Two pieces of news from the same week in late March pull in opposite directions and yet, read together, are more honest about what AI is doing to the workforce than either optimistic or pessimistic takes on their own. Meta cut several hundred jobs, explicitly framing the move as part of an efficiency push enabled by its AI investments. And a report from an AI company found that a measurable skills gap is opening between workers who have deeply integrated AI tools into their work and those who have not, with the gap showing up in productivity, compensation, and hiring outcomes. Both things are happening at the same time, in the same economy, to different people.
The dominant framing in public debate treats the AI jobs question as a binary: either AI is displacing workers and the pessimists are right, or it is creating new kinds of work and the optimists are right. Neither frame has much predictive power for individuals, because both are true in aggregate. The real story is distributional. AI is concentrating productivity gains among a narrowing group of workers who have the skills, the organisational support, and the inclination to use it effectively. At the same time, it is enabling organisations to reduce headcount in functions where AI substitution has crossed a cost-effectiveness threshold. These processes are happening in parallel, not in sequence.
The "AI power user" category that the skills-gap research describes is real and worth taking seriously. Workers who have built genuine fluency with AI tools are producing more output, taking on work that previously required more senior roles, and making themselves harder to replace. But the report's framing invites a conclusion that its own data should complicate: it comes from an AI company with an obvious interest in establishing that AI skills are valuable and that people who lack them are falling behind. The prescription it implies, learn to use AI or get left behind, is convenient for vendors of AI tools. It is also probably correct.
The Meta situation is different in character. The cuts are small relative to Meta's overall headcount and are being attributed partly to performance management rather than pure automation. But the framing matters: Meta is explicitly connecting AI investment to reduced headcount as a feature of its business model, not as an unfortunate side effect. When a company this large makes that connection public and casual, it signals that the calculation has become unremarkable. That normalisation is worth watching.
What is missing from most coverage of these two stories is the connection between them. The emergence of AI power users who are dramatically more productive is not just a personal career story. It is also an organisational story: a company that employs one very productive AI-augmented worker where it previously employed three does not simply increase output by 3x. More commonly, it holds output roughly constant and reduces the workforce. The productivity gains go to the company's margins, not to additional employment. The power users and the laid-off workers are part of the same economic transition, not separate phenomena.
The honest version of where we are is that both narratives have real-world support, and neither explains the whole picture. AI is creating genuine new capability for workers who engage with it seriously. It is also enabling headcount reductions that would not have been made otherwise. The question that neither story addresses well is where the net balance falls, for different kinds of workers, over what timeframe, and in which sectors. That question does not have a tidy answer in Q1 2026. What is becoming clear is that the effects are not uniform, not slow, and not going to wait for policymakers to develop a coherent response.