Thank you, Keith, and while it is always an honor to represent the Federal Reserve and speak at public events, being here today is truly a special privilege for me. In the 17 years that I have spent in different roles for the Fed, one of the things that I am most proud of is the modest contribution I made, while here in St. Louis, to the growth and success of FRED (Federal Reserve Economic Data).
Among all the Federal Reserve’s accomplishments over the years, I consider FRED to be the greatest public good that we have ever created. As a trusted and reliable resource available to all, FRED is the goodwill ambassador for the Federal Reserve around the world. A crucial reason for that reputation and success has been the willingness of the leaders of the Federal Reserve Bank of St. Louis to keep their eyes on the horizon of what FRED could be, as a tool for users around the world to improve understanding of the economy through data access and charting, as technology marches forward.
We open a new chapter in that endeavor at this conference, which is a chance for all of us to explore trust in data, data accessibility, and storytelling with data. The backdrop for our discussion is the technological earthquake of artificial intelligence (AI), which is bringing new opportunities and challenges for FRED and others in this audience.
Technological leaps, in fact, have always driven the evolution of FRED, so before I get into AI, for the larger audience outside this room, let me start with a little of that history. In 1961, St. Louis Fed Research Director Homer Jones started to mail out a typed weekly digest of economic data from a variety of government agencies to Fed policymakers and staff, academics, journalists, and other members of the public. A telephone answering machine later fielded calls from those who couldn’t wait for the mail, and then in 1991, data were updated continuously on a computer bulletin board, accessed with dial-up modems. FRED became a website in 1995, initially with 865 data series and about 6,000 users per week, and graphics capabilities have steadily improved over time. Currently, FRED hosts more than 850,000 data series from governments and other sources around the world and it reached 18 million unique users in 2025.
Through all that time, the core value of FRED has remained the same: It is a trusted conduit for data from diverse sources because it offers a common platform to access all data in a consistent and reliable way, without a commercial or other parochial interest, following best practices for transparency and reliability. The ultimate credibility of the data lies with those sources that produce the data. But FRED’s distribution of the data is a guarantee that it has been accurately reproduced—that what you see is what it claims to be. In recognition of this preeminence, and FRED’s success in presenting and combining data in engaging ways, I can announce today that users seeking certain data published by the Federal Reserve Board will soon be routed to FRED and its tools.
Like past technological advances that have improved FRED, and perhaps more than most of them, AI is empowering users to do more with data. AI can find, summarize, and visualize data more quickly than one person—or even many people—can. AI can work with much larger quantities of information. And AI significantly broadens the audience for data through its searches and by explaining complex concepts to lay users. We are seeing these improvements already, and the possibilities for much more are enormous.
But that is not all we are seeing. Unlike past technological advances that primarily aided in the distribution of data, AI can create content as well, and this capability creates new risks. As I think is well known, AI can hallucinate, introduce bias, and hide key assumptions, often with confidence and polish. This risk has implications for FRED and its role as the reliable source of data produced by other organizations. AI may not accurately cite the sources for data. AI may also make a variety of errors in interpreting data, such as that old bête noire for economists and statisticians—confusing correlation with causation.
Another challenge that we are dealing with is an onslaught of traffic growth. FRED traffic has always tended to grow each year, but the pace has picked up since the advent of AI. It is presently growing 150 percent a year, with most of that growth from AI and bots. Already about half of visits to FRED are from AI agents retrieving data, with the other half from flesh-and-blood visitors. One challenge for us in understanding these AI visits is that FRED’s managers don’t know whether AI agents are correctly attributing data to the original source, or to FRED, as we have sometimes seen in testing. Despite occasional problems of misattribution, I can report that AI agents have generally been very accurate in describing visualizations and explaining complex terms and concepts in simple ways.
In response to these challenges and opportunities, FRED has introduced the FRED MCP Connector to simplify connecting FRED data with AI systems. Using the Model Context Protocol (MCP), the FRED MCP Connector enables users to find and retrieve data from FRED using an AI agent of their choice—which the FRED staff calls “BYOAI.” Why is this a big deal? Up till now, FRED has been designed to be viewed and understood by humans. But an AI bot sees and interprets data very differently. And that means the FRED team must adapt how FRED is presented to this new type of viewer so that the data are conveyed back to the agent’s master accurately and in a well-documented manner. There will be a demonstration of the FRED MCP Connector later today, and I encourage you all to attend.
In addition to creating that new architecture, the FRED staff is developing skills to help AI agents that visit FRED to understand and work with economic data more effectively and accurately. And we are working with the St. Louis Fed’s other data and archive repositories to more readily link to the data in FRED. For example, when researching data related to monetary policy or banking, we hope to link to relevant historical documents in the FRASER (Federal Reserve Archival System for Economic Research) library of Federal Reserve material. Likewise, while FRED holds current data, including revisions, ALFRED (ArchivaL Federal Reserve Economic Data) is a library of data as it was first reported and in other earlier forms, and we are working to establish links there also, for AI inquiries to include how data have evolved.
Panels today will get into all of this and, I suspect, unearth questions and issues and use cases that we haven’t thought of, which is why gatherings like this one are so valuable. It is an opportunity to explore the “now and next” for data as a community of practitioners who have faced—and have successfully addressed—challenges related to access, trust, and interpretation of economic data.
Our market economy depends on information, and in ways that most people take for granted, it depends on the quiet and dependable presence of a trusted source of economic data. With FRED, ALFRED, and FRASER, the St. Louis Fed and the Federal Reserve System perform a vital public service, and this conference today represents our commitment to continue that service, as AI presents new opportunities and challenges. On behalf of the FRED team, we are looking forward to learning from all of you in building the next generation of data access and distribution tools.
Thank you.