AI Wage Growth: Experts Warn of Reversal

For months, the narrative surrounding artificial intelligence has been almost uniformly triumphant.

AI, we’re told, is the ultimate productivity booster, a technological tide lifting all boats – or at least, the wages of those savvy enough to harness its power.

It’s a vision of enhanced human capability, where intelligent tools amplify our output, making us more valuable, and thus, better paid.

Yet, a disquieting counter-narrative is emerging from the ivory towers of academia, suggesting that this golden age of AI-fueled wage growth might be far more fleeting than many anticipate.

Ioana Marinescu, an associate professor at the University of Pennsylvania’s School of Social Policy & Practice, is sounding a warning bell that cuts through the prevailing optimism.

Her research, conducted in collaboration with the Brookings Institution, suggests we may already be approaching a critical inflection point.

This is a moment where the very technology celebrated for its wage-boosting potential could begin to exert downward pressure on paychecks.

It’s a sobering thought for a workforce still grappling with the dizzying pace of technological change.

Marinescu’s model posits a concept she calls “intelligence saturation” .

In essence, there’s a sweet spot where AI enhances human productivity, leading to higher wages.

But push beyond that point, and the dynamic reverses.

Her research indicates that once roughly 37% of cognitive or “intelligence” tasks are automated, wage growth could begin to decline.

This isn’t some distant, theoretical future; Marinescu notes that we are currently over 14% automation, a figure that has been steadily climbing since the 1980s.

This places us firmly on the upward slope of what she and coauthor Konrad Kording term the “AI pay curve“.

This is a trajectory where wages initially rise, then flatten, and ultimately decline as machines increasingly take over human work.

The initial phase of this curve is where we’ve largely been operating.

AI, in its early iterations, acts as a force multiplier.

It helps workers process information faster, analyze data more efficiently, and automate tedious, repetitive tasks, freeing them to focus on higher-value activities.

This increased productivity naturally translates into higher wages.

But as AI becomes more sophisticated and capable, its role shifts from augmentation to substitution.

When machines can perform tasks not just faster, but entirely independently, the need for human input diminishes, and with it, the economic leverage of the human worker.

So, what are the harbingers of this impending peak?

Marinescu points to two key indicators: a noticeable slowing in overall wage growth and a subtle but significant shift in employment patterns.

We should, she argues, begin to see a movement away from intelligence-heavy jobs towards more physical ones.

The logic is simple: if AI is displacing cognitive work, the remaining opportunities might increasingly lie in areas requiring embodied presence, manual dexterity, or complex interpersonal skills that remain challenging for machines to replicate.

Consider the historical context.

The share of routine cognitive jobs, which serves as a proxy for intelligence-based work, fell from 49% in the late 1970s and 1980s to 35% in 2018 – a 14-point drop that Marinescu attributes to automation.

This steady erosion of cognitive tasks has been underway for decades, but the advent of generative AI promises to accelerate this trend dramatically.

Jobs once thought safe, like translation, basic marketing copywriting, or even aspects of software development, are now squarely in AI’s crosshairs.

If these roles are displaced without a commensurate creation of new, high-value intelligence jobs, the overall share of intelligence-sector employment will inevitably decline, setting the stage for a broader wage downturn.

The speed and severity of this decline are not predetermined, however.

They hinge on how easily the physical and intelligence sides of the economy can substitute for one another.

If workers can readily transition from cognitive roles to physical ones, the impact might be softened.

But if these sectors become increasingly distinct, or if the physical sector itself becomes saturated with displaced intelligence workers, the wage losses could accelerate.

Similarly, within the intelligence sector, if tasks become less substitutable with each other, wages might grow for longer, but the eventual decline could be steeper once it begins.

This paints a picture of a complex economic dance, where the nuances of task substitutability dictate the future of human labor.

Marinescu’s research offers a crucial lens through which to view the ongoing AI revolution, urging a move beyond simplistic narratives of inevitable progress.

It compels us to ask difficult questions about the nature of work, the distribution of economic gains, and the policies needed to navigate a future where intelligent machines are not just tools, but potential competitors.

The single most important metric to track, she concludes, is the evolving share of workers in physical versus intelligence-sector jobs.

If that balance begins to shift decisively, with more humans gravitating towards roles that AI finds difficult to automate, it will be a clear signal: the AI pay boost may have already peaked, and a new, more challenging economic reality could be just around the corner.