Thousands of business segments can benefit from artificial intelligence (AI), but perhaps nothing more than residential mortgages. It is almost always a slow process, but most people who apply for mortgages want this process to be much faster.

AI Foundry and Radius Financial Group Collaborate to Automate Home Mortgage.
AI Foundry and Radius Financial Group Collaborate to Automate Home Mortgage.

It is highly information-intensive and comes from a variety of sources: banks, employers, real estate agents, appraisers, and more. The costs involved are high, so there is money to invest in improvement, and there is a need to make things cheaper.

So, I wasn't surprised when I heard that AI Foundry, located near Boston, was taking a long way toward automating the mortgage process with AI. Since 2015, continuous technology entrepreneur Steve Butler has been working to solve this problem.

Mortgage processing is a document-rich process. Fortunately, artificial intelligence excels at extracting information from documents. But mortgages involve thousands of different types of materials, and they come in a variety of media and formats - paper, fax, electronics, and more.

A few years ago, Butler and his colleagues decided that a deep learning model might be the key to improving accuracy when extracting information from documents.

They determined that they were correct about this assumption, but training all models was just a slogan. So far, they have been able to identify and extract information from 300 types of documents, including loan estimates, settlement cost disclosures, W-2, and assessments, but Butler hopes to reach at least 600.

Each type of model training requires a lot of information. AI Foundry obtains anonymous, curated, and tagged data from its customers. They have hundreds of thousands of training data documents, but you need more to master other document types.

Workflow and analysis of mortgage decisions

The deep learning model is mainly to extract thousands of data elements from the document, but other aspects of the mortgage process are more workflow-oriented. Once the report is received, it must be identified, marked as received, checked for signature, multiple compliances, and quality checks, and then entered into the mortgage company's loan origination system.

The Robotic Process Automation (RPA) system is well suited for this type of work and is a combination of a workflow engine and non-artificial users of multiple information systems. Based on the data of the challenges faced by most commercial RPA systems, there are many occasional branches, so AI Foundry developed its RPA capabilities (and interfaces to other RPA systems). The company has developed a loan processing robot,

On average, it takes 40 days to process and approve mortgages today, compared to a slight improvement in 47 days a few years ago. The average cost per lease is about $9,000, nearly half of which comes from back-end labor costs.

AI Foundry's Butler believes that they can process mortgages for at least 1-2 weeks at the expense of 10% of the back-end costs of the mortgage bank. It is a risk-free proposal for banks that provide mortgages. They paid the loan to AI Foundry. Because mortgage demand is cyclical, they don't want to keep people around when they don't need it.

Post a Comment

Previous Post Next Post