OpenAI's AI Jobs Transition Framework: Why Only 18% of Jobs Face High Automation Risk and What It Means for Europe

OpenAI's Chief Economist Ronnie Chatterji and labor economist Alex Martin Richmond published the AI Jobs Transition Framework in April 2026, analyzing over 900 occupations covering 153.7 million U.S. jobs (99.7% of employment) . The core finding: outright automation risk is lower than many earlier forecasts predicted, with reorganization and augmentation being far more common outcomes than job elimination .

The Four-Dimensional Analytical Model

The framework moves beyond simple "AI exposure" measures by combining four dimensions :

• Technical AI capability — Whether current AI systems (GPT-4 class models) can perform the occupation's core tasks at or above median human level. GPT-4 class models perform at or above median human level on 60–80% of white-collar tasks .
• Human necessity — Whether the occupation requires regulatory mandates, relational trust, or physical presence that insulates it from full automation. Regulatory (judges, lawyers in courtrooms), relational (teachers, therapists, hospice nurses), and physical (plumbers, electricians, hands-on caregivers) factors create protective moats .
• Demand elasticity — Whether the occupation faces inelastic demand (fixed need regardless of cost) or elastic demand that could expand as AI lowers costs .
• Observed ChatGPT usage — Validated against real-world adoption patterns; workers in the most automatable roles (data-entry clerks, accountants) are using AI for three times more of their relevant responsibilities compared to those in less vulnerable roles .

The Four Transition Archetypes

Combined and validated, these dimensions sort every occupation into four buckets :

1. Higher automation risk (~18% of U.S. jobs) — High AI exposure, weak human necessity, inelastic demand. Includes data entry, bookkeeping, customer service, cashiers, legal support work, and classification clerks. These are the roles most likely to see job elimination in the near term.

2. Will reorganize (~24% of jobs) — High AI exposure, strong human necessity. The task composition shifts significantly, but human workers remain essential for core relational, regulatory, or physical functions. Employment may decline in headcount even as the role persists.

3. Will grow/expand (~12% of jobs) — High AI exposure combined with elastic demand means AI lowers costs and expands the market, creating more jobs. Examples include roles where AI acts as a productivity multiplier in growing sectors.

4. Will be relatively unaffected (~46% of jobs) — Low AI exposure to begin with, often due to physical, relational, or regulatory barriers. These occupations see minimal near-term AI impact.

U.S. vs. EU Automation Risk: 18% vs. 14%

The OpenAI framework estimates that approximately 18% of U.S. jobs face a relatively higher risk of short-term automation . For the EU, a separate analysis by CEDEFOP (the European Centre for the Development of Vocational Training, an EU agency) finds that about 14% of EU adult workers face a very high risk of automation . This risk is concentrated in routine jobs with low demand for transversal and social skills, and is higher among male workers .

Important caveat: The 14% EU figure comes from CEDEFOP's independent skill-needs methodology, not directly from OpenAI's framework. However, the direction is consistent: the EU's more regulated labor markets, stronger social protections, and different occupational mix produce a lower near-term automation share than the U.S.

Country-Level Divergences

A Coface analysis of AI labor exposure across European countries (published April 2026) maps automation-exposed task content :

• Higher automation-risk countries: Luxembourg, the UK, Switzerland, the Netherlands, Norway, and Denmark show the highest shares of automation-exposed task content, ranging from ~12% in Turkey to nearly 20% in the UK .
• Luxembourg, Sweden, and the Netherlands tend to have larger shares of growth-oriented, high-skill service employment, but also high AI exposure in their financial and professional services sectors.
• Germany, Greece, and Italy face a different risk profile: larger shares of manufacturing, routine-based administrative work, and lower digital-service penetration mean that automation potential is higher in tradable goods sectors .

The OpenAI EU Economic Blueprint 2.0 (January 2026) further segments countries by AI penetration and intensity of use. Lithuania and Latvia stand out for combining both high population penetration and high-intensity use of advanced AI thinking capabilities, while other EU states lag significantly in diffusion .

Policy Context: Chatterji in Brussels, the EU AI Act, and Job Displacement

Ronnie Chatterji was appointed OpenAI's first Chief Economist in October 2024, previously serving as an economic adviser in the Biden and Obama administrations. His core message is that the economy is in the "in-between times" — most people currently use AI as a complement, not a substitute . As of late June 2026, Chatterji is in Europe for a series of high-level conversations, including a POLITICO Europe event titled "AI and the future of work: Is the EU prepared for the transition?" .

Key policy tensions:

• OpenAI was the first U.S.-based AI developer to sign the EU AI Act General Purpose Code of Practice .
• The Carnegie Endowment argues that what is needed alongside the AI Act is a "dedicated European labor transition framework, ideally embedded in the bloc's 2028–2034 budget" .
• The European Commission notes that AI and digital technologies are "set to increase employment overall in Europe, but the gains will be uneven, benefiting mainly high skilled, prime-age workers and women, while low skilled and young workers, and structurally weaker regions remain more exposed" .

ECB Survey: Limited AI Intensive Use

The ECB's Survey on the Access to Finance of Enterprises (SAFE) for Q4 2025, covering over 5,000 euro area firms, provides a reality check on AI adoption depth :

• More than 70% of firms report using AI in some form by late 2025 .
• However, only about 7% of firms use AI intensively enough to change how they produce, innovate, and compete .
• The breakdown: 27% do not use AI at all; 33% use it very infrequently; 31% use it moderately; only 7–9% use it intensively .
• A sizable investment gap exists: while most firms use AI, few actually invest in dedicated AI infrastructure. Firms that do invest plan to allocate an average of 9% of total investment to AI in 2026 .
• The ECB concludes that "diffusion is increasing quickly" but that the gap between casual use and transformative integration remains very wide .

EU Efforts Toward Technological Sovereignty

The EU Economic Blueprint 2.0 from OpenAI frames European AI competitiveness explicitly around sovereignty . It segments EU countries by AI penetration and high-intensity use of advanced "thinking" capabilities, finding that only a small set of countries (Lithuania, Latvia) combine broad access with deep diffusion. The broader strategic push — "technological sovereignty" — is reflected in EU investment plans for AI infrastructure, the AI Act's regulatory ambition, and calls from bodies like the Carnegie Endowment for a ring-fenced EU labor transition budget embedded in the 2028–2034 Multiannual Financial Framework . The European Commission's own analysis projects that AI will increase overall employment in Europe but warns that without targeted policy, structurally weaker regions and lower-skilled workers will be left behind .