Mortgage Cadence EVP of Services Jim Rosen: Seeing the Return From Our AI Investments
When we think of the investments that companies are making in new AI-powered software, the big tech firms spring to mind first. With Google investing over $300 billion and Facebook and Microsoft close behind, it’s clear that someone thinks “thar’s gold in them there mountains.”
We think that too, which is why our teams have been busy finding ways to make the most of these new technologies for the mortgage industry. But ultimately, the new uses we come up with for these powerful tools will come with additional costs tied to them.
Every lender that gets excited about what AI will finally allow them to do–and that list is long–also needs to carefully consider how they will earn those investments back, with a return.
For the big tech companies, ROI from AI will likely come from increased advertising revenue driven by better targeting and advertising messages, all delivered via the AI’s advanced algorithms.
But where will the ROI come from for the mortgage industry? We have some thoughts on that.
Why AI Is the Answer to our Efficiency Problem
With the cost to originate a mortgage loan now hovering around $12,000 per loan and the technology expenses still around 5-10% of total costs, it’s clear that people and process are costing lenders far too much.
We know the industry is currently overstaffed, and we’re hearing stories daily about lenders going through the painful process of correcting their staffing levels. Efficiency, on the other hand, is still an elusive target.
It’s hard to move the needle on efficiency when the lender is still tied to workflows that require humans in the process. A perfect example of this is document management.
During the loan origination process documentation flows in from the borrower and must be identified, indexed and then evaluated by humans in order to underwrite and process the loan. Over the years, we’ve seen this process go from a fully manual, paper report to loan file process through digitalization, where the processor accesses another software system to download the report and then uploads it into a new system so they can manually pull the data off and enter it into the LOS.
Today, most lenders are using powerful loan origination software that can automatically order and receive reports from third parties and pull them right into the lender’s system. But, in almost every case, a human is still staring at these digital reports and comparing them to the data in the LOS.
We haven’t been able to trust our systems to put the right information in the right place and so we’ve kept our human staff on guard to keep the process going. This is especially true for documentation provided by the borrower.
This can only become so efficient. Most lenders have reached the limit of what they can do with the automation they have today.
To solve the efficiency problem, we will have to employ AI.
Why Doc Management Will Be the First Place Lenders See an AI ROI
In almost every other part of the process, automation is capable of keeping the loan moving toward the closing table. Where it falls short is in dealing with the documentation, and especially those documents provided by the borrower. But AI based technology can change that.
When AI tools that have been trained on the various documents in use in the mortgage industry are employed to study scanned borrower documents, it can determine exactly what it’s looking at about 80% of the time. And it’s getting smarter every day.
When these new tools are used to classify incoming documents, the efficiency gain for the lender is significant and immediate.
When humans are only processing 20% or less of the actual workload, the time savings are significant. And that’s before the AI is actually processing the information it identifies. We’re just talking about recognition at this stage.
How the Story Gets Even Better
It starts with AI that’s capable of recognizing income-related documents, things like W-2s and 1099s. Virtually every loan the lender makes will require this documentation, so building AI into loan origination systems makes great sense.
But every additional document set AI tools can process just increases the ROI the lenders get from using the tools.
Eventually, lenders will decide that they want to add some workflow to their AI-powered document management. Maybe they’ll just want to validate data between the document images and the LOS and send discrepancies to an exception queue for human processing.
Maybe the lender will want the AI to let them know if the data on a document exceeds a certain threshold, which may trigger a new workflow and send the loan down a new process. Once the AI knows what data it’s dealing with and what the lender wants the ultimate outcome to be, these systems are powerful enough to make certain the loan goes down the right path.
There are many potential use cases and that’s what’s going to be the ultimate ROI from the addition of AI to LOSes: the ability for lenders to finally decide exactly how they want to operate and know that their technology can achieve that end.
When you consider that, today, lenders are locked into whatever workflows their technology partners have built into their loan processing software, this opens up a whole new world of possibilities.
Unlike big tech and VC-funded startups with a great appetite for risk and the willingness to throw tens or even hundreds of billions of dollars into AI in the hope that great things will result, lenders will move carefully into this new world, as they should. They’ll start slowly, verifying compliance and return as they go.
We expect most lenders to start with document management and grow from there. They won’t spend as much as the big tech firms, but the returns they see could ultimately be just as great.
(Views expressed in this article do not necessarily reflect policy of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes your submissions. Inquiries can be sent to NewsLink Editor Michael Tucker at mtucker@mba.org or Editorial Manager Anneliese Mahoney at amahoney@mba.org.)
