The numbers: what each report actually says

Four major reports from 2025-2026 are repeatedly cited in the debate about AI and employment, frequently out of context. Each answers a different question.

The WEF Future of Jobs Report 2025 projects that 92 million jobs will be displaced globally by 2030, while 170 million will be created — a net balance of +78 million. The question it answers: what is the aggregate effect on the total number of jobs in the world? Crucial limitation: the 92 million displaced and the 170 million created are not the same people, are not in the same sectors, do not require the same qualifications. A global net gain does not distribute well-being individually.

The McKinsey Global Institute (November 2025) found that 57% of working hours in the US are technically automatable with current technology — 44% by AI agents, 13% by robots. The question it answers: what is the technical potential for automation? Explicit limitation: McKinsey itself describes this as technical potential, not as an unemployment forecast. Automating a task does not eliminate the job — it reorganizes what the worker does.

Goldman Sachs (April 2026) estimates that 300 million global jobs are exposed to some AI impact, but that only 2.5% of American employment is at risk of direct elimination. The potential for productivity growth is estimated at $7 trillion in global GDP over 10 years. The question it answers: what is the scale of economic exposure? Exposure is not elimination.

The Stanford HAI AI Index 2026 brings a concrete and disturbing data point: the employment of software developers aged 22 to 25 has dropped nearly 20% since 2024, while the employment of developers over 30 has continued to grow. The entry ladder is being pulled up — not the entire building.

What aggregate data hides

The central problem with all these numbers is aggregation. A global or national average number does not arrive uniformly at any specific career, region, or demographic group. Three asymmetries are especially important.

Sectoral asymmetry: Administrative functions face a 26% direct risk of displacement, according to Second Talent's analysis (May 2026). Customer service, 20%. In contrast, healthcare, education, and specialized manual labor show growing demand. AI does not impact all sectors equally — it specifically impacts work that consists of processing structured information and following explicit rules.

Entry asymmetry: The Stanford HAI data on junior developers is the most dramatic example of a broader trend: AI is squeezing the entry-level job market in many cognitive professions. Drafting emails, producing basic reports, doing initial analysis — tasks that traditionally absorbed early-career workers — are being automated. Experienced professionals, who add judgment and context, remain in demand. The ladder that allowed them to get there is being removed.

Reskilling asymmetry: The WEF and McKinsey converge on one point: 56% of displaced workers in highly automated sectors report difficulty transitioning to new jobs, even when those jobs exist and are geographically close (McKinsey, 2024). Reskilling is not an abstraction — it is an individual transition crisis that net gain statistics do not capture.

What is growing: the creation side

The debate about displacement frequently obscures the creation side. Job openings that require AI skills grew 134% above 2020 levels. In the US, 275,000 openings in January 2026 required AI fluency. Demand for AI governance skills grew 150%. Prompt engineering, 90%.

Jamie Dimon confirmed in February 2026 that JPMorgan Chase has already displaced workers due to AI, but has "big plans for redeployment". The bank's total headcount did not fall — it changed composition. This is the pattern that emerges in large organizations: not mass layoffs, but a reorganization of roles with intensive training of existing employees to work with AI systems.

The question the reports don't answer

What WEF, McKinsey, and Goldman Sachs answer well: the aggregate effect on the total number of jobs, the technical potential of automation, the scale of economic exposure. What none of them answer: what happens to your specific job, in your company, in your sector, in your city, with your current qualifications.

This individual question is the one that matters to most people — and it has no answer in aggregate data. It has an answer in specific occupational, sectoral, and qualification profile analysis. PrezenceAI monitors exactly these sectoral analyses — because the impact of AI on work is not a global statistic, it is a series of specific impacts that add up to the global.