Today, I\u2019m going to talk about artificial intelligence (AI). Now standing here in Silicon Valley, many of you might find AI exciting, full of promise and possibility. For others, it may feel more worrisome, a potential disrupter to lives, livelihoods, and what it means to be human (see, for example, Kennedy et al. 2025 and Lange and Alper 2025).<\/p>\n
This is not surprising. People also had mixed feelings about electricity and the automobile, with some fearing them as dangerous, supernatural, even evil, while others imagined their potential, quickly adopting the novelty of driving or using electricity in their homes (for example, see Sullivan 1995 and Norton 2008). Irrespective of where they stood, almost everyone wondered how these technologies would change the role of humans in the world.<\/p>\n
But knowledge gives us power\u2014to separate facts from fears, speculation from reality, and worst-case scenarios from more likely outcomes. And this is my topic for today. What we know about AI, what remains uncertain, and how monetary policy should respond in the near term and the longer run.<\/p>\n
Transformations take time<\/p>\n
Let me start with a simple maxim\u2014transformations take time. Think about electrification. Sitting here with lights, computers, and fully electric vehicles, it\u2019s easy to imagine it just happened with the flip of a switch. The truth is, it took nearly 100 years to move from Michael Faraday\u2019s and Joseph Henry\u2019s generation of the electric current in the 1830s to electricity boosting productivity growth and transforming the economy (see Al-Khalili 2015, Du Boff 1967, and Devine 1983).<\/p>\n
The evolution is instructive (Figure 1). The figure shows the timeline of electric innovation and productivity growth in the United States. Many things had to happen to turn Faraday\u2019s discoveries into practical applications. Innovators had to develop the light bulb, the electric motor, and the electric unit drive. Builders had to lay out the electrical grid, install transmission lines, and ensure sufficient access for households and businesses to make electricity a worthwhile investment (for example, see Devine 1983, Du Boff 1967, and David 1990).<\/p>\n
Figure 1
Key electricity innovations and U.S. labor productivity growth<\/p>\n
![\"\"](\"https:\/\/www.newsbeep.com\/us-ca\/wp-content\/uploads\/2026\/02\/el2026-06-figure1.png\")
Note: Productivity growth shown as the filtered year-on-year change in output per hours worked based on data collected by Bergeaud, Cette, and Lecat (2016) and made available through the Long-Term Productivity project. The growth trend is obtained using a Hodrick-Prescott filter with a coefficient of 1000, in line with the approach in Bergeaud (2024).<\/p>\n
And this was just the beginning. To create sustained gains in productivity growth associated with general-purpose technologies (GPT), the very essence of work had to change. Production had to be redesigned, factory floors reengineered, and workforces retrained (see Devine 1983). Firms had to shed the constraints of the previous steam-powered world and reimagine business in an era of electricity.<\/p>\n
The scale of change was enormous, and it took time.<\/p>\n
Is AI different?<\/p>\n
The question is, should we expect a similar long period of access, adoption, learning, and transformation for AI? Maybe, but we are already well into it. The origins of AI date back to the 1940s and \u201950s, and it has evolved much like electricity did (Figure 2).<\/p>\n
Figure 2
Parallel timelines of AI and electricity innovations<\/p>\n
![\"\"](\"https:\/\/www.newsbeep.com\/us-ca\/wp-content\/uploads\/2026\/02\/el2026-06-figure2.png\")
<\/p>\n
AI began with foundational research on neural networks and thinking machines. It was soon recognized as a field of study. (For a detailed history, see Nilsson 2009 and Stanford\u2019s One Hundred Year Study of Artificial Intelligence by Littman et al. 2021.) Then came the experiments, use cases, and eventually applications like applied machine learning and robotic processing. These demonstrated AI\u2019s value but fell short of creating sustained gains in productivity growth. (This is consistent with recent analysis by Fernald et al. 2024, Foerster 2024, and Kahn and Rich 2026.)<\/p>\n
Then in 2022, again nearly 70 years from AI\u2019s beginnings, ChatGPT and other large language models, with natural language processing, became available to anyone who wanted to use them. (For earlier examples of key innovations, see He, Cao, and Tan 2025.) Their accessible and humanlike interface encouraged people to try them and made businesses want to harness their capabilities (see, for example, Bick, Blandin, and Deming 2025 and McKinsey 2025). Their arrival signaled a major change in AI technology and access.<\/p>\n
