Net job creator or net job destroyer? 92 million displaced, 170 million new roles — a net gain on paper. Both sides, steel-manned, with every claim traced back to a named source. Undated on purpose — this issue gets refreshed as the argument moves.
Eight parts, one cover. Same frame every issue in The Debate; only the words and the two filled grid cells change.
The most visible form of AI displacement — the mass layoff announcement — is not the most common. The more pervasive version is quieter: a company integrates AI tools and simply does not hire the people it would previously have hired. The headcount stays flat, the output rises, and the person who never got the offer never knows what they missed.
Goldman Sachs describes this as AI being used to avoid adding headcount rather than to immediately fire existing workers. It is real displacement. It is just invisible to the standard measures of unemployment. The 22-to-25-year-old software developer or customer service representative entering the market in 2025 is not losing a job they already have — they are not getting one they would have got. The employment data for that cohort already shows a 16% drop. That is not a projection; it is observed.
The optimistic case is also real: McKinsey documents 60 years of technology creating more jobs than it destroyed, and the WEF projects 78 million net new roles by 2030. Both things can be true. The question is not the long-run net figure. It is who makes the transition, at what speed, and who pays for it.
Each side, put as well as its own advocates would put it. Sources named. No figure here is offered as settled fact; each is a claim by a named author, linked in full at the end.
Goldman Sachs (2023, updated 2025): generative AI could automate tasks equivalent to 300 million full-time jobs worldwide. 26% of office roles and 20% of customer service positions are highly exposed. Goldman's own framing: "automate tasks", not "eliminate jobs wholesale" — a distinction that matters less to the worker whose tasks are gone.
Workers aged 22–25 in AI-exposed roles — software developers, customer service representatives — have already seen a 16% drop in employment. Experienced workers remain more stable. This is not a projection; it is observed data from the cohort entering the labour market now.
WEF Future of Jobs Report 2025: 92 million roles projected displaced globally by 2030. The transition requires workers to move into roles that don't yet exist at scale, using skills they don't yet have — and the training infrastructure for that transition is not yet built.
AI is suppressing hiring rather than firing existing workers. Employers integrate AI to avoid adding headcount rather than immediately cutting staff. Real displacement; invisible to the person who never receives the offer and whose absence never registers in unemployment statistics.
McKinsey: 60% of today's US workforce is employed in occupations that didn't exist in 1940. Technology has consistently generated more jobs than it eliminated over the medium term — a pattern that has held through mechanisation, electrification, computing, and the internet.
WEF projects 170 million new roles by 2030 against 92 million displaced — a net gain of 78 million, primarily in AI-related, green economy, and care economy roles. The optimistic headline figure is real; the question is whether the workers displaced are the same workers who take the new roles.
McKinsey: 30–50% of work activities could be automated, but partial automation of tasks often increases output and wages for the remaining human workers. Augmentation, not replacement, is the dominant observed pattern so far in sectors where AI has been deployed at scale.
AI handles genuinely dangerous, repetitive, or dull work — reducing injury rates in logistics, fatigue in medical imaging review, and error rates in data processing. New roles emerging include AI trainer, AI auditor, and AI-human workflow designer. None of these existed five years ago.
— WEF Future of Jobs Report 2025
House rule: any percentage about jobs or wages is attributed to its author and study. We reference claims; we do not manufacture certainty. The true long-run figure is unknown and contested.
The same AI tool, deployed in two different organisations, can produce two entirely different outcomes for workers. In one, productivity gains flow to wages and headcount — the work becomes better, not fewer. In the other, the same gains flow to shareholders and the headcount falls. The technology does not determine which happens. The contract and the bargaining position do.
This is not a new observation — the same was true of every previous round of automation — but it is worth stating plainly because the public debate tends to run on a deterministic track: AI will either destroy jobs or create them, full stop. The honest answer is that AI creates the conditions for both, and the outcome is decided by institutions. Where workers have effective bargaining power, automation tends to produce augmentation. Where they do not, it tends to produce displacement. The technology is the variable that changes the stakes. Governance is the variable that determines who wins.
The transition cost is real and unevenly distributed. The people most affected are not those with decades of experience and accumulated skills to fall back on. They are the youngest and least established workers — people entering the market in 2024 and 2025 who are already seeing it in their employment figures. The medium-term net-gain forecast is true and simultaneously inadequate as a response to someone who cannot get their first job.
The most honest position is that both the displacement and the creation scenarios are well-evidenced, both are happening simultaneously in different sectors, and the outcome for any individual worker depends heavily on factors — geography, industry, education level, age — that the aggregate forecasts obscure.
Three pressure points are where the practical answer is being decided. Each is moving at a different speed.
The observed data is clearer on the entry-level and task-exposed side than on the new-roles side. The 16% drop for 22-to-25-year-olds is documented. The 78 million net new roles are a projection. Projections have been reliably optimistic in the past — and reliably correct in the very long run. Neither of those facts is useful to the 24-year-old in front of you now.
Even if the net figure is positive, displacement and creation are not distributed evenly across the same populations, the same geographies, or the same timelines. A green-economy role created in 2030 does not help a customer service representative displaced in 2025. The transition requires active redistribution — of training resources, of income support, of time — to prevent the net gain from being a gain only for those who were already winning.
The policy levers are known: invest in transition training, strengthen bargaining rights, consider how productivity gains are shared through wage floors or profit-sharing requirements. None of these are exotic ideas. The constraint is not knowledge; it is political will. The countries that moved quickly on skills investment after previous automation waves fared better than those that waited for the market to self-correct. That pattern is not guaranteed to hold, but it is the strongest available guide.
Every claim above, traced to its author. Read the originals; argue with them, not with us.
The Debate — No. 06 · AI and Work · Craig Stanley Studio · CS·PRESS · Undated by design, refreshed as the argument moves · One red, used once · All colour pairings meet WCAG 2.1 AA.
Nine debates, each titled and never dated, each refreshed in rotation as the argument moves. Plain English. Both sides put at their strongest. Every claim about the world attributed to its author with a link. No hype, no banned words, no manufactured certainty.
AI and Creativity · AI and the Environment · AI and Ethics · AI and Power · AI and Truth · AI and Young People · AI as a Force for Good · AI in Education.