{
  "id": 20907083,
  "title": "We lead MIT’s FutureTech program on the economics of AI: Don’t ask if AI can do the job, ask how much it costs",
  "url": "https://urgent.news/2026/11/20/we-lead-mits-futuretech-program-on-the-economics-of-ai-dont-ask-if-ai",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-20T13:00:00.000Z",
  "source": {
    "name": "Fortune",
    "slug": "fortune",
    "url": "https://fortune.com/2026/11/20/ai-automation-cost-jobs-mit-futuretech-neil-thompson/"
  },
  "original_language": "en",
  "account": "Every week brings major advances in AI’s ability to do tasks previously exclusive to humans, raising concerns of widespread job losses for knowledge workers. Research at MIT’s FutureTech program suggests the impact of AI will likely be more nuanced, with cost playing a major role instead of just capabilities. The team studied computer vision first, finding that only 23% of wages paid for those tasks were worth automating at the time. The reason: fixed costs. Building a system costs the same regardless of how many workers it replaces. For example, a large bakery may choose to automate ingredient spoilage detection, while a smaller corner bakery might not be able to afford the upfront fixed costs, keeping its bakers.\n\nGenerative AI, the technology powering large language models, is different. A general-purpose model sold by the token is cheaper to start using than a custom vision system, making the fixed-cost barrier lower. However, experienced workers found that the most capable models produced usable output about two-thirds of the time. That success rate can mask two realities: AI excelling at some tasks every time and failing at others. Regardless, firms must determine where AI can be trusted and pay people to review its work or complete the remaining tasks. This additional cost is often the most significant step in moving from \"usually right\" to \"reliably right.\" AI is generally a collaborator rather than a replacement, steadily improving across most tasks, not overnight. Nevertheless, gradual automation changes the worth of work, as seen when GPS accessibility reduced taxi drivers' expertise and wages, while software taking over routine proofreading increased remaining work expertise and wages.\n\nBusiness leaders should consider AI's full cost per task, including review and correction, not just demos. Policymakers should track AI's cost and reliability, not just capabilities. Workers and employers should also assess which part of a job AI takes, as that determines whether the remaining work becomes more or less valuable. Firms will hand work to AI when it's cheaper than a person, but the question should be about the cost, not whether AI can do the job. This decision will vary task by task and firm by firm, deciding whose pay rises or falls.",
  "summary": "Firms will hand work to AI when it is cheaper than a person, not when a demo says it can do the job.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}