Two dishonest camps dominate the AI and jobs conversation.
The first camp — usually composed of AI company executives and economists with think-tank positions — insists that AI will create more jobs than it destroys, just as every previous technological revolution did. They point to the industrial revolution, the computer era, the internet. New technologies create new categories of work. Net positive. Nothing to see here.
The second camp — a coalition of anxious workers, populist politicians, and contrarian journalists — insists that this time is fundamentally different, that AI will eliminate most knowledge work, and that mass unemployment is imminent unless governments intervene dramatically.
Both camps are wrong. Here is what the evidence actually shows.
What AI Is Actually Replacing Right Now
AI is not replacing jobs wholesale. It is replacing tasks. This distinction matters enormously and gets lost in almost every popular discussion of the topic.
A customer service representative's job involves dozens of tasks: responding to common queries, escalating complex issues, managing difficult emotions, identifying sales opportunities, navigating internal systems, documenting interactions. AI can handle the first task reliably. The others? Not yet, and for some of them, not soon.
The result is not that customer service representatives are being fired en masse. The result is that each representative handles more calls per day. Headcount grows more slowly, or stays flat while output grows. That is a real economic impact on employment — just not the dramatic "robots taking jobs" narrative.
The Roles That Are Actually at High Risk
Honest analysis identifies specific task profiles — not job titles — that AI genuinely threatens. Roles that involve: primarily processing and reformatting structured information, following explicit rules to reach binary decisions, producing standardised written content at high volume, translating between data formats.
These task profiles exist across many job titles: junior data analysts, entry-level copywriters, basic legal document review, standard financial reporting, routine code documentation. People in these roles need to be developing skills in adjacent areas now, not when the displacement arrives.
The Roles That AI Makes More Valuable
Equally important and less discussed: AI significantly increases the economic value of people who can direct, evaluate, and apply AI outputs. The prompt engineer, the AI output editor, the person who knows enough about a domain to tell when the AI is confidently wrong — these people are more productive and therefore more valuable with AI than without it.
The doctor who uses AI diagnostics to see more patients is more valuable than the doctor who does not. The developer who uses AI coding assistance to ship features three times faster commands a higher salary. The analyst who uses AI to process a thousand documents in the time it used to take to process ten produces more insight per hour.
What Governments and Companies Should Actually Do
The policy response to AI displacement should be retraining infrastructure and portable benefits, not technology restrictions. You cannot slow AI adoption without sacrificing the economic gains that adoption creates — gains that fund schools, hospitals, and social safety nets.
Companies have an obligation to their workforces that most are not meeting: proactive retraining, honest communication about which roles are at risk, and genuine investment in helping people develop the skills that AI cannot replicate.
The Honest Bottom Line
Some people will lose jobs because of AI. This is happening now and will accelerate. The dishonest comfort of "it will all work out" does not serve those people.
Equally, the catastrophist narrative of mass unemployment does not match the evidence. What matches the evidence is significant structural disruption, concentrated in specific task profiles, requiring real and rapid adjustment — by workers, companies, and governments — that most are not yet taking seriously enough.
The truth is uncomfortable and requires action. That is exactly why so few people want to say it out loud.
