Accenture Credit Services’ Jason Wilborn: What Will the Mortgage Industry Do with Generative AI?
Jason Wilborn is Operations Delivery Manager with Accenture Credit Services, Rancho Cordova, Calif.
You know those moments where you have a random thought and that thought stays with you? It re-surfaces every now and then, but the details are so clear no matter how long ago it was. One of mine takes me back to July 2005, I was 27 years old, two years into the mortgage industry and working for a major bank in their wholesale division. I worked my way up from a Doc Drawer to an Underwriter and I was in the middle of an underwrite when it hit me, “This was too easy!” There had to be something wrong; every appraisal comes in at value, no or stated income documentation, no or stated asset documentation needed, (NINA and SISA anyone?), Option ARMs with neg am and incredibly low qualifying rates? Sure thing! Escalations only happened when I tried to ‘decline’ a loan. That thought passed and I moved on as duty called, yet I remember it so clearly. I could take you to the exact spot that I was sitting in 18 years later if you asked me.
Around that same time our industry talked a lot about automation and technology (we all went ‘paperless’ about this time). I remember two senior executives visiting the site talking about how soon title and appraisal would all be automatically reviewed and approved, DU and LP were the gold standard, CLUES and other proprietary AUS were popping up for all the non-agency paper and the need for Underwriters in our industry would soon be eliminated. Anyone else remember these themes? Talk about famous last words! Fast forward and we all know what happened next. The stock market crash, AIG, Dodd/Frank, Great Recession, Quantitative Easing, OCC consent orders, Helicopter Ben…
There were a lot of casualties in the next few years. People let go and left the industry, effectively all the sub-prime lending programs, channels, divisions, entire companies, market liquidity, and unfortunately along with it, our collective trust in automation and technology. A figurative nuclear winter settled over our industry. The pendulum swung from checking for a pulse and funding the loan to overtly conservative lending guidelines (which continued to tighten until 2013), deep scrutiny on every underwriting decision, and excessive stipulations, “we’re going to need a letter of explanation on that”. We put human eyes on everything, not the least of which was other human eyes (ha!), and no longer trusted automation. There were tough conversations with qualified borrowers on why the answer was “no.” The wagons were circled, and we proceeded with caution. Rightfully so, as I remember pausing for just a moment when people asked me what industry I worked in.
Here we are 15 years later, valuations have more safeguards, our guidelines and programs are sound, credit standards have normalized, the NMLS and CFPB patrol our industry relentlessly, and non-QM still feels like a bad word. What about automation and technology though? We talk about it a lot; I see more and more start-up companies offering innovative technology products. Why have we been so tepid to embrace it? Why has innovation in our industry taken so long to catch back up? The one casualty from the financial crisis that should have revived first was our faith in automation and technology, yet it is the one that we continue to keep at arm’s length. Any technology or automation is only as good as the underlying corpus of data, business rules and process management they are meant to support (throw reporting and analytics into that concept as well). We sacrificed technology and automation after the financial crisis to soothe our own conscience. It was a reminder of what we had told it to do.
I started this article with a trip down memory lane for a reason. As mentioned above, technology and automation are only as good as the fundamental business rules and corpus of data that underpin them. This is where I want to layer in Generative AI. Had Gen AI in its new current form (not neural networks or machine learning) been around from 2001-2005, it would have made the financial crisis worse, not better. I can say that because Gen AI would have referenced the same policies and procedures, the same investor requirements, the same AllRegs and the same seller guide that other decision engines referenced and it would have given the same recommendations, only faster.
Gen AI generates content or responses from the corpus of data based on the feedback and training given to it by human interaction and other models or datasets. Those inputs are the underlying data and business rules referenced by large language models that then generate content to the end user. Let us use ChatGPT for a minute. The average human being has a vocabulary of 20,000-45,000 words. ChatGPT has data corpus (vocabulary) of over 3 billion words and countless human interactions. ChatGPT has a vocabulary ~66,000 times greater than the best of us and a massive amount of context from which to base its responses. It references that dataset to respond to our questions; it is interactive and intelligent, but it is also extraordinarily wrong sometimes. It is only as good as the underlying data, feedback loops and training it references. As those references get better, so do its responses.
We are on the cusp of a generational leap in knowledge management; from being in awe of a person that types the fastest, knows search engine, and has the audacity to click on page two of the results, to one honoring the human that is simply asking the right questions. Which leads us to asking ourselves the right questions. Do we trust ourselves to automate a process or task? Do we trust our data? Do we trust our process? Are we willing to let technology supplement our people in the ways we know will be useful? Are we going to allow technology, like Gen AI, to improve our knowledge management in ways that enhance process efficiencies, improve the customer experience, help our underwriters navigate complex scenarios or our call centers to summarize calls and control test them?
Given time and the right foundational interface, Gen AI WILL change how we do business in ways we have yet to foresee. Will we let it, or will we stay stuck in our own world of manual underwrites and reactionary management?
Accenture announced a $3 billion investment over three years in its Data & AI practice to help clients across all industries rapidly and responsibly advance and use AI to achieve greater growth, efficiency and resilience. If feels like Julie Sweet was talking to us when she said “Companies that build a strong foundation of AI by adopting and scaling it now, where the technology is mature and delivers clear value, will be better positioned to reinvent, compete, and achieve new levels of performance. Our clients have complex environments, and at a time when the technology is changing rapidly, our deep understanding of ecosystem solutions allows us to help them navigate quickly and cost effectively to make smart decisions.”
Note: Opinions are my own and not necessarily the views of my employer
(Views expressed in this article do not necessarily reflect policies 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 Editor Michael Tucker or Editorial Manager Anneliese Mahoney.)