How Could AI Impact Economic Productivity?

Bill Gates once remarked that “we always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.”

That insight feels particularly relevant today as artificial intelligence dominates headlines and investor presentations. AI promises to revolutionize industries, fundamentally alter work, and unlock trillions in economic value. Goldman Sachs predicted that AI could boost global GDP by 7% over the next decade. McKinsey suggests productivity gains could reach $4.4 trillion annually. The forecasts are staggering.

But economic productivity doesn’t happen overnight. History shows that even the most transformative technologies (electrification, computers, the internet) took years, sometimes decades, to translate into measurable productivity gains. Early adoption is messy. Implementation is expensive. Organizational change is slow.

While rapid development and extensive funding all but guarantee AI will boost productivity eventually, the more elusive questions are when, how much, and where. More importantly, how should portfolios be positioned once productivity gains materialize?

Let’s walk through what economic productivity actually means, what the data shows so far, and what history teaches us about technology-driven growth cycles.

What Is Economic Productivity and Why It’s Important

Simply put, economic productivity is how efficiently we convert resources and labor into goods and services. It’s often expressed as a ratio of outputs to inputs. When productivity rises, the economy can produce more with the same amount of effort — or produce the same amount with less.

That might sound abstract, but productivity growth is the foundation of rising living standards. It’s what allows wages to increase over time and companies to expand profit margins. An economy that becomes more productive can support higher incomes and better quality of life without simply working longer hours.

History bears this out. The electrification of factories in the early 20th century dramatically increased manufacturing output per worker. The personal computer revolution of the 1990s transformed office work and data processing. The internet enabled entirely new business models and global supply chains. Each of these technological waves drove sustained productivity booms that catalyzed the economy.

For investors, productivity growth matters because it influences most metrics that drive long-term returns. Higher productivity supports corporate earnings growth, stronger GDP expansion, and can help keep inflation in check even as wages rise. Conversely, when productivity stagnates, economic growth slows, wage gains compress, and central banks face tougher trade-offs between supporting employment and controlling prices.

The Current AI Hype vs. Economic Reality

If you believe the forecasts, AI is poised to transform the global economy at unprecedented scale and speed.

That said, to date, AI’s measurable impact on aggregate productivity tells a more tempered story.

According to a recent analysis from MIT: “AI’s economic impact has been more modest than the hype suggests, at least so far.” While AI has demonstrated clear productivity gains in specific use cases (e.g., software development, customer service automation, certain types of data analysis), that hasn’t yet translated into broad, economy-wide productivity growth.

US labor productivity growth in the current business cycle (since late 2019) has averaged 2.2% annually — an improvement over the 1.5% rate of the previous cycle from 2007-2019. Recent quarters have shown even stronger performance, with year-over-year growth reaching 2.8% in late 2025.

So productivity has improved. But how much of that stems from AI specifically?

According to St. Louis Federal Reserve research, AI-related investment categories like information processing equipment, software, R&D, and data centers contributed 0.97 percentage points to real GDP growth in the first three quarters of 2025, accounting for 39% of total GDP growth. That’s substantial, and actually exceeds the contribution of IT components during the dot-com boom in 2000.

Therein lies a critical nuance though: these GDP contributions are derived from investment in AI infrastructure versus businesses actually using AI to produce more efficiently. Companies are building the foundation, spending heavily on the tools and infrastructure. The economy-wide productivity payoff from deploying those tools is mostly ahead of us.

Historical Parallels: What Past Technology Waves Teach Us

Major technological advancements have, historically, followed a similar pattern: early optimism, underwhelming near-term results, and outsized long-term impact. The timeline from adoption to productivity gains (at least that meet rampant expectations) is almost always longer than expected.

The Productivity Paradox (1970s-1990s)

Computers began appearing in offices throughout the 1970s and 1980s. By the mid-1980s, they were ubiquitous in American businesses. Surely, all that technology would supercharge productivity, right?

Not quite, at least initially.

From the early 1970s through 1995, productivity rose about 1.5% per year, well below the strong growth rates of earlier decades. Economist Robert Solow captured the frustration in 1987 with a now-famous observation: “You can see the computer age everywhere but in the productivity statistics.”

What went wrong? Nothing, actually. The problem was timing and implementation.

Businesses needed to reorganize workflows, retrain employees, develop new software, and rethink entire business processes. Installing computers on desks didn’t automatically make workers more productive — it took years of complementary investment and organizational change before the gains solidified.

Between 1995 and 2000, productivity growth finally accelerated to about 2.5% per year, which was comparable to the strong rates seen before the 1970s slowdown. The technology had been around for decades, but the productivity payoff didn’t hit until businesses learned how to use it effectively.

