For years, the dominant narrative around AI and jobs has been a reassuring one: yes, AI will change work, but it will augment more than it displaces. Workers will become more productive. New tasks will emerge. The pie grows.
A new survey from Epoch AI and Ipsos chips a significant crack in that narrative. Of 2,000 U.S. adults polled, 20% of full-time workers say AI has already replaced parts of their job. Only 15% say AI has created new on-the-job tasks they wouldn’t have taken on otherwise. For the first time in a major survey, displacement is outpacing augmentation — and the gap isn’t theoretical. It’s happening now.
The Numbers
The Epoch AI/Ipsos survey, published April 12, 2026, asked 2,000 U.S. adults about their direct experience with AI at work. The key findings:
- 50% of all respondents used AI in the past week for personal or work reasons
- 20% of full-time workers reported AI had taken over tasks they used to do themselves
- 15% reported AI had created new on-the-job tasks
- The net displacement effect: 5 percentage points in favor of replacement
Five percentage points may sound modest. But this is the first time a major survey has measured AI replacement exceeding AI augmentation in real-time worker experiences — not projections, not models, not think-tank forecasts. Workers’ own reports of what is happening to their jobs right now.
What “Replacement” Actually Means
The survey doesn’t claim that 20% of workers have lost their jobs entirely. What it captures is more granular and, in some ways, more telling: AI is taking over specific tasks within existing roles. A marketing analyst whose data queries are now automated. A paralegal whose document review is handled by an AI tool. A customer service rep whose Tier 1 tickets are routed to a chatbot.
This is the “frog in warming water” scenario that labor economists have warned about. No single task displacement feels like a job loss. But as the tasks accumulate, the role erodes. The job title remains. The work inside it shrinks.
Nichols Miailhe, an AI policy leader at the Global Policy on Artificial Intelligence, told NBC: “When one in five workers say AI is already replacing parts of their job, we can start talking about labor market restructuring happening in real time. The fact that replacement seems to be outpacing augmentation should draw our attention: the policy window to shape how AI transforms work is probably closing faster than most governments realize.”
The Counterweight
The survey arrives in a landscape where the AI automation narrative is far from settled. There are legitimate reasons for skepticism:
- AI in the workplace remains error-prone. Amazon’s drive to replace human workers has reportedly slowed overall productivity.
- Automation experiments haven’t held up. Klarna’s widely publicized AI replacement of customer service staff ended with the company hiring humans back after an 11-month experiment failed.
- AI critic Gary Marcus has argued that the math on AI-driven unemployment doesn’t add up, and that productivity gains from AI remain unproven at scale.
These counterpoints matter. The Epoch AI survey measures perception — what workers believe is happening to their tasks. It doesn’t measure economic output, company financials, or whether the AI-displaced tasks were actually being done well. The gap between “AI took my task” and “AI did that task competently” remains a live question.
The Bigger Economic Picture
The Epoch AI findings don’t exist in isolation. A separate, sweeping economic survey from researchers at the Federal Reserve Bank of Chicago and several top universities has found that economists are increasingly revising their models to account for serious labor market disruption.
The convergence is what makes this moment notable. When worker self-reports, academic modeling, and institutional economic research all point in the same direction — toward displacement exceeding augmentation — the burden of proof shifts. It’s no longer “prove AI is taking jobs.” It’s “prove it isn’t.”
What Workers Can Do With This Data
For individual workers, the survey reinforces what many already feel: AI is not a future concern. It’s a present one. The practical implications:
- Audit your tasks. Which parts of your role are most automatable? Which require judgment, creativity, or relationship-building that AI can’t replicate?
- Invest in augmentation skills. The 15% who report AI creating new work are early indicators of where human value shifts — toward managing, evaluating, and directing AI outputs rather than competing with them.
- Track your industry. Replacement rates vary dramatically by sector. The 20% average masks much higher numbers in data-heavy, rules-based roles.
- Don’t wait for policy. Miailhe’s warning about the closing policy window is pointed at governments, but workers facing displacement can’t afford to wait for regulatory frameworks that may arrive too late.
The Inflection Point
This survey doesn’t settle the debate. But it does shift it. For years, the optimistic case rested on a claim that augmentation would outpace displacement — that AI would make workers more productive faster than it made them redundant. That claim now has a data point working against it.
One survey doesn’t make a trend. But one survey where replacement beats augmentation is the kind of data point that demands a follow-up. If the next Epoch AI survey shows the gap widening — not narrowing — the “AI augments more than it displaces” narrative will need more than theory to sustain it.
SOURCES
- Epoch AI / Ipsos — April 2026 U.S. AI & Work Survey
- Federal Reserve Bank of Chicago — Economic Survey on AI Labor Market Impact
- NBC News — Coverage of Epoch AI findings
