Every two years, the World Economic Forum publishes a document that large companies, government policymakers, and workforce development organizations treat as a benchmark for planning: the Future of Jobs Report. It synthesizes surveys of hundreds of major employers across dozens of countries and industries to estimate how automation, AI, demographic shifts, and other macro-forces are changing what work exists and what skills it requires.
The report is not a prediction of a fixed future — it is a structured synthesis of current employer expectations, which are themselves shaped by technology deployment timelines, economic conditions, and strategic choices that remain uncertain. But it has proven consistently directional: the trends it identifies tend to be underway, even when specific timelines slip.
This article explains what the report is, what the major findings across its editions have established, what the 2025 report adds, how researchers outside the WEF understand the same dynamics, and what any individual should do with this information.
What the Future of Jobs Report Is
Background and Methodology
The World Economic Forum launched the Future of Jobs Report in 2016, with subsequent editions in 2018, 2020, 2023, and 2025. Each edition surveys executives and HR leaders at large companies — typically those with over 50,000 employees — covering topics including:
- Which job categories they expect to grow or decline over the next five years
- Which skills they expect to become more or less important
- What share of tasks they expect to be automated
- What their plans are for reskilling and workforce transitions
- What barriers they face in preparing their workforces
The 2025 edition surveyed employers representing more than 14 million workers across 55 economies and 22 industry clusters — the broadest coverage of any edition to date.
The data is supplemented by partnership with LinkedIn (for skills demand data derived from job postings and career transitions), Coursera (for learning trend data), and several specialized labor market research firms. This triangulation of survey data, job posting analysis, and actual learning behavior provides a richer picture than any single source.
What It Measures and What It Does Not
The report captures employer intent and expectation, not actual labor market data. It is forward-looking in a way that administrative employment statistics are not. This is its primary value: labor market statistics describe what has already happened, while the Future of Jobs Report describes what employers currently plan and anticipate.
Its limitation is the same: employer expectations are subject to systematic biases. Executives consistently overestimate the pace of technology adoption (automation projections have historically been optimistic about timelines), and they underestimate resistance from workers, regulatory constraints, and implementation challenges.
Carl Benedikt Frey and Michael Osborne of the University of Oxford made the most widely cited prediction about automation risk in their 2013 paper "The Future of Employment," estimating that 47% of U.S. occupations were at high risk of automation. That figure was subsequently critiqued by Melanie Arntz, Terry Gregory, and Ulrich Zierahn (OECD, 2016), who argued that the Frey-Osborne analysis overestimated automation risk by treating occupations as monolithic bundles rather than disaggregating the tasks within each occupation. Their task-level analysis produced a much lower estimate: approximately 9% of jobs in OECD countries at high automation risk. Both analyses were directionally informative; neither was precisely correct. The WEF approach acknowledges this uncertainty more explicitly than most.
The directional findings of the WEF report are more reliable than the specific timelines. When the 2016 report identified AI and machine learning specialists as a fast-growing role, that was correct. When it suggested that 65% of children then entering primary school would end up in jobs that did not yet exist, that figure was speculative and has not been validated.
The Core Findings Across Editions
Jobs Are Transforming Faster Than They Are Disappearing
A persistent and important finding across all editions is that job transformation is more significant than job elimination. Most jobs are not being replaced wholesale by automation; rather, the mix of tasks within jobs is changing. Tasks that are routine, predictable, and data-processing-intensive are shifting toward machines, while the remaining tasks become more heavily weighted toward human judgment, social interaction, and non-routine problem-solving.
The 2023 report estimated that 44 percent of workers' core skills would need to change by 2027. The 2025 edition raised this further, projecting that 40 percent of core skill sets will be disrupted by AI alone in the same timeframe.
This distinction matters for how we think about workforce policy. "Automation is coming for your job" suggests displacement as the primary challenge. The more accurate framing from the data is: "The nature of most jobs is changing substantially, and the skills required for them are shifting faster than they have in previous technological transitions."
