Subheadline: Updated and expanded for late 2025: AI is no longer “coming for jobs” in theory, it’s quietly rewriting tasks, career ladders, and bargaining power in practice. The question is not only what disappears, but what kind of work survives.
By Carlos Taylhardat | 3 Narratives News | December 29, 2025
Editor’s note (revisited): We first published an earlier version of this piece on May 22, 2025. We have revisited it, expanded it, and updated the sources to reflect where AI actually is now, including 2025 labor projections, new exposure research, and the first real bite of workplace AI rules in Europe.
Intro
Matías S. Zavia thought he had one of those modern jobs that would always exist. Not glamorous, not viral, but quietly essential. He translated. He adapted tone. He made English stories sound like they were born in Spanish, not poured through a machine.
Then Gizmodo’s Spanish-language operation was replaced by automated translation. His role, and the human judgment inside it, became a “cost center.” The company moved on. The internet barely paused. Zavia did what displaced workers have done for centuries, he tried to figure out whether this was a temporary shock or the beginning of a different economy entirely. The Verge and Ars Technica covered the shift, and Business Insider told Zavia’s story in blunt, human terms.
Now zoom out. That same replacement logic is spreading, not only through translation and media, but through customer service, marketing, legal support work, basic coding, scheduling, billing, and everything that used to be “entry-level” office oxygen.
And yet, at the same moment, companies are hiring new kinds of people, paying startling salaries for AI safety and reliability roles, and promising that the real story is productivity, not unemployment. The fight is not only about jobs. It’s about who gets leverage in the next decade.
Context
There are two facts that can both be true. First, AI is already changing work, mostly by automating pieces of jobs rather than deleting entire occupations overnight. Second, this can still devastate people, because a life is not made of “pieces,” it’s made of rent, routine, pride, and the fragile confidence that you are still needed.
In 2025, the World Economic Forum surveyed over 1,000 large employers and projected major churn by 2030: 170 million new roles created, 92 million roles displaced, a net gain of 78 million, paired with an urgent warning that upskilling is not optional anymore. World Economic Forum (Future of Jobs 2025 press release).
The IMF has warned that AI could affect around 40% of jobs worldwide, sometimes replacing tasks, sometimes complementing them, often widening inequality if policies do not catch up. IMF (Georgieva, 2024).
Meanwhile, the International Labour Organization has refined its measurement of which jobs are exposed to generative AI and keeps landing on a point that makes both optimists and pessimists uncomfortable: the dominant effect is often augmentation, not full automation, but augmentation can still mean layoffs if one worker with tools replaces three without them. ILO working paper (2025).
Regulators are finally stepping into the room. The EU’s AI Act entered into force in 2024, with workplace-relevant obligations rolling out in phases. In 2025, bans and literacy obligations began applying, and general-purpose AI model obligations came online in August. European Commission (AI Act timeline).
So the question we are revisiting is not “Will AI replace workers?” That phrasing is too clean. The real question is: What happens to a society when the ladder rungs change faster than people can climb?
Narrative 1 (Side A): AI as a catalyst for innovation and better work
In this worldview, the story is not a theft, it’s a leap. The same way electricity reorganized everything without “ending work,” AI reorganizes work without ending ambition. The biggest misunderstanding, the optimists argue, is imagining jobs as fixed objects instead of living systems.
They point to what AI already does well in 2025: drafting, summarizing, translating, scheduling, spotting anomalies, generating code scaffolds, and reducing friction that used to eat whole afternoons. Many organizations are not replacing workers, they are compressing the boring parts of jobs so humans can spend more time on judgment, relationships, and creative decisions.
The OECD’s 2025 research on small and medium-sized businesses contains a detail optimists love: most SMEs reported that generative AI had no effect on overall staff needs so far, and only modest shares reported staffing increases or decreases. In this telling, companies are experimenting cautiously, not swinging an axe. OECD (Generative AI and the SME Workforce, 2025).
Optimists also argue that the new jobs are real, not speculative. Not just “machine learning engineers,” but the sprawling ecosystem around AI: model evaluation, safety testing, incident response, data governance, AI procurement, workplace training, and compliance. When a company posts a high-stakes role to assess catastrophic AI risks, it’s a signal that the technology has moved from novelty to infrastructure. Business Insider (Dec. 2025).
Then there is robotics, the physical form of the same argument: if AI can move from text to hands, whole categories of dangerous and punishing labor can shrink. Elon Musk has said Tesla’s humanoid robots could become “the biggest product ever,” and Reuters has reported his claims that robots could eventually sell for roughly the price of a used car. In the optimistic narrative, this is not a nightmare, it’s liberation, fewer injuries, fewer broken backs, fewer night shifts that grind people down. Reuters (Oct. 2024).
And this side has a moral argument: technology has always displaced some work while raising the ceiling for what humans can build. The goal is not to freeze the economy in amber. The goal is to steer the transition so the benefits show up as shorter workweeks, higher productivity, better services, and new paths for people whose talent was previously trapped in repetitive tasks.
If you want to see how we think about AI as a tool rather than a replacement in publishing, read our related explainer: Truth and Lies About AI Assistance in the Newsroom.
