Artificial intelligence in banking has evolved over decades, long before the recent surge of interest in generative AI and machine learning. Financial institutions have relied on increasingly sophisticated decisioning technologies to improve underwriting, credit risk management, fraud detection, and customer service, demonstrating that AI has long been a foundational component of modern banking operations.
From early credit scoring models to today’s adaptive fraud detection systems, AI has consistently helped financial institutions make faster, more informed decisions. While the technology has become more advanced, one principle remains unchanged: AI is most effective when it augments human expertise, allowing organizations to improve accuracy while maintaining the oversight needed to manage risk.
Here are two good examples. FICO, a company built on helping many different verticals has a long history in applied technology. In the early 1980s they were refining their business model to help the underwriting process. As a way to make underwriting more consistent, faster, and efficient, the FICO score provided a tool that enabled large issuers to deal with massive paper volumes. Integrating this into underwriting, then later credit management, is as good as an example as you will find on early artificial intelligence.
At Citi in 1985, when autodialers were considered rocket science and PCs were beginning to roll out, we used a room-full of Dec Rainbows to screenscrape data into one from CICS machines to move into an analytic file. One room had about 50 of these “leading edge” CP/M 8088 processors and staff would keep hitting the enter button to get to the next file, then run macros to save and store the data. We’d move the data over to a DEC Vax/VMS on a floppy, then scrub the file through an algorithm that would use a “best time to call” strategy across multiple US timezones. That was AI, as primitive as it may be.
A key point about AI is that it should help in the decisioning process. The more you use it, the smarter it gets, but it still needs human input.
In today’s fraud management, some complaints about false positives.
A lot of that goes on the human side of the equation. In this role, you are attempting to tighten point of sale decisioning to ensure irrefutable transactions. You can shut everything off and let no-one in; that’s one way to stop false positives, but it also ends “true negatives”. So, with the help of AI, and some operational forethought, you tweak the parameters until you get a balance.
This way, when you are on a business trip to San Francisco, your cardholder will not set off alarms when they go to buy a cup of coffee at Starbucks. Now if that same transaction occurs in an eastern bloc country, expect bells and whistles to go off.
As artificial intelligence in banking continues to mature, its greatest value remains its ability to support better decision-making rather than replace human judgment. Whether evaluating credit risk, detecting fraudulent transactions, or optimizing customer interactions, AI enables financial institutions to process more data and respond more quickly than traditional approaches.
The challenge is finding the right balance between automation and oversight. Organizations that continuously refine their AI models while incorporating human expertise will be best positioned to improve operational efficiency, reduce false positives, and deliver a stronger customer experience.
