Reading Time : 0 Mins

Cloning 120-Volt Batteries: The Inexpensive Route To Build Better Underwriting Efficiency

DP_Vasu

Bibliophile, Movie buff & a Passionate Storyteller.

How do you identify and offer credit to the right borrowers? How can you help your underwriters approve more of good borrowers and less of the bad ones?

We discussed with a global bank recently where the bank’s 25-member underwriting team has about 3 senior underwriters who review qualified applications that need their experience for intense evaluation. In contrast, the job of eliminating the bad leads and passing on qualified applications to the senior underwriters lies with the rest of the 22 member team of junior underwriters.

The bank has no qualms with this approach since the rule engine they currently have is working well for them. The team of junior underwriters follow the business rules and are doing a good job of identifying good applications. But with 400 inbound credit requests coming in every day, they are looking at ways to increase the number of “good” borrowers and eliminate the bad ones.

Are Rule Engines helping your lending process be faster and less costly?

The scenario above is common for any financial institution such as credit unions, specialty lenders, or cash advance companies that does lending. Most of these institutions are using rule engines today, which have been built carefully considering the needs of their lending market, and they are performing well. However, the goal is to see an increased number of “good” borrowers who can be repeat customers, the revenue source that leads to their very existence and growth. How can they identify them?

Should they continue to add more underwriters to the team who can follow the rule engine and identify new borrowers? “The human body generates more bioelectricity than a 120-volt battery and over 25,000 BTUs of body heat” ruefully observes Morpheus in the Sci-Fi movie “The Matrix” that released 20 years ago, and goes on to explain the terrible solution(use of human power) the machines resorted to in order to survive.

Whether it’s a myth to be debunked or a truth to wait and watch for, the point is about the “human power” involved. For example, in the aforesaid bank scenario, building a system that “emulates” what the team of underwriters can help achieve the goal. As the ‘imitation’ system starts offering new borrowers, all that the non-senior underwriters have to do is validate and see if they would have approved them in the absence of the system.

Machine Learning systems to build a decision-making matrix

Now, how do we build a system like that? A “supervised” machine learning (a subset of Artificial Intelligence) system is a good approach to build such kind of systems. How can a supervised machine learning system help here? Remember “The Imitation Game” movie where Alan Turing builds a machine to crack the codes?

Jack English—© 2014 The Weinstein Company. All Rights Reserved.

Is AI expensive?

And as most of us think, the use of Artificial Intelligence in a scenario like what we are talking about is not going to be very expensive. In fact, using a rule-based workflow tool can actually be expensive commercially and otherwise in the long run for the following reasons:

  • Rule-based systems are built on a set of facts about a situation and a set of rules for how to deal with those facts.
  • Rule-based systems are deterministic systems and not having the right rules can be challenging
  • As more and more rules get added, rule-based systems can become unwieldy.

On the other hand, unlike rule-based systems, machine learning is probabilistic and uses statistical methods rather than deterministic rules. As mentioned above, historical context plays a critical role in what machine learning says about future outcomes.

Having said that, although Artificial Intelligence systems can build in the efficiencies that lending institutions and Chief Revenue Officers crave, and the transparency that consumers demand, humans such as underwriters will always play a crucial role in decision-making.

Further Read:

To understand why the traditional credit scoring is not enough for lending businesses; the challenges involved in ML adoption, how to overcome them and how similar institutions like yours are leveraging AI/ ML to improve their ROI and reduce default rates, download our white paper on “Applying Data Science to Financial Lending.

Leave A Comment

Related Posts