{"slug": "ai-leaves-many-supposedly-automatable-jobs-intact", "title": "AI Leaves Many Supposedly Automatable Jobs Intact", "summary": "Nearly four years after the mainstream adoption of AI, many jobs previously labeled as \"automatable\" remain intact because AI typically automates only specific tasks rather than entire roles. Benjamin Todd, president of the nonprofit 80,000 Hours, told Business Insider that the \"defined, clean bit is actually a minority of the work,\" noting that predictions of full automation have proven overly optimistic. The report highlights that radiologists, for example, spend only 36.4% of their time interpreting images, with the remainder devoted to tasks like patient care and administration that AI cannot easily replace.", "body_md": "# AI Leaves Many Supposedly Automatable Jobs Intact\n\nBusiness Insider reports that, after nearly four years of mainstream AI, many 'automatable' roles remain active because AI typically automates parts of jobs, not entire roles. Benjamin Todd, president of the nonprofit **80,000 Hours**, told Business Insider that the \"defined, clean bit is actually a minority of the work,\" adding \"We should caveat all of this with 'survived so far.'\" The article notes high-profile predictions that proved optimistic, citing **Dario Amodei**'s March prediction about code generation and **Geoffrey Hinton**'s 2016 comment about radiologists. Business Insider also cites a **2013** observational study in the **Journal of the American College of Radiology** that found radiologists spent **36.4%** of their time interpreting images, with the remainder on consults, supervision, patient care, and administration.\n\n### What happened\n\nBusiness Insider reports that, despite wide deployment of AI over the last four years, many roles widely labeled \"automatable\" remain intact. The article quotes **Benjamin Todd**, president of the nonprofit **80,000 Hours**, saying the portion of work that is \"defined\" and cleanly automatable is often a minority of a job's tasks. Todd is quoted: \"We should caveat all of this with 'survived so far.'\" The piece also references predictions by **Dario Amodei** (a March forecast about AI writing all code) and **Geoffrey Hinton** (a 2016 comment about radiologists becoming obsolete within five years). Business Insider cites a **2013 observational study** in the **Journal of the American College of Radiology** that reported radiologists spent **36.4%** of their time interpreting images, with the rest of their time spent on consults, supervision, patient care, and administrative duties.\n\n### Editorial analysis - technical context\n\nThe article highlights a practical point about task decomposition in real-world jobs versus benchmark tasks. Industry reporting often finds that machine learning systems excel on narrow, well-specified components of work, for example, image interpretation or code generation prototypes, but struggle with coordinating the heterogeneous, interactive, and exception-heavy tasks that occupy much of many roles. This is a pattern observed across deployments where partial automation yields productivity gains without wholesale role elimination.\n\n### Context and significance\n\nFor practitioners, the Business Insider reporting underscores that head-fake automation headlines can misrepresent operational reality. Observed patterns in comparable transitions show organizations frequently adopt AI to assist or augment incumbents rather than replace them outright. That dynamic affects project scoping, evaluation metrics, and deployment risk assessments: success metrics tied only to isolated tasks can overestimate end-to-end impact.\n\n### For practitioners, what to watch\n\nTrack measurements of end-to-end workflow time and exception rates rather than focusing solely on component accuracy. Watch reporting that breaks down time budgets for target roles, third-party evaluations of human-AI handoffs, and studies that quantify non-routine coordination work. Business Insider's piece provides a contemporaneous media snapshot, but more granular operational data will be necessary to assess long-term labor impact.\n\n## Scoring Rationale\n\nThis story is broadly relevant to AI practitioners because it tempers expectations about full job automation and highlights measurement priorities. It is not a technical advance or regulatory shock, so its practical importance is moderate.\n\nPractice interview problems based on real data\n\n1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.\n\n[Try 250 free problems](/problems)", "url": "https://wpnews.pro/news/ai-leaves-many-supposedly-automatable-jobs-intact", "canonical_source": "https://letsdatascience.com/news/ai-leaves-many-supposedly-automatable-jobs-intact-692e3222", "published_at": "2026-05-27 17:21:02.434069+00:00", "updated_at": "2026-05-27 17:21:05.282175+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "ai-research"], "entities": ["Business Insider", "Benjamin Todd", "80,000 Hours", "Dario Amodei", "Geoffrey Hinton", "Journal of the American College of Radiology"], "alternates": {"html": "https://wpnews.pro/news/ai-leaves-many-supposedly-automatable-jobs-intact", "markdown": "https://wpnews.pro/news/ai-leaves-many-supposedly-automatable-jobs-intact.md", "text": "https://wpnews.pro/news/ai-leaves-many-supposedly-automatable-jobs-intact.txt", "jsonld": "https://wpnews.pro/news/ai-leaves-many-supposedly-automatable-jobs-intact.jsonld"}}