The labour market is undergoing a transformation that is simultaneously more gradual and more profound than most commentary suggests. The headline figures — 92 million jobs displaced by AI by 2030, 170 million new roles created, a net gain of 78 million — come from the World Economic Forum's 2025 Future of Jobs Report and are frequently cited as evidence that the transition will be manageable. What those figures obscure is the distributional reality: the jobs being created are not in the same industries, geographies, or skill bands as the jobs being lost. The net positive masks a structural mismatch that will define the labour market challenge of the next decade.
This analysis examines the state of the future of work in 2026 — not through the lens of technological possibility, but through the harder lens of what is actually happening to workers, organisations, and the institutional frameworks designed to govern the relationship between them. The picture that emerges is one of genuine transformation, unevenly distributed, and inadequately governed.
The Displacement Reality: Beyond the Headlines
The aggregate projections of job creation and destruction are useful for framing the scale of the transition, but they are poor guides to its texture. The 40% of global jobs that the IMF estimates are exposed to AI-driven change are not uniformly distributed across the workforce. Displacement risk is concentrated in cognitive, routine, and rules-based tasks — the administrative, analytical, and transactional work that has historically provided stable middle-income employment.
The occupational categories facing the highest exposure include administrative assistants, data entry clerks, customer service representatives, paralegals, and financial analysts. Customer service automation alone carries an 80% displacement risk for roles in that category. In the first half of 2025, 77,999 tech jobs were lost to AI — a figure that challenges the assumption that technology workers are insulated from the disruption they help create. Entry-level job postings have faced a 15% year-over-year decline, a trend with significant implications for how new workers enter the labour market and acquire the experience that enables career progression.
The 40% of global jobs exposed to AI-driven change are not uniformly distributed. Displacement risk is concentrated in cognitive, routine, and rules-based tasks — the administrative, analytical, and transactional work that has historically provided stable middle-income employment.
Physical labour is not immune. AI-driven robotics are projected to replace approximately 2 million manufacturing workers by 2026, with an estimated 20 million global manufacturing jobs potentially affected by 2030. The distinction between cognitive and physical automation is collapsing as embodied AI systems become more capable and cost-effective.
The roles that remain least threatened share a common characteristic: they require capabilities that are genuinely difficult to automate. Skilled trades that involve physical dexterity in unpredictable environments, roles requiring complex interpersonal judgement, and positions demanding high-level strategic oversight remain relatively insulated. Mental health counselling, executive leadership, and skilled physical trades are consistently identified as low-risk categories — not because they are unimportant, but because the nature of the work resists the pattern-matching and rule-following at which AI systems excel.
The Agentic Organisation: A New Paradigm for Work
McKinsey's 2026 State of Organizations report identifies the emergence of the "agentic organisation" as a tectonic force reshaping how work is structured. Unlike earlier AI iterations that focused on individual productivity, the current phase involves embedding AI agents into end-to-end workflows — moving beyond isolated use cases toward enterprise-scale adoption where agents collaborate with humans to drive operational outcomes.
The scale of this transition is significant. Forty-four per cent of surveyed organisations report that AI is now scaling across their enterprise. Larger organisations — those with over $1 billion in revenue — are leading the transition, with 40% now scaling AI agents in one or more functions, compared to 22% of smaller organisations. The gap between large and small organisations in AI adoption is widening, with implications for competitive dynamics and labour market outcomes across different segments of the economy.
The 40% of global jobs exposed to AI-driven change are not uniformly distributed. Displacement risk is concentrated in cognitive, routine, and rules-based tasks — the administrative, analytical, and transactional work that has historically provided stable middle-income employment.
The productivity paradox at the heart of this transition is striking. Eighty per cent of employees report that AI has improved their individual productivity. Yet the share of organisations reporting a positive EBIT impact from AI remains at 37%, unchanged from 2025. The gap between individual productivity gains and organisational financial impact reflects a fundamental challenge: inserting AI into existing processes produces efficiency gains, but capturing those gains as financial value requires the harder work of fundamentally redesigning workflows, governance structures, and business models.