MIT economist David Autor has tracked this dynamic for two decades. His research on labor market polarization (Autor, Levy, and Murnane, 2003; Autor, 2019) documents how automation consistently displaces middle-skill, routine-intensive work while increasing demand for both high-skill cognitive work and low-skill manual work that resists automation. The result is a hollowing out of the middle of the wage distribution — jobs that paid middle-class wages in manufacturing and clerical work have been partially automated, pushing workers toward either upskilling into higher-wage positions or accepting lower-wage service work. The WEF findings are consistent with Autor's longer-running analysis.
The Growing vs. Declining Job Categories
Growing roles across editions consistently cluster in three areas:
Technology: Software developers, AI and machine learning specialists, data analysts, cybersecurity professionals, cloud computing engineers, and increasingly, AI prompt engineers and AI trainers.
Green transition: Renewable energy engineers, sustainability specialists, environmental engineers, and electric vehicle specialists reflect the expected labor demand from energy transition investments. The 2025 edition specifically identifies electrification and energy storage engineers as new high-growth categories not prominently featured in earlier editions.
Caregiving and social services: Nurses, social workers, personal care workers, and education professionals are consistently projected to grow, driven by demographic aging in high-income countries and persistent labor demand that automation has not significantly disrupted.
Declining roles cluster around tasks that are information-processing, routine clerical, and transaction-based:
- Bank tellers and cashiers
- Data entry clerks
- Postal workers
- Administrative assistants in document-intensive roles
- Accounting clerks
The 2025 report identifies clerical and secretarial roles as the single largest category of decline, with approximately 7.3 million fewer jobs in these roles expected by 2030. The arrival of capable large language models — which can draft, summarize, and manage correspondence effectively — has accelerated this projection substantially compared to earlier editions.
"The labor market is not being destroyed by technology. It is being pulled in two directions at once: toward high-skill, high-judgment roles that leverage technology, and toward high-contact, human-presence roles that technology cannot replicate. The middle is being hollowed out." — David Autor, MIT, paraphrasing the consensus in labor market polarization research
The Skills That Matter Most
The 2025 Top Skills List
The 2025 Future of Jobs Report's ranking of skills by employer priority shows a striking pattern:
| Rank | Skill | Category |
|---|---|---|
| 1 | Analytical thinking | Cognitive |
| 2 | Resilience, flexibility, agility | Personal attributes |
| 3 | Leadership and social influence | Social |
| 4 | Creative thinking | Cognitive |
| 5 | Motivation and self-awareness | Personal attributes |
| 6 | Technological literacy | Technical |
| 7 | Empathy and active listening | Social |
| 8 | Talent management | Social |
| 9 | Service orientation | Social |
| 10 | AI and big data | Technical |
The pattern is notable: only two of the top ten are specifically technical skills. The remaining eight are cognitive and social-emotional capabilities that have appeared on these lists for years. What has changed is the urgency: employers now list AI and big data as skills they expect to prioritize, whereas earlier editions did not, and they are simultaneously emphasizing that social-emotional capabilities are becoming relatively more valuable as routine cognitive tasks are increasingly automated.
How This Compares to Historical Skill Demand
Frank Levy and Richard Murnane (2004, The New Division of Labor) introduced a framework distinguishing between routine tasks that follow explicit rules — and are therefore automatable — and non-routine tasks requiring pattern recognition in ambiguous situations, which remain human-dependent. Their framework predicts the exact skill shifts the WEF data documents two decades later: as routine cognitive tasks automate, non-routine judgment, creativity, and interpersonal skills become relatively scarcer and more valuable.
The persistence of social-emotional skills at the top of the WEF list reflects something deeper than current technology limitations. Amy Edmondson of Harvard Business School has documented that organizations perform better when their members trust each other, communicate openly, and take interpersonal risks. These capabilities are valuable independent of any technological shift — and they become organizationally critical precisely as other work becomes more automated and interdependent.
