How Financial Institutions Can Prepare for the Future of Fraud with Responsible AI Deployments – with JoAnn Stonier of Mastercard

As financial organizations navigate an increasingly sophisticated landscape of cybercrime, many are looking to generative AI and agentic systems for a competitive edge. However, as the World Economic Forum notes in a 2025 white paper, while AI adoption promises to enhance security and streamline operations, it also introduces significant complexities related to data privacy and responsible deployment.

According to the Federal Trade Commission, consumers reported losing more than $12.5 billion to fraud in 2024, a 25% increase over the previous year, a staggering figure that underscores the urgent need for a robust, multi-layered security approach.

In the financial services sector, tackling the inherent problem of fraud while balancing the risks of AI deployment is the central challenge for the sector at large. It also serves as a representative case for virtually any business handling sensitive data.

This article unpacks how financial services leaders can overcome these challenges to build a robust, responsible approach to AI applications. Based on insights shared by JoAnn Stonier, Fellow at Mastercard, on a recent episode of the ‘AI in Financial Services’ podcast, readers will gain a clearer perspective on how to build the foundation for a future where sophisticated AI systems work in tandem with human oversight.

This article delivers three key insights for financial leaders seeking to bring AI into their own businesses responsibly:

• Choosing maturity over hype: While cutting-edge applications like generative and agentic AI grab headlines, more established, deterministic AI capabilities can better improve real-time fraud prevention.
• Using data to understand patterns, not people: Responsible AI in fraud detection isn’t about collecting more personal data, but about using existing data more intelligently to identify complex behavioral patterns.
• Adopting a team-sport approach to governance: Deploying AI responsibly requires a collaborative effort across an organization, with a clear purpose and an iterative process that considers a wide range of risks.

Listen to the full episode below:

Guest: JoAnn Stonier, Fellow of Data and AI, Mastercard

Expertise: Privacy, Data Governance, Data Ethics, and Responsible Data Practices

Brief Recognition: JoAnn, formerly Chief Data Officer at Mastercard, leads enterprise-wide data governance, analytics, and innovation efforts. Her career spans privacy, risk, and AI strategy across finance and technology ecosystems. JoAnn is also a seasoned professor with influence in regulatory and academic circles.

Choosing Maturity Over Hype

Much of the media attention on AI focuses on the latest advancements, such as LLMs, generative AI, and emerging agentic systems. While these technologies are powerful, Stonier points out that the real, measurable business value in financial services often comes from more established, deterministic AI.

JoAnn explains that Mastercard, for example, has been using data and analytics for over 17 years to monitor its global network 24/7. Initially, this involved looking at patterns based on past behavior to predict fraud. However, with the evolution of AI, the company can now analyze billions of transactions in real time, enabling faster, more accurate fraud detection.

Stonier continues, explaining that for the customer, this means fewer false positives — the dreaded moment when your card is declined for a legitimate purchase. Stonier gives an example of a customer with a summer home: a few years ago, a purchase in a second, distant location might have been flagged as suspicious.

Now, AI can recognize spending patterns in both locations over time, allowing transactions to go through without interruption. This improved experience is a direct result of advanced, deterministic AI and analytics that have been refined over time.

“What we’re able to do now is look at patterns on a very different kind of scale, right? And we’re also able to future-proof the way we look at things.

So, while in the past we were able to look at patterns based on past behavior and then apply them to current transactions. Now it’s much more real-time, and our pattern analysis is better. We can use AI to understand your pattern. So the experience has gotten better, and that’s all because of AI.”

– JoAnn Stonier, Fellow of Data and AI at Mastercard

While these foundational AI systems are the workhorses of fraud prevention, Stonier acknowledges that new capabilities, like agentic AI, will change how these systems are deployed.

She describes “agent-ish” AI as a precursor to fully autonomous systems, noting that conversational bots have evolved into sophisticated task-doers. However, true agentic AI, such as a self-driving Waymo vehicle, is highly complex and requires significant oversight.

As financial institutions begin to deploy agents, they will be able to analyze and strengthen different parts of their network, but Stonier stresses that humans must remain in the loop.

Using Data to Understand Patterns, Not People

The use of personal data in fraud detection can be a sensitive topic for customers. Stonier clarifies that payment networks do not collect an individual’s personal information for fraud analysis.

Instead, the company receives a minimal set of data points for each transaction: date, time, location, merchant name, and transaction amount. From this limited information, AI can deduce spending patterns, such as a customer’s usual grocery store or gas station, without knowing their name or exact address.

When a suspicious pattern is detected, a payment network will work with its bank partners, who then contact the cardholder. The succession of inquiry, which Stonier refers to as “data minimization,” ensures that no more information is collected or used than is necessary to achieve the specific purpose of fraud prevention.

“We don’t use that for other purposes. We’re using it for the purpose of preventing fraud,” Stonier says. Her distinction is crucial to maintaining consumer trust and complying with privacy regulations. The objective is not to know everything about a customer but to use data responsibly to secure the entire payment ecosystem—for the benefit of banks, merchants, and cardholders.

Adopting a Team-Sport Approach to Responsible AI Governance

For financial leaders looking to deploy AI responsibly, Stonier emphasizes that it must be treated as a team sport. It requires collaboration among product, innovation, and risk teams, with a shared understanding of both business goals and potential challenges.

“When it comes to AI, it’s really a team sport. The leaders in developing products and solutions really understand that there’s lots of different risks that need to be navigated [and] combine their understanding of all the risks with their imaginative thinking and what they’re trying to achieve. 

First of all, if you don’t have a clear purpose defined in what you’re trying to achieve, how are you evaluating model drift, bias? And how can you really make sure you’re deploying an iterative process to drive design thinking so that you really get innovation, but that you’re also constantly doing the learning loops, you really want everyone involved?” 

– JoAnn Stonier, Fellow of Data and AI at Mastercard

For leaders looking to deploy AI effectively in their organizations, JoAnn outlines a clear framework. She defines a successful AI governance process as one that starts with a clearly defined purpose and a deep understanding of the intended outcomes. Leaders must ask a series of questions:

• Do we have the correct data, and is it of the right quality?
• How are we building and evaluating our models to account for things like model drift and bias?
• How can we manage the various risks—from data security to intellectual property—that accompany AI innovation?

Stonier concludes that while AI presents a landscape of “unknown unknowns,” the best approach is to be open to these challenges and keep people at the center of the design process. Products and solutions should be built with the individual user in mind, ensuring outcomes that benefit all communities. Her human-centric, collaborative approach is the key to navigating the future of AI in financial services.