The AI boom promised to transform businesses, slash costs, and supercharge productivity once and for all.
But here’s the thing: the numbers still aren’t showing that happening in practice.
A recent survey of thousands of executives around the world revealed something few expected to hear — most CEOs admit that artificial intelligence still hasn’t generated any real impact on either employment or productivity at their companies.
And you know what’s even more interesting?
This scenario is leading economists to dust off a classic paradox from the 1980s, when the world went all in on computers and also sat around waiting for a revolution that took decades to show up in the data.
The question is: are we repeating the same story, or is something different this time around?
Let’s break down what’s behind this economic paradox and what it might tell us about the future of AI in business. 👇
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
What the Survey Actually Showed
The study surveyed thousands of business leaders from major global corporations, and the result was, to put it mildly, surprising for anyone following the hype around artificial intelligence: the overwhelming majority of executives interviewed said they still can’t see concrete productivity gains within their operations. We’re not talking about small businesses trying to take their first steps with technology — we’re talking about CEOs of organizations that have already invested billions of dollars in AI tools, data infrastructure, and team training. And still, the needle on operational performance has barely moved.
What stands out isn’t just the data point itself, but the context in which it appears. Never before in corporate history have so many companies adopted a technology so quickly. ChatGPT took just two months to reach 100 million users — something that took Instagram two and a half years to pull off. Tools like Microsoft Copilot, Google Gemini, and dozens of generative AI platforms are already embedded in the daily workflows of companies worldwide. Yet when executives look at their performance reports and try to attribute productivity growth directly to AI usage, the number that shows up is disappointingly close to zero.
Another finding from the survey was that most leaders also didn’t see a significant impact on employment levels at their companies — neither up nor down. Contrary to what many feared or expected, AI hasn’t triggered a massive wave of layoffs nor created a substantial volume of new jobs. The labor market, at least so far, seems to have absorbed the technology far more quietly than the most pessimistic or optimistic economic models predicted. That alone is a key data point for understanding why the paradox is taking shape.
The Productivity Paradox Isn’t New
To understand what’s happening with AI today, it helps to take a step back and look at a similar moment in economic history. In the late 1980s, economist Robert Solow made an observation that remains famous to this day: you can see the computer age everywhere except in the productivity statistics. That quote became known as the Solow Paradox — or the productivity paradox — and it summed up a collective frustration that economists, managers, and governments were feeling at the time. Companies had invested heavily in personal computers, software, and office automation, but the macroeconomic productivity indicators simply weren’t responding. The gains everyone expected to see within a year or two ended up taking nearly a full decade.
What happened next was interesting: the productivity revolution did arrive, but on its own schedule. During the 1990s, especially the second half of the decade, the United States experienced one of the most significant productivity growth periods in modern history. Companies finally learned to reorganize their processes around the new digital tools, workers developed real technological fluency, and the supporting infrastructure — networks, data standards, systems integration — finally matured enough for the benefits to show up in the numbers. The lesson was clear: transformative technologies rarely produce immediate impact. They require adaptation time, organizational restructuring, and a learning curve that doesn’t show up in quarters, but in years.
Economists like Erik Brynjolfsson from MIT, who studied the Solow paradox extensively, are already drawing direct parallels with the current AI moment. According to him, we’re living through a period of investment without visible returns — exactly like what happened with computers. Companies are buying tools, hiring specialists, and building data pipelines, but they haven’t yet reorganized their workflows deeply enough for AI to generate measurable gains at scale. The difference, according to Brynjolfsson, is that generative AI has a much broader transformation potential than the personal computer — which could mean both that the impact will be larger when it arrives and that the waiting period could be even longer.
Why Executives Still Aren’t Feeling the Difference
There are a number of practical reasons that explain why AI‘s impact hasn’t yet reached the productivity reports of major companies — and understanding those reasons is essential for knowing what to expect going forward.
Fragmented adoption and one-off usage
The first one is the issue of fragmented adoption. Even at companies that have already implemented AI tools, usage is still very sporadic. A marketing team uses ChatGPT to draft copy, an IT team uses a code copilot, a finance department experiments with spreadsheet automations. But these isolated uses rarely translate into systemic productivity gains — and it’s precisely systemic gains that move the needle in macroeconomic data. As long as AI remains a supplementary tool scattered across disconnected pockets within an organization, the effect on overall results will continue to be marginal.