Skills Hardest to Automate
The question of which skills AI and automation cannot easily replicate is central to individual career strategy. The consensus across the WEF report, Oxford's Frey and Osborne (2013), McKinsey Global Institute analysis, and MIT work by Autor and colleagues converges on several categories:
Novel situation reasoning: The ability to reason effectively in genuinely new situations — where no prior example applies directly — remains extremely difficult for current AI. Large language models are trained on past text; they extrapolate patterns rather than reasoning from first principles about genuinely novel configurations.
Social and emotional intelligence: The ability to read emotional context, build trust, navigate conflict, and provide the specific kind of care that humans want from other humans remains resistant to automation. This matters not just for social work and therapy but for management, sales, and any role where the interpersonal relationship is part of the value being delivered.
Complex judgment under ethical uncertainty: Situations requiring genuine moral reasoning, weighing competing values, and taking responsibility for judgment calls in ambiguous situations are poorly handled by AI systems, which lack the accountability and contextual judgment that such decisions require.
Embodied and physical craft skills: A surprising finding from multiple analyses is that the automation of physical craft — skilled trades, surgery, fine arts and crafts — is slower than the automation of knowledge work. The robotic systems required to replicate fine motor skill, situational physical adaptation, and haptic precision are not yet economically competitive with human labor for many tasks.
"The skills most protected from automation are not necessarily the most prestigious or highly compensated today. A skilled plumber, electrician, or surgical nurse works in a more automation-resistant domain than many white-collar roles that are more susceptible to large language models."
The Skills Half-Life Problem
What Skills Half-Life Means
The concept of skills half-life describes the rate at which a given skill's relevance in the labor market decays. The IBM Institute for Business Value has estimated that the half-life of a professional skill — the time before half of what you know is obsolete or insufficient — has contracted from approximately 30 years in the 1980s to approximately 5 years for many technical skills today.
This figure varies enormously by domain. Skills in rapidly evolving technology areas (specific programming frameworks, particular software platforms) may have a half-life of 2-3 years. Fundamental reasoning skills, interpersonal competencies, and deep domain expertise in slow-moving fields may have half-lives of decades. The aggregate shortening of skills half-life reflects the acceleration of technical change, particularly in digital domains.
The World Economic Forum's own analysis in the 2023 report cited that approximately 44% of workers' core skills will be disrupted within five years — a figure that, when annualized, implies roughly a 7-9% annual obsolescence rate for professional skills. This is dramatically faster than traditional educational investments were designed to accommodate.
The T-Shaped and Pi-Shaped Professional
Organizational learning researchers have responded to the skills half-life problem with structural models for professional competency. The T-shaped professional model — deep expertise in one domain (the vertical bar) combined with broad literacy across multiple adjacent areas (the horizontal bar) — became influential in the 2010s as a framework for navigating rapid change.
More recently, the pi-shaped professional model (two deep vertical competencies with broad horizontal literacy) has gained traction as a more resilient configuration: a professional who is expert in data analysis AND human behavior, for example, can navigate a wider range of changing contexts than one who is expert in data analysis alone.
Google's Project Aristotle data, published by Julia Rozovsky (2016), found that the highest-performing teams at Google were not necessarily teams of pure specialists but teams with varied expertise profiles that could address problems from multiple perspectives. This team-level finding has individual career implications: building a second deep competency is valuable not just for individual resilience but for organizational contribution.
Implications for Education and Training
If technical skills require substantial update every three to five years, the traditional model of education — invest heavily in a credential at age 18-22, then apply that credential for a 40-year career — is structurally inadequate. The credential depreciates too quickly relative to the career length it was meant to support.
The WEF 2025 report estimates that 120 million workers in the world's 12 largest economies may need to be reskilled in the next three years. This is not a problem that any individual organization or education system has the capacity to solve alone. It requires system-level redesign of how learning, credentialing, and career development interact across an entire working life.
