Ensuring AI-readiness of Data Is a Long-term Commitment — JPMorganChase Firmwide CDO

(US & Canada) Mark Birkhead, Firmwide CDO at JPMorganChase, speaks with Sajid Khan, Partner/Principal, AI and Data Financial Services at EY, in a video interview about how JPMC uses AI and ML and its future uses, and how the company ensures AI-readiness of its data.

JPMorgan Chase & Co. is an American multinational finance company. It is the largest bank in the U.S. and the world's largest bank by market capitalization as of 2023.

JPMorganChase has been utilizing AI and ML for over a decade, says Birkhead, and one of its earliest use cases involves fraud detection. To date, this serves as a major source of investment when it comes to AI and ML, he reveals.

Adding on, Birkhead states that the current use cases in production drive a wide variety of business outcomes across all lines of business. He continues that approximately 400 of these AI/ML use cases cover everything from marketing and personalization to trade optimization and efficiency.

JPMC has held the top spot in the Evident AI Rankings for the past two years, which underscores its commitment to AI and ML as business priorities, says Birkhead. Regardless of the increasing attention towards GenAI and LLMs, he notes that traditional analytics, machine learning, and LLM play a critical role in the company’s operations.

However, Birkhead maintains that GenAI will be unlocking numerous new use cases for organizations in the next five years. Elaborating further, he envisions GenAI use cases from three horizons.

1. The first horizon portrays the present, where most people think of GenAI in terms of generic AI language tools, answering responses, and summarizations.

2. The next horizon will happen in a year or two when models will be customized with proprietary data. 

3. Speaking of the third phase, Birkhead affirms that it will arrive sooner than expected. In that usability phase, agents will become more capable of writing code, analyzing datasets, and eventually powering LLMs to interact with databases.

According to Birkhead, GenAI will have a significant material impact in the future. He further affirms that at JPMC, approximately two dozen GenAI use cases are in production today.

The company leverages GenAI to efficiently answer client questions, create marketing content and personalization, plan travel itineraries in the Chase travel businesses, and summarize meetings for client advisors.

When asked how JPMC ensures AI readiness of its data, Birkhead states that, firstly, safeguarding the data is paramount for the company. For instance, he says, the company would not allow using its data and its client or customers’ data to train external models.

With extensive parameters and guardrails in place, the data used is governed, controlled, and of high quality, says Birkhead. Also, the company guarantees that if the data does leave the organization, it is an absolute necessity.

In continuation, Birkhead states that safeguarding and governance give the license to do more with data. He adds that while data has always been critical, it feels like oxygen for many people, and in his perspective, the importance of data has now increased multifold.

Data becomes an intellectual property when one enters the world of GenAI, and it is the way with which one can customize algorithms to reflect the brand voice and deliver great client services.

Keeping the scenario in mind, Birkhead states that modernizing data and ensuring its AI-readiness is a long-term commitment. While organizations can make incremental progress year after year, building an analytic factory to produce AI models that support the business takes strategy, investment, and an enabling leadership team.

Highlighting JPMC’s data strategy, Birkhead states that the components include data design principles, operating models, principles around platforms, tooling, and capabilities. Additionally, talent, governance, data, and AI ethics also come into play, but the ultimate goal is to have incredibly high-quality data that is self-describing and understandable by both humans and machines.

From Birkhead’s standpoint, to be AI-ready with data, organizations have to get data to a state where a data scientist, user, or AI researcher can go into a marketplace and understand everything about the data.

Similarly, organizations have always converted data, algorithms, and models into machine language, with numbers represented in zeros and ones. However, now that data needs to be translated into formats that LLMs and other models can understand.

Ultimately, it boils down to ensuring solid alignment on data principles, ontologies, classification vocabularies, robust metadata, publishing standards, and more, to make data AI-ready.

AI-ready data also hinges on discoverability, and JPMC ensures that data scientists or analysts do not spend the majority of their time finding things but rather create value for customers and clients.

Delving deeper, Birkhead also shares that the company is focused on building unstructured and structured data products that can be merged and are designed to be interoperable. This enables the organization to move the needle forward regarding discoverability.

In conclusion, Birkhead says that, along with other key considerations, computing resources and platforms are also crucial for data readiness. JPMC makes certain that the data scientists have a seamless and efficient experience when working within its platforms.

CDO Magazine appreciates Mark Birkhead for sharing his insights with our global community.