The change was dramatic. We heard this in our business and community outreach. In addition to longstanding outreach to businesses and communities, the San Francisco Fed and the Federal Reserve System Innovation Office launched the EmergingTech Economic Research Network (EERN)<\/a> in 2024 to support a better understanding of how new technologies like generative AI (GenAI) are shaping the economies of today and the future. And we experienced it with our own employees. AI went from science and sci-fi to everyday work and conversation.<\/p>\n We now know that this surge moment was more than simple enthusiasm. Excitement has turned into real investments. Businesses of all sizes are working to leverage AI. Around the Twelfth Federal Reserve District\u2014the nine states in the West\u2014we\u2019ve heard about firms using AI for consumer research, back-office operations, sales, and product development, saving time and money. Some examples include using AI to research and develop new crop varieties in agriculture, help scale tasks more effectively in IT and finance, and automate important but time-consuming routine tasks in the health-care profession. (For more insights on how our business contacts are utilizing AI, see, for example, Federal Reserve Bank of San Francisco 2024, 2025a,b and 2026.) Research case studies find similar results\u2014cost-savings when firms use AI to automate\u2014in everything from call centers, software development, and financial management, to marketing and even health care (see Imas 2026 and Sukharevsky et al. 2025).<\/p>\n As Figure 2 shows, commercial applications and expanding adoption are critical steps for AI to produce more systematic increases in productivity growth.<\/p>\n The question is: Will it be enough? Is GenAI a sufficient catalyst to change the nature of production and business?<\/p>\n From possible to transformative<\/p>\n The truth is, no one\u2019s sure. It is easy to see the possibilities but harder to know when and how they will evolve.<\/p>\n So far, most macro-studies of productivity growth find limited evidence of a significant AI effect (see, for example, Imas 2026, Acemoglu 2025, and Aghion and Bunel 2024.) Even firms that say it\u2019s useful find little evidence of transformative gains (see, for example, Yotzov et al. 2026 and McKinsey 2025). Some of this could be timing. AI adoption and use are still evolving, so it may be too soon to see results in aggregate measures. AI technology is also changing rapidly. So, firms could be focused on acquiring and learning new tools, like agentic AI, rather than rethinking business.<\/p>\n But it could also be that we are simply not there yet. That while GenAI and related applications are useful, they are not the innovation that spurs broad-based reorganization of the economy.<\/p>\n Take the financial sector. Several firms have reported using AI in the loan application process (see Ng 2026 and Board of Governors 2025a,b,c.) Uses range from initial document review to checking the final application. While automating these steps saves time and money, it falls short of transforming the overall process. It is analogous to replacing a steam-powered motor with an electric one but leaving the factory floor unchanged\u2014good progress but not transformative.<\/p>\n So, what will that take? Likely, something you don\u2019t expect.<\/p>\n The factors that made electricity a general-purpose technology were not solely technologies, but ideas. Innovative firms used imagination and creativity to start fresh and build a world shaped by electricity, rather than leverage electricity in a steam-powered world.<\/p>\n The same path is likely to hold for AI. Technology will enable, but ideas will determine when it transforms.<\/p>\n Monetary policy in a time of transformation<\/p>\n And this brings me to monetary policy and how we should think about AI as we work to achieve full employment and price stability, our congressionally mandated goals. Fortunately, this is not the first time the Federal Open Market Committee (FOMC) has had to manage policy in a time of transformation.<\/p>\n In fact, I began my career at the San Francisco Fed during a similar period. It was the mid-1990s\u2014the beginning of the computer and internet revolution. Businesses were ramping up investment in information technology equipment and software to take advantage of these emerging technologies, but there was little impact on official measures of U.S. productivity growth. (As Anderson and Kliesen 2006 point out, incoming data during 1995 and 1996 were not signaling an increase in labor productivity growth. Subsequent data revisions showed that productivity began accelerating before 1995.) At the same time, FOMC policymakers were concerned about elevated inflation and a tightening labor market. Standard macro models and monetary policy thinking suggested interest rates should rise to ward off overheating in the labor market and a more significant run-up in inflation.