spends its day fixing errors that never needed to exist.
Here is what that guessing game costs, for every 1 million invoices. Based on Google’s own 2025 disclosure that a median AI prompt uses 0.24 Wh of electricity and 0.26 ml of cooling water:
- ⚡ 5 million AI model calls. Up to 8 billion tokens.
- ⚡ Up to 10 MWh of electricity. Three households, for a whole year.
- 💧 Up to 100,000 litres of clean water. That is 600 bathtubs, evaporated.
Now picture the hundreds of millions of B2B invoices flowing through German AP departments every single year. Gigawatt hours. Olympic pools of drinking water. All spent reconstructing data that arrived perfectly structured in the first place.
Every one of those litres is optional. You choose it, or you refuse it.
Germany rightly holds itself to high climate standards. So let’s be honest: OCR on born-digital documents is not innovation. It is waste with a GPU attached.
Our ERP-independent AP automation Docnova Flow refuses the detour. It takes the XRechnung, ZUGFeRD or Peppol invoice as native XML, runs it through approval workflows and fraud detection, and posts it straight into SAP or any ERP.
No PDF. No OCR. No hallucinated amounts. Zero tokens.
100% of the data. 0% of the water.
The greenest AI is the one you never need to run.
How many invoices does your AP process each month? Multiply. Then decide.
#eInvoicing #XRechnung #ZUGFeRD #Peppol #SAP #Sustainability #GreenIT #APAutomation #DocnovaFlow #Melasoft