The AI Moment: What History, Data, and the Fed Tell Us About AI’s Impact on Productivity and Economic Policy

AI at a Crossroads Between Promise and Uncertainty

Artificial intelligence stands at a fascinating crossroads in 2026. Public sentiment oscillates between excitement about AI’s transformative potential and deep-seated fears about its disruptive impact on jobs, livelihoods, and human identity. This duality isn’t unprecedented — similar mixed feelings emerged during the adoption of electricity and automobiles, technologies some once viewed as “dangerous, supernatural, even evil.”

The Federal Reserve Bank of San Francisco’s latest economic analysis cuts through this emotional turbulence with a fundamental question: What do we actually know about AI’s impact on productivity and the economy, and what remains uncertain? As the Fed’s research demonstrates, “knowledge gives us power — to separate facts from fears, speculation from reality, and worst-case scenarios from more likely outcomes.”

This analysis is particularly crucial as the Federal Open Market Committee (FOMC) navigates monetary policy in an era of technological transformation. The stakes couldn’t be higher: getting policy right during potential paradigm shifts affects employment, price stability, and economic well-being for all Americans.

Transformations Take Time — The Electrification Analogy

History offers a powerful lens for understanding AI’s current trajectory. Consider electrification, perhaps the most analogous general-purpose technology (GPT) transformation. The journey from scientific discovery to economic transformation spanned nearly a century and unfolded in distinct phases:

Foundational Research (1820s-1840s): Michael Faraday and Joseph Henry established the principles of electric power generation, laying the theoretical groundwork for what would eventually revolutionize the global economy.

Key Inventions (1870s-1880s): Thomas Edison engineered commercial electric lighting in 1879, while Nikola Tesla developed the alternating current motor in the 1880s. These breakthroughs provided the technological building blocks for widespread adoption.

Infrastructure Development (1890s-1900s): The electrical grid emerged, transmission lines spread across landscapes, and access expanded. This critical phase created the foundation for mass adoption.

Unit-Drive Adoption (1900s-1910s): Individual electric motors began replacing steam-powered equipment in factories, marking the beginning of direct productivity applications.

Transformation and Productivity Gains (1920s-1940s): The era of strong productivity growth finally materialized, with U.S. labor productivity growth reaching 3-4% during this period. Crucially, this came decades after the key inventions.

The key insight is that sustainable productivity gains required more than technological adoption — they demanded fundamental business transformation. As economic historians note, “the very essence of work had to change.” Production processes were redesigned, factory floors reengineered, and workforces retrained. Firms had to shed the constraints of the previous steam-powered world and completely reimagine their operations.

Is AI Different? Parallel Timelines of AI and Electricity

AI’s development timeline reveals striking parallels to electrification. The foundational research phase began in the 1930s-1950s with early work on neural networks and thinking machines. Commercial applications and expanding adoption emerged in the 1990s as data-driven machine learning found industrial applications in process automation.

The ChatGPT release in 2022 represents a pivotal moment — approximately 70 years from AI’s beginnings, large language models with natural language processing became broadly accessible to businesses and consumers. This accessible, human-like interface catalyzed widespread adoption in ways that earlier AI applications, despite demonstrating value, had failed to achieve.

However, the critical question remains: Are we at the equivalent of the 1900s (early adoption phase) or the 1920s (transformative business redesign)? The Federal Reserve’s research suggests we may already be well into the transformation timeline, but the productivity transformations that would definitively answer this question remain marked with question marks.

The Post-ChatGPT Surge — From Enthusiasm to Real Investment

The 2022 ChatGPT release triggered more than simple market enthusiasm — it catalyzed real investments across businesses of all sizes. The San Francisco Fed’s EmergingTech Economic Research Network (EERN), launched in 2024, has documented extensive AI adoption across the Twelfth Federal Reserve District’s nine western states.

