Artificial Intelligence: Hype vs. Reality and What it Means for Community Banks

There’s been no shortage of media coverage about the pitfalls and possibilities of artificial intelligence, or AI. Whether predicting the technology could replace 300 million jobs or bring about the very extinction of the human race, sensational predictions often provide few details about how the technology actually works, how it differs from existing software, and how businesses—specifically community banks and our business customers—can put it to use.

To help separate fact from fiction, let’s start with a working definition of AI and related terms, before discussing a few notable use cases for community banks.

Defining AI

Put simply, AI is any process that allows computers to simulate human intelligence. Famously, British computer pioneer Alan Turing’s imitation game (now called the Turing Test) judged whether a machine could demonstrate learned behavior through natural language responses to human-generated questions. Artificial intelligence could potentially be achieved through a variety of techniques like formal logic and decision trees.

The foundational technology that enables modern AI as we know it today is machine learning or a process by which computers can create their own algorithms to achieve a desired result through trial and error. In other words, machine learning algorithms learn from data. They are not explicitly programmed with rules, but they can identify patterns in data and use those patterns to make predictions about the relationship between inputs most likely to reach the desired outcomes.

One advantage of machine-generated algorithms is the ability to process much more input data than humans and to potentially spot non-obvious correlations between data points and a desired outcome. Additionally, while a human-created algorithm remains fixed, a machine-learning algorithm could self-improve over time by adjusting the weighting of inputs based on past results.

Think of an algorithm designed to determine a loan applicant’s creditworthiness that applies fixed weights to a variety of factors—such as the applicant’s income, credit score, and collateral—to determine whether to approve or deny the application and how to price the loan. If a machine-learning algorithm found that any given applicant characteristic was less likely to predict default than originally expected, it could assign less weight to that characteristic for future credit decisions.

Additional AI Use Cases

Like most new technologies, AI presents a mix of risks and opportunities for community banks. Potential use cases for AI include using chatbots to offer 24/7 customer support, monitoring transaction data to spot suspicious activity, and underwriting loans. The use of AI in these functions may decrease costs and free up staff resources to allow more time for direct interaction with customers.

While the technology can be leveraged to address basic customer service requests or provide after-hours support, these benefits must be balanced against the technology’s limitations in providing adequate responses to more complicated or multi-part inquiries, which could lead to customer frustration.

The ability to analyze vast amounts of data and identify non-obvious, non-intuitive relationships has another potential benefit: to identify more creditworthy borrowers. This could be accomplished by expanding traditional customer data sources to include information—such as personal cash flow, rent history, utility and cell phone payment history, employment history, and property ownership—that may correlate positively with the ability to repay a loan.

This ability to identify underserved borrowers could be a way for banks to reach borrowers that were previously underserved or credit invisible. According to the Consumer Financial Protection Bureau, 26 million consumers—about one in 10 U.S. adults—could be considered credit invisible because they do not have any credit record at the nationwide credit bureaus. Another 19 million consumers have too little information to be evaluated by a widely used credit scoring model. AI has the potential to build upon community banks’ efforts to expand credit access by enabling lenders to evaluate the creditworthiness of millions of consumers who are hard to score using traditional underwriting techniques.

It is likely too early to predict what the ultimate impact of artificial intelligence on the banking industry will be. This technology, while capable of producing impressive results, is still in its nascent stage.

And while AI has the potential to reduce some operational costs or to improve the accuracy of loan underwriting, it will never replace the local knowledge and personal relationships--the value of which is inestimable and are clear differentiators for community banks.