5 ‘extraordinary turnarounds’ for global well-being

Close proximity is the key to knowledge spillovers among startups

6 ways to nurture IT talent

Credit: Igor Kutyaev / iStock

Ideas Made to Matter


Financial services’ deliberate approach to AI


Financial service companies have been using artificial intelligence in various forms for decades. So it’s not surprising that employees and executives alike are showing early interest in generative AI-based tools such as ChatGPT.

What is surprising, industry executives said at the 2024 MIT FinTech conference, is the rapid pace of adoption in the financial sector, where generative AI is proving valuable for a wide variety of employees, not just data scientists.

“AI is basically top of mind for everybody. It’s been a few months of transition and very rapid acceleration,” said Jose Lobez, PhD ’12, vice president of global AI and data innovation at Visa.

During a panel discussion, Lobez and Asim Tewary, the former chief AI officer at PayPal, discussed early adoption of AI in financial services, why the industry will be slow to adopt AI in some customer-facing cases, and the benefits of partnering with fintechs. The panel was moderated by  MBA ’16, a research scientist at the MIT Initiative on the Digital Economy.  

AI is assisting financial workers

As in other industries, in financial services AI is largely augmenting the tasks performed by employees rather than replacing human workers. Early use cases include back-office automation, data aggregation and visualization, and fraud prevention.

Far from taking jobs away, rolling out AI tools at Visa has led to increased hiring in recent years, Lobez said. One reason is the company’s emphasis on using AI to accelerate innovation and high-quality work instead of to cut staffing, he said.

Another reason is the current state of AI. “Even if companies wanted to be able to reduce head count, I don’t think the technology is 100% there yet,” he said. “I always tell the team, ‘We don’t need to solve everything with AI, because we probably can’t.’ Sometimes you just can solve things with a really simple model.”

Tewary likened the potential long-term impact of AI in financial services to the ATM in banking. When tellers were no longer needed to dispense cash or check account balances, they could take on higher-skill responsibilities, such as advising customers and cross-selling products.

Even so, practical AI applications may affect some roles and functions in the industry more than others, Tewary said. In particular, content generation, marketing, communication, and paralegal services may not need as many human workers.

Regulatory concerns stall some uses

The financial industry has been slow to put AI in front of customers. Lobez and Tewary said that financial service firms are experimenting with products such as AI-enabled chat agents, but this work is happening slowly.

Customer-facing applications require much more effort than internal use cases, in part because they’re subject to far more regulatory scrutiny, Tewary said. AI models must be more explainable (showing how a system came to a decision) and traceable (showing what data, processes, and artifacts went into the system). Additionally, outputs must be validated to avoid hallucinations, or inaccurate answers based on fabricated data.

“You have to be able to explain why certain decisions were made — why a credit limit was set at that amount, for example,” Tewary said. “There’s an absoluteness that’s expected from regulators about being able to explain how the decision was made. Anytime you impact the consumer or introduce a system risk, regulators get very concerned.”

Also concerning is the increased sophistication of deepfakes that could mimic either virtual or human agents and convince customers to do things such as transfer their life savings into what they think is a legitimate account. That heightens the importance of not only getting customer-facing AI products right but also keeping interactions secure. “It’s a matter of making sure we stay a couple of steps ahead of the game,” Lobez said.

AI will become the operating system of work

Ruane said that the state of AI adoption in financial services is consistent with many previous types of general-purpose technology. What begins as a point solution addressing individual tasks continues as a wave of process innovations capable of reimagining entire industries — much like manufacturing transformed when electricity came to factories.

Related Articles

4 ways AI is changing finance
5 predictions for fintech in 2024
Why finance is deploying natural language processing

Tewary said he believes that in the future, AI will likely become so mainstream that, despite running key applications, it will be regarded much like the operating systems on today’s mobile devices.

“As I think of it, AI will just go away and become pervasive, like an operating system you build applications on top of,” he said. “AI is underlying everything that empowers how people do things.”

Amid this push for innovation, large companies can benefit from working with fintechs and startups. Visa has approximately 2,000 partnerships with fintechs and startups, Lobez said, and launched a $100 million generative AI fund to work with startups that are rethinking the future of payments and commerce. “It’s impossible for a big company to solve all its problems on its own and to maximize value for its clients and customers,” he said.

Inertia and resource allocation can slow things down in the enterprise setting, and startups bring unique skills and flexibility to the table, Tewary said. That said, companies like PayPal look for mature startups with a proven value proposition.

“There’s a complex technology ecosystem,” Tewary said. Established finance companies “want to work with startups with an architecture that’s API-driven and can be integrated with that ecosystem.”

Read next: Can generative AI provide trusted financial advice?

For more info Sara Brown Senior News Editor and Writer