Narrative 2 (Side B): AI as a threat to livelihoods, wages, and dignity
In this worldview, the optimistic story skips the part where people get hurt. The average worker cannot “pivot” as quickly as a boardroom can sign a new software contract. The question is not whether society becomes richer overall, it often does, the question is who absorbs the shock.
Skeptics point out that when companies say “AI will augment workers,” the next sentence is often “so we can do more with fewer people.” Goldman Sachs estimated that generative AI could expose the equivalent of 300 million full-time jobs globally to automation, not necessarily all eliminated, but meaningfully disrupted. Goldman Sachs research (2023 PDF).
They also argue that “jobs created” statistics can be emotionally dishonest. A newly created AI compliance job in a major city does not help the translator in another country who was replaced by an automated pipeline, or the office worker whose role is reduced to proofreading machine output for less pay. Work does not only provide income, it provides identity. When people are “retrained,” they often lose status as well as stability.
There is also a trust crisis. Newsrooms that quietly publish AI-generated work, then issue corrections or face plagiarism allegations, teach the public that “automation” is not neutral. It is a management choice with consequences. The CNET episode became a cautionary tale about speed, errors, and transparency, and it helped fuel labor pushback inside media organizations. The Washington Post (2023).
Then there is surveillance. Workers are not only being replaced, they are being measured more aggressively, sometimes by AI systems that pretend to read emotion, attention, or intent. Europe’s AI rules explicitly move against parts of this future, including bans on certain manipulative or emotion-tracking practices in workplaces, because the risk is not just unemployment, it’s a workplace where humans are treated like data exhaust. European Commission (AI Act).
And the deepest fear is inequality. The IMF’s warning is not abstract: if higher-income workers get “complemented” by AI while lower-income workers get “replaced,” the wage gap becomes policy, not accident. The story stops being about technology and becomes about power, who owns the tools, who captures the productivity gains, and who is told to be grateful for the opportunity to adapt. IMF (Georgieva, 2024).
If you want the darker, broader version of this debate, see our larger 2025 explainer: ChatGPT: Humanity’s Greatest Tool, or Its Most Dangerous Experiment?
Narrative 3 (Silent Story): The missing rung problem
Both narratives talk about the destination. The silent story is about the staircase.
Most people do not enter the middle class through genius. They enter through ladders, internships, junior roles, assistant positions, and the patient accumulation of competence. AI’s most disruptive effect may be that it removes the easiest “practice jobs,” the ones where you learn by doing the simple parts before you earn the right to do the complex parts.
A world where AI drafts every memo and designs every first-pass presentation may sound efficient. But it raises a quiet question: where do new professionals learn to write, think, and persuade if the apprenticeship layer disappears?
The same logic applies everywhere. Junior developers learn by fixing small bugs. Young journalists learn by reporting small stories. Junior paralegals learn by reading and summarizing documents. Translators learn by doing the hard, human work of tone. If AI takes the first rung, society may still have “jobs,” but fewer pathways into them.
This is why policy and culture matter as much as technology. AI literacy is becoming a baseline expectation. Worker protections are being rewritten in real time. Some governments are regulating harmful uses. Some companies are quietly restructuring work without telling the people underneath it. The future will not be decided by a model release. It will be decided by whether societies build bridges for ordinary workers across the transition.
And if we want a sober conclusion, it is this: the AI debate is not a single argument. It is a thousand local stories of people trying to remain useful in an economy that rewards speed more than stewardship.
Key Takeaways
- This story is updated and expanded from our May 22, 2025 version, with 2025 labor projections and regulatory developments.
- Major forecasts suggest huge churn by 2030, not a simple job apocalypse, but rapid disruption that demands upskilling.
- Research increasingly shows AI will augment many jobs, yet augmentation can still drive layoffs through productivity compression.
- Governments are beginning to regulate workplace AI, including limits on manipulative and emotion-tracking practices in Europe.
- The overlooked risk is the missing rung problem, AI removing entry-level tasks that used to train the next generation.
Questions This Article Answers
- Is AI actually replacing jobs right now, or just changing tasks?
In most sectors, AI is first changing tasks, compressing time, reducing junior workloads, and increasing output expectations. In some areas, like translation and basic content production, it has already contributed to direct displacement.
- What do major forecasts say about jobs by 2030?
Leading employer surveys project significant churn: many roles created, many displaced, and a large net change that depends on how quickly workers can reskill and how companies redesign work.
- Which workers are most at risk?
Jobs with high volumes of routine digital tasks are often most exposed. But exposure does not always mean elimination, it can also mean wage pressure, role shrinkage, or new monitoring.
- What does regulation have to do with the job story?
Regulation shapes which AI uses are allowed in workplaces, what transparency is required, and whether workers can contest harmful automated decisions.
- What is the “missing rung problem” and why does it matter?
If AI removes the beginner tasks that used to teach people skills, entry paths into middle-class careers narrow. That can be more destabilizing than any single round of layoffs.
Alt text: A late-night office scene showing a worker at a desk facing a glowing AI interface, suggesting tension between human labor and automation.
Process & AI-Use Disclosure
Process & AI-Use Disclosure: This article was researched and written by a human editor. AI tools were used to help locate public sources, cross-check dates, and structure the update, then the final wording was rewritten and verified by the editor. Our AI policy is published at How We Use AI. If we make an error, we will correct it transparently in our Corrections log.
