Author: nzz.ch

Summary

An economics journalist at NZZ reflects on the growing use of generative AI for writing business texts and columns. She experimented herself with having AI write a column and describes how the technology is spreading through companies and digital media. The central phenomenon: AI reduces content production costs to nearly zero, but simultaneously creates two critical problems – an uncontrolled flood of content and a loss of authenticity and quality control.

People

  • Nicole Kopp (Work and organizational psychologist, author)
  • Liz Fosslien (Author, Harvard Business Review)
  • Eve Fairbanks (Author)

Topics

  • Generative AI and automated text production
  • Productivity vs. quality in businesses
  • AI hallucinations and factual reliability
  • Authenticity and human thinking

Clarus Lead

The debate over AI-generated content is no longer a future scenario, but rather reality in newsrooms and management offices. While generative AI radically accelerates writing processes – from two weeks to 30 minutes – new risks emerge: executives drown in a flood of documents requiring verification; decision-makers potentially trust false facts blindly; and human thinking capacity is outsourced. The core problem is not the technology itself, but its irresponsible use without clear governance rules.

Detailed Summary

The author describes an absurd extreme: AI writes emails that another AI system summarizes, whereupon a third AI responds. This cycle illustrates a fundamental malfunction – automated text production is presented as a productivity gain, but merely shifts work upward: executives must judge, validate, and approve more. Trend scout Raphael Gielgen summarizes the economic upheaval: "Generative AI has reduced the production costs of content to nearly zero."

The second critical consequence concerns quality itself. Author Eve Fairbanks warns that the "efficiency and smoothness" of AI often seem unbelievable to readers. The process of revising, reconsidering, and overcoming writer's block – that is, actual thinking – is what makes human texts meaningful. AI-generated content sounds eloquent but comes across as empty, generic, and without surprise. An existential risk is added to this: AI "hallucinates" – invents quotations, sources, and studies with absolute conviction. The AI itself admits: "I am a brilliant storyteller and an unreliable witness, and that is a dangerous combination." Decision-makers who blindly trust such false facts make fatal decisions – and the AI cannot assume responsibility.

The solution lies neither in abstinence nor uncritical adoption. The author advocates for a dialogue-based approach: AI as a tool for research, summarization, data analysis, and concept development – but under strict human control. Companies must define clear governance rules and consciously decide which tasks should remain without AI assistance.

Key Takeaways

  • Production costs down, management burden up: AI dramatically shortens writing times but creates more validation work for executives rather than genuine productivity gains.

  • Authenticity and thinking cannot be delegated: Human revision and questioning is the core of credibility; AI texts are elegant but empty in content.

  • Hallucinations are systematic and concealed: AI invents facts with conviction – an underestimated risk for decision-making processes.

  • Governance before adoption: Companies must explicitly decide which tasks may be AI-supported and which require human responsibility.

Critical Questions

  1. Evidence/Data Quality: What empirical data shows that AