Data in the age of AI: A conversation with Mark Birkhead of JPMorganChase

JPMorganChase is a global financial institution with $4 trillion in assets. With over 300,000 employees, the firm has long focused on collecting and leveraging data to better serve its millions of customers and clients. But in 2023, JPMorganChase embarked on a new data transformation journey. McKinsey Senior Partner Kevin Buehler recently sat down with Mark Birkhead, firmwide chief data officer, to hear more about how the firm is leveraging its vast data set in innovative ways, including using AI, while still meeting the strictest data governance and privacy standards.

This transcript has been edited for length and clarity.

Kevin Buehler: What does the data estate look like at JPMorganChase?

Mark Birkhead: If you think about what this firm does and our footprint, we operate in almost 100 countries. In the United States, we are known not only for our consumer businesses, like Chase, but also for our investment banking and asset and wealth management businesses. Globally, we are known for many things, including payments. When you put all that together—and you think about the volume of transactions that we move in any given day—we have over an exabyte of data across the firm. That data comes in many forms and modalities, including structured and unstructured data and voice and video files. We have a big data estate at the firm, and it’s something we think is an incredible strategic asset.

Kevin Buehler: Can you talk more about how AI is influencing your data strategy?

Mark Birkhead: Our data strategy is part of our mission to provide great service to our clients and customers. Jamie Dimon, our chairman and CEO, is incredibly passionate about data, analytics, and AI. The mission to use it comes from the top, and we are always looking for ways to use data to solve strategic problems. All of this comes together, and we have a really cohesive approach to data and analytics across the firm. Overall, our data strategy centers on how we can best deliver all types of data assets and curate them in a way that is discoverable, highly accurate, and highly governed and controlled. That last part is really critical for us as a bank because our customers count on us to keep their information private.

Kevin Buehler: How do you manage data centrally while also distributing it in a federated fashion?

Mark Birkhead: In late 2024, Jamie Dimon and the operating committee recognized the need to create an organization whose only job was to focus on data. That’s when we created the firmwide chief data and analytics office, pulling all our data initiatives together under one umbrella organization. At the top of the house, we have Teresa Heitsenrether, our chief data and analytics officer. She reports to Jamie and sits on the operating committee, which shows how critical this function is. In my role as chief data officer, I oversee a team that sets the central data strategy for the firm. And then we partner with business lines and corporate functions to implement that strategy. Because these teams are closest to the data, we believe they should have responsibility for it and ownership of it. We provide them with data strategies around privacy and governance, and then with the capabilities and tooling to extract value from our data.

Kevin Buehler: What are some key components of JPMorganChase’s data strategy?

Mark Birkhead: Our top priority is to keep our customers’ and clients’ data safe above all else. Beyond that, there are several additional shifts we are working on right now. The first is modernizing our data so that it’s AI-ready. To that end, we have stood up data strategy teams to examine the implications of AI and gen AI. We are focused on pulling humans out of the cycle when it comes to fixing data, and that is going to be a multiyear effort. To do that, we are working with our technology partners, data users, and data teams to publish data in a way that’s very consistent and understandable by LLMs [large language models]. The second pivot is that we are leaning into data products. We have launched several data products that have been very successful. Now we are working to ensure those products are interchangeable, interoperable, and reusable. And, finally, we are focused on ensuring our data is available in milliseconds.

Kevin Buehler: How do you make sure your data is used in risk-appropriate ways?

Mark Birkhead: We believe managing data risk is critically important. The key thing about managing data risk is that you have to understand what those risks are. So at the firm, we are fixated on a number of specific data risks. This includes monitoring quality, storage, protection, and destruction. But in addition to that, we are firmly invested in ensuring that people use data for the right purposes, and that those purposes are good and aligned with all our expectations when it comes to privacy and the commitments we’ve made to our customers and clients. We ensure that any use cases that go into production at the firm leverage the right data in the right format in the right locations. We are not going to put forward a capability that doesn’t have the right accuracy, the right explainability, or the right permissions.

Kevin Buehler: What kind of investments has JPMorganChase made recently in the data space?

Mark Birkhead: We have spent a lot of time investing in platforms for data governance to ensure we can manage data risks in ways that respect privacy. We built a lot of these platforms over time, and now we are investing in pulling them all together into a central platform. We are also ensuring that all our teams have access to capabilities beyond just data management. To that end, we have invested in something called LLM Suite, which is a vehicle that delivers LLM models into the hands of our employees. Right now, we have 200,000 users on LLM Suite. It allows us to put the right guardrails in place to ensure our employees have access to LLMs in a safe way, so they can be more efficient in their jobs. And then beyond that, we are investing in building other data platforms that require more engineering and data work, including ways to deploy AI agents.

Kevin Buehler: Looking to the future, what excites you most about your data portfolio and potential use cases for it?

Mark Birkhead: In the data space, we have long been waiting for the kind of compute that exists today. We have been waiting for this kind of capability. And we have been waiting for this type of investment. And it’s all coming together right now, so it’s a pretty magical time. One technology we are particularly excited about is AI agents. AI systems with more reasoning capabilities will create endless possibilities. And we are not going to be able to deploy AI agents or other emerging technologies without continual investment in data. So being able to put data and AI together is going to be really meaningful and impactful, not only for our company, but also for our customers and clients.

Kevin Buehler: What advice would you give someone stepping into the role of chief data officer for the first time?

Mark Birkhead: We have likely all seen the statistics that the average tenure for a CDO or CDAO [chief data officer/chief data and analytics officer] is not much more than two years. And that has been a consistently stubborn number. So while it has been more than a decade since the role of CDO gained prominence, there is still high turnover. I have had both stumbles and successes in my career, but I have been able to beat that number in several roles, and I think that’s for one main reason: I focus on setting a really solid strategy. If you cannot communicate what has to be done to your team and senior stakeholders—and if you cannot explain to them what investment is needed to get that done—you won’t get very far. Think through all the challenges for the organization. Think through what the business wants to do and where it wants to go. Think about your clients’ needs and about the data you are responsible for. Then, put together a data and analytics strategy based on that thorough thought process.