How can lenders fully maximize the value that new AI tools can bring to our industry, and the borrowers we serve?
Over the past few weeks, we’ve watched AI-based tools continue to explode into the marketplace. Everyone is getting in on this game, from Open.ai’s fourth iteration Chat-GPT tool, to Google launching its Bard AI tool and Microsoft rolling out its AI assistant Sydney.
Those planning to attend our upcoming Ascent annual users conference have likely already picked up on the news that we’ll be presenting our own AI tools and functionality and demonstrate how they are integrated within the Mortgage Cadence Platform (MCP).
But to fully maximize the value these new tools can bring to our industry and the borrowers we serve, we must find a better way to share industry data so we can train up our models.
If we take the next step in bringing AI-powered technology that will deliver more mortgage lending efficiency, it will be imperative to feed a significant amount of data at the problem.
AI models require a great deal of information to be effective. This month, Adobe announced its new Firefly AI tool to help graphic designers quickly create new images from descriptions of what they’d like to see. To make it work, Adobe trained its model on hundreds of millions of images in its Adobe Stock library.
We can do that in some parts of our business where we have plenty of existing data. Mortgage documents constitutes one such case. If we ever hope to apply AI tools for higher functions, like decisioning or post-close QC, we’re going to need access to significantly more data.
To get to the next level, we must develop a strong data governance model that protects the privacy and intent of borrowers, supports various data security and protection protocols, sufficiently anonymizes data to make it useful and provides a reliable path to sharing data that lenders, technology providers and regulators can agree upon. This is required to create real solutions that will deliver real benefits to the industry.
As providers in the mortgage origination space, there is an incredible amount of consumer and transactional data processed by and stored in technology platforms. With such a massive treasure trove of data, advanced technologies like AI and ML can find patterns, trends and opportunities for tech enhancements that we can only guess at now.
The problem is that most companies in our space are sensitive about data, and for good reasons. We’re paranoid on behalf of our customers when it comes to data because lawmakers and their regulators require us to be, and our customers expect it from us.
No one wants to be responsible for losing their borrowers’ personal financial information. As a result, getting access to this information has been problematic and continues to be so.
But imagine if it wasn’t.
If you’ve seen what today’s AI models are doing in other businesses, you sense that the limitations are undefined -- if we can create and educate the models. If we can’t learn to share our data, securely and confidently, the promise of AI is always going to be circumspect and limited.
Like most technology trends, we’re of the opinion that a trend this impactful can’t be stopped. To get there, the industry must evolve in terms of its understanding of data governance, control and sharing. We may even have to deploy some new risk mitigation techniques to inspire confidence in lender leadership. We can get there. When we do, the results could be very exciting, for all of us.
By Jim Rosen, EVP, Services at Mortgage Cadence
Follow us on LinkedIn to be notified when our next article is released.