Real-world applications span diverse sectors:

• Agriculture: Using AI to research and develop new crop varieties that can adapt to climate change
• IT and Finance: Scaling tasks more effectively through intelligent automation and analysis
• Healthcare: Automating important but time-consuming routine tasks, freeing medical professionals for patient care
• Business Operations: Deployment across consumer research, back-office operations, sales, and product development

Research case studies have documented measurable cost savings from AI automation in call centers, software development, financial management, marketing, and healthcare. These micro-level success stories create a compelling narrative of AI’s utility and business value.

The Evidence Gap — Why Macro Data Doesn’t Show AI Productivity Yet

Despite these promising firm-level examples, most macro-studies of productivity growth find limited evidence of a significant AI effect. Even firms that report AI as useful find little evidence of transformative gains at the aggregate level. This disconnect between micro-level adoption and macro-level transformation presents two possible explanations:

Timing Explanation: AI adoption is still evolving, and it may be too soon to see results in aggregate productivity measures. Firms are currently focused on acquiring and learning new tools, including emerging agentic AI capabilities, rather than fundamentally rethinking their business processes.

Transformation Gap: Current generative AI and related applications may be useful but not yet the innovation that spurs broad-based organizational transformation. Many implementations simply automate existing processes rather than reimagine them.

The financial sector provides a telling example. Firms are using AI in loan application processes for document review and final application checking. While these implementations save time and money, they fall short of transforming the overall lending process. As the Fed’s analysis notes, “It is analogous to replacing a steam-powered motor with an electric one but leaving the factory floor unchanged — good progress but not transformative.”

From Possible to Transformative — What AI Needs to Become a True GPT

The electrification parallel reveals that technology alone is insufficient for transformation — ideas are what made electricity a general-purpose technology. Innovative firms didn’t just adopt electricity; they used “imagination and creativity to start fresh and build a world shaped by electricity, rather than leverage electricity in a steam-powered world.”

The same principle likely applies to AI: “Technology will enable, but ideas will determine when it transforms.” The critical question remains whether generative AI is a sufficient catalyst to fundamentally change the nature of production and business. As the Fed’s analysis candidly acknowledges, “The truth is, no one’s sure.”

For AI to become truly transformative, firms need to move beyond automation to fundamental business process redesign. This requires the “something you don’t expect” factor — breakthroughs in how work is organized, not just improvements in the tools used to perform existing tasks.

Lessons from the 1990s — Greenspan, Computing, and Monetary Policy

The mid-1990s computer and internet revolution offers valuable precedent for today’s AI moment. Businesses were rapidly increasing IT investment, yet official U.S. productivity growth measures showed little impact. Standard macroeconomic models suggested interest rates should rise to prevent economic overheating.

Federal Reserve Chairman Alan Greenspan took a contrarian approach, distrusting the official productivity numbers and positing that the computer revolution would spur sustained productivity growth, allowing faster economic expansion without inflationary pressure. Crucially, incoming data during 1995 and 1996 were not signaling productivity increases — subsequent data revisions revealed that productivity acceleration had already begun before the official statistics captured it.

Greenspan’s evidence came from disaggregated micro data and direct business intelligence:

• Wholesale and retail firms using inventory management systems to reduce warehouse stockpiling
• Trucking companies leveraging GPS technology to minimize deadhead hauling
• Manufacturing firms using computer-assisted designs for mass customization and waste reduction

The San Francisco Fed’s own research during the 1990s involved direct firm engagement — talking to businesses and walking factory floors. They discovered numerous examples of computers and software making operational differences, but more importantly, they found firms asking how to fundamentally alter how they produced goods and services. This focus on factory design and business processes with computers at the foundation mirrored electricity’s transformation pattern and “began to deliver sustained gains in output, revenue, and longer-run profitability.”

The FOMC’s patience during this period proved prescient — the “roaring ’90s” followed with a strong labor market and sustained economic growth.

Three Lessons for Monetary Policy in the Age of AI

The 1990s experience yields three actionable lessons for contemporary monetary policy:

1. Look Beyond Aggregate Data: “We won’t find all the answers in the aggregate data on productivity, the labor market, or inflation.” Seeing developments before they fully emerge in macroeconomic statistics requires digging deeper into disaggregated information that foreshadows transformation.

