One of the most common themes in the mortgage industry is to point out its failure to adopt technology. It’s a complaint among pretty much everyone. Yet, as soon as ChatGPT launched to mainstream use this year, our industry (like everyone else) began the AI hype cycle. Today is a good day to ask the question – why is this time different?
Sure, AI is as transformational to work as the Internet was to communication. As a result, it is easy to assume our companies and partners will all jump at the chance to harness the benefits of on-demand answers to almost any question through large language models and/or the ability to contextualize information, data and documents through agentic AI. If you saw the second Internet coming, you’d jump at the chance to be on the cutting edge too, right?
Dig in beyond the hype, however, and adoption of AI starts looking more and more like adoption of every other advancement in the industry.
Fannie Mae’s roadmap for eClosing adoption was first printed on a three-ring binder and distributed on a VHS (I’ve seen it!) in the late 90s and we’re still struggling to get eNote adoption in 2025. A similar story plays out in document ingestion, direct source data collection and appraisal modernization. While Plaid and similar tech might not be AI-level innovation, it can accomplish what OCR never could. Yet, we spent hundreds of millions (billions?) on OCR and we cannot effectively work with direct bank or asset data yet.
Now, AI appears claiming to solve for challenges that span the mortgage ecosystem. Some claim the result will be the same as OCR – money spent without much to show. Most claim that AI will revolutionize our business.
If we accept that premise, AI will revolutionize the mortgage business. How soon can we expect adoption followed by results following by ROI?
Here are the 3 things I believe will dictate whether we see adoption and ROI for AI:
1. Know your customer. Product-market fit requires clarity and honesty about who the customer is and how your solution solves their pain point. As Google puts it, “eliminate the toil.” Most mortgage companies serve the Mortgage Professional (either MLO or mortgage broker) or referral source (online lead gen, in the case of consumer direct). AI confuses the issue by claiming to make things easier for the customer but is often referring to the consumer.
The consumer is not the customer for most mortgage companies. In determining how to evaluate AI, be clear on who it serves and how that fits into your specific strategy and roadmap. Buying consumer-facing AI does not deliver ROI if everything else in your business serves the mortgage pro.
2. Delegate to technology. Leaders understand delegation. I’ll bet everyone in our organizations wishes they could delegate more of their work to a trusted colleague or team member. Whether you implement AI or not, this should be a question you are asking throughout your daily activities: what can I offload for the highest return in time and efficiency? With AI capabilities, the answer continues to come back as more and more of our work can be done; however, tech partners using AI will never get a chance to prove it without a delegation plan.
At the same time, it’s expensive to maintain both an AI workforce and a human workforce offloading single tasks at a time. We know people are comfortable delegating to people and resist trusting technology. Do not add AI without a corresponding reduction and expect ROI. Either plan for the overlap and taper down accordingly or just know the moment is coming to have to decide where to save and how much.
One sidenote: Another way to achieve actual ROI is growth. Moving people from simple, repeated tasks to more complex work is another way to delegate to AI while still growing your top line revenue. Example – moving Ops staff from Fannie & Freddie Refi workflow to a new condo product or new renovation product. AI can reduce cost AND unlock product expansion, if your company is built for change management.
3. Opt out versus opt in (also known as default to digital). One of the most powerful tools for adoption is changing the default. For instance, a company looking to expand the adoption of single source data collection (i.e. link bank accounts to a mortgage application) can make the primary “to-do” on the digital application portal to sign into bank accounts using an income and asset verification tool. Only when the consumer proactively opts out for an alternative PDF upload are documents accepted.
Thanks to AI, our companies are rethinking how Ops and CX interact to produce better results. The largest, digital lenders were doing this 5 years ago even before AI made the roadmap smoother. Consider where you can explore Opt out-first processes in your CX and in team member “set-up” functions.
AI for AI’s sake is expensive and probably no more efficient than what you are doing today.
In order to believe AI can drive adoption, a company must identify the customer who will use the tool, create the delegation plan to get that user to actually try it and remove any obstacle or choice (to the degree possible) from reverting back to the old way.
If you think AI will actually deliver on both the adoption and cost savings promises that technologists have been making for years, you have to answer the adoption question. Will this time be different?
Jeremy Potter is the founder of Next Belt Strategies.
This column does not necessarily reflect the opinion of HousingWire’s editorial department and its owners. To contact the editor responsible for this piece: [email protected].
English (US) · 