<\/p>\n Then Fed Chair Alan Greenspan offered a different view. He distrusted the official productivity numbers, finding them at odds with the surge in information technology investment we were seeing. He posited that the burgeoning computer revolution would spur sustained productivity growth, allowing the economy to grow faster without putting upward pressure on prices (see Board of Governors 1996). In this scenario, the economy would not need to be bridled, and interest rates would not need to increase.<\/p>\n He found evidence for this possibility in the disaggregated micro data, including what he was hearing from executives. And there he saw considerable possibilities. Wholesale and retail firms were using inventory management systems to reduce warehouse stockpiling. Trucking companies were leveraging GPS to reduce deadhead hauling. Manufacturing firms were using computer-assisted designs to do mass customization and reduce waste.<\/p>\n We asked the same questions at the San Francisco Fed. We talked to firms and walked factory floors, looking for evidence that computers and software were making a difference. We found numerous examples. But we also discovered something bigger and more significant. Businesses were asking questions about how they could fundamentally alter the way they produced goods and services. Their focus was on factory design and business process, with computers and software at the foundation. This looked like the pattern with electricity, and it began to deliver sustained gains in output, revenue, and longer-run profitability.<\/p>\n Policymakers took this information and debated whether and when these micro examples could lead to the type of economy-wide productivity growth that Greenspan considered (see Board of Governors 1997 and Greenspan 1999). In the end, the FOMC remained patient on policy and the roaring \u201990s followed, with a strong labor market and sustained growth (see Fernald 2015). (As Frankel and Orszag 2001 and Mankiw 2001 highlight, there were other factors behind the strong economic performance in the 1990s, including deregulation, globalization, and stable food and energy prices in global markets. For more information on policymakers\u2019 actions in the 1990s and the conditions that supported the roaring \u201990s, see Daly 2025 and Blinder and Yellen 2001.)<\/p>\n So, what do these lessons teach us about monetary policy today? First, we won\u2019t find all the answers in the aggregate data on productivity, the labor market, or inflation. Seeing developments before they fully emerge requires digging deeper, relying on disaggregated information that foreshadows transformation. Second, there is no matrix that tells us exactly what data to follow at any moment in time. Greenspan\u2019s innovation wasn\u2019t looking at more data; it was looking at the right data\u2014finding inconsistencies in what he saw and working to resolve them. Third, talking to businesses matters. Businesses invest, experiment, and learn long before we see it in aggregate productivity data. Incorporating this information and seeing how it supports or negates what we expect is essential to making appropriate policy decisions.<\/p>\n In the end, it wasn\u2019t just good luck that allowed Greenspan to navigate the 1990s, it was also good practice. The willingness to listen to businesses, hear their ideas, use data and evidence to test them, and invite others to do the same.<\/p>\n Turning back to today, none of this is a playbook for managing the economy and AI, but it is a foundation. It is a reminder that monetary policy is a forward-looking business. To get it right, we must be grounded in what we see and open-minded to what\u2019s possible. The willingness to confront what we know and what we don\u2019t is essential to making appropriate and durable policy that serves all Americans.<\/p>\n References<\/p>\n Acemoglu, Daron. 2025. \u201cThe Simple Macroeconomics of AI.\u201d Economic Policy 40(121, January), pp.13\u201358.<\/p>\n Aghion, Philippe, and Simon Bunel. 2024. \u201cAI and Growth: Where Do We Stand?<\/a>\u201d Manuscript, EmergingTech Economic Research Network, Federal Reserve Bank of San Francisco.<\/p>\n Al-Khalili, Jim. 2015. \u201cThe Birth of the Electric Machines: A Commentary on Faraday (1832) \u2018Experimental Researches in Electricity\u2019<\/a>.\u201d Philosophical Transactions, Series A: Mathematical, Physical, and Engineering Sciences 373(2039), 20140208.<\/p>\n Anderson, Richard G., and Kevin L. Kliesen. 2006. \u201cThe 1990s Acceleration in Labor Productivity: Causes and Measurement<\/a>.\u201d Federal Reserve Bank of St. Louis Review 88(3, May), pages 181\u2013202.<\/p>\n Bergeaud, Antonin. 2024. \u201cThe Past, Present, and Future of European Productivity<\/a>.\u201d Proceedings from \u201cMonetary Policy in an Era of Transformation,\u201d ECB Forum on Central Banking, Sintra, Portugal, July 1\u20133.<\/p>\n Bergeaud, Antonin, Gilbert Cette, and R\u00e9my Lecat. 2016. \u201cProductivity Trends in Advanced Countries between 1890 and 2012<\/a>.