McKinsey wonders how to sell AI apps with no measurable benefits

Original Article ↗ Hacker News Discussion ↗

AI and Headcount Reduction

  • Many commenters note a gap between the sales pitch (“copilots reduce staff”) and reality: companies adopt AI but rarely cut headcount.
  • Engineers often see only modest productivity gains (5–15%), which don’t map cleanly to firing discrete people; extra capacity tends to get absorbed as more work.
  • Where AI does replace people, it’s often outsourced or low-status roles (e.g., translation, L1 support, basic data entry) rather than core internal staff.

Management Incentives and Organizational Politics

  • Multiple comments argue managers are structurally disincentivized to reduce headcount because power and status correlate with team size.
  • Headcount cuts are more likely via top-down mandates (layoffs, RTO) than through careful AI-driven efficiency projects.
  • Some suggest AI projects fail mostly for the same political and organizational reasons traditional IT projects did, not because of technical limits.

Who Is Actually Replaceable?

  • A recurring theme: AI is most capable of mimicking middle management (emails, meetings, slideware), but those are the people who decide what gets automated.
  • Others push back, saying executives make many non-visible decisions and must be held legally liable, so “AI C-suite” is unrealistic.
  • Several argue real inefficiency lies in the “big fat middle” of organizations rather than individual contributors.

Real vs Hype Use Cases

  • Practically useful cases mentioned: coding assistance, content moderation pre-filtering, customer-support deflection, document and email summarization, and “grunt work” automation.
  • Many embedded AI features in mainstream tools (PDF readers, Notion, Office, contract signing) are described as intrusive, low quality, or simply in the way.
  • AI is framed as paradoxical: clearly useful in some workflows, yet often not measurably improving overall productivity or justifying its cost.

Vendor Strategies and AI Everywhere

  • Commenters see a rush to “stuff AI into every feature” as investor- and marketing-driven more than user-driven.
  • Concerns include vendor lock-in, future price hikes once customers depend on AI workflows, and “enshittification” after adoption.

Consultants, Measurement, and Bubble Risk

  • Several point out the irony of McKinsey warning about unmeasurable AI benefits, given consulting value is itself hard to quantify.
  • There is skepticism that reported “ROI” figures are methodologically sound; if only 30% can even claim quantified ROI, true returns may be much lower.
  • Some foresee an AI bubble deflating via a series of disappointments as promised cost savings fail to appear, even while AI tools persist as everyday utilities.