92 Million Displaced. 170 Million Created. The Number Nobody Wants to Sit With Is in the Middle.

What the Data Actually Shows

Start with what is already happening. Between January and June 2025, company globally reported 77999 tech job losses directly attributed to AI adoption. That is not a projection. That is a confirmed figure from the first half of last year, and it represents hundreds of people losing jobs to automation every single working day. AI was directly linked to 4.5% of all job losses reported across 2025. Companies in the United States using ChatGPT report that 49% of them have already replaced workers as a result.

The International Monetary Fund estimates that 40% of jobs worldwide are currently exposed to AI, meaning AI could already perform a significant portion of what those roles involve. In advanced economies, the IMF puts that figure at 60%. Goldman Sachs' analysis goes further, estimating that approximately 300m full time jobs globally could be affected by generative AI, though this uses a broader definition that include significant task change rather than outright elimination. The WEF 92 million figure and Goldman 300m are measuring different thing, which is why they look so different. Neither is wrong. They are answering different questions about the same underlying transformation.

The McKinsey Global Institute adds a third dimension: by 2030, at least 14% of employees globally could need to change not just their tasks but their entire careers because of digitisation, robotics and AI. That is not a number about jobs disappearing. It is a number about the inadequacy of current skills for the jobs that will exist. The WEF own report find that 39% of existing skill set will become outdated between 2025 and 2030, down from 57% in 2020 but still representing a fundamental shift in what competency looks like across industry.

Which Jobs face the highest risk

Clerical and secretarial roles sit at the top of nearly every displacement risk list. The WEF identifies these as the fastest declining category. Administrative assistants, data entry clerks, accounts payable processors, legal researcher and medical transcriptionists are all roles where AI can already do most of what the job requires faster, cheaper, and without lunch breaks.

Paralegals face an 80% automation risk by 2026 according to one analysis. Legal researcher face a 65% risk by 2027. Medical transcription is already 99% automated. In banking, major institutions are projecting average workforce reductions of around 3% driven by AI, while 80% of bank executives expect a 5% productivity boost from the same technology. The productivity gain and the job loss are two sides of the same event.

The picture for manufacturing is different in character but similar in direction. MIT and Boston University research indicates that AI driven robotics will have replaced approximately 2 million manufacturing workers globally by the end of 2026. Global manufacturing could lose 20 million jobs to automation by 2030. These are not knowledge workers in offices. These are people doing physical work that machines are increasingly capable of replicating at lower cost and higher consistency.

Research from the University of Pennsylvania and OpenAI introduced a counterintuitive finding that has been widely discussed since its publication: educated white collar workers earning up to 80,000 dollars a year are among the most likely to be affected by workforce automation. This is not what most people assumed when they thought about AI and jobs. The conventional assumption was that AI would take the routine, lower paid work and leave the skilled, higher paid work alone. The University of Pennsylvania and OpenAI research suggests the opposite is partly true: the more codifiable a role is, regardless of how much education it required to reach it, the more vulnerable it is to AI replacement.

What AI Is Creating and Who Gets Those Jobs

AI and machine learning specialists, data analysts, cybersecurity professionals and renewable energy technicians are all growing faster than any other category in the labour market right now.

AI and ML engineer roles grew 41.8% annually in 2025. The jobs AI is creating pay well, grow quickly, and reward the people who can work alongside the technology rather than those who compete with it.

The problem is access. Getting from a clerical role that AI is automating to an AI engineering role that AI is creating requires education, retraining, time, money, and in many cases a relocation. The WEF finds that 63% of employers identify skills gaps as the primary barrier to business transformation, and that 85% plan to prioritise workforce upskilling in response. Almost half of employers expect to transition staff exposed to AI disruption into other parts of their business rather than simply letting them go. The intention is there. Whether the execution matches the intention, across millions of companies and hundreds of millions of workers, in the five years between now and 2030, is the central uncertainty that no projection can resolve.

Mental health professionals. Surgeons. Social workers. Primary school teachers. Skilled tradespeople who diagnose and fix problems in physical environments where every situation is slightly different. Research by Harvard Business School Suraj Srinivasan, published in February 2026, finds that employers are seeking more AI related skills in some fields while demand for structured and repetitive tasks is declining. The roles growing fastest are those where AI augments human capability rather than replaces it, where the human being in the role is doing something the AI feeds into rather than something the AI renders redundant.

WEF Managing Director Saadia Zahidi described the situation plainly in the 2025 report launch: the landscape of work is evolving at a pace that requires collective action in the public, private and education sectors to address growing skills gaps. That is not a warning about some future event. That action is needed now, for a transition that is already underway in every major economy.

The Honest Answer to the Question

Will AI replace jobs? Yes. It is already doing it, in specific roles, in specific industries, in specific countries, at a pace that is visible in the job loss data being reported every quarter. Will AI create jobs? Yes. It is already doing that too, faster than almost any other technology in economic history, in roles that pay better and grow more quickly than the ones they are replacing.

The net number, 78 million more jobs in 2030 than there are now, is probably roughly accurate as a macro projection. It is also nearly useless as a guide to what any individual worker, employer, educator or government should do right now. Because the transition between the 92 million displaced and the 170 million created is not a smooth escalator. It is a gap that requires skills, investment, time and policy support that do not automatically materialise because the net figure looks positive.

Goldman Sachs noted that over 85% of employment growth in the United States since 1940 has come from technology driven job creation, and that 60% of workers today are in occupations that did not exist in 1940. Every major technology transition in modern economic history has produced net job growth while causing genuine, painful disruption for the workers caught in the middle of it. AI is not an exception to that pattern. It is the latest, fastest, and most broadly distributed version of it.

The question is not whether AI will replace jobs. It already is. The question is whether the systems built to manage that transition, the education systems, the retraining programmes, the social safety nets, the corporate upskilling investments, are moving fast enough to catch the people the transition will otherwise leave behind.

With 39% of current skills projected to become outdated by 2030 and 63% of employers already citing skills gaps as their primary barrier to growth, is the pace of workforce transformation currently fast enough to match the pace of the AI adoption driving it?