2. Follow the Right Data, Not More Data: “There is no matrix that tells us exactly what data to follow at any moment in time.” Greenspan’s innovation “wasn’t looking at more data; it was looking at the right data — finding inconsistencies in what he saw and working to resolve them.”

3. Business Intelligence Matters: “Talking to businesses matters. Businesses invest, experiment, and learn long before we see it in aggregate productivity data.” Incorporating qualitative and micro-level information is essential for appropriate policy decisions.

As the Fed’s analysis concludes, “It wasn’t just good luck that allowed Greenspan to navigate the 1990s, it was also good practice.”

The Current Policy Stance — Patience, Openness, and Forward-Looking Judgment

While no playbook exists for managing the economy during an AI transformation, the 1990s experience provides a solid foundation. Monetary policy is inherently “forward-looking business” that must be “grounded in what we see and open-minded to what’s possible.”

The Federal Reserve’s dual mandate — full employment and price stability — remains the guiding framework for policy decisions. The FOMC has successfully managed through technological transformations before and can draw upon institutional knowledge and experience.

Current institutional efforts include the EERN launched in 2024, ongoing business and community outreach across the Twelfth District, and active monitoring of AI deployment across industries. This multi-layered approach enables the Fed to maintain awareness of both micro-level developments and macro-level implications.

The analysis emphasizes that “the willingness to confront what we know and what we don’t is essential to making appropriate and durable policy that serves all Americans.”

What to Watch — Early Indicators and Signals of AI-Driven Transformation

Based on the Fed’s analytical framework, several key indicators merit close monitoring:

Business Process Redesign: Are firms merely automating existing tasks, or fundamentally rethinking how they produce goods and services? This represents the crucial distinction between steam-motor replacement and factory floor transformation.

Disaggregated Productivity Data: Micro-level and sector-specific productivity measures may signal transformation before aggregate data captures the change.

Business Investment Patterns: Monitor capital expenditure on AI infrastructure, tools, and complementary organizational changes.

Workforce Evolution: Look for evidence of retraining programs, new role creation, and organizational restructuring beyond simple job displacement.

The Ideas Factor: Evidence that firms are using creativity and imagination to build AI-native processes rather than retrofitting AI onto legacy systems.

Next-Generation AI Tools: How quickly firms move from learning agentic AI and advanced tools to deploying them transformatively.

The Bigger Picture — AI, Uncertainty, and the Future of the Economy

AI stands at a critical but uncertain juncture in 2026. Commercial applications continue expanding, but the productivity transformation has not yet materialized in macroeconomic data. Historical precedent suggests this is normal and expected during general-purpose technology transitions.

Multiple future scenarios remain possible:

• GPT Pattern: AI follows the general-purpose technology pattern and delivers broad-based productivity gains, though timing remains uncertain and could span years or decades
• Incremental Progress: AI remains useful but incremental, providing good progress without transformative economic impact
• Unexpected Breakthrough: Innovation in ideas and business models catalyzes transformation in ways not currently anticipated

Policy humility is appropriate given that “what we know about AI and its impact on productivity growth and the economy remains uncertain.” However, the stakes are consequential — getting monetary policy right during potential transformation affects employment, prices, and economic well-being for all Americans.

The Federal Reserve’s approach balances this uncertainty with actionable frameworks. By monitoring micro-level developments, maintaining business intelligence networks, and learning from historical precedent, the Fed positions itself to recognize transformation signals before they appear in aggregate statistics.

The San Francisco Fed’s title — “The AI Moment?” — deliberately includes a question mark. Whether 2026 represents AI’s truly transformative moment or merely a precursor remains an open and consequential question. What’s certain is that the Federal Reserve, armed with historical perspective and contemporary analysis, is prepared to navigate whatever unfolds.

As we stand at this inflection point, the lesson from electrification, computing, and now potentially AI is clear: transformations take time, require imagination alongside technology, and ultimately reshape not just how we work, but how we think about work itself. The AI moment may indeed be upon us — but its full realization likely lies in the creative applications we haven’t yet imagined.