AI Job Replacement Statistics: Which Jobs Are Most Exposed in 2026

Artificial intelligence is changing how work gets done. Most research says it changes tasks more than it removes whole jobs.

A lawyer can use AI to review contracts and still advise clients. A software developer can generate routine code and still design systems and solve hard problems.

An administrative worker may face more risk if most of the role is repetitive digital work that AI can do on its own. That is why researchers separate exposure from replacement.

The International Labour Organization (ILO) estimates that about one in four workers worldwide holds a job with some exposure to generative AI. Only 3.3% of global employment falls into its highest exposure category.

Early signs of change are also showing up in the data. Stanford’s 2026 AI Index reports that employment among software developers aged 22 to 25 has fallen almost 20% since 2024.

This article covers the latest AI job replacement statistics, the occupations most exposed and where automation may have the biggest effect.

AI Job Replacement Statistics at a Glance

The table below pulls together the main figures from the ILO, the International Monetary Fund (IMF), the World Economic Forum (WEF), Stanford and Anthropic.

---

---

These numbers do not forecast the same number of layoffs. Exposure, automation potential, job displacement and unemployment each measure something different.

Exposure means AI could perform or help with some tasks in a job. Displacement means a job is expected to shrink or end. Unemployment means a person is out of work and looking for a job.

How Many Jobs Could AI Affect?

The broadest estimate comes from the IMF. It says AI could affect about 40% of jobs worldwide.

Exposure is higher in wealthy economies because more workers there do cognitive, digital and professional tasks. The IMF puts exposure at about 60% of jobs in advanced economies.

The gap makes sense when you look at the type of work. Low-income countries have a larger share of farm and manual work, which AI tools reach less easily.

Wealthy countries have more office, finance, legal and technology jobs. Those jobs rely on reading, writing and analysis, which are tasks AI handles well.

Exposure Does Not Mean a Job Will Disappear

The IMF says roughly half of exposed jobs in advanced economies may benefit from AI through higher productivity. The other half could face lower labor demand if AI can do important parts of the role.

The difference comes down to how much a human is still needed. Judges are highly exposed because AI can analyze documents and case law. Still, legal accountability and social norms make it far less likely that judges will be replaced.

Clerical jobs can have similar technical exposure but much lower complementarity. That means AI can do more of the work without a person guiding it, which makes substitution more likely.

One in Four Jobs Is Exposed to Generative AI

The ILO offers a narrower view focused only on generative AI. Its 2025 study looked at nearly 30,000 tasks across 436 detailed occupations.

It found that about 24% of global employment has some degree of GenAI exposure. Only 3.3% falls into the highest exposure category.

Exposure rises with national income. That may seem backward, since earlier waves of automation often hit factory and manual work first.

Generative AI is different because it is good at cognitive and language-based tasks. Accountants, programmers, financial analysts and administrative staff can be more exposed than construction workers, electricians or cleaners.

Which Jobs Are Least Exposed?

Jobs that involve unpredictable physical work remain much less exposed to generative AI. Examples include electricians, plumbers, construction workers, carpenters, care workers, cleaners, farm laborers, mechanics and hospitality staff.

AI can help these workers with scheduling, diagnostics, instructions and paperwork. But a language model cannot fix a leaking pipe.

Full automation of this work would need robotics as well as AI. That is generally more costly and technically harder than automating digital office work.

Which Jobs Are Most Exposed and Which Are Set to Decline?

The ILO continues to find the highest exposure among clerical occupations. The common feature is not low skill. It is digital task structure.

Jobs become easier to automate when much of the work involves reading, writing, summarizing, sorting, searching, moving information or following set rules.

The WEF Future of Jobs Report surveyed more than 1,000 major employers representing over 14 million workers. Employers expect the fastest declines by 2030 in the five roles shown below.

Accountants and auditors, printing workers and other clerical jobs also appear among declining roles. AI is not the only cause.

Digital payments reduce demand for cashiers and bank tellers. E-commerce affects postal and administrative work. Robotics and wider business digitization add to the trend. The WEF still names AI and information processing as one of the three most important technologies driving both growing and declining jobs.

Higher-Skilled Jobs Are Becoming More Exposed

Exposure is also rising in more skilled work. The ILO points to financial analysts, investment advisers, web and multimedia developers, application programmers and media professionals.

Better AI coding, reasoning, image creation, speech and video tools are moving automation beyond office administration. Bookkeeping, tax preparation, contract review, legal research, due diligence and compliance all include tasks AI can now do in part.

Higher-level roles are more likely to be supported than replaced because they involve accountability, client relationships, judgment and regulation. The likely change is fewer hours spent producing information and more time spent interpreting it.

Creative work is changing too. The WEF now lists graphic designers among declining occupations, linked partly to generative AI. Writers, editors, translators and designers also show up in real-world AI usage data.

AI can produce generic marketing copy very cheaply, which puts pressure on commodity content. Original reporting, brand strategy and creative direction are harder to automate because their value depends on expertise and trust.

Will AI Destroy More Jobs Than It Creates?

Current forecasts say no, although they expect major disruption. The WEF estimates that changes in technology, demographics, trade and the green transition could create 170 million jobs by 2030.

In the same period, it expects 92 million jobs to be displaced. That leaves a net gain of about 78 million jobs.

The 92 million figure should not be read as 92 million jobs lost to AI. It covers displacement from several large structural changes, not only AI.

Employers also expect strong growth in some fields. The WEF lists software and application developers among the fastest-growing occupations through 2030, and it expects care, education, delivery and construction roles to add large numbers of jobs.

