AI and generative AI applications continue to dominate conversations within the financial services industry, but implementing generative AI is near impossible if bank data isn’t standardized and accessible. 

What makes good data? 

“Good data is usable data,” Brendan Grove, chief technology officer at fintech Prizeout, told Bank Automation News, noting that it starts with creating data structures, organizing data and identifying business uses for the data. This is all necessary in today’s world in which data serves as the foundation for AI-driven insights and personalized banking experiences.  

This is “the year of financial data intelligence,” fintech MX Technologies stated in its 2024 Predictions: Financial Data Intelligence report, released in January. 

The report glorifies the value of data that sits within financial institutions, including its ability to power AI and machine learning, provide predictive insights and personalize the overall banking experience.  

However, data that is not understood cannot be used in a meaningful way. 

Investing in AI “starts with good, clean, quality data; if you haven’t started your journey, that’s where it starts,” Terrence Thomas, chief information officer at First Bank, said at Bank Automation Summit U.S. 2024 in Nashville, Tenn in March. 

“You can build on top of that foundation of good, clean quality data to get the best results from your AI journey,” he said. “If you don’t have good, clean quality data. … You’re making a mistake.” — Terrence Thomas, CIO at First Bank

According to business software provider G2, the “good, clean, quality data” Thomas described can be defined by the following components: 

• Valid; 
• Complete; 
• Consistent; 
• Uniform; and 
• Accurate. 

“Clean data is essential for [business intelligence] and big data teams, business leaders, marketing managers, sales representatives, and operational employees, especially in retail, financial services, and other data-intensive businesses,” according G2’s website. 

A solid foundation

But only 5% of data is typically structured, Abrar Huq, co-founder and chief revenue officer of fintech Arteria AI, tells BAN on a recent episode of “The Buzz” podcast. Unstructured data sits in documentation as a PDF or Word document that’s not organized in a way that a dataset could consume.  

To create a solid data foundation and achieve clean data, banks must standardize their data, or transform it into a standard format that is easily understood and used.  

This lack of standardization is a problem that many banks are aware of, Bentzi Aviv, global head of fintech solutions at Amdocs, told BAN. Banks are now investing heavily in data standardization to create a more holistic view of customer data, he said.  

Huntington, for example, increased its spending on outside data processing and other services 10% year over year to $166 million in the first quarter, according to the bank’s April 19 earnings supplement.  

Other institutions, like JPMorgan Chase, are working through cloud migration to accelerate their data standardization strategy and their ability to manage data. The bank plans to move 75% of its data to the public or private cloud by the end of 2024, JPMorgan Chief Executive Jamie Dimon said in his April 2023 letter to shareholders.

“Getting our technology to the cloud — whether the public cloud of the private cloud — is essential to fully maximize all of our capabilities, including the power of data.” JPMorgan CEO Dimon. 

Dimon added that moving to the cloud presents four benefits to JPM: 

1. Enhances the delivery of new services to market; 
2. Reduces cost of compute power; 
3. Provides compute capability across all JPM data; and 
4. Allows for the quick adoption of new technology. 

Building a clean database

Building a strong data foundation is a job for data engineers rather than data scientists, Prizeout’s Grove said. 

“You want to start by building a foundation with which others can work off of,” he said. 

Data scientists can analyze data, drive insights, create data products and make the end consumer happier through added value, data engineers can create data structures that make an organization’s data accessible, he said. 

JPMorgan, which leaned into its data engineering hiring strategy in 2022, now has 2,000 AI and machine learning experts and data scientists on its team, Dimon said in his letter to shareholders. 

To build a data structure, financial institutions must first address how the data will be used. They must ascertain what challenged they are solving for, whether that is more online banking logins, deposit or loan growth, or more cashback for members, for example.

After use cases are decided comes the “fun” part for fintechs like Prizeout, Grove said. 

The fintech, which services multiple credit unions, figures out how to use “the treasure trove of data to create value for [credit union] members,” Grove said. 

To achieve this, Prizeout has created pipelines and an extract, transfer and load structure that takes a credit union’s data and transforms it into the fintech’s “clean database,” Grove said. Once the data is in the database, all the fintech’s APIs, algorithms and recommendation systems can work off one set of assumptions that have already been transformed. 

To tap into a credit union’s data, Prizeout goes directly to bank cores, like Q2, COCC, Alkami and Lumen, and behind the online banking walls, Grove said.  

“We have the core suite of APIs already built so that what we’re doing is we’re making the connection,” he said. 

Prizeout has channels built into cores and has connections with 18 credit unions and credit union organizations, including: Suncoast Credit Union, MSU Federal Credit Union and Langley Federal Credit Union.  

Creating value through personalization

Once their data is in a standardized format, banks can start tapping it to provide personalized insights to their clients — an offering that is in high demand. 

Personalization is not just preferred, it’s expected. 

Amdocs’ April 8 Multigenerational Banking Personalization by the Numbers report surveyed 1,000 working-age adults in North America about their banking personalization preferences and found:  

• 78% are interested in tailored financial planning tools;  
• 66% are interested in building better money habits with tools that gamify personal finances; and 
• 53% of bank clients are interested in customer real-time recommendations generated by AI. 
• Personalized services used to be exclusive to the private wealth banking sector and required a lot of manual interference from relationship managers, Amdocs’ Aviv told BAN. Now, banks can look to technology and generative AI to support the roles of relationship managers.  

Banks “can use AI to analyze in real time the actual situation of the customer,” Aviv said, noting consumers expect the same personalization from their banks that they get from Netflix and YouTube. 

That personalization and use of generative AI isn’t possible without standardized data, he said.  

Keeping compliance at the forefront

While standardized and clean data present an opportunity for personalization and a foundation for future innovation, applying that data to new technology comes with caution. 

When tapping data sources for AI and generative AI, FIs must consider compliance in how their data is being used. 

Banks need to have confidence in the data that AI taps into, BankUnited Chief Technology Officer Michael Lehmbeck said at Bank Automation Summit U.S. 2024.

To ensure compliance, FIs are not only vetting the tech providers they work with but are also looking at how data is handled internally. 

Wells Fargo has implemented an internal council on generative AI deployment made up of the following bank leaders: People officer, head of digital strategy, head of technology, general counsel and chief operating officer, Steve Hagerman, chief information officer of consumer banking at Wells Fargo, said at the summit. 

Similarly, KeyBank set up a center of excellence to navigate generative AI ahead of set regulation, Dominic Cugini, chief transformation officer at the bank, said at the summit. 

While FIs set their own safeguards around AI, they can look to the Biden administration’s Executive Order on the Safe, Secure and Trustworthy Development and Use of AI, published in October 2023, for guidance on the importance of clean data. 

“In the end, AI reflects the principles of the people who build it, the people who use it, and the data upon which it is built,” the order states.