Google's new report says AI won't kill all the jobs

Generative AI is a general-purpose technology on par with steam and electricity, but with a diffusion timeline compressed by the internet.

Google published an economic report authored by its inaugural Technology and Society Visiting Fellow, MIT economist Andrew McAfee, analyzing how generative AI will shape labor markets, productivity, and future growth.

Faster diffusion than steam, 80% of jobs touched

Economists reserve the term general-purpose technology for foundational breakthroughs that rewrite how entire economies function. McAfee argues that generative AI belongs in that rare tier alongside electrification and steam power.

Unlike historical transformations that took decades to crawl through industrial supply chains, generative AI travels instantly across the web. Its natural language interface also removes technical friction, letting non-technical workers adopt it immediately.

According to the report, roughly 80% of jobs across the U.S. economy contain at least 10% of tasks that could be executed twice as fast without any loss in output quality.

As organizations build complementary software and processes around these capabilities, baseline productivity gains are expected to compound into broader macroeconomic acceleration.

The phantom unemployment wave

Panic around widespread technological unemployment is likely overstated, McAfee writes. Historically, the productivity booms driven by general-purpose tech have triggered stronger, not weaker, demand for human labor.

The catch is where those jobs end up. The report highlights that over 85% of total U.S. employment growth since 1940 has materialized in entirely new job categories that did not exist before.

There is also a silver lining for junior talent. Because generative models synthesize institutional knowledge and serve it on demand, McAfee notes they are especially effective at leveling up entry-level workers, potentially narrowing wage disparities.

Why it matters

For individual workers, particularly those early in their careers, the takeaway is clear: generative AI acts as a floor-raiser rather than an outright replacement. Waiting on the sidelines to adopt these tools is a losing trade, because baseline task expectations are already shifting upward.

For the tech ecosystem, this report outlines the playbook Google and its peers are selling to enterprise buyers and regulators. By framing AI as an economic engine rather than a labor-destroying liability, frontier labs are trying to secure social license for massive infrastructure buildouts.

We think McAfee's historical framing is economically sound, but it glosses over the friction of real-world transitions. Even if 85% of future jobs are net new, the retraining window between task obsolescence and job creation can be painful. The winners will not just be the foundational model builders, but the organizations that figure out the reskilling pipeline before their existing workflows break down.