True story from this week.
A Fortune 500 vendor sent us an invoice. Here’s the journey it took to reach my inbox:
PDF → printed → faxed → photographed → emailed as JPEG.
Five formats. One document.
I called them. “Why are you doing this?”
Nobody knew. “That’s just our process.”
Before I tell you why this one invoice matters, here are three things I came across recently that made me realise this isn’t a quirky vendor issue. It’s a billion-dollar problem:
Anyway, back to my story. 80% of your company’s data is in formats like the invoice I got, and most companies can use only about 20% of it.
So Here’s What I See
Last month, a CFO told me his AP team forwards the same invoice 4.7 times before processing. He tracked it for a week.
Here’s the journey:
9:00 AM - Generic AP email
9:45 AM - Dallas forwards to Mumbai
2:30 PM IST - Printed for signature
4:00 PM IST - Scanned, sent to ERP
Next day - Wrong GL code (scan cut off digit)
Two weeks later - Vendor calls, nobody finds it
This costs 20-30% of operational expenses before counting missed discounts or duplicates.
Small companies have messy docs because they’re scrappy. Big companies have 200 vendors, five ERPs, and nobody remembers why it works this way. You’re processing 5,000 invoices from 200 vendors in 47 formats.
Every Time I Talk To Someone About This, They Say The Same Two Things
“We’re looking into automation”, which means someone saw a demo 18 months ago and decided it was too complex.
“Our documents are too messy for automation”, which is backwards. Messy documents are exactly why you need it.
Enterprise ops teams manually process unstructured data for 80% of their work hours. Four days a week, typing numbers from PDFs into systems.
What we’ve built at Docsumo handles the mess. Handwritten notes. Faxed documents. Any photo-of-a-printout situation from Monday. We’ve trained our models on 20+ million documents specifically so they work when formats are all over the place.
The companies getting this right aren’t asking “can we automate?” They’re asking “what should our team actually be doing?”
What Changes When You Fix It. FOR REAL
National Debt Relief processes debt settlement letters - which vary “a lot” and maintains 95%+ accuracy. They’re saving 2,500+ hours yearly. That’s human attention redirected from “is this a 6 or an 8?” to “does this application make sense?”
Biagi Bros handles 11,000+ bills of lading monthly. The system processes. Humans review exceptions.
Arbor settles 75,000+ insurance claims with 99% accuracy on ACORD forms. If you know ACORD forms, you know there are about 47 variations, and they’re all slightly different.
None of these are clean operations. They’re dealing with real chaos that doesn’t disappear. The bottleneck does.
Look, Here’s What This Comes Down To
Your AP team shouldn’t spend 32 hours weekly typing invoice data. They should spend 8 hours on exceptions and 24 on analysing vendor terms, catching discounts, negotiating contracts.
We’re focused on the space between receiving a document and making a decision from it. Getting the right docs to the right people at the right time - processed, validated, flagged if something’s off.
One Last Thing About That Invoice
That photo-of-a-fax invoice from Monday? When I called them, nobody knew why they did it that way. They just did.
Sometimes the system’s broken because nobody questioned it in ten years.
Rushabh
P.S. If you’re dealing with this and want to talk through what’s possible, just reply. Genuinely curious what patterns you’re seeing.