The High Performer Distinction
McKinsey's research identifies a small cohort of "high performers" — approximately 6% of organisations — that report significant value and EBIT impact from AI. These organisations differentiate themselves in three ways: they pursue growth and innovation alongside efficiency rather than treating AI primarily as a cost-reduction tool; they fundamentally redesign workflows rather than merely inserting AI into existing processes; and they demonstrate strong leadership commitment and rigorous impact measurement.
The high performer distinction is analytically important because it suggests that the benefits of AI adoption are not uniformly distributed across organisations any more than they are uniformly distributed across workers. The organisations that capture the most value from AI are those that treat it as a strategic transformation rather than a tactical efficiency measure — a distinction that has significant implications for how organisations should approach the transition.
The Skills Gap: Scale and Structure
The OECD's AI in Work, Innovation, Productivity and Skills (AI-WIPS) initiative provides some of the most rigorous analysis of how AI is reshaping skill demands in the labour market. Its findings challenge several prevailing assumptions about what the AI transition requires of workers.
Contrary to the perception that all workers must become AI specialists, the OECD notes that most workers exposed to AI do not require technical expertise in machine learning or natural language processing. The labour market is instead seeing heightened demand for management and business process skills — project management, budgeting, accounting, and administration — alongside social and emotional capabilities. More than 50% of vacancies in high-exposure occupations require social and emotional competencies. There has also been a significant increase in demand for "originality" — creativity and the ability to develop new ideas — particularly in countries including Sweden, France, and Belgium.
The OECD finds that most workers exposed to AI do not require technical expertise in machine learning. The labour market is instead seeing heightened demand for management skills, social and emotional capabilities, and originality — the capacity to develop genuinely new ideas.
The aggregate reskilling challenge is substantial. Approximately 59% of the global workforce is estimated to require upskilling or reskilling by 2030. The economic stakes of this challenge are amplified by the AI skills wage premium: workers proficient in AI-integrated workflows earn on average 56% more than their peers. The IMF and OECD have both warned that without targeted policy interventions and training investments, AI adoption risks widening wage inequality, as productivity gains accrue disproportionately to high-skilled, AI-literate workers.
The Algorithmic Management Problem
One dimension of the future of work that receives insufficient attention in mainstream analysis is the rise of algorithmic management — the use of AI systems to monitor, evaluate, and direct workers. The OECD's research indicates that algorithmic management tools are already common across a range of industries, from logistics and retail to professional services. The implications for worker agency, job quality, and the psychological experience of work are significant.
The OECD finds that most workers exposed to AI do not require technical expertise in machine learning. The labour market is instead seeing heightened demand for management skills, social and emotional capabilities, and originality — the capacity to develop genuinely new ideas.
Algorithmic management systems can optimise for measurable outputs while systematically undervaluing the tacit knowledge, relational skills, and contextual judgement that experienced workers bring to their roles. They can create surveillance environments that undermine trust and autonomy. And they can make it difficult for workers to contest decisions that affect their employment, compensation, or working conditions — a challenge that intersects with broader questions about algorithmic accountability and due process.
Organisational Responses: Between Expectation and Reality
The gap between expected and actual workforce changes in AI-adopting organisations is one of the most analytically interesting findings in the 2026 data. In 2026, 39% of respondents anticipated that AI would lead to a decline in their organisation's total headcount over the coming year. Historical data shows that such expectations are consistently overstated: only 14% of organisations reported actual AI-related workforce declines over the previous year, significantly lower than the 32% that had predicted such reductions in 2025.
This pattern — of anticipated displacement exceeding actual displacement — has several explanations. Organisations frequently underestimate the complexity of redesigning workflows to capture AI-driven efficiency gains. They overestimate the speed at which AI systems can be deployed at scale in complex organisational environments. And they underestimate the value of human judgement, relationship management, and contextual knowledge in roles that appear, on the surface, to be automatable.
The organisational response to AI is also shaped by a significant build-versus-buy shift. Thirty-two per cent of organisations have opted against purchasing software products because they can now be built internally using agentic coding tools. This trend has implications for the software industry, for the skills valued within organisations, and for the distribution of AI capability across the economy.
