Artificial intelligence has become more of a buzzword in mortgage technology recently but new research raises questions about whether lender use has actually increased.
Mortgage lender familiarity and experimentation with AI are higher but full deployment is another matter, a recent Fannie Mae survey finds.
Limited or trial use of artificial intelligence or machine learning technologies has increased to 22% from 13%. Familiarity also is up at 65% compared to 63%.
But the percentages for lenders either reporting that they've fully rolled out the technology or are planning to have gone down since Fannie last conducted a survey on this topic.
Just 7% of senior mortgage executives responding to the government-related mortgage investor's survey have deployed AI, compared to 14% in 2018. The share that said they have a rollout planned for some time in the next two years dropped to 29% from 38%.
The industry's struggles with profitability and related changes in priorities may be one reason why the majority of lenders seem to be aware of artificial intelligence technologies but are less likely to be investing in full use of them.
"Given the rising costs of today's business environment, AI applications intended to improve operational efficiency are clearly highly valued by lenders and could function as a starting point," Peter Gahvami, vice president, modeling and data science, Fannie Mae, said in a blog.
Nearly three-fourths of lenders identified this as the primary motivation for the use of artificial intelligence, compared to only 42% five years ago. But borrower or consumer experience fell dramatically as a motivation to just 7% from 41% in the same time span.
Technology used to identify fraud and defects early in the origination process ranked No. 1 most frequently among respondents (26%). Also 20% of surveyed lenders, especially depositories, put compliance reviews at the top of the list of seven uses for AI mentioned in Fannie's survey.
More than one-fourth of lenders indicated fraud and defect detection was the most appealing category for them, followed by loan offerings (18%), property valuation (11%), virtual assistants (9%), borrower prepayment and default assessments (8% for each).
Interestingly, although AI-driven chatbots and other virtual assistants have gotten high marks in some consumer surveys and some mortgage companies have found they produce efficiencies in call centers, respondents overall took mixed views of them.
More than a third considered chatbot technologies to be the least appealing, followed by AI that handles prepayment assessments (21%), property valuation (19%), loan offerings (11%), default risk management (10%), compliance and fraud/defect detection (2% each.)
Of the 10 categories Fannie gave lenders to review and rank as challenges, the complexity of integrating the technology with existing infrastructure was No. 1, by 28% of respondents, followed by high costs (18%).
Other challenges topping the list for some were: lack of proven success (12%), data availability (8%), lack of overall tech strategy or staff skills (7% each), consumer consent or uncertainty around where to start (6% each), data security/privacy (4%) and possible bias concerns (3%).
