{
  "id": 3212051,
  "title": "AI’s productivity paradox: We are electrifying the old factory",
  "url": "https://urgent.news/2026/08/25/ais-productivity-paradox-we-are-electrifying-the-old-factory",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-25T06:04:24.000Z",
  "source": {
    "name": "ITWeb",
    "slug": "itweb",
    "url": "https://www.itweb.co.za/article/ais-productivity-paradox-we-are-electrifying-the-old-factory/KWEBb7yLbopvmRjO"
  },
  "original_language": "en",
  "account": "Artificial intelligence (AI) is rapidly infiltrating the business world, yet productivity gains are not keeping pace. According to research from the National Bureau of Economic Research, 69% of businesses are actively utilizing AI, but a staggering 89% report no impact on labour productivity over the past three years. Despite this, executives expect AI to boost labour productivity by an average of 1.4% over the next three years. This discrepancy suggests that business leaders believe the productivity benefits of AI are still forthcoming.\n\nThe rise of AI capabilities, adoption, and investment is undeniable, while research is increasingly demonstrating productivity improvements in individual tasks. However, these improvements have yet to translate into measurable productivity across organizations. This does not necessarily indicate that AI is failing, but rather that the technology may be advancing faster than the organizations adopting it.\n\nEconomist Paul A. David explored a similar issue three decades ago, comparing AI to electrification. Industrial factories were built around steam power, and when electric motors arrived, manufacturers initially introduced them into factories designed around steam limitations. Initially, the technology change outpaced organizational adaptation, leading to limited productivity gains. However, when manufacturers redesigned factories around electricity, larger productivity gains emerged.\n\nThe parallels between AI and electrification provide valuable insight into today's AI productivity debate. While AI can undoubtedly enhance individual productivity, translating these gains into broader enterprise productivity is proving more challenging. McKinsey's global AI research found that while there is widespread organizational adoption of AI, only 39% of respondents attribute any enterprise-level impact to AI. This highlights an important distinction: AI can make tasks more efficient without translating into significant enterprise-level productivity gains.\n\nIf AI reduces the preparation time for a four-hour management report to two hours, task productivity has improved dramatically. However, if the same report, approvals, meetings, and workflows remain unchanged, it remains unclear how much of those two hours the enterprise actually captures. This may explain why AI productivity appears impressive yet disappointing.\n\nThe \"electrifying the old factory\" analogy may not be entirely accurate. AI is not electricity, but it can change the cognitive organization of production by generating and interpreting information, making recommendations, and increasingly taking actions. Additionally, measurement problems, early general-purpose technology diffusion, and limitations in reliability, verification, and accountability in high-value environments may contribute to the apparent productivity gap.\n\nOrganizational design is likely one factor executives can directly influence to bridge the gap. McKinsey's research found that workflow redesign has the strongest relationship with an organization's ability to see EBIT impact from generative AI. However, only 21% of respondents using generative AI reported that their organizations had fundamentally redesigned at least some workflows.\n\nEnterprise AI maturity can be understood through three stages. The first stage is AI-assisted work, where humans remain central to existing workflows, and copilots and assistants improve individual performance. The unit of productivity remains the employee. The second stage is AI-enabled processes, where AI begins operating across workflows, retrieving information, making recommendations, triggering actions, and coordinating systems. Productivity increasingly becomes measured in terms of cycle time, quality, throughput, and cost-to-serve. The third stage is the AI-native operating model, where organizations question whether existing processes must endure. For example, instead of merely asking where AI can be inserted into eight steps of a procurement process containing multiple systems and approvals, the AI-native question is: Why do we need eight steps at all?\n\nThis shift—from automating existing work to reconsidering why the work is structured that way—may contain the larger productivity opportunity. However, this requires a fundamental reevaluation of organizational design, workflow structure, and the role of humans in the AI-driven enterprise.",
  "summary": "AI productivity appears simultaneously impressive and disappointing, as it is advancing faster than the organisations adopting 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."
}