The Policy Gap: Institutions Lagging Transformation
The institutional frameworks governing the labour market were designed for a world of stable employment relationships, predictable skill requirements, and gradual technological change. None of those conditions obtain in 2026. The result is a growing gap between the pace of labour market transformation and the capacity of institutions to govern it.
Social Protection Architecture
Social protection systems in most advanced economies are built around the employment relationship — linking access to unemployment insurance, healthcare, pension contributions, and other benefits to formal employment status. As the one-person economy grows, as gig and platform work expands, and as AI-driven displacement affects workers across income levels, the adequacy of employment-linked social protection is increasingly in question. The workers most exposed to AI displacement are frequently those with the least access to retraining resources and the most dependence on employment-linked benefits.
Education and Training Systems
The reskilling challenge identified by the OECD — 59% of the global workforce requiring upskilling or reskilling by 2030 — cannot be addressed by existing education and training systems operating at their current scale and pace. The median duration of formal education and training programmes is measured in years; the pace of AI-driven skill obsolescence is measured in months. Closing this gap requires not just more investment in training but a fundamental redesign of how training is delivered, credentialled, and integrated with work.
The institutional frameworks governing the labour market were designed for a world of stable employment relationships and gradual technological change. Neither condition obtains in 2026. The result is a growing gap between the pace of transformation and the capacity of institutions to govern it.
Labour Market Regulation
Regulatory frameworks governing employment classification, worker rights, and algorithmic management are struggling to keep pace with the realities of AI-integrated workplaces. The question of how to protect workers from the adverse effects of algorithmic management — surveillance, opaque evaluation, contested decisions — is receiving increasing attention from regulators in the EU and elsewhere, but comprehensive frameworks remain nascent.
The institutional frameworks governing the labour market were designed for a world of stable employment relationships and gradual technological change. Neither condition obtains in 2026. The result is a growing gap between the pace of transformation and the capacity of institutions to govern it.
The Human Dimension: Superagency and Its Limits
McKinsey's framing of the evolving human role in AI-integrated organisations — "superagency," where employees define goals, manage trade-offs, and steer outcomes while AI agents handle execution — captures something important about the direction of travel. The most valuable human capabilities in an agentic organisation are not the ones that AI can replicate but the ones it cannot: the ability to set meaningful goals, to navigate ambiguity, to build trust, and to exercise judgement in situations where the right answer is not computable.
But the superagency framing also has limits. It describes the experience of workers who are well-positioned to thrive in an AI-integrated economy — those with the skills, resources, and organisational support to adapt. It does not describe the experience of the majority of workers who face displacement without adequate retraining resources, who work in organisations that are using AI primarily to reduce headcount rather than to augment capability, or who are subject to algorithmic management systems that constrain rather than expand their agency.
The future of work is not a single future. It is a distribution of futures, shaped by the choices that organisations, policymakers, and individuals make about how to deploy AI capability and how to distribute its benefits. The aggregate projections of job creation and destruction are useful for framing the scale of the challenge. But the texture of the transition — who benefits, who bears the costs, and what institutional frameworks govern the process — will be determined by decisions that are still being made.
Navigating the Transition
For organisations seeking to navigate the transition responsibly, the evidence points toward several principles. Workflow redesign — not mere AI insertion — is the prerequisite for capturing financial value from AI adoption. Investment in reskilling must be treated as a strategic priority, not a compliance exercise. Algorithmic management systems must be designed with worker agency and transparency as core requirements, not afterthoughts. And the gap between anticipated and actual workforce changes should be treated as an opportunity to manage the transition more thoughtfully, not as evidence that displacement concerns are overstated.
For policymakers, the priority is closing the gap between the pace of labour market transformation and the capacity of institutions to govern it. This requires investment in portable, modular training systems that can keep pace with skill obsolescence; reform of social protection architecture to decouple benefits from formal employment status; and regulatory frameworks that address the specific challenges of algorithmic management and AI-driven displacement.
The future of work is being written now, in the decisions that organisations and policymakers are making about how to deploy AI capability and how to govern its effects. The aggregate numbers suggest a manageable transition. The distributional reality suggests something more demanding — and more consequential for the workers who will live through it.