The weight of internal reorganization
Another important factor is what’s known as the cost of reorganization. Implementing AI for real isn’t just about buying a software license and handing it out to employees. It means rethinking processes, redefining roles, creating new approval workflows, training teams, and often facing considerable internal resistance. This process takes time and organizational energy — and during it, productivity can actually drop before it rises. Companies that understood this are investing heavily in change management and culture before expecting results. Those that haven’t are still frustrated with expensive tools that deliver very little.
There’s a pretty illustrative parallel here with what happened during the transition to electricity in the early 20th century. When factories swapped steam engines for electric motors, the first thing they did was simply replace one with the other in the same factory layout. The result? Virtually zero productivity gain. It wasn’t until engineers completely redesigned the factories — taking advantage of the flexibility electricity allowed to reorganize production lines, create independent workstations, and distribute power in a decentralized way — that real gains appeared. With AI, we’re seeing exactly the same pattern: most companies are plugging new technology into old processes and wondering why nothing has changed.
The measurement problem
There’s also a measurement problem that economists call data lag. A good portion of the value AI is generating today is hard to capture using traditional productivity metrics. An employee who uses AI to respond to emails faster, prepare presentations in half the time, or research information with far more agility is clearly being more productive — but that gain doesn’t show up directly in GDP or in the indicators executives use to measure corporate performance. This creates a curious situation where the technology is working at the individual level but hasn’t yet consolidated into collective, measurable impact.
On top of that, many AI benefits are qualitative rather than quantitative. The quality of a report produced with AI assistance can be significantly better, the response time to a customer can improve, the depth of a strategic analysis can increase — but none of those improvements are easily translated into spreadsheet numbers. Traditional macroeconomic indicators were designed to measure the production of tangible goods, not quality improvements in complex cognitive tasks. This creates a statistical blind spot that’s likely underestimating the true impact of AI in the corporate environment.
What Could Change in the Coming Years
Despite the current paradox, there are concrete reasons to believe the picture will change — and probably faster than it did with computers in the 90s.
Exponential model evolution
The main one is the speed at which AI models themselves are evolving. While personal computers evolved at a relatively linear pace during their first decades, large language models are advancing at an exponential rate. The leap in capability between GPT-3 and GPT-4, for example, was stunning — and the models on the way promise even more robust performance in complex tasks involving reasoning, planning, and autonomous execution. This means the tools available to companies two or three years from now could be fundamentally different from what exists today, which completely changes the return-on-investment equation.
The era of AI agents
Another element that could speed up the turning point is the development of AI agents — systems that don’t just answer questions but execute tasks autonomously, chaining actions together, making decisions within defined parameters, and interacting with other digital systems. Companies like OpenAI, Anthropic, Google, and Microsoft are investing heavily in this direction, and the first enterprise agents are already being tested in real-world environments. When this technology matures, AI will stop being a support tool and become an active agent within business processes — and that’s when the productivity impact should finally become visible in the numbers.
It’s also worth thinking about the cascading effect that could emerge when different layers of AI start integrating within the same company. Today, most use cases are isolated: a chatbot here, an automation there, a code assistant over there. When these systems begin talking to each other, sharing context, and operating as an intelligent network distributed throughout the entire organization, the compounding gains could far exceed the sum of their parts. That’s the inflection point where the curve tends to take off, exactly like what happened when the internet connected computers that had previously operated as islands inside companies in the 90s.
Sectors that could lead the shift
Some sectors are closer to that inflection point than others. Areas like financial services, healthcare, and software development have characteristics that favor deeper, more integrated AI adoption: large volumes of structured data, repetitive processes with high added value, and professionals with above-average technological fluency. It wouldn’t be surprising if the first clear signs of productivity gains attributable to AI started showing up in these segments before spreading to the rest of the economy.
The Historical Message for Today’s Decision-Makers
For executives caught in the dilemma right now — investing in AI without seeing clear returns — the historical message from the Solow Paradox is both a warning and a reassurance. The warning is that technology alone doesn’t transform any organization: processes, culture, and management models need to change along with it. The reassurance is that the companies that invested early in computers, even without seeing immediate returns, were exactly the ones that reaped the biggest rewards when the turning point arrived.
The historical pattern suggests that the gap between adoption and concrete results can take five to fifteen years for general-purpose technologies — and AI fits that category perfectly. That doesn’t mean current investments are being wasted. Quite the opposite: organizations that are building solid data infrastructure, developing teams with AI competency, and experimenting with diverse use cases are accumulating a type of intangible capital that will convert into real competitive advantage when the ecosystem matures.
AI will likely follow the same path as the great technological waves that came before — and those who are best prepared when the data finally confirms what intuition already suggests will come out ahead. 🚀
