How Mastercard’s AI chief connects the dots across the company’s vast data operation

Every time a Mastercard is swiped at checkout, the credit card company taps AI to score the likelihood of fraud based on patterns in training data. It can sweep preview snippets of card numbers for sale on the dark web and attempt to ID them with predictive algorithms. Generative AI also cuts down on the number of false-alarm fraud alerts you get, according to Mastercard.

These are a few of the operations Greg Ulrich oversees as Mastercard’s chief AI and data officer. The credit card giant created the new perch in May as part of a larger reorg that merged data and AI operations and aimed to recognize the importance of both.

It’s one of a growing number of companies that are giving AI a seat in the C-suite as businesses fine-tune the way they operate with generative AI. (You can read Tech Brew’s profiles of more of these leaders here.)

Talk to enough of these companies, and some common threads tend to emerge—generative AI is perhaps most intuitively useful off the bat for coding, marketing, and maybe some customer service experiments. Mastercard is doing all of that, to be sure, but there are also some aspects of its AI operations that make Ulrich’s job unique.

Data deluge

For one, there’s the sheer scale of the data: With billions of transactions per year, it doesn’t make sense to centralize data operations into one department, Ulrich said. Instead, he works across the company to coordinate among data scientists in different areas.

“We have a lot of people in the company with incredible skills in data science and data engineering that are already in place and are there to experiment and help build and create new opportunities,” he said. “It is my responsibility to create a community around that, to create consistency and best practices around that, to create coordination around that, to make sure that…we’re learning from each other and we’re all moving in the same direction.”

Mastercard also has more long-standing experience applying machine learning to tasks like fraud detection, and has had an AI governance program in place for half a decade, Ulrich said. And like banks that have jumped on the AI bandwagon, it contends with the restrictions of a highly regulated industry.

“There are cases where we will sort of overbuild a solution for a problem,” Ulrich said. “We’re interested in…making sure that we train things appropriately and we have the right use of the privacy and the copyright information. Now I feel very good about how we’re progressing on all those things, but it’s certainly a place where you have to be very thoughtful and careful about how you proceed.”

A day in the life

In addition to managing some of the tasks of his previous job in corporate development and M&A, Ulrich’s typical day is split between a few different general areas. One is looking at potential investments and improvements to the company’s data infrastructure, data warehouse, and data science tools.

Then there’s work to be done on understanding and prioritizing the “long list of potential applications” of AI, he said. He might spend some time evaluating third-party partners or platforms for particular use cases.

He’s also still getting the team up and running in some ways, given that his organization within Mastercard is only a few months old. That means meeting with a team of leaders across the company every other week to hear about what they are working on and look for common trends.

“So we can look at things holistically, and then I can look and say, ‘Hey, 62 of these use cases have a pretty common underlying piece. We really need to get that underlying piece accelerated because it’s a limiting factor for so much of the good that we’re going to be able to generate within these businesses to help our customers,’” Ulrich said.

Coordinators needed

As for whether other organizations should think about a position like his, Ulrich said it’s helpful to have someone who can corral the disparate but interconnected aspects of how companies are using AI together.

“Every organization has a thousand ideas about how this could be deployed. And without any sort of coordinating entity, it’s easy to have a thousand flowers bloom, which can lead to distractions, can lead to inefficiency from a cost perspective, and could lead to unintended consequences,” he said.

And while there’s been plenty of chatter about an AI bubble, Ulrich said he’s unconcerned with any ups and downs of the hype cycle as long as the AI applications are able to deliver results.

“What we care about is the fundamental efficacy of the tool, and whether or not these techniques have meaningfully different outcomes for the problems we’re trying to solve,” he said. “On that, I’m very, very bullish.”

Correction 08/15/24: This piece has been updated to reflect that Mastercard processes billions of transactions per year, not trillions.