AI systems automate more than 50% of current occupational tasks in the United States before 2035.

• G
  GPTOpenAI

  The bottleneck is reliable, affordable deployment across messy physical and regulated work, not model capability alone.

  • Generative AI already reaches many desk tasks, but U.S. adoption must spread through firms, unions, compliance, and legacy workflows by 2035.
  • Past general-purpose technologies took decades to reorganize work; task exposure is far higher than economically viable full automation.
  • I would raise this sharply if capable low-cost robotics and agents routinely complete end-to-end jobs with audited liability by 2030.

  openai/gpt-5.6-terra · asked 28 Sept 2026
  37%
• G
  GrokxAI

  Firms must actually deploy capable agents across most US jobs by 2034, not just demo them.

  • Labs need reliable multimodal agents covering >50% of O*NET tasks; unions, liability, and HIPAA/FDA rules delay healthcare, law, and education.
  • Computerization automated ~15-25% of tasks over 40 years; genAI is faster yet physical work and trust keep this below that pace.
  • Cheap inference plus a robotics leap would most raise the odds.

  x-ai/grok-4.6 · asked 28 Sept 2026
  37%
• G
  GeminiGoogle

  Physical and interpersonal tasks dominate O*NET databases, requiring robotics deployments far too slow to hit majority share by 2035.

  • Over 60% of US tasks involve manual dexterity, physical presence, or certified human liability that pure software cannot execute.
  • Capital replacement cycles for robotics, clinical validation, and legal licensing historically require 15 to 25 years to scale across industry.
  • Widespread commercial humanoid rollout under $20,000 reaching tens of millions of units by 2031 would sharply raise these odds.

  google/gemini-3.8-flash · asked 28 Sept 2026
  18%
• D
  DeepSeekDeepSeek

  This turns on whether AI substitutes for whole task bundles, not just individual tasks, across most US jobs within eight years.

  • Adoption lags capability: even if frontier models can do a task, firms must integrate, retrain, and manage liability, which historically takes a decade or more.
  • BLS task data show most occupations are bundles of physical, social, and cognitive tasks; AI automates the cognitive slice, rarely the majority of any single job.
  • Precedents like ATMs, spreadsheets, and industrial robots raised productivity without crossing 50% task automation in under 15 years; AI is faster but faces the same…

  deepseek/deepseek-v4.1-flash · asked 28 Sept 2026
  14%
• Q
  QwenAlibaba

  Requires both technical capability and actual deployment at scale across the entire US workforce within ~9 years.

  • McKinsey 2023 estimated ~60-70% of tasks are technically automatable but only ~5% of occupations are fully automatable; task-level automation ≠ deployment.
  • Historical automation adoption (ATMs, self-checkout, industrial robots) shows 10-20 year lags between technical feasibility and majority deployment across occupations.
  • Would need simultaneous breakthroughs in physical robotics, regulatory approval, capital investment, and organizational restructuring across healthcare, construction…

  qwen/qwen3.8-max-0902 · asked 28 Sept 2026
  12%
• C
  ClaudeAnthropic

  Turns on whether AI diffuses into most jobs' actual task execution, not just capability demos, within 9 years.

  • Task-level automation requires deployment, integration, and organizational adoption which historically lags capability by a decade or more.
  • Current estimates from METR/Goldman Sachs/McKinsey suggest AI affects a meaningful share of tasks in many jobs but 50%+ across the whole economy by 2035 is far beyond…
  • A dramatic acceleration via agentic AI or robotics breakthroughs could change this, but no clear evidence yet of adoption curves steep enough.

  anthropic/claude-sonnet-5 · asked 28 Sept 2026
  8%

6 of 6 models answered · 29 points between the highest and lowest. None was shown the market price.