Ryan Craig, in A New U: Faster + Cheaper Alternatives to College (2018), argues that the traditional four-year college credential is increasingly mismatched with the speed of labor market change, and that coding bootcamps, apprenticeship programs, and microcredentials represent a structurally more appropriate response to skills that change faster than four-year degree programs can track. The WEF data supports the direction of this argument, though the evidence on the labor market outcomes of specific alternatives to four-year degrees is mixed.
Reskilling vs Upskilling: The Critical Distinction
What Each Means
Upskilling means learning additional skills or deepening existing ones within your current role or occupational domain. A software developer who learns a new language or framework is upskilling. A nurse who gains competency in a new diagnostic technology is upskilling. The core role remains; its requirements expand.
Reskilling means acquiring the skills needed for a fundamentally different role, typically in response to your current role being displaced or substantially changed by automation. A data entry clerk reskilling to become a data analyst is undertaking reskilling. An assembly line worker reskilling to become a maintenance technician for robotic equipment is reskilling.
Reskilling is dramatically more difficult, costly, and uncertain than upskilling. It requires not just skill acquisition but identity transition, credential reacquisition in many cases, and network rebuilding in a new occupational community.
Sociologist Arne Kalleberg (2011) documented in Good Jobs, Bad Jobs that the United States has systematically underinvested in worker transition support relative to comparable economies, leaving individual workers to bear the cost and risk of occupational transitions that are driven by structural economic changes they did not cause and cannot individually prevent. The WEF's employer surveys consistently show that while the majority of employers say they expect to offer reskilling support, the actual investment in reskilling programs is substantially lower than the stated commitment.
The Reskilling Funding Gap
The 2023 and 2025 WEF editions both document a significant gap between the reskilling need and employer investment. Employers report that only 42 percent of employees who need reskilling are expected to be reached by employer programs. The remaining 58 percent will need to self-fund or rely on public programs.
Public reskilling programs have a mixed record. Short-term credential programs for displaced workers show inconsistent labor market outcomes. The programs that show strongest results are those with direct employer partnerships, where training is calibrated to specific job openings and participants receive placement support rather than just instruction.
The most successful large-scale examples share a common feature:
Germany's apprenticeship system (Dual System): Approximately 50% of German youth enter apprenticeships combining classroom education with structured workplace learning. The system produces workers with recognized, industry-verified credentials and employers who have invested directly in their workforce. David Finegold and David Soskice's 1988 analysis of the German system remains the foundational academic treatment; subsequent research has consistently found that the dual system is a source of Germany's unusual combination of low youth unemployment and high manufacturing productivity.
Singapore's SkillsFuture program: Launched in 2015, SkillsFuture provides every Singaporean citizen over 25 with S$500 in skills credits (top-ups provided for workers over 40), a national skills framework mapping competencies across industries, and subsidized mid-career transition programs. The 2023 evaluation found participation rates above 50% for the target working population.
Denmark's flexicurity model: Combines flexible employment relationships with generous social insurance — workers can be hired and fired easily, but comprehensive unemployment benefits (up to 90% of previous salary for two years), active labor market programs, and universal healthcare mean that individual workers are protected from the economic shock of job loss.
All three systems integrate employer, government, and worker investment in a coordinated system rather than leaving the responsibility entirely to any single party. This integration is what the WEF report identifies as the key design feature missing from most countries' workforce development infrastructure.
What the 2025 Report Adds: AI as the Defining Variable
Earlier editions of the Future of Jobs Report identified technology as an important driver of change. The 2025 edition treats AI as the dominant variable, with its own section on AI-specific impacts distinct from general technology adoption trends.
Key 2025 findings specific to AI:
AI augmentation is outpacing AI replacement in near-term employer plans. The majority of employers report planning to use AI to augment workers in their current roles — taking over specific tasks rather than entire jobs — rather than replacing roles wholesale. This finding is consistent with the task-displacement rather than job-displacement pattern documented by Autor and others.
