This piece originally appeared in the February 2024 edition of MortgagePoint magazine, online now.
Generative AI is the buzziest technology of recent times—and many boards are pressuring mortgage leaders to develop a game plan for adoption.
That’s for good reason. Generative AI is already catching on throughout the broader financial services ecosystem. And tech-forward lenders like Zillow expect the technology to be “a platform shift on par with the introduction of … the touch interface on the first smartphones,” per comments from the company’s Q1 earnings call.
That means lenders who don’t embrace generative AI now will soon be disadvantaged versus those who do. Lenders need a clear and informed strategy to embrace this new technology impactfully. Here, I’ll explain three practical steps before investing in this new technology.
But first, let’s take a closer look at what generative AI is.
What Exactly Is Generative AI?
Generative AI uses models trained on extensive data to create original text, images, and audio. That data can be public (i.e., from across the internet) or proprietary, depending on the use case.
Today, most people who have interacted with generative AI use a conversational, chat-based interface such as ChatGPT.
When a user enters a query (like “What documents do I need for a mortgage application?”), the AI draws on its training data to predict and generate the response most likely to satisfy the user’s needs. The “generative” part is new to the AI landscape. But AI itself has been around for years.
Most lenders have been using non-generative AI to analyze loan documents and automate decisions, sometimes achieving sub-second response times.
The most exciting capability that generative AI brings is its ability to produce new and coherent material—from chatbot replies to full documents—using only a simple, natural-language prompt.
The question is where generative AI can have the most significant impact. The answer depends greatly on your organization’s existing technologies, management structure, and needs. Assessing these factors is a critical first step in building your generative AI strategy—more on that in the next section.
Step 1: Perform an AI Readiness Check
Before you invest in generative AI, it’s essential to know whether your organization is ready for maximum impact. That involves considering two crucial areas: your AI maturity and efficiency needs. An experienced partner can help you evaluate both. The right partner will start by taking stock of the AI you already use, other than generative AI, the newest flavor.
If your lending operation is familiar with nongenerative AI, you can likely implement a generative component with minimal disruption. But if your lending operation uses little AI, adopting generative AI could put the cart before the horse.
Instead, a partner might suggest developing more institutional knowledge about AI (say, when it comes to proper data querying or responsible use) so your teams can make the most of generative AI tools.
A partner can also gauge your organization’s current efficiencies. They might spend time with call center agents to see if they’re drowning in calls or speeding through their queues. And they may check in with product teams about software development timelines.
This evaluation can help you understand where generative AI could impact most. Put simply: the more inefficiencies exist, the more value you stand to realize.
Both assessments give you a clear picture of your organization’s readiness for generative AI. Now, it’s time to think about use cases.
Step 2: Scope the Highest-Value AI Use Cases
Every generative AI strategy must highlight specific use cases with the highest impact in the shortest time. One promising thing for many lenders is using generative AI as a virtual loan officer assistant.
Imagine an AI-powered assistant that loan officers can ask to pull up specific loan documents, identify market trends, or distill complex regulations into simple explainers. Loan officers can use this chatbot while communicating with customers to complete otherwise manual tasks, deliver guidance faster, and serve more customers.
Another high-value generative AI use case for lenders is a software development copilot.
Picture conversational AI that can help software developers draft specifications, write and debug code, and automate application testing and deployment—all in seconds. That accelerated workflow is something we call digital efficiency. With an experienced partner, lenders can build a custom AI copilot to drive faster software development, lower development costs, and shorter timelines to launch new products or features.
Those are just two use cases for a rapidly evolving technology. It’s also possible to train AI models on internal datasets to develop dozens of other functionalities.
One lender has already done that.
It trained an AI chatbot on over 70,000 internal messages from an earlier bot iteration. The bot can answer basic mortgage questions from potential customers. It also uses sentiment analysis to gauge when a human agent should step in (e.g., if a customer seems frustrated or a question is on the bot’s “do not answer” list).
As more generative AI use cases emerge, a partner can help you quickly build, test, and tweak proofs of concept to identify the most practical applications for your organization.
Step 3: Roadmap Your AI Implementation
Once you’ve identified a few generative AI use cases, start mapping your implementation. The specifics will vary for every lender, of course. But your roadmap should do two things:
1. Prioritize each generative AI use case by cost and impact.
2. Establish a timeline for implementation.
First, I recommend prioritizing use cases with the highest impact and lowest development and implementation effort and cost.
In your AI readiness check, for instance, you might have learned that your customer service agents struggle to keep up with call volumes. Here, generative AI could deliver a significant increase in efficiency.
With the help of a partner, you can gauge the cost and complexity of implementation (i.e., whether you need a custom system or an off-the-shelf tool).
If you’re leaning toward custom-built software, a partner can help you identify ways to lower costs (say, by contracting out the development work so internal teams can focus on their most essential projects).
Next, build out your implementation timeline. If you’re in a downturn but expect volumes to stabilize in a few months, seize the moment to fast-track work on a new generative AI tool so you are ready to operate more efficiently when activity bounces back.
With a robust AI roadmap, you can give your organization a structured way to gain from generative AI.
The Best AI Strategy Is the One That Starts Now
Generative AI holds significant promise for mortgage lenders. Market leaders have recognized this promise and are working earnestly to find applications that will translate to meaningful product and service improvements.
Such applications could position leaders to gain market share and boost profit margins when the market picks up. But that trajectory is anything but guaranteed. Lenders who act now to implement generative AI solutions in their business will be best prepared to compete in this market cycle and beyond.
