AI In Finance

A call comes in at 9:30 PM.
It’s from your bank’s fraud department.

“Sorry to bother you this late,” the agent says. “Did you just make a purchase at a restaurant in Dallas, Texas?” Since you’re sitting on your couch at home (which is not in Dallas), watching “The Walking Dead,” the answer is no. The agent thanks you and goes on to deny the transaction.

You’re starting a month-long European vacation next week, so you decide to send next month’s car payment this month. Your online banking app asks if you need to double check the numbers: “This payment is much larger than your usual payment. Are you sure?”

No matter how terrific your bank or credit union is at customer service, it doesn’t have the manpower to look in on your spending habits at 9:30 on a Tuesday night. What it does have is Artificial Intelligence doing the data dive based on your money management history.

Artificial Intelligence (AI) has been transformative in the finance industry, allowing institutions to monitor and find patterns in millions of diverse microtransactions. About 32 percent of financial service providers are already using AI technologies like Predictive Analytic and Voice Recognition, according to a research conducted by the National Business Research Institute and Narrative Science.

AI is being used to find errors and inconsistencies in account records and change the way banks and insurance companies underwrite policies and loans. AI bots are serving as digital payment advisers and biometric fraud detection mechanisms.

The modus operandi of money laundering is to funnel money through a long series of legitimate transactions so that its criminal origins are overwritten. Artificial intelligence is capable of analyzing and finding patterns in millions of transactions, so it can detect money laundering and other crimes, preventing banks millions of dollars of fraud losses.

A recent article in Risk and Insurance online said that “Artificial intelligence has the potential to take underwriting from a detect-and-repair mindset to a predict-and-prevent philosophy. That means fewer educated guesses, more accurate information and not only making sense of treasure troves of data but using it as a competitive advantage… From a restaurant’s health inspection performance to a factory’s OSHA violation history, an incredible amount of data is currently available to underwriters — and machine-learning technology can analyze it to find red flags and help make more accurate underwriting decisions.” 

If you’ve used the chat function for customer service, chances are you were chatting with a bot. About 80 percent of repetitive transactions and queries can be handled by Robotic Process Automation (RPA). This allows knowledge workers to dedicate their time in value-add operations that require high level of human intervention.

Just as humans can be specialized, there is a version of AI called “weak AI” or “narrow AI.” This is a program that’s not designed to learn as it gathers more data; its job is to stay within a specific set of guidelines and accomplish a single task. For all their humor and responsive answers, Apple’s Siri and Amazon’s Alexa are considered weak AI.

According to a Machine Design article, these systems “are looking for things similar to what they know and classifying them accordingly. If you ask Alexa to turn on the TV, the programming understands key words like On and TV. The algorithm will respond by turning on the TV, but it is only responding to its programming. In other words, it does not comprehend any of the meaning of what you said.” But they will turn on the TV every time, not another appliance. To err, after all, is uniquely human.