putilov_denis - stock.adobe.com

Finance industry giants disclose AI challenges

Education, explainability, privacy and integration are some of the problems institutions face when implementing machine learning tools and technology.

NEW YORK -- From credit cards to title insurance to loans and even fraud and risk management, the finance industry is making use of AI tools and technologies. But implementing these tools and working through their inherent challenges has proved difficult for financial institutions.

That was the sentiment shared by major credit card providers, insurance companies and banks during the Ai4 2022 Finance Summit here on March 1. The application of AI tools in finance is necessary for many institutions, and many plan to increase their budget to continue to implement the tools.

Sixty-five percent of 706 senior IT professionals in the finance sector plan to increase their IT budget in 2022, according to a November 2021 report by Enterprise Strategy Group.

Among that increase, 62% of respondents say they will likely increase spending in artificial intelligence and machine learning.

However, since finance involves sensitive data from consumers and large corporations, some enterprises find themselves trying to find a balance between the benefits and risks of AI tools.

The first problem for companies using AI technologies in finance is a lack of education, said Priya Rajan, CMO at DataVisor, during a panel discussion about AI and credit cards at the conference.

DataVisor, a fraud and financial crime detection company, uses AI and machine learning to identify fraud attacks. But, Rajan said, so much is still unknown about AI that it's difficult to identify what's true AI versus what's not.

"Education is such a big part of this transformation in this industry and I do expect that to continue in the next decade because we're just scratching the surface of what the technology is and what it can do," Rajan said.

Image of a panelist discussion at Ai4 Finance conference
Panelists discuss ways AI has impacted the credit card industry.

AI challenges in finance

Another challenge is explainability, said Clay Jackson, vice president of product management for small business cards at Capital One, during the panel discussion.

Credit card providers are put into tricky situations where not giving more credit to someone might mean that person can't pay for funeral fees or get a job.

"We're taking actions that impact customers' lives," Jackson said.

When a customer is denied credit, institutions must be sure why they're saying no, Jackson said.

"I feel like the explainability problem slows us down," he said. "And for the right reasons."

AI tools also pose privacy and bias challenges to the credit card sector, said Rick Ballmann, vice president of engineering, data intelligence and customer experience at American Express, during the same panel discussion.

While banks and financial institutions can offer consumers loans based on where they currently spend their money, it doesn't mean they should, Ballmann said.

"It's like knowing the boundary between going too far and what the customer would appreciate as good customer service," he continued. "It's not exactly what's possible, [but] what's the right thing to do for the customer."

The use of AI tools in fraud detection also presents data privacy issues, said Besa Abrashi, senior project manager at American International Group (AIG), a finance and insurance company.

As someone who works on AIG's fraud detection team, she said there are different data privacy regulations and procedures that must be followed when exchanging data from one place to another – on premises, off premises or in the cloud.

"You have to go through all these data privacy regulations and procedures to get all the approvals," Abrashi said.

To be able to stay competitive in the game, you have no choice but to develop your AI lab space within your organization.
Michelle WangSenior lead management officer, Wells Fargo

AI software vendors need to be flexible about where they install their AI platforms to ensure financial institutions can avoid data privacy issues, Abrashi continued.

"Especially now with all the regulations that are becoming stricter and stricter in terms of what type of data you can share," she said. "Everything now is PII [personally identifiable information]. Not only the first name and the last name, but they consider everything PII."

In the reinsurance industry, AI tools have the potential to solve many problems, but integrating them is the problem, said Dean Marcus, an actuary at Guy Carpenter & Company, a reinsurance company based in New York.

"Partially because of data challenges … but also just because of calibrating the models and ensuring that they're doing what you want them to do in a timely way," Marcus said.

Despite all of these challenges, financial institutions have little choice but to use AI tools to stay competitive, said Michelle Wang, a senior lead management officer at Wells Fargo.

Wang works in the risk management department at Wells Fargo, which uses AI technologies for things like data analytics, she said.

"As newer adopters, there are always challenges to understand how to use the tools that are available to us," she said. "To be able to stay competitive in the game, you have no choice but to develop your AI lab space within your organization."

Enterprise Strategy Group is a division of TechTarget.

Next Steps

Problems with generative AI for firms such as JPMorgan Chase

Dig Deeper on AI business strategies

Business Analytics
Data Management