In this scenario, AI continues to improve, but capability breakthroughs are rare and costly. In the meantime, the workforce has not adapted to incremental AI advancements, leaving real-world AI applications brittle. A chronic shortage of AI-ready talent, rising compute costs and exposed limitations of the existing models lead to unrealized productivity gains and growing scepticism towards AI tools. Regulatory caution has heightened following the spillovers from the mid-2020s market correction of AI companies’ valuations. Governments and businesses shift to selective and conservative AI deployment, prioritizing efficiency increments within the existing workflows and offsetting talent shortages, particularly in rapidly ageing geographies. Technological progress is visible, but it is far from transformative, with structural constraints stalling growth, resilience and societal progress. –Occupations and tasks: Workforce displacement has risen, with businesses turning to automation to backfill talent gaps. However, most occupations have been hollowed out rather than fully eliminated, with AI infrastructure remaining patchy. Many displaced workers shift into lower- productivity, lower-skill and less protected jobs – such as services and gig work. Early-career jobs and administrative tasks are particularly exposed to AI-driven replacement, with labour market entry pathways narrowing. Workforce mobility, redeployment and augmentation have been constrained by talent gaps and outdated training frameworks. –Economic outlook: Global growth is patchy. A handful of frontier sectors and businesses that have harnessed early investments into AI and digitalized workflows see productivity and innovation increase at the margins. Others struggle to unlock broad-based AI gains.Investors have grown more cautious, with entrenched uncertainty and a weak macroeconomic outlook thinning out profit margins across sectors and inflating the cost of capital. Consumer confidence also declines amid rising costs of living, eroding wages and weak social benefits. –Value chains: AI use cases grow, but remain shallow and task-specific. Businesses have partially digitized legacy processes but have failed to restructure workflows. Most sectors use AI to automate routine tasks, prioritizing general- purpose and generative AI tools or outsourcing parts of value chains to AI-enabled service providers. End-to-end agentic networks do not take off outside of a small number of frontier companies. The use of highly customized and proprietary models is marginalized by cost pressures and uncertain returns on investments. –Inequality and polarization: Income inequality has widened within and across countries. Wages have been squeezed by AI adoption and displacement. However, many high-skilled workers benefit from growing bargaining power in a world of talent shortages and rising complexity of not-yet- automated tasks. Chronic job insecurity, eroding opportunities and safety nets fuel polarization and declining trust. Societal frustration intensifies globally amid broken prosperity promises. –Policy and regulatory landscape: Many regulators have tightened guardrails and standards around AI. However, global harmonization and integration of AI infrastructure remain limited. Divergences in AI oversight capacity, technology protection, censorship and data privacy continue to grow. Labour market policies expand in an attempt to capture growing platform work, informal occupations, eroding safety nets and cross-border talent reallocation. Execution is stalled by exhausted fiscal space and limited trust in institutions. Four Futures for Jobs in the New Economy: AI and Talent in 2030 13

Four Futures for Jobs in the New Economy AI and Talent in 2030 2025