The analysis of AI-driven job displacement focuses almost exclusively on income. Work in advanced economies is not primarily an income mechanism. It is a primary source of identity, social structure, daily purpose, and the sense of being legible to other people.
The McKinsey and WEF frameworks for assessing AI’s impact on employment track aggregate job creation and destruction. Goldman Sachs measures wage changes in re-employment. These are important measurements. They are also insufficient — and their insufficiency is not marginal. It is structural.
Work in advanced economies is not primarily an income mechanism. For most people who hold a professional category — doctor, lawyer, financial analyst, architect, journalist — that category is a significant component of who they understand themselves to be. When Erik Erikson mapped adult psychological development, he identified productive work as central to the resolution of the core adult tension between generativity and stagnation. He was not writing about income.
The identity loss the data doesn’t capture
Consider a 34-year-old financial analyst whose role is restructured around an AI copilot, and whose salary is consequently cut by 20%. The Goldman Sachs data captures this: a worker displaced by AI, re-employed at lower wages. What the data does not capture is that she has also lost the professional category through which she was recognised by colleagues, the peer group that gave her professional life social meaning, and the developmental pathway — junior analyst to senior analyst to portfolio manager — that organised her understanding of her own future.
The concept emerging in labour economics as the “AI precariat” acknowledges income disruption. It does not adequately address identity disruption. Richard Sennett’s The Corrosion of Character (1998) documented, in the context of earlier labour market restructuring, how the elimination of stable professional narratives — the storied career — produces not just economic precarity but psychological disorientation at a deep level.
Historical precedent
The deindustrialisation of steel, textiles, and coal communities in the United States and United Kingdom during the 1970s and 1980s provides the most fully documented precedent. The economic analysis of those transitions focused, as AI analysis does now, primarily on income and employment figures. The longitudinal public health data told a different story: elevated mortality, accelerated cognitive decline, and substance dependency that tracked not employment status but occupational identity loss.
The Harvard Study of Adult Development — the longest longitudinal study of adult life — consistently finds that the quality of social relationships and sense of purposeful contribution are stronger predictors of late-life health and cognition than income, above a basic threshold. The professional category provides both: the peer group and the sense of contribution. When AI eliminates the profession, it eliminates both simultaneously.
“A net gain of 78 million jobs is irrelevant to the person whose profession has been restructured. Aggregate figures and individual outcomes are different things. The policy frameworks being built for the AI transition have not yet distinguished between them.”
What policy is not measuring
No policy framework currently being developed in any major economy includes a measure of professional identity disruption as a welfare indicator in AI displacement analysis. The WEF, IMF, and OECD frameworks address employment and income. The psychological consequences — measured not in wages but in identity, meaning, and social belonging — are unaccounted for.
Keynes, in Economic Possibilities for our Grandchildren (1930), anticipated technological unemployment and observed that the human problem of occupying one’s time and energies purposefully, once economic necessity was reduced, would be harder than the economic problem. He was, in the current transition, closer to the mark than the frameworks his successors built.
SOURCES
— Erikson, E. — Identity and the Life Cycle, Norton, 1959
— Harvard Study of Adult Development
— Sennett, R. — The Corrosion of Character, Norton, 1998
— Goldman Sachs — AI displacement wage data, 2026
— Stanford HAI AI Index 2024
— Keynes, J.M. — Economic Possibilities for our Grandchildren, 1930
— Susskind & Susskind — The Future of the Professions, Oxford, 2015
