There is endless hype and little real information about how to put genAI safely into practice in the mortgage industry. I have done about 40 interviews and feedback sessions since May and in addition to being fun, it has really opened my eyes to what’s going on in mortgage since the genAI frenzy started. I have a lot of gratitude for each person or team that has taken the time to meet with me. This topic has refilled my technology love tank.
Let’s break it down – what’s the deal with genAI in mortgage?
First, let’s acknowledge that “AI” has been in mortgage in some form or fashion for about 30 years. Built-for-purpose AI, sometimes called “narrow” or “weak” AI has dozens of practical applications, widely adopted in the industry. What’s already out there falls in four broad AI subspecialties.
I think the mad scramble to come up with AI strategies is unnecessary and fails to give the industry enough credit. We’ve been doing AI for years, we just didn’t call it AI.
Next, what’s all the hype about? The hype is about a form of AI that is now widely available called generative AI (genAI). In mortgage especially, the form of genAI we are really excited about combines machine learning (ML) with natural language processing (NLP) to generate new content. Generative AI refers to a type of artificial intelligence that can create new content, such as text, images, music, or code, by learning from existing data. Unlike traditional AI, which is typically designed to recognize patterns or make predictions, generative AI models are capable of generating new, original outputs that are similar to the data they were trained on.
Why should we be so excited about this? What is everyone losing their shirt about? This type of technology is truly disruptive. Occasionally, just a few times in a lifetime, a tool comes along that revolutionizes the way we live and work – and genAI is one of those tools. For those of us old enough to remember it, think back to when the internet became available – or when your mom lugged home the first personal computer. Everything changed. That’s what this is like. So we should definitely all be paying attention.
And finally, where is this going in mortgage?
My prediction is that the uses of AI today will continue. Our applications of machine learning will evolve, we will see more and more supervised and unsupervised learning applications (think smarter models). We will see companies being to (responsibly) use the decades of transactional data that has been sitting idle in new ways to drive new value and enrich the lives of homeowners. We will see (and, in fact, are seeing already) the next levels of technology in OCR and intelligent content recognition (ICR) that will finally perfect document recognition and data extraction. Expert systems will be, well, more expert. We will still have numerous uses for business rules based deterministic systems. This is mortgage after all.
What we specialize in at my company, however, is genAI. Where is that going? We will continue to see operator-facing chat and “ragbots”, or bots that use retrieval augmented generation (RAG) to enrich requests to the large language models (LLMs), constrain the answers provided, and improve operator experiences. More and more, we will see the lives of our operators enriched and simplified with immediate and relevant answers to all sorts of questions.
We are going to see more and more assisted customer email responses that combine sentiment analysis with language, image, and video-based generative content, most uses with a “human in the loop” to control for model hallucinations and ensure only accurate and appropriate content is going to homeowners. We will see the most creative companies invest in experimentation with multimodal enrichment approaches, again with a human in the loop. Without naming names, whole classes of technology will become obsolete, as what they do will be replaced with a much cheaper, simpler alternative.
We will not see unguarded, wholly automated applications of genAI put in front of customers. The industry will not accept it, and the technology is not ready. The cleverest companies will embrace the uncertainty and go into the lab to run lots and lots of science experiments. They will scour their organizations for use cases (don’t forget, innovation without a use case is just a hobby). They will take those use cases and figure out which ones can be solved or enabled with genAI. They will filter these lists and take this tipping point moment in time as a chance to revisit their technology roadmaps and invest in technology, much of which will have nothing to do with AI all.
It's a great time to be in mortgage.