The Internet Boom (1995-2005)

The internet followed a similar trajectory. Early commercial adoption began in the mid-1990s, accompanied by breathless predictions about revolutionizing business and society (of course, those proved right in the long run). The dot-com bubble inflated rapidly, with investors pouring money into companies with “.com” in their names but undefined business models.

The bubble burst spectacularly in 2000-2001, wiping out trillions in market value and prompting skepticism about whether the internet would ever deliver on its promise.

But the productivity gains were real, they just took time to take root. Productivity growth stayed elevated through the early 2000s as businesses integrated internet technologies into their operations.

The timeline from early adoption to measurable economy-wide productivity impact was roughly 10-15 years. Early investors and enthusiasts overestimated what would happen in the first few years, but likely underestimated the long-term picture.

What This Means for AI

The parallels are hard to ignore. AI is following a familiar script: massive investment, sky-high expectations, legitimate but narrow use cases, and an implementation lag between widespread adoption and productivity gains.

If history is any guide, we should expect AI’s economic impact to unfold over the ensuing decade.

Current AI Headwinds

Even with significant investment flowing into AI infrastructure, several obstacles stand between current adoption and the transformative impact forecasted by consulting firms and investment banks.

Implementation Costs

Integrating AI software effectively is a tall task. Companies must invest not only in software licenses but also in reorganizing workflows, retraining employees, upgrading legacy systems, and building internal expertise. These complementary investments are expensive and time-consuming.

A bank implementing AI-powered fraud detection, for instance, can’t simply flip a switch. It needs to integrate the system with existing databases, train staff to interpret AI recommendations, establish protocols for overriding errors, and ensure regulatory compliance.

Quality and Reliability Concerns

AI still makes mistakes. Language models hallucinate facts. Image recognition systems misclassify objects. Autonomous systems fail in edge cases. Human oversight is still imperative, especially for high-stakes decisions (e.g., medical diagnoses, legal rulings, financial transactions).

This limits AI’s ability to fully automate tasks and constrains productivity gains. Instead of replacing workers, AI typically augments them, which is valuable but less impactful than full automation.

Regulatory and Legal Uncertainty

Governments are still figuring out how to regulate AI, and that uncertainty slows corporate deployment. Questions around data privacy, algorithmic bias, intellectual property, and liability are unresolved in many jurisdictions.

Labor and Political Pushback

As you’d expect, AI’s potential to displace workers foments political and social resistance that could limit adoption. People may support regulations that slow deployment or restrict AI use in certain sectors.

Already, we’ve seen pushback against AI in creative industries (writers, artists), customer service (call centers), and transportation (autonomous vehicles). Whether this resistance materializes into permanent impasses or fades as adoption spreads is uncertain — but it’s a variable nonetheless.

What the AI Boom Means for Investors

The path from today’s investment boom to tomorrow’s economic boost is long and uncertain. Here’s how to think about positioning portfolios in the meantime:

• Expect a long timeline. History suggests technology-driven productivity takes 10-15 years to fully materialize. The internet and personal computers both followed this pattern. Even if AI proves to be on an accelerated schedule, that’s more likely to be 5-10 years versus 1-3.
• Beware the hype cycle. AI-related stocks have already priced in significant optimism. If productivity gains lag expectations or take longer than forecasted, valuations could face pressure. Markets tend to overreact to both excitement and disappointment.
• Focus on implementation. Every company claims to be “leveraging AI,” but few are successfully integrating it into operations at scale. Look for firms demonstrating actual productivity improvements and revenue growth.
• Maintain diversification. AI’s impact will be uneven across sectors and timeframes. Some industries will benefit quickly; others will see minimal change. Broad exposure hedges against the uncertainty of which companies and sectors will capture the value.

Stay realistic. Gates was right: we tend to overestimate near-term change and underestimate long-term implications. The next two years may disappoint investors expecting rapid, economy-wide gains. The next ten could prove more transformative than current forecasts suggest. That said, investors have a bad habit of chasing returns in hot sectors – which virtually assures that they will compromise the long term returns their portfolios might otherwise earn. Even though we can’t know with certainty how the future will play out, we do know from history that stock prices often overshoot in both directions – on the upside, as well as on the downside. Investors who aggressively chase the hot sectors on the upside inevitably perfect the art of buying high and – after a string of disappointments that the lofty stock prices could never fulfill – selling low. For this reason, chasing hot stocks is a time-tested formula for reducing your own returns, whatever the future holds. Investors that stay disciplined in their allocations participate in the upside, protect the downside and recover sooner than those chasing performance. These are the fundamentals of long-term performance. So, don’t let your enthusiasm for specific stocks or sectors mettle with your discipline. In the long run, discipline wins.