\u201d Review of Income and Wealth 62(3, September), pp. 420\u2013444.<\/p>\n Bick, Alexander, Adam Blandin, and David Deming. 2025. \u201cThe State of Generative AI Adoption in 2025<\/a>.\u201d St. Louis Fed On the Economy, November 13.<\/p>\n Blinder, Alan S., and Janet L. Yellen. 2001. The Fabulous Decade: Macroeconomic Lessons from the 1990s. Century Foundation Press.<\/p>\n Board of Governors of the Federal Reserve System. 1996. \u201cMeeting of the Federal Open Market Committee, Sept. 24, 1996: Transcript<\/a>.\u201d<\/p>\n Board of Governors of the Federal Reserve System. 1997. \u201cMeeting of the Federal Open Market Committee, May 20, 1997: Transcript<\/a>.\u201d<\/p>\n Board of Governors of the Federal Reserve System. 2025a. \u201cTranscript: AI in Payments<\/a>.\u201d Washington, DC, October 21.<\/p>\n Board of Governors of the Federal Reserve System. 2025b. \u201cTranscript: Bridging traditional finance with the digital asset ecosystem<\/a>.\u201d Washington, DC, October 21.<\/p>\n Board of Governors of the Federal Reserve System. 2025c. \u201cTranscript: Tokenized Products<\/a>.\u201d Washington, DC, October 21.<\/p>\n Daly, Mary C. 2025. \u201cPolicymaking Amid Change<\/a>.\u201d SF Fed Blog, November 10.<\/p>\n David, Paul A. 1990. \u201cThe Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox.\u201d The American Economic Review 80(2), pp. 355\u2013361.<\/p>\n Devine, Warren D., Jr. 1983. \u201cFrom Shafts to Wires: Historical Perspective on Electrification.\u201d The Journal of Economic History 43(2), pp. 347\u2013372.<\/p>\n Du Boff, Richard B. 1967. \u201cThe Introduction of Electric Power in American Manufacturing.\u201d The Economic History Review 20(3), pp. 509\u2013518.<\/p>\n Federal Reserve Bank of San Francisco. 2024. \u201cHow Twelfth District Businesses are Rapidly Embracing GenAI<\/a>.\u201d SF Fed Blog, May 3.<\/p>\n Federal Reserve Bank of San Francisco. 2025a. \u201cRoundtable on AI and Medical Service Delivery<\/a>.\u201d SF Fed Blog, July 15.<\/p>\n Federal Reserve Bank of San Francisco. 2025b. \u201cRoundtable on AI and Product Development<\/a>.\u201d SF Fed Blog, May 14.<\/p>\n Federal Reserve Bank of San Francisco. 2026. \u201cThe AI Investing Landscape: Insights from Venture Capital<\/a>.\u201d SF Fed Blog, February 11.<\/p>\n Fernald, John G. 2015. \u201cProductivity and Potential Output Before, During, and After the Great Recession<\/a>.\u201d NBER Macroeconomics Annual 29, pp. 1\u201351.<\/p>\n Fernald, John, Huiyu Li, Brigid Meisenbacher, and Aren S. Yalcin. 2024. \u201cProductivity During and Since the Pandemic<\/a>.\u201d FRBSF Economic Letter 2024-31 (November 25).<\/p>\n Foerster, Andrew. 2024. \u201cThere\u2019s a lot of hope that we\u2019ve entered a period of higher productivity growth, maybe spurred by AI. My work suggests that remains unlikely<\/a>.\u201d LinkedIn post (May 2024, accessed February 12, 2026).<\/p>\n Frankel, Jeffrey, and Peter R. Orszag. 2001. \u201cRetrospective on American Economic Policy in the 1990s<\/a>.\u201d Brookings Research.<\/p>\n Greenspan, Alan. 1999. \u201cHigh-Tech Industry in the U.S. Economy<\/a>.\u201d Testimony before the Joint Economic Committee, U.S. Congress, June 14.<\/p>\n He, Ran, Jie Cao, and Tieniu Tan. 2025. \u201cGenerative Artificial Intelligence: A Historical Perspective<\/a>.\u201d National Science Review 12(5).<\/p>\n Imas, Alex. 2026. \u201cWhat Is the Impact of AI on Productivity? Reconciling the Micro and the Macro Evidence<\/a>.\u201d Ghosts of Electricity Substack, January 29 (accessed February 12, 2026).<\/p>\n Kahn, James, and Robert W. Rich. 2026. \u201cTrend Productivity Growth Update.<\/a>\u201d November 2025 update to the Kahn-Rich trend productivity model, posted January 26 (accessed February 12, 2026).<\/p>\n Kennedy, Brian, Eileen Yam, Emma Kikuchi, Isabelle Pula, and Javier Fuentes. 2025. \u201cHow Americans View AI and Its Impact on People and Society<\/a>.\u201d Pew Research Center Report, September 17.<\/p>\n Lange, Jason, and Alexandra Alper. 2025. \u201cAmericans Fear AI Permanently Displacing Workers, Reuters\/Ipsos Poll Finds<\/a>.\u201d Reuters, August 20.<\/p>\n Littman, Michael L., Ifeoma Ajunwa, Guy Berger, Craig Boutilier, Morgan Currie, Finale Doshi-Velez, Gillian Hadfield, Michael C. Horowitz, Charles Isbell, Hiroaki Kitano, Karen Levy, Terah Lyons, Melanie Mitchell, Julie Shah, Steven Sloman, Shannon Vallor, and Toby Walsh. 2021. Gathering Strength, Gathering Storms: The One Hundred Year Study on Artificial Intelligence (AI100) 2021 Study Panel Report<\/a>. Stanford University.<\/p>\n Mankiw, N. Gregory. 2001. \u201cU.S. Monetary Policy During the 1990s<\/a>.\u201d Paper presented at Kennedy School of Government conference on \u201cEconomic Policy During the 1990s.\u201d<\/p>\n McKinsey. 2025. \u201cThe State of AI in 2025: Agents, Innovation, and Transformation<\/a>.\u201d Survey, November 5.<\/p>\n