Your Documents Have Mistakes in Them
98%+ of Filings Have at Least One Error
Gain Compliance examined more than 100 regulatory filings; 98.2% failed to conform in at least one area. In the same study, 30.9% of companies filed data that didn't match between the printed PDF and the electronic version of the same filing. Two versions of one document went to the same regulator carrying different numbers.
A failure rate that high rules out carelessness as the explanation. The people preparing these filings are experienced, and most of them check their work repeatedly. The filings fail because the process itself, a chain of Word documents fed by Excel workbooks with no verification layer between the draft and the regulator, is structurally incapable of reliably producing a clean result.
We've watched how these filings actually come together. An analyst pulls figures from a stack of operational spreadsheets, pastes them into a Word master, reconciles a late adjustment the night before the deadline, then regenerates the PDF and hopes the electronic filing still agrees with it. Every step is manual, and every manual step is a chance for the document to drift away from its sources.
The inputs are broken before assembly starts
Nearly all, if not all, numbers in filings trace back to a spreadsheet, which is generally fed raw source data. Spreadsheets are inherently manual, and historically highly error-prone. A 2024 study found that 94% of business spreadsheets contain errors. An earlier study series conducted by Panko from 1995 and 2000 found that 91% of spreadsheets analyzed contained at least one serious error. The two studies use different error thresholds, so they aren't a strict apples-to-apples comparison, but together they span thirty years and could indicate an increase in error rate.
Manual document creation is inherently prone to human error.
Nothing checks the draft before it ships
The verification layer between first and final draft, if any, also suffers from a high error rate borne from human error. The clearest proof is the mismatch figure: 30.9% of companies filed data inconsistently between the printed PDF and the electronic filing, per the same Gain Compliance study. The threshold there is simply that two versions of one filing disagreed with each other.
Reconciling numbers across a PDF, an electronic filing, and the source spreadsheets behind both, while correctly ignoring intentional formatting and rounding differences, is a harder problem than it sounds. That difficulty is exactly why no reliable verification layer has existed for this until now.
Filings are heavily scrutinized. Your documents have mistakes too
Similar studies are not able to be conducted on private or confidential documents that you probably work on in your day to day. Millions of proprietary documents are created in high volume, high intensity industries like investment banking, private equity, and insurance each year; often by sleep deprived staff who are juggling multiple clients and projects simultaneously.
Artificial Intelligence makes the problem worse
The proliferation of artificial intelligence has made document creation abundant. Naturally, this shifts the bottleneck to the quality control layer. Not only does it take many people out of the creation seat, but it drastically increases the need to review because large language models are just as error prone as humans, if not more so, especially without the proper guardrails (buzz word used today is harness).
