Two Things Are True at Once
Walk into most modern workplaces today and you will find AI quietly doing work that used to take a team an afternoon - drafting emails, summarising meetings, writing first-pass code, generating marketing copy. Ask the people doing that work how they feel about it and you will get two very different answers depending on who you ask. Some will tell you it has made their job easier and more creative. Others will tell you it has made their job precarious, or has already replaced it entirely. Neither group is wrong.
Where AI Is Genuinely Helping
For a large number of knowledge workers, AI tools have become something closer to a very capable junior assistant than a replacement. Developers use AI coding assistants to handle boilerplate and catch bugs, freeing up time for the architectural thinking that actually needs a human. Doctors use AI transcription and summarisation tools to spend more face-to-face time with patients instead of typing notes. Small business owners who could never have afforded a marketing team now use AI to draft campaigns, translate content, and analyse customer feedback. For people in under-resourced roles - a single HR person doing the work of five, a solo shop owner managing accounts and inventory - AI has been a genuine equaliser.
Where AI Is Genuinely Hurting
At the same time, entire categories of entry-level work are shrinking. Roles built around repetitive, well-defined tasks - basic copywriting, first-line customer support, simple data entry, junior-level translation - are the easiest for AI systems to absorb, and companies under margin pressure are absorbing them quickly. This is not a hypothetical future problem. Hiring for several entry-level knowledge-work categories has already softened in recent years, and new graduates in some fields are finding it harder to get the first job that used to be the standard way into a career. The people hit hardest are rarely senior decision-makers - they are the workers just starting out, who no longer get the chance to learn on the job the way earlier generations did.
The Middle Is Where It Gets Complicated
Global labour research bodies have tried to put numbers on this shift. Widely cited industry projections suggest tens of millions of jobs will be displaced by AI and automation this decade, while a larger number of new roles - many of them not yet clearly defined - will be created in the same period, implying a net positive for total employment. That net figure is real, but it hides the transition cost. A displaced customer service agent does not automatically become an AI systems trainer. The new jobs tend to require different skills, sometimes in different cities, often with a gap in between where a real person has to pay real bills.
The Skills Holding Up Best
Work that depends on physical presence, hands-on trust, or emotional judgement - nursing, skilled trades, teaching, therapy, in-person sales - has proven far more resistant to automation than most office-based knowledge work. So has work that requires taking responsibility for a decision, not just producing an output: a doctor signing off on a diagnosis, a lawyer deciding how to argue a case, a manager deciding what to prioritise. AI can draft the options. It is much worse, so far, at being accountable for the choice.
What This Actually Means for You
If your job involves producing a fairly predictable output from fairly predictable inputs, it is worth actively learning how to use AI tools well before your employer decides to use them instead of you. If your job depends on judgement, relationships, or physical skill, the near-term risk is lower - but the tools are still worth learning, because the people who use AI well are increasingly out-competing the people who refuse to touch it. The uncomfortable truth is that "AI won't take your job, but someone using AI might" has become less of a slogan and more of a description of what is already happening.
