Mastercard is a global payments technology company that operates one of the world's largest payment networks, connecting consumers, financial institutions, merchants, governments, and businesses to enable secure transactions.
Mastercard has been using AI for many years. What experiences have most shaped the way you and the company think about AI?
From a company perspective, AI and data have been foundational to Mastercard for decades. We operate in over 200 countries and territories and processed over 175 billion transactions last year. We are more than payments; about 40% of our net revenue comes from services such as fraud prevention, business intelligence, marketing solutions, and open finance, all of which rely on large volumes of aggregated and anonymized data.
There is simply no way to operate at that scale and speed without AI. It first became central to fraud detection, helping us determine in milliseconds whether a transaction is likely legitimate or fraudulent. Personally, I joined Mastercard through the acquisition of a big-data SaaS company, which gave me an early perspective on how powerful data-driven systems can be and reinforcing why principles, governance, and humility are so important when deploying AI.
Given Mastercard’s scale and the sensitivity of the data you handle, how do you decide when an AI innovation is ready for deployment?
We take a very thoughtful and purposeful approach. One example is fraud scoring. Every transaction receives a score that helps issuers assess the likelihood of fraud. We have used generative AI to bring in additional signals and improve those assessments, but only after extensive testing to ensure we identify more fraud without creating friction or false positives.
We build solutions internally, test them on historical data, roll them out in a single market, gather feedback, and then expand. We followed that approach with a fraud solution that began in Canada before becoming a global offering. We also apply the same discipline to new technologies such as our large tabular model, starting small, validating results, and only deploying once we are confident in the model’s effectiveness and safety.
How do you balance innovation with maintaining trust?
Trust is one of Mastercard’s most valuable assets. It has been built over decades with consumers, merchants, and financial institutions around the world. That trust is supported by deeply embedded principles around data responsibility, technology governance, and privacy.
Trust is also the currency of innovation. There is enormous pressure to move quickly because AI capabilities are advancing rapidly. However, we cannot sacrifice trust in pursuit of speed. That is why we focus heavily on testing, governance, and responsible deployment, ensuring that any solution launched at scale is robust, secure, and aligned with our standards.
Can you tell us about Agent Pay and how you see AI agents making payments and purchases on behalf of consumers and businesses?
Agentic commerce is a natural extension of Mastercard’s role as a network that enables safe and secure commerce. Our objective is to ensure that transactions conducted by AI agents have the same security, transparency, and reliability as traditional e-commerce transactions.
Agent Pay combines capabilities such as tokenization, agent identification, intent capture, and authentication. We have also introduced technologies like Verifiable Intent, which creates a tamper-resistant record of what a user has authorized an agent to do. This helps provide transparency and accountability across the ecosystem, ensuring that agent-driven commerce can scale safely and securely.
Building on this, we recently introduced Agent Pay for Machines, extending these capabilities to support autonomous, machine-to-machine transactions that are continuous, programmatic and executed at machine speed across our network.
What is your take on the so-called Mythos experiment?
My view is that these models are demonstrating capabilities to identify vulnerabilities that were previously difficult to detect. The additional visibility improves speed and efficiency, allowing the ability to prioritize and address potential issues faster in areas that matter most. Whether it is Anthropic’s work, OpenAI’s cyber-focused models, or other developments, we are still at the beginning of a very steep trajectory in AI capability.
I do not believe progress is about to slow down. We will continue to see rapid advances, including from open-source models. That makes it critical that organizations invest appropriately in cybersecurity and work together through public-private partnerships to ensure ecosystems remain secure as these capabilities evolve.
How do you approach educating consumers about AI-driven changes and personalization?
It begins with trust, transparency, and consent. We have our published Data and Tech Principles that make it clear that our use of technology such as AI should be transparent and allow for individual control. We also want consumers to understand how their data is being used and to have control over it, which we do with My Data Portal.
Our principles are centered on transparency, consent, and consumer value. Because Mastercard operates in a B2B2C model, communication often happens through our partners and customers. However, regardless of who delivers the service, we remain focused on ensuring data is handled responsibly and that consumers are treated in a way that maintains trust in the ecosystem.
What are your top priorities for the next 12 months, and what excites you most about your work?
Our focus is execution. We have a clear strategy centered around building our data and AI foundations, driving productivity, scaling AI solutions to drive value for our customers, and enabling the broader ecosystem. The challenge now is continuing to execute while staying humble and adaptive as the technology evolves.
One major priority is expanding our large tabular model, which has already been trained on 15 billion transactions and is now being tested across use cases such as fraud detection, cybersecurity, personalization, and optimization. We are also continuing to enhance our data commercialization platform, improve productivity across workflows, advance agentic commerce, and prepare for future machine-to-machine payment ecosystems.
