1. What AI is really doing to jobs right now
Tens of thousands of people lost their jobs to artificial intelligence in the US alone last year. At least, that is the line doing the rounds across headlines, LinkedIn posts and conference panels. It is the kind of statistic that sticks because it feeds a familiar fear: the machines have arrived for our pay cheques.
Look a little closer though and the picture is much murkier. The just under 55,000 losses cited were job cuts explicitly labelled as AI-related in official filings. Even Challenger, Gray & Christmas, which tracks this, shows the number is still small when set against the much bigger forces driving redundancies: restructuring, closures, cost cutting and broader technology change. AI is in the story but it isn’t the whole story.
That gap between rhetoric and reality runs through much of the evidence. From economists debating whether GDP is missing a hidden AI dividend to companies like Klarna quietly reintroducing humans after trumpeting their AI gains, the facts are far less clear-cut than the hype.
The current AI jobs debate is becoming too blunt to be useful. It encourages us to ask whether AI is taking jobs when the more valuable questions are these: where is pressure really showing up first, what kind of work is changing and what should leaders be doing about it now?
And that’s where I start in an analysis on The People Space today: not with speculation but with the data, the history and the questions the evidence still leaves unanswered.
So, how many jobs has AI really taken?
The direct answer is so far, not that many. The strongest evidence still does not show a clear wave of AI-driven unemployment. Anthropic’s recent labour market analysis found no systematic rise in unemployment among workers in highly AI-exposed roles. If you are looking for a neat, measurable spike in people being pushed out of work by AI the data does not yet support it.
But that does not mean nothing is happening. Where the pressure seems to be showing up first is in hiring, especially for younger workers entering highly exposed occupations. Anthropic found a 14% drop in the job finding rate for workers aged 22 to 25 in those roles compared with the 2022 baseline. That is a subtler signal than a redundancy headline, but it may be the more important one.
A profession can look stable while becoming much harder to enter. Existing employees stay put. Fewer new people get in. The ladder remains standing, but fewer people are allowed onto it.
That should concern HR. Entry-level roles are where people build judgement, context and the messy human understanding that work still depends on. Strip out too much junior work too soon and you do not simply save cost but risk weakening your future talent pipeline.
So, is AI value real or overstated?
Again, the answer is nuanced. There is credible evidence that AI is improving performance at task level. Writing, summarising, coding and customer support all show gains in the 5% to 25% range in various studies. That is real. But task gains are not the same as wholesale job replacement and they are definitely not the same as business transformation.
This is where many organisations are getting ahead of themselves. AI capability is running far ahead of actual workplace adoption. The technology can do more than most organisations have yet embedded into live workflows. That gap between what is theoretically possible and what is operationally real is where a lot of the hype sits.
It is also why leaders need to stop mistaking visible AI activity for proven value. Some employers are already building AI use into performance expectations and hiring criteria. That may sound progressive but it may also be premature if organisations are still unclear on where AI genuinely improves quality, speed or outcomes.
And what about jobs themselves?
Here the smarter lens is job redesign, not simple job loss. AI can replace parts of a role without replacing the role itself. A job is not just a list of tasks – it is judgement, accountability, relationships, edge cases and context bundled together. If AI takes away one slice of that bundle the role may narrow, shift or be reinvented but it does not automatically disappear.
That is also why the Brookings Institution’s recent work matters. Rather than looking only at which jobs are most exposed to AI, Brookings asked a better question: which workers are equipped to cope if disruption comes? Its analysis found that many highly exposed workers still have relatively strong adaptive capacity. They tend to have more transferable skills, more financial resilience or better access to alternative jobs.
But it also identified a sizeable group with high exposure and low adaptive capacity, concentrated largely in clerical and administrative roles, and mostly women. In other words, exposure alone tells you very little. Two people can sit in similarly exposed roles and face very different futures.
This has big implications for HR and workforce planning. The real issue is who inside exposed jobs has the support, mobility and opportunity to adapt.
So, is AI really going to take your job?
Today’s evidence says: not yet. What AI is doing right now is changing tasks, tightening hiring, reshaping expectations and forcing a more serious conversation about how work should be designed. Some roles will shrink, some will grow and many will change shape.
The bigger unknown is whether the jobs created will be decent, stable and human-centred – or whether we end up with more of the precarious work that has defined much of the 21st century.
The future of work is still being written. This is why it pays to read the fine print, not just the headlines.
👉 Read my full analysis where I look at the latest research and who is most exposed: Is AI really going to take your job 7-minute read
