Does AI Increase Productivity? An Analysis of the Solow Paradox

As AI companies become increasingly dominant in the global economy and pour billions of dollars into data centers, whether these technologies actually deliver on one of their most fundamental promises, increased productivity, remains a big question mark. Even though the tech world argues that artificial intelligence (AI) will revolutionize the way we work, economic data does not yet support that narrative.

Technology Does Not Always Translate Linearly into Productivity

One of the people who views these economic expectations with skepticism is JP Gownder, vice president and principal analyst at global market research firm Forrester. According to Gownder, there is not yet any strong sign in the current data that AI has created a noticeable jump in productivity.

In an interview with The Register, Gownder said they have not seen an AI-driven productivity increase, and stressed that information technology often does not translate into productivity as linearly as people assume. In other words, the tools in our hands becoming more powerful does not always mean work gets done faster or more efficiently. Sometimes the integration and learning curve of new technologies can stall or even reduce productivity in the short term.

The Solow Paradox: "The Computer Age Is Everywhere, Except in the Statistics"

This does not mean AI cannot be a transformative technology. But even if its impact is large, it may not show up in macroeconomic indicators. This phenomenon is explained by the Solow Paradox, put forward in 1987 by Nobel-winning economist Robert Solow and summed up as "You can see the computer age everywhere but in the productivity statistics."

Historical data also supports this paradox. Gownder points to data from the US Bureau of Labor Statistics:

* Between 1947 and 1973, before personal computers became widespread, productivity grew by an average of 2.7% a year. * In the 1990–2001 period, when PCs went mainstream, this rate fell to 2.1%. * Between 2007 and 2019, the growth rate dropped to 1.5%. * Only between 2001 and 2007, with the full integration of the internet into business processes, was a temporary productivity increase of 2.8% achieved.

Today we may be going through a similar process with AI. The technology is everywhere, but the productivity results are not yet visible.

Research Does Not Confirm Expectations: 95% of Companies Saw No Profit

Academic studies and workplace practice also show that AI has not yet matured and has not created the expected "magic wand" effect.

For example, a striking study by MIT (Massachusetts Institute of Technology) found that 95% of companies that integrated AI did not achieve a meaningful increase in revenue or profit margins. AI departments set up with high expectations and huge investments have not yet delivered a return on investment (ROI).

In software development, another area presented with big expectations, the situation is even more complicated. There are findings that programmers using AI-assisted coding tools, although they produce code faster, spend more time debugging errors and integrating the code, and therefore end up slower in total time to finish the job.

Autonomous Agents and the "Sloppy Work" Problem

"AI Agents," which aim to automate every task, are also still far from what was promised. AI agents tested by Center for AI Safety researchers on remote-work tasks were able to complete less than 3% of the complex jobs given to them.

In addition, some studies suggest that AI is hurting workplace relationships and work quality. Employees handing off low-quality or draft-like work output to others or to the next stage, expecting that "AI will fix it anyway," creates a "sloppy output loop" in business processes. This lowers overall quality while increasing the effort spent fixing errors.

According to Gownder, much of generative AI "isn't really working" right now. He notes that the problem is not limited to the end-user experience; the vast majority of projects are not delivering concrete profit. This picture also explains why mass layoffs have not happened yet: AI is not yet productive enough to replace humans.

The 2030 Outlook: Are Structural Job Losses Coming?

Despite the current picture of inefficiency, long-term projections suggest AI could have devastating effects on the workforce. According to Forrester's forecasts, automation technologies such as AI and physical robots could eliminate about 6% of jobs by 2030. That corresponds to roughly 10.4 million positions.

JP Gownder stresses that these losses are structural. That is, these are not jobs that temporarily decline due to economic fluctuations, but positions that will permanently disappear as technology advances. This will have a significant impact on the economy and create the problem of reskilling the workforce.

The "AI" Narrative and the Hidden Reality of Outsourcing

Perhaps one of the most striking analyses is about how companies use the "AI" narrative. Over time, it is inevitable that some employers will realize AI has not met expectations. Indeed, there are reports of companies that laid off employees in the name of the "AI revolution" and then quietly rehired them when things went wrong.

But according to Gownder, the "layoffs because of AI" narrative is often used as cover to hide other cost-cutting methods.

> "They say they're laying people off because of AI, but three weeks later they set up a team in India because it's much cheaper."

This shows that tech companies are still getting their productivity gains not from technology but from traditional outsourcing and cheap labor.

The Gap Between Hype and Reality

AI is undoubtedly a powerful technology and will shape the future. But today's data shows that a magical productivity leap has not happened yet, and that the Solow Paradox is repeating itself once again. The real test for companies lies in positioning AI not merely as a "hype" tool but as a strategic tool that will genuinely improve business processes. Otherwise, billions of dollars in investment could turn into disappointments that never show up in the productivity statistics.