Since 1980, something odd has been happening with automation. Companies aren't using it to make things faster or better. They're using it to pay people less.
MIT economists Daron Acemoglu and Pascual Restrepo tracked forty years of automation decisions and found a pattern nobody was talking about. Firms targeted automation at workers earning above-market wages - not at tasks where machines would genuinely boost productivity. The goal wasn't efficiency. It was wage control.
The numbers are stark. Automation accounts for 52% of income inequality growth over the past four decades. That's more than half of the widening gap between high and low earners, driven by machines deployed to suppress wages rather than improve output.
The Productivity Paradox Nobody Saw Coming
Here's where it gets worse. This approach didn't just hurt workers - it hurt productivity too. The study found that wage-targeting automation offset 60-90% of potential productivity gains. Companies automated tasks that weren't ready for automation, creating bottlenecks and inefficiencies that cancelled out most of the benefit.
Think of it like this: if you automate a factory line to cut labour costs but the machines can't handle edge cases, you've just traded skilled problem-solvers for rigid systems that break when anything unexpected happens. The wage bill drops. Output doesn't rise. Sometimes it falls.
The research distinguishes between two types of automation. Productivity-enhancing automation takes tasks machines genuinely do better - faster, more precise, more consistent. Wage-targeting automation replaces workers who cost too much, regardless of whether the machine actually does the job better.
Most automation over the past forty years falls into the second category.
What This Means for the Next Wave
We're heading into a new automation cycle powered by AI. The question is whether companies will repeat the same mistake.
Early signs aren't encouraging. The MIT study suggests that without policy intervention, firms default to wage-targeting. It's rational from a single company's perspective - labour costs drop immediately, productivity concerns are someone else's problem.
But collectively, it's a disaster. Suppressed wages mean suppressed demand. Lower productivity means slower growth. Income inequality creates political instability. The short-term win for individual firms becomes a long-term loss for everyone.
The researchers argue for policy that steers automation toward productivity rather than wage suppression. Tax incentives for genuine productivity gains. Penalties for automation that just replaces workers without improving output. Investment in retraining programmes that help displaced workers move into roles where human skill still matters.
None of this is happening yet. Which means the next wave of AI automation is likely to follow the same pattern as the last forty years - targeting wages, suppressing productivity, widening inequality.
The technology isn't the problem. The incentive structure is. Until that changes, automation will keep making things worse for most people while concentrating gains at the top.
For business owners, the message is clear: automate for productivity, not for wage control. The short-term saving isn't worth the long-term damage - to your business, your sector, and the economy that sustains both.