As those of us in the lending industry know, unsecured loans inherently carry with them the highest level of risk. Because they’re not backed by any piece of collateral like an auto loan or a mortgage, they’re often an afterthought in the innovation process. 

In this episode, Upstart co-founder Paul Gu shares two reasons why unsecured loans were a great place to apply Artificial Intelligence:

1. To demonstrate efficacy

If you want to demonstrate the efficacy of machine learning when applied to something that’s been done for a very long time, like the lending industry, you want to apply it in the place where it’s going to matter the most - and be the hardest.

And that’s unsecured loans. 

Every other asset class like car loans, home loans, and even credit cards to some degree are backed by "something." For car and home loans, they’re of course backed by the physical assets of cars and homes. And the credit card is backed by the further utility of being able to use the card. 

But when you give someone $20,000 and say, “Please pay me back,” you’re not really backed by anything. And so you’ve got to be VERY good at deciding who does and does not get those loans.

2. It’s the Most Broad Loan

From the consumer perspective, it’s the most flexible kind of loan. Whereas a car loan or a home loan can only be used for certain things, the unsecured personal loan has no limits, and so it’s a natural starting place. 

Once you’ve nailed down secured loans, it’s a natural transition to other types of loans as well.

A Shift in Approach

So how did we Upstart shift from the traditional lending approach to a more AI and Machine Learning-driven approach? 

First, we use an immense quantity of data. Over 1,000 variables and millions of rows of repayment data are used to train our AI model to make the decisions needed, both through traditional and non-traditional data points. 

And secondly, we changed the way we learn from the data.

Traditionally, a person looks at the data and plots points on a line. That doesn’t work in the real world because we all know that the world isn’t a straight line. To actually make individualized predictions, we implemented state of the art Machine Learning algorithms, and combined the big data approach with the modern algorithm approach, resulting in what we call AI lending.

Reducing Friction

So why the focus on AI and Machine Learning? Wasn’t the old way of lending working just fine? Well, yes and no. 

At the end of the day, underwriting a loan is just figuring out everything needed to ascertain the risk of the loan. And there are multiple ways to do that. You can follow a person around all day and make sure that they have the ability to pay back the loan, but very few consumers are going to sign up for that. 

A highly developed AI and ML-driven lending process results in much lower friction for the end user, and in today’s world that cannot be overstated. The truth of the matter is that the more work you ask people to do, the less likely they are to complete your application process.

As soon as you ask for any kind of document upload, you’re going to lose a portion of people who don’t want to have to go through that process. 

So really, the AI and ML-driven process exists not only to make life easier for those of us in the lending industry, but to remove the friction for the end user, allowing more consumers to originate loans through our organizations. 

And isn’t making life easier for our consumers why we’re all in this business to begin with?