AI literacy is now a top-five skill priority for employers, and organizations are investing in training programs to bring their entire workforces to a functional level of AI tool competency. This is a significant shift from 2023, when AI skills were primarily relevant to technical roles.
The fastest-growing specific job categories in the 2025 edition are AI and machine learning specialists (ranked first), sustainability specialists, and fintech engineers. Interestingly, care economy roles — nursing, social work, education — also appear strongly in the growing category, confirming the bifurcation between AI-leveraging technical roles and human-contact roles.
Where AI-driven displacement is most concentrated: The 2025 report identifies the following as most at risk from AI specifically (as opposed to general automation):
| Role Category | Primary AI Risk | Severity |
|---|---|---|
| Administrative and secretarial | Document generation, scheduling, correspondence | High |
| Customer service (tier 1) | Conversational AI, chatbots, routing | High |
| Basic legal research | Document review, precedent search | Medium-High |
| Financial analysis (routine) | Data aggregation, standardized reporting | Medium |
| Translation (common language pairs) | Machine translation quality | Medium |
| Programming (junior level, standard tasks) | Code generation tools | Medium |
| Creative content (commodity level) | Generative AI for templated output | Medium |
The pattern across high-risk categories is consistent: high volume, low novelty, well-defined outputs, limited judgment requirements. Jobs with these characteristics are more vulnerable than jobs requiring contextual judgment, interpersonal trust, physical presence, or genuinely novel problem-solving.
Country-Level Variation: The Skills Gap Is Not Universal
One of the most important aspects of the WEF data that summary coverage often misses is the significant variation by country, sector, and skill level. The global skills disruption narrative masks very different situations in different contexts.
High-income economies with strong education systems (Northern Europe, Canada, Australia) face the skills challenge primarily as a mismatch between existing credentials and evolving employer needs — a reskilling problem.
Middle-income economies with rapidly expanding labor forces (India, Indonesia, Mexico, Brazil) face a more fundamental challenge: creating enough formal employment to absorb workers moving out of agriculture and informal labor, at a pace faster than automation is eliminating formal jobs. For these countries, the WEF findings about AI risk are relevant but secondary to foundational labor market development.
Low-income economies where most work is informal and where technology adoption is slower face a different timeline: the skills gap is real but less acute than in higher-income countries, because the pace of automation is slower and formal employment itself remains the primary development challenge.
For individuals in high-income economies, the WEF findings are most directly applicable. For workers in developing economies, the relevant policy framework is different and the WEF report should be read with appropriate adjustment for local context.
What Individuals Should Do
Audit Your Role for Task Displacement, Not Job Displacement
Rather than asking "will my job be automated?", ask "which tasks within my job are most likely to be automated over the next five years?" The more accurate framing produces more actionable intelligence. Tasks involving structured data processing, standard document production, and routine information retrieval are candidates for automation in most roles. Tasks involving client relationship management, complex judgment, creative problem-solving, and team coordination are more durable.
Use this audit to identify which aspects of your current role to develop more deeply — the human-judgment-intensive parts — and which to learn to use AI tools to perform better — the automatable parts that will still need to be managed and reviewed by humans.
Invest in Foundational Skills Over Point Skills
The WEF's top skills list consistently reflects a pattern: foundational cognitive and social capabilities (analytical thinking, creative reasoning, communication, adaptability) retain their value across many different technological contexts, while specific technical skills are more volatile.
A developer who deeply understands distributed systems design can adapt to many different implementation languages and frameworks. A marketer who deeply understands consumer psychology and persuasion can operate effectively as the specific tools and channels change. The investment in deep understanding compounds; the investment in specific tool familiarity may depreciate.
