How lenders are using AI now

It seems inevitable that artificial intelligence will play a larger role in the mortgage industry going forward. But exactly what form this will take and how quickly AI technology will be implemented in the origination space is unclear.

The mortgage industry is notorious for its slow adoption of technology. Despite that, a June Insights report from consulting firm Startmor Group found that 22% of lenders surveyed have already started to use AI.

So far, AI has mainly been applied in marketing rather than in operations and fulfillment, the report found. Lenders that have thus far experimented with such technologies are typically on the larger side and have the capital and the scale economies to do so.

Per the report, 35% of mortgage shops are using AI to identify opportunities with current customers, 30% use AI to target potential customers, while 25% of lenders use AI to make better product offers to borrowers. Close to 100 lenders responded to the survey questions on AI, while 350 lenders participated in all parts of the consulting firm's survey.

Using AI for more nuanced things like underwriting and workflow rank at the very bottom, with only 5% of lenders responding that they use AI for these tasks.

"We still need a human to do the thinking when the system cannot accommodate situations that are not a clean pass or fail," said Jennifer Fortier, principal at Stratmor, in a written statement.

According to the report, one of the main benefits of AI that lenders are taking advantage of is its ability "to sift through reams of data" to determine who is the most likely to close on a loan. 

"And if the AI can then help the consumer prioritize what they should focus on to get the best deal on their mortgage, it frees up the lender's staff and moves more prospective borrowers into the market," the consulting firm's report said.

AI technologies can also streamline the process of verifying data provided by borrowers during the home buying process and make sure that a loan checks off all the boxes prior to being sold to an investor.

"One of the biggest issues in mortgage banking is that the consumer shows up with a bunch of data and nobody believes them. There is no trust in the process," said Garth Graham, senior partner at Stratmor, in a written statement. "The lender asks the borrower for data, then they pay a third party to verify it, but not until after they have asked the borrower to give them the income, verify it with W2s and copies of their tax returns and then recent pay stubs to make sure the borrower has the 'ability to pay.' What a mess."

Graham sees AI as the key to removing "all the checking and the checking of checkers that has created an environment where the cost to originate is over $10,000."

Putting such technology in place is not likely to replace all human interaction between the mortgage lender and the borrower, the report noted.

"Where AI cannot replace a loan officer, a place digital lending has also failed, is by providing a sounding board and voice of reason and reassurance to a borrower," said Jennifer Smith, principal at Stratmor. "When — heaven forbid! — a borrower is irate, they do not want to talk to a non-human. I can't imagine AI trying to placate a borrower whose closing date has been moved or who has received a foreclosure notice."

Per the report, challenges lay ahead for the implementation of AI technologies. One such issue is user adoption.

"Every lender is familiar with old manual processes that just will not go away, despite deploying solutions to streamline the work," Fortier said. "User trust and 'old habits die hard' factors are big obstacles and not easy to overcome, particularly in the absence of workflow redesign."

Lenders may also be wary of implementing AI for underwriting because there are no clear regulations in place for it. Industry stakeholders expect some sort of regulatory framework to be implemented in the next couple of years. Companies that build automated systems may be required to be more transparent about what data they're relying on and how their systems make the decisions.

Uncertain regulatory status and the cyclicality and fragmentation of the mortgage industry will keep adoption at a very leisurely pace, the report concluded. "The industry change with AI will be significant, but it is going to take some time," Graham said.