AI and Robotics: Two Different Job Effects

The WEF also looked at specific technologies. For AI and information-processing technologies, it estimates about 11 million jobs could be created and about 9 million displaced by 2030.

That is a small net gain of about 2 million jobs. Robotics and autonomous systems look different.

For robotics, the WEF expects about five million more positions to be displaced than created. Robotics is therefore expected to be a bigger net job destroyer than AI over this period.

The result fits the physical nature of the work. Robots affect warehouse, factory and delivery tasks, while AI mostly affects digital tasks that already run on computers.

Early Employment Effects Are Appearing Among Young Workers

The most important new evidence comes from real employment data rather than forecasts. Stanford’s 2026 AI Index highlights research showing that employment for software developers aged 22 to 25 has fallen almost 20% from 2024 levels.

Young customer-support workers show a similar pattern. Employment among many older workers in the same occupations has held up better.

Researchers also compared workers aged 22 to 25 across jobs with different AI exposure. After controlling for changes within firms and industries, employment in the most exposed jobs fell about 16% relative to the least exposed jobs.

The gap began to show in 2024 and continued through 2025. This does not prove that AI caused every lost job.

Technology companies have also faced higher interest rates, slower hiring and post-pandemic restructuring. Still, the declines are concentrated among young workers doing highly exposed tasks, which supports the view that AI is starting to change hiring choices.

Why Entry-Level Roles Face Risk First

Junior staff often do the tasks that today’s generative AI handles well. A senior lawyer is still needed to advise a client, but a firm may need fewer junior lawyers to review thousands of documents.

The likely result is not the end of whole professions. It is narrower hiring pipelines.

That could cause a longer-term problem. If firms hire fewer junior workers, industries need new ways to train future senior staff.

Software Developers: Exposed but Still Growing

Programming is one of the most common real-world uses of AI. Anthropic’s Economic Index found AI use is heavily concentrated in computer and mathematical work, especially coding and debugging.

Yet the WEF still lists software and application developers among the fastest-growing jobs through 2030. AI may reduce the number of workers needed for a set amount of software while also making software cheaper and raising demand.

Customer Service Is Easy to Automate

Customer-support work has many repeat patterns. AI systems can answer common questions, find account details, troubleshoot basic problems, summarize chats, draft replies and pass hard cases to humans.

People are still needed for angry customers, unusual cases, negotiation and decisions that need discretion. Even so, a company that once needed 1,000 agents may handle the same volume with fewer people backed by AI.

Automation Versus Augmentation

How people actually use AI offers some of the best evidence against a simple replacement story. Anthropic’s initial Economic Index found that about 57% of AI usage involved augmentation and about 43% involved automation.

Augmentation means a person and AI work together, such as asking AI to suggest code, improve writing or explain data. Automation means the task is handed to AI more directly.

Newer Anthropic research finds AI appears in the task mix of nearly half of sampled occupations. But effective AI coverage can differ a lot once researchers look at how well AI completes tasks and how much working time those tasks take up.

That matters because a job where AI can help with 70% of tasks is not a job where AI can do 70% of the work alone.

Companies Expect More Workforce Reductions

Employers increasingly expect AI to affect staffing. Stanford’s 2026 AI Index found that 32% of organizations expect AI to decrease their workforce over the next year.

Another 13% expect AI to increase staffing, 43% expect little or no change and 12% are unsure.

Expected reductions are highest in service operations, supply chain and software engineering. Reductions reported so far are much lower than what employers expect in the future.

This suggests the largest labor-market effects may still be ahead.

Women Face Higher Generative AI Exposure

The ILO also found a gender gap. Globally, about 4.7% of female employment falls into the highest GenAI exposure category, compared with 2.4% of male employment.

The gap is much larger in high-income economies, at 9.6% for women and 3.5% for men. The main reason is how jobs are spread.

Women are more heavily represented in clerical and administrative work, where generative AI exposure is especially high.

The Bottom Line: Job Transformation, Not Mass Unemployment

Current evidence does not support claims that AI has already caused mass unemployment. Stanford notes that large job losses have not yet appeared in overall employment data, despite early effects among younger workers and exposed jobs.

The ILO also says job transformation is more likely than full replacement for most exposed workers. The effects are expected to show up in several ways.

This process can still disrupt many lives even if total employment keeps growing. About one-quarter of workers hold jobs with some generative AI exposure, and the IMF puts broader AI exposure at about 40% of global employment.

Clerical and administrative jobs face the greatest risk of substitution. Data entry, bookkeeping, secretarial work, bank clerks, cashiers and routine customer support are the most exposed.

The WEF forecasts more jobs created than displaced through 2030. The ILO expects transformation to be the main effect.

The key question is not whether AI will replace jobs. It is which tasks AI will take over, how many people are still needed afterward, and how fast workers can move into the new work that grows around the technology.

Sources

• International Labour Organization: Generative AI and Jobs: A Refined Global Index of Occupational Exposure
• ILO: Generative AI and Jobs: 2025 Update
• ILO: How Might Generative AI Impact Different Occupations?
• Stanford University: 2026 AI Index Report, Economy
• International Monetary Fund: AI Will Transform the Global Economy. Let’s Make Sure It Benefits Humanity
• IMF: Leveraging Artificial Intelligence and Enhancing Countries’ Preparedness, 2026
• World Economic Forum: Future of Jobs Report 2025
• World Economic Forum: Jobs Outlook
• Anthropic Economic Index
• Anthropic: Introducing the Economic Index
• Anthropic: Economic Index: New Building Blocks for AI Use