Google’s New AI Jobs Initiative Is Creating Work — but Not the Kind of Work Most Displaced Workers Can Do

This month, Kate Alessi chose a bold stance to take. In an interview with The Times, Google’s managing director for UK and Ireland contended that concerns about mass unemployment brought on by AI were exaggerated, that history rarely cooperates with doom, and that new jobs typically emerge where old ones decline. The thesis is well-organized. It’s also the kind of thesis that sounds much cozier in a glass office in King’s Cross than it does on an open-plan administrative floor in Leeds.

Google has been quietly promoting a number of initiatives, including partnerships with higher education institutions, AI literacy courses, and skills programs. These initiatives are real and, in some respects, actually helpful. However, there’s a sense that the company’s optimism is outpacing the evidence as it rolls out. For the most part, a displaced accounts clerk cannot easily transition into the new roles being created around AI governance, prompt design, model oversight, and implementation engineering after a weekend of online modules.

The mood was uncomfortably accurately captured in Anthropic’s recent survey of 81,000 Claude users. Concerns regarding displacement were expressed by 25% of respondents. The story’s purported winners, software engineers, acknowledged that they were anxious “pretty much 24/7.” In the belief that AI would take over, project managers were giving developers more difficult tickets. The fear is no longer abstract. When someone discovers their team has completely stopped hiring juniors, it appears in Slack channels and on quiet Tuesday afternoons.

Then there’s the budget story, which is rarely presented with the optimism. This month, Quartz provided direct documentation of it: Challenger, Gray & Christmas recorded over 15,000 layoffs in the United States in March that were directly related to AI, Dell eliminated 11,000 positions in fiscal 2026, and Cisco publicly reallocated hundreds of millions into AI. The funds did not disappear. It shifted. into power contracts, data centers, and GPUs. The simplest lever to pull was payroll.

The mismatch is difficult to ignore. It is not realistic to expect a secretary in Birmingham or an office clerk in Manchester, the very categories identified as highly exposed with low adaptive capacity, to become an AI oversight specialist by next summer. There is training. For most, the pathways don’t. Retraining initiatives typically benefit employees who were previously in close proximity to the new field. For everyone else, the time between the old and new jobs is measured in years rather than weekends.

Beneath the optimistic headline, Alessi’s deeper point—that most people are merely scratching the surface of what these tools can do—is likely the more truthful one. That is accurate. Furthermore, “scratching the surface” isn’t a career. In more straightforward terms, the OECD has stated that AI changes tasks more than it eliminates roles, but those who fall between the cracks typically remain there for longer than the policy papers indicate.

The entire shape of it has a hint of familiarity. Each automation wave comes with a chorus of assurance, and each wave leaves behind a cohort that the assurance was unable to reach. The fact that new work is being produced is accurate, according to Google. It might simply be incorrect about who is allowed to do it. As this develops, the question is not whether AI will create jobs. The question is whether the jobs it creates will be anywhere close to those who recently lost theirs.