The buzz around artificial intelligence (AI) is a cultural phenomenon. While the body of true experts in the field is relatively small, there’s an intense demand for education on the topic. Leaders in financial services are especially keen to learn how AI can augment their business models or workflows. 

Financial services firms know they’re sitting on petabytes of financial data that could create revenue opportunities if only they could process it. They also read the headlines about how AI will change the world, and it triggers a fear of being left behind (something traditional financial institutions were worried about before AI hit the scene). 

Sorting through the hype is tricky. It’s much easier for firms to write articles stuffed with keywords and tech jargon about the possibilities of AI than it is to talk about real-world applications. This is especially true because the differences between algorithms, AI, machine learning (ML), and subtopics like generative AI are challenging for casual readers to keep straight. 

Data hygiene and enrichment 

The humble spreadsheet is a workhorse of financial services and information-based work. Anyone can learn to use a spreadsheet within a few hours, and the online resources for performing advanced tasks boggle the mind. Unfortunately, spreadsheets are also prone to catastrophic errors, such as when an Icelandic bank sold a tranche of shares that were undervalued by about16 million pounds. As the data was moved between spreadsheets, it wasn’t properly cleaned and formatted, resulting in a massive and irreversible error. 

Generative AI can help analyze, clean, and organize complex datasets, which is an important first step because many of the other potential AI applications require high quality (i.e., clean, accurate, organized, formatted) to deliver meaningful results. 

This applies to payment processing, where accurate customer data is vital. It will only grow more important as institutions embrace real-time payments where erroneous payments can’t be stopped or reversed. 

Speaking my language 

Discussions of AI in financial services often reference the opportunity held in large volumes of numeric data. AI models are tuned to digest data much faster and more precisely than humans. Which industry has loads of this data warehoused and waiting for a use case? Financial services. In a presentation at the 2022 MIT Technology Review EmTech Digital Conference, one of JPMorgan Chase’s top AI researchers, Sameena Shah, busted this common myth, 

“When we think about financial firms, we think it’s mostly about numeric data, but actually from an AI perspective it’s mostly about textual data.” 

She went on to talk about Know Your Customer (KYC) protocols as an example. Virtually all financial institutions and most financial services firms participate in KYC at some level. KYC generates a lot of textual data with high variability, making it a perfect use case for AI analysis. 

An automated fraud referee 

Another domain related to KYC is anti-money laundering (AML) and sanctions. Complying with these rules is necessary for financial institutions, and the rules change regularly because the lists of prohibited entities change regularly and unpredictably. 

AI models can be programmed to enforce compliance with these frameworks, acting as a kind of mechanical referee that can update the model when the rules change. 

It’s also capable of integrating disparate data points into meaningful fraud analysis. 

Criminals are proficient at exploiting human and system-based vulnerabilities. As discussed on a November 20203 episode of the Payments Professor, AI and ML can be trained to examine hundreds of data points to assess whether a transaction is anomalous. These can include obvious data points such as IP address, but they can also look at the physical position of a smartphone to check if it’s sitting on a fraudster’s charging shelf or in the hand of the account holder. 

Don’t confuse possibility for practical application 

A brief word of advice for anyone evaluating the opportunities of AI at your financial institution: the technology has tremendous potential if applied to certain tasks. It’s likely to become a technological boondoggle if your only criteria for success is “we don’t want to get left behind our competitors.” 

Ask your team to spend time researching AI and how it might be used to improve tasks they perform daily or weekly. Estimate the value of time saved or opportunities you can capitalize on, then compare it to the estimated cost of implementing an AI solution. 

Building in-house competency with AI is possible, but you may find it much easier and economical to partner with vendors already invested in it. In short, learn what you can on your own and partner with trusted experts to unlock the full value of AI at your institution. 

LeackStat 2024