AI Labor Displacement and Augmentation

The central labor market question of the AI era is not simply whether jobs disappear, but whether AI augments human capability or substitutes for it — and the answer varies by task, role, and deployment choice.

What It Is

McKinsey Global Institute (November 2025) estimates that currently demonstrated technologies could theoretically automate activities accounting for about 57% of US work hours — but explicitly frames this as a measure of technical potential, not a forecast of job losses. Adoption takes time, and the key variable is how organisations choose to deploy AI: as a replacement for workers or as a complement that raises their productivity.

Jones (2026) provides the framing: AI’s labor impact depends on whether it is used to (a) automate tasks, displacing workers; or (b) augment workers in existing tasks, raising their productivity and wages; or (c) create new tasks that only humans can do. Historical technological transitions involved all three simultaneously, with the net outcome depending on relative rates.

Why It Matters (for the Economy)

The 57% automation potential figure is striking, but McKinsey’s own analysis reveals something more nuanced: more than 70% of skills sought by employers today can be applied in both automatable and non-automatable work. This means most skills don’t become worthless — they get redirected. Workers will spend less time preparing documents and doing basic research, and more time framing questions and interpreting results. The labour market concern is therefore less about skills becoming obsolete wholesale, and more about whether the transition is manageable, especially for workers with limited adaptive capacity.

The Brookings Institution (April 2026) flags a more structural concern: AI disrupts not just individual jobs but the pathways connecting them. Career progression depends on workers moving through sequences of roles that build skills incrementally. If AI erodes key “Gateway” occupations — the stepping stones between entry-level and high-wage roles — the mobility ladders that workers without four-year degrees depend on may collapse.

Evidence & Examples

  • McKinsey estimates 57% of US work hours are theoretically automatable with current technology; the $2.9 trillion in annual US economic value that could be unlocked by 2030 depends on workflow redesign, not just task automation (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)
  • Demand for “AI fluency” in US job postings has grown sevenfold in two years — faster than any other skill — indicating employers are already adjusting expectations (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)
  • 15.6 million STARs (workers skilled through alternative routes, without four-year degrees) — one-fifth of the 70 million STARs total — work in occupations in the top quartile of AI exposure; 23 million STARs have low adaptive capacity (How AI may reshape career pathways to better jobs.md)
  • Nearly 11 million STARs are in Gateway occupations (the stepping-stone roles for economic mobility) that are highly AI-exposed, concentrated in administrative and clerical roles (How AI may reshape career pathways to better jobs.md)
  • Only 51% of pathways between Gateway and Destination occupations are not highly exposed to AI (How AI may reshape career pathways to better jobs.md)
  • Acemoglu, Autor, and Johnson (cited by Brookings) argue AI’s potential as a “collaborator” may enable new tasks, create new work, and increase the value of human expertise — but lower-wage roles subject to automation may lose pathways to advancement (How AI may reshape career pathways to better jobs.md)

Tensions & Open Questions

  • Augmentation vs. automation is a deployment choice, not a technological inevitability. The same AI system can be set up to replace a worker or to make that worker more productive. This makes policy and firm-level incentives critical.
  • The deskilling risk: Shen & Tamkin (Feb 2026) find that AI assistance impairs conceptual understanding, code reading, and debugging abilities in novice developers — without delivering significant efficiency gains on average. If AI assistance erodes the skill-building that Gateway occupations provide, long-run mobility may suffer even without direct job losses (see AI Skill Formation and Deskilling).
  • Regional concentration: Brookings documents that AI exposure of Gateway occupations varies sharply by region — Sun Belt metro areas (Florida in particular) and Northeast state capitals show the highest concentration of AI-exposed STARs. This points to uneven geographic impacts that national averages mask.
  • ⚠️ CONTRADICTION: McKinsey (2025) argues most skills “remain relevant” and will be redirected; Brookings (2026) argues the pathways that let lower-wage workers develop and demonstrate skills may collapse even if skills themselves don’t become worthless. These views are not mutually exclusive but emphasise different levels of analysis.
  • 📅 POTENTIALLY STALE — early-career displacement now empirically confirmed: The framing of “augmentation vs. displacement is a deployment choice” is accurate at the macro level, but the Stanford Canaries in the Coal Mine paper (Brynjolfsson, Chandar & Chen; Oct 2025; in Raw/) provides empirical evidence that displacement IS already occurring for entry-level workers. Using ADP payroll data, workers aged 22–25 in the most AI-exposed occupations saw a 16% relative employment decline since late 2022, while senior workers in the same occupations saw 6–9% growth. Critically, declines are concentrated in occupations where AI automates (not augments) tasks. This is the strongest large-scale empirical evidence to date that AI is substituting rather than augmenting at the entry level. [Raw/Canaries in the Coal Mine — pending ingestion]

AI Career Pathways and Workforce Mobility · Skill Partnerships Human-AI · AI Skill Formation and Deskilling · AI as a General Purpose Technology