The new CEO at one of my startups hung a motivational poster in his office with a bald eagle and the words, “If you chase two rabbits, both will escape.” An ancient proverb that might date back to Confucius, the meaning was immediately clear to our product team: stop chasing after every market opportunity and focus on the one we had the best chance to catch.
In the intelligent document processing (IDP) market, many vendors are chasing after all the rabbits in the forest. You can tell by the marketing messages and statements from the founders: “we can process any and all unstructured data, no matter the industry”. Look at their website solutions page where they will list several different industries and use cases.
This strategy is typical with most technology-driven companies who sincerely believe their AI engine can process any document. Theoretically this universality may be possible, given enough time and money. Practically, it never works out that way; we’ve seen the evidence and we always advise vendor clients on the urgent need to focus product development on industry-specific pain points. But many still waste enormous sums of sales and marketing resources on a horizontal technical approach that customers are not buying. Let’s be honest: they suffer from massive FOMO* on a big sale.
That’s why it was refreshing to meet last week with Ocrolus, an IDP company whose laser focus on the U.S. lending market has resulted in a product that is fine-tuned for the needs of the users. Based in New York City, Ocrolus has over 200 corporate employees plus an army of over 800 on-demand data validators who help it deliver 100% data accuracy to clients. The company has raised over $125 million from investors and its cloud platform has processed over 265 million lending-related documents. Customers include PayPal, Plaid, and Funding Circle.
Today’s consumer expects near-instant decisions when applying for a loan, and lending companies are eager to provide that because the competition for borrowers is incredibly fierce. The goal is to get to a Yes or No as quickly as possible. This is the loan underwriting stage and until recently it required a human to manually validate every document.
Now when someone applies for a loan with an Ocrolus lender and uploads supporting documents to verify income, Ocrolus is embedded within the loan underwriting process. The documents are automatically uploaded to the Ocrolus cloud for document classification and data extraction, and the validated data is returned quickly to the process workflow for decision-making. This can take place in less than a minute.
For example, in mortgage lending Ocrolus is tightly integrated with Encompass, the leading loan origination system. Mortgage loan files are notoriously difficult to automate due to the dizzying diversity of structured, semi-structured and unstructured documents involved. The loan file is uploaded into Ocrolus, where income verification documents such as bank statements and W-2 tax forms are validated. The platform can also split individual documents within a single file and can identify the presence or absence of documents on the underwriter’s checklist (such as appraisals, correspondence, surveys, etc.).
Ocrolus recently announced a new offering for its customers called “Instant”, a machine-only processing solution that can cost less and deliver results faster than the human in the loop (HITL) solution. Ocrolus AI models have matured to the point where the company can now offer this option with a high degree of confidence in the accuracy outcomes. Early indications from the customer base are this will be very popular. The company will continue to offer the HITL option for more complex use cases such as low-quality or poorly scanned images where the machine can use a bit of human help.
Ocrolus is not the only IDP vendor going after the lending market; there are dozens of “generalists” who list financial services as one of the vertical markets they serve. What sets Ocrolus apart from the generalists are two key competitive advantages.
First, the company possesses a deep domain knowledge of the lending process and the pain points encountered by loan underwriters. I was briefed by David Snitkof, the SVP for Growth. Prior to joining Ocrolus, David co-founded Orchard, a pioneering data, analytics, and transaction platform for Online Lending. He also worked in the financial services industry as SVP for Analytics & Strategy in the Small Business & Commercial Banking group at Citi, and as Senior Manager, Risk Management, at American Express. David also publishes the Fintech AI Review. So yeah, David knows lending.
The second competitive advantage comes from focusing its AI model training on lending market documents. Ocrolus told me that in the last 18 months alone, it has processed bank statements from over 6,500 unique banks and credit unions from across the United States. I think this is a major reason why the company can confidently launch its Instant product, knowing the machine will recognize any bank statement and process them at the highest levels without the need for annotations, HITL, or training iterations.
I plan to go into more detail about Ocrolus in a future vendor research report. I will close with this: Ocrolus chose to chase one rabbit, and its discipline to stay focused has made it the leader in document automation for the lending market.
*Fear Of Missing Out