Cal Newport, in Deep Work (2016), argues that the ability to focus intensely on cognitively demanding problems — and to continuously acquire hard skills through deliberate practice — is both rare and increasingly valuable. Newport's argument aligns with the WEF data: what protects workers from automation is not the specific things they know, but their capacity to learn, reason, and create at high levels.
Build AI Tool Fluency Now
The 2025 WEF data shows that employers are actively prioritizing AI literacy in their workforce training budgets. Workers who understand what AI tools can and cannot do, how to prompt and evaluate them effectively, and how to integrate them into professional workflows are significantly more productive in roles where these tools apply.
This is not a call to become an AI engineer — that is a specific technical specialization. It is a call to develop working fluency with AI tools that are relevant to your field, which is now table stakes for professional effectiveness in most knowledge work domains.
Take Continuous Learning Seriously as an Operating Principle
The skills half-life data suggests that passive career maintenance — continuing to do what you were trained to do without deliberate skill updating — results in accelerating obsolescence. This is not a permanent source of anxiety to be managed; it is a practical operational reality to be incorporated into professional planning.
Concretely: allocating 5-10 percent of professional time to deliberate skill development — reading, courses, projects outside your current role, mentorship, professional community engagement — is increasingly the difference between staying professionally current and falling behind.
Peter Cappelli of the Wharton School, in Will College Pay Off? (2015), documented that the return on specific educational credentials is declining as skill requirements change faster than credentials track them. His prescription, consistent with the WEF findings, is to prioritize learning capacity and adaptability over credential accumulation — to become the kind of professional who can learn the next thing, not just the professional who learned the right thing at one point in time.
The Future of Jobs Report is most useful not as a source of anxiety about what is coming but as a structured input to professional planning. Its findings are directional, not deterministic. The skills it identifies as growing in importance are skills worth developing. The role categories it identifies as declining are worth noting if you work in them or are considering entering them. And the reskilling challenge it documents at the system level is a genuine policy problem — one that individuals can partially address through their own choices but that ultimately requires systemic solutions in education and workforce development.
Frequently Asked Questions
What is the World Economic Forum's Future of Jobs Report?
The Future of Jobs Report is a biennial publication by the World Economic Forum that surveys hundreds of large employers globally to assess how jobs, tasks, and skills are expected to change over a five-year horizon. It identifies which roles are growing and declining, which skills are becoming more or less important, and what employers plan to do about workforce transitions through reskilling and upskilling programs.
What does skills half-life mean in the context of workforce skills?
Skills half-life refers to the rate at which a given skill's relevance decays over time. Technical skills, particularly in fast-moving technology areas, may have a half-life of two to five years before they become outdated. The concept is used to argue that a one-time education followed by a static career is no longer viable and that continuous learning is a professional necessity across all fields.
Which skills does the WEF say are hardest to automate?
The WEF and complementary research from MIT, Oxford, and McKinsey consistently identify social and emotional skills as the hardest to automate: complex reasoning in novel situations, empathy and interpersonal care, creative ideation, ethical judgment, and leadership in ambiguous contexts. These require capabilities — context-sensitivity, emotional attunement, genuine understanding — that current AI systems do not possess.
What is the difference between reskilling and upskilling?
Upskilling means deepening or extending skills in your current role or domain — learning new tools, advancing in your specialty, or taking on broader responsibilities. Reskilling means acquiring the skills needed for an entirely different role, typically in response to your current role being displaced or significantly changed. The WEF emphasizes that both are needed at unprecedented scale, with the challenge of reskilling being substantially more difficult and expensive.
What does the 2025 WEF Future of Jobs Report say about AI's impact on employment?
The 2025 WEF Future of Jobs Report found that AI and automation are expected to displace approximately 85 million jobs while creating around 97 million new ones by 2030, for a net positive but highly disruptive transition. The report identified analytical thinking and AI literacy as the top skills employers prioritize, and found that 40 percent of the core skills required for jobs will change by 2027, requiring urgent reskilling across the global workforce.
