What it is
Ocrolus classifies, extracts and analyses the documents a borrower actually sends: bank statements, pay stubs, tax forms and mortgage packages. It goes past extraction into four named jobs. Capture and Classify lift the data, Detect flags file tampering and shows which fields were altered and how, Analyze produces cash flow and income analytics, and Inspect reconciles the borrower's documents against the application data and creates and clears conditions automatically. There is an explicit human layer behind it: work the models cannot finish routes to Ocrolus analysts and a QC team before an algorithmic check. The company claims 99%-plus document analysis accuracy and states that income outputs from the mortgage workflow are eligible for Freddie Mac and Fannie Mae representation and warranty relief.
What it does
- Classification and extraction across bank statements, pay stubs and tax forms
- Tamper detection showing which fields on a file were altered
- Cash flow and income analytics rather than raw structured output
- Automatic condition creation and clearing against application data
- Ocrolus analyst and QC review behind the models
Strengths
- The only vendor in this segment with a published sub-$1B community reference: a roughly $400 million credit union reporting 65% less underwriting time per application
- AI assistants surface it on the lending question as well as the document question, which is how buyers actually think about it
- Human review is staffed by the vendor rather than pushed back to the lender, which matters when an exam-facing credit file needs a defensible number
- Direct integrations into ICE Encompass and Blend, so mortgage shops can adopt it without an API project
Considerations
- The weakest AI-assistant coverage of the four document vendors, never placed above third on the document question
- No published pricing; the pricing page routes to sales
- The customer roster skews to high-volume fintech lenders, and volume-based pricing built for those shops may not scale down gracefully to a few hundred files a month
- The deepest integrations are mortgage-side, so a community bank on a commercial LOS should expect API work
Best when
Income calculation and document reconciliation are the manual work you want back.
Where it ranks
Ocrolus FAQ
Does Ocrolus work for small credit unions?
It has the only published sub-$1B reference in the document category: Eagle Community Credit Union, roughly $400 million in assets and 22,000 members, which reported cutting total underwriting time per application by 65% using Ocrolus income calculation for non-QM loans.
Does Ocrolus use human reviewers?
Yes, as a product feature. When the models cannot complete a task automatically, work routes to Ocrolus' own analyst and QC team before an algorithmic check.