{
  "id": 1839543,
  "title": "Offloading work tasks to AI comes with a cost – to our brains",
  "url": "https://urgent.news/2026/08/19/offloading-work-tasks-to-ai-comes-with-a-cost-to-our-brains",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-19T00:38:11.000Z",
  "source": {
    "name": "The Conversation AU",
    "slug": "the-conversation-au",
    "url": "https://theconversation.com/offloading-work-tasks-to-ai-comes-with-a-cost-to-our-brains-289384"
  },
  "original_language": "en",
  "account": "In a world where high-stakes work tasks are increasingly completed using generative artificial intelligence, the cost to our cognitive abilities is becoming increasingly apparent. Take, for example, South Africa's Draft National AI Policy, which had to be withdrawn due to fabricated citations and non-existent sources, errors that were introduced by AI. Similarly, Deloitte faced a significant financial setback when their use of generative AI to produce a commissioned document led to the creation of fabricated citations and quotes, costing Australian taxpayers A$440,000. These incidents serve as stark reminders of the hidden costs of relying on AI to do our work.\n\nThe issue lies in our modern workplace culture, which prioritises the final deliverable over the process of obtaining it. This outcome-driven approach rarely rewards the actual thinking that goes into producing a quality output. By skipping the process, we lose the ability to fully grasp the meaning and implications of our work, making it difficult to critically evaluate the information we consume and discern facts from partial truths and lies.\n\nEmpirical evidence further highlights the disconnect between our perception of AI's benefits and its actual impact on the quality of our work. A 2025 study by the University of Melbourne and KPMG found that while 66% of respondents use AI regularly, more than half believe their performance improves with its use. However, 61% of these respondents have had no AI training, and 60% report using AI inappropriately, complacently, and without transparency. Moreover, a study from July 2026 by the Centre of AI Safety found that even the top-performing AI agents fail to complete approximately 85% of projects to a standard acceptable for commissioned work.\n\nTo address this problem, we need to redefine our working relationship with AI and take a more human-centric approach. This involves setting realistic time frames for deliverables that allow for genuine thinking rather than quick, algorithmic production. Encouraging traditional brainstorming sessions, such as whiteboard sessions, can help facilitate this process by encouraging human synthesis rather than relying solely on algorithmic prediction.\n\nAdditionally, fostering a culture of constructive criticality is essential. This means actively questioning the logic, structure, references, and assumptions behind all outputs, whether generated by humans or AI. By expecting teams to defend the reasoning and methodology behind their work, we can quickly identify who is actually doing the thinking and who is merely copying and pasting from an AI agent. Creating critique groups can provide a safe and comfortable environment for open feedback and discussion.\n\nWhile technology can certainly assist with certain processes, it is crucial to remember that the most valuable asset we possess is our ability to think critically. The polished final page may be impressive, but the ebbs and flows of the cognitive process required to genuinely write and evaluate content are what truly make knowledge work valuable.",
  "summary": "We risk replacing genuine knowledge with “work slop”.",
  "key_points": [
    "South Africa's Draft National AI Policy withdrawn due to AI errors",
    "Deloitte's AI-generated document cost Australian taxpayers A$440,000",
    "61% of AI users report inappropriate, complacent, untransparent use"
  ],
  "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."
}