Nobel laureate economist Daron Acemoglu doesn’t sugarcoat it. AI, he warns, is going to increase inequality between labour and capital. “That is almost for sure,” he told the Financial Times. And the rhetoric about AI democratising opportunity? He calls it what it is — rhetoric.
The Survey
A new Financial Times survey cuts through the marketing narrative that’s dominated AI coverage. The findings are blunt: AI primarily benefits those who are already advantaged. People with education, abstract and quantitative skills, and familiarity with computers and coding can leverage AI tools effectively. Everyone else? They’re on the wrong side of a widening gap that no amount of “AI for everyone” branding can bridge.
Acemoglu’s assessment is particularly pointed: “The rhetoric out there is that the tools are going to be democratizing. But the reality is… AI is going to increase inequality between labour and capital.”
Who Gets the Gains
The mechanism is straightforward once you strip away the hype:
- Capital owners gain from AI-driven productivity increases — higher output, lower costs, wider margins.
- Highly skilled workers gain from AI-augmented productivity — they can do more, faster, and command higher wages.
- Workers without strong digital skills gain little or lose — their tasks get automated, their bargaining power erodes, and the new roles created by AI require skills they don’t have.
This isn’t a prediction. It’s already observable. The companies seeing the biggest productivity gains from AI are those with sophisticated engineering teams who can deploy and maintain AI systems. The workers being displaced are disproportionately in roles that require less formal education and technical training.
The Skills Trap
Acemoglu identifies the core problem: AI tools require significant existing skills to use effectively. The people who need AI upskilling most — workers in routine jobs facing automation — are precisely the least positioned to access and leverage these tools.
It’s a trap that policy responses haven’t addressed. When governments talk about “AI literacy,” they typically mean introductory workshops and online courses. But becoming genuinely productive with AI tools requires the kind of quantitative and abstract reasoning skills that take years to develop — the very skills that the workers most at risk often lack.
The result: AI upskilling programmes mostly help people who were already going to be fine.
The New Zealand Context
For New Zealand, the warning is especially sharp. NZ’s economy has a relatively small tech sector and a large proportion of workers in agriculture, tourism, and traditional services — exactly the roles most vulnerable to AI automation.
The questions Acemoglu raises land differently in a small economy:
- Can NZ’s education system pivot fast enough? Current curricula are still catching up to basic digital literacy. AI fluency is a generation away for most students.
- Is the government’s AI strategy addressing the right problem? Incentivising AI adoption without parallel investment in worker retraining may accelerate the very inequality Acemoglu warns about.
- What happens to regional economies? Auckland and Wellington may adapt. The regions — where digital infrastructure and training access are already limited — face a compounding disadvantage.
The Uncomfortable Question
If Acemoglu is right — and his track record on technology and inequality is strong — then the current approach to AI policy is backwards. Most efforts focus on getting AI tools into more hands. The harder, more important work is building the foundation of skills and institutional support that makes those tools genuinely useful rather than merely available.
An AI coding assistant given to someone who can’t code isn’t democratisation. It’s a paperweight with a subscription fee.
The “shitshow” Acemoglu predicts isn’t inevitable. But avoiding it requires honesty about who AI actually helps right now, and serious investment in the people it doesn’t.
SOURCES
- Financial Times survey on AI and inequality
- MIT — Daron Acemoglu research on technology and labour
