{
  "id": 13119799,
  "title": "Nobody has been in charge for decades of the single most important part of the AI revolution: documentation",
  "url": "https://urgent.news/2026/10/09/nobody-has-been-in-charge-for-decades-of-the-single-most-important",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-09T12:20:00.000Z",
  "source": {
    "name": "Fortune",
    "slug": "fortune",
    "url": "https://fortune.com/2026/10/09/ai-data-documentation-biggest-problem/"
  },
  "original_language": "en",
  "account": "Decades ago, nobody held the power to oversee the most crucial aspect of the AI revolution: documentation. Today, AI agents engage with company knowledge at a rate nearly twice that of human developers. However, this does not imply that developers have stopped consulting documentation; instead, they are increasingly delegating tasks to agents. Developers may request assistance in wiring up an API, buyers could ask a chatbot for vendor comparisons, or customers might inquire about payment failures via a support bot. The truth remains that these agents rely on your company's documentation, help center, and policies, taking whatever is found at face value to respond to requests. The shift in who your documentation serves is fundamental, and many companies continue to write for an outdated audience. Every company possesses authoritative knowledge, but much of it is outdated or contradicting. For instance, a setup guide written a month ago and never updated may still be used, or a pricing page may contradict the updated help center. While a human might notice something feels amiss and seek clarification from a colleague, an AI agent would not. It retrieves the version it found, writes it into code or a support answer, and repeats the mistake thousands of times before anyone can correct it. Developers are already experiencing this frustration, with 66% of them citing AI output that is close to correct but still wrong as their top issue. This is not solely the fault of the models, but also the material being fed to them. Therefore, the question arises: who is responsible for this problem? At most companies, nobody currently bears this responsibility. Documentation has traditionally been viewed as a publishing task, which sufficed when readers could fill in the gaps. However, with the rise of AI agents, this approach no longer works at scale. Instead, a new approach resembling infrastructure work is emerging. Someone must determine which source is authoritative when conflicts arise, someone must connect product changes to the corresponding content, ensuring that documentation updates automatically when the product evolves. Someone must structure information in a way that allows agents to retrieve it cleanly, and monitor which queries agents struggle with, as these gaps directly indicate areas where customers receive inaccurate answers. This new role is known as knowledge engineering. It entails more than just producing pages; it involves maintaining accurate, current, and usable knowledge for both humans and machines. This shift in roles mirrors the changes occurring in radiology. In 2016, AI pioneer Geoffrey Hinton warned against training radiologists, predicting they would become obsolete. However, Mayo Clinic, recognizing the significance of AI, increased its radiology staff by 55% between 2016 and 2025, and U.S. radiology residency programs offered a record number of positions in 2025. Jensen Huang emphasized that the misconception of AI merely being a scanning reader is misguided. In reality, radiologists now spend less time recognizing images and more time attending to patients, and demand for their services continues to grow. The practical shift lies in how success is measured. Page views and time spent on a page indicate human engagement but offer little insight into whether an AI agent provides the correct answer. Companies excelling in this transformation monitor different signals: the queries agents run, instances where retrieval fails, situations where sources contradict each other, and the speed at which product changes are reflected in the knowledge base. Once, documentation served as a reference consulted by individuals. Now, it functions as an input that machines act upon, often without human intervention. Companies that anticipate success will treat their knowledge as the infrastructure it has become. The views presented in this commentary are solely those of the author and do not necessarily reflect the opinions or beliefs of Fortune.",
  "summary": "Every company has knowledge that looks authoritative, but actually isn’t. And now we're training AI models on it.",
  "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."
}