In the lean years that followed the stock market crash of 2008, successive inquests examined the role of administrative overload in the US subprime mortgage crisis.
In order to properly vet the surge of people suddenly clamouring for credit, lenders were faced with a mountainous mish-mash of paperwork. Consider that a typical US mortgage loan file contains between a few dozen and several hundred unique document types; often inaccurately named; routinely missing elements; and with little or no consistency between formats.
For asset managers, this chaotic information landscape made accurate decision-making extremely difficult. The secondary mortgage market packages and sells loans from different originators, with an array of buyers, servicers and resellers all involved to carry out due diligence, assess risk and price existing/potential assets.
Having access to clear, consistent documentation or key indices within documents can reveal a discrepancy between perceived and actual value. Identifying this discrepancy quickly and accurately enables maximum resale value – without accuracy, on the other hand, things get murky.
And back in 2008, they did – with catastrophic results. Unfortunately, the difficulty of extracting meaning from diverse physical documents remains a challenge for financial institutions today. And the challenge is not confined to the financial sector -– education, law, medical, accounting, and other professions all have their struggles with paperwork.
Wasting time on paperwork
Around the world, time-poor financial executives are trying to understand the value of assets they control (loans, securities) before purchasing them, or even pricing a bid, all based on physical documents that are completely disorganised. This is an incredibly time-consuming and error-prone ordeal, as asset managers wrestle with multiple file types and formats, trying to see what is missing from files or even what is the ‘best’ version of any given document.
What if those loan managers and financial institutions had the ability to autonomously catalogue, organize and standardize all those documents in an accurate and reliable way?
What if they had a tool that could index, rename and sort all documents within a file so they could be quickly and transparently understood, say according to keywords, tags or links?
What if they could scrape specific content from within individual documents instead of having to go through each one by hand? How much easier life would be.
There is a solution at hand. Altada has developed an AI solution that includes a specialist document intelligence capability designed to help organizations leverage their data into a competitive edge. Document intelligence provides people with quick, accurate analysis of massive datasets – such as a pile of mortgage loan documents sitting on someone’s desk.
AI gives a clearer picture
The tool works by ‘wrangling’ components of multiple documents; extracting, structuring and sorting the usable data into a format that can be used by the institution’s own investment models. This gives the investment professional a much clearer picture of the true value of assets they are managing.
Altada’s document intelligence tool uses optical character recognition (OCR) to extract the information in a machine-readable format, while natural language processing (NLP) is used to understand context, patterns, and other factors that in turn improve the OCR output.
There’s a fair bit to take in there. But if you imagine how the AI tool can wrangle – working the documents, wrestling with them, stubbornly teasing out meaning – then you’ve pretty much got it.
Altada’s platform has an 86% accuracy rating on novel document types (formats that are unique or that have not been encountered before) and over 90% accuracy on pre-trained document types. This is a level of accuracy, especially on highly varied and complex US mortgage document types, that has not been seen before; that no-one else is capable of.
The time-saving, meanwhile, is colossal. Altada’s solution reduces the loan portfolio processing time to 48 hours and can reduce the cost of processing a loan file by 90%. The machine quickly and efficiently gets on with its job while the investment professional has time to look at what really matters – value, risk, profitability, decisions, meaning – in order to make informed, strategically sound decisions.
Want to find out more? Talk to us and we’ll take you on a deeper dive into the future of document intelligence and sentiment analysis for the financial sector.
Thanks for reading and please stay tuned for more blog insights!
Follow us on Twitter @altadasolutions
Connect with us on LinkedIn @AltadaTechnologySolutions
Visit our website www.altada.com