Consider the massive size of real estate lending. The Fed’s latest report shows mortgage debt topping $9 trillion. When including mortgages from businesses, it tops $15 trillion. Over 10 million homes and commercial properties sell each year.
Equally staggering is how much data exists on the borrowers. Lending is big business and big data, and while banks are harnessing that data, private lenders now need to follow suit.
Powered by the rise in computer processing power and individual data, artificial intelligence can find patterns that predict borrower behavior, helping lenders make more money. AI is not quite the stuff of movies like Terminator. Think of it as software analyzing statistics at scale.
How AI Works
There are two kinds of AI: supervised and unsu
pervised. With supervised AI, humans create certain “rules,” and software sorts data based on the set rules. The AI learns the lender’s underwriting rules — collateral value, borrower experience, etc. The difference is AI can examine thousands of applications in short order, grouping them by profitability or default risk. Supervised AI is helpful for big banks, but most of us don’t need that heavy lift.
For those of us in the hard money lending space, unsupervised AI is more relevant. In unsupervised AI, humans don’t create rules at the beginning. Rather, a data scientist feeds it a massive amount of data and essentially flips a switch to let the AI identify patterns across millions of variables.
Imagine AI analysis of borrower records shows that people who frequently post messages on Facebook at night default at a higher rate than others. Perhaps these borrowers are sleep deprived, disorganized or bad at time management. What matters is that you now have a predictive variable and can screen future applicants for night owl posts.
There are three groups of AI applications, both supervised and unsupervised, now used in the lending process:
AI that determines creditworthiness for borrowers with limited credit history:
As in our example about Facebook, many companies are using AI to sift alternative data to predict creditworthiness. This has been important for markets like Africa where a growing middle class uses smartphones but lacks traditional credit or FICO scores.
With AI, a lender could examine a borrower’s digital footprint for creditworthiness by having the applicant download an app to their phone. With the app feeding data to a credit scoring platform, variables such as social media, browsing, geolocation and more are used to get a fuller picture of the borrower. One company called Lenddo has done this across Africa and Asia.
China is another market where consumer credit scores are underdeveloped. Technology companies now draw data from behavior online and elsewhere to analyze people’s search, location and payment data to compute creditworthiness in what’s called a “social credit” system.
Some pioneering lenders in North America are experimenting with search history data. Many car buyers, especially young people who haven’t taken out much credit before, don’t qualify for auto loans. Auto lenders are now getting comfortable extending loans despite “thin” credit scores when the borrowers’ search characteristics are favorable. Mortgage lenders are also trying this. While not official yet, the Wall Street Journal reported that Freddie Mac partnered with tech company ZestFinance to expand underwriting to make mortgages more available for first-time home buyers and minorities using search data.
AI that streamlines the existing loan process:
Large lenders use AI to reduce underwriting overhead and delays, which increases profits per loan. Recently, some tech companies have gone further by using AI to automate the entire loan process. Fans of this technology say it leads to less bias and better loans.
A company called Upstart claims to use AI to automate all steps from the application to the final decision for a loan. In addition to FICO scores and job history, this AI examines a borrower’s education, SAT scores, GPA and other factors. Although no humans are involved in credit decisions, Upstart claims that full automation has not led to any spike in defaults.
AI that finds and delights customers:
When it comes to e-commerce, the behemoth Amazon loans billions of dollars to small businesses reselling on its platform. AI is used to identify borrowers who are low credit risks based on their inventory turnover and profitability. Amazon relies completely on AI, so no humans are involved, not even with filling out an application, and it offers unsolicited loans on “take it or leave it” terms.
In the field of customer support, AI is mostly used for things like chatbots. But I know of one company that has started using AI to help customers pay off loans faster by sending borrowers a no-pressure analysis of whether they can pay faster to save interest and fees. Other startups help consumers with their entire financial picture, including increasing take-home pay, reducing expenses and consolidating debt. Bank customers seem to appreciate the service, which makes them more reliable borrowers for the future.
Using AI In Hard Money Lending
It’s tempting to think hard money lenders don’t need AI since we deal with far fewer applicants than banks, and we always have property as collateral. But AI could be used to find new potential borrowers, streamline the lending process, identify risks and opportunities and more. You don’t need to go out and hire a bunch of coders and data scientists; some AI companies offer software on a subscription basis.
In the old days, some hard money lenders would claim they didn’t need the internet for their jobs. Today, few would get anything done without it. That’s how AI will evolve in our industry: from a novelty to an indispensable tool that lets you earn more. I think first movers in AI will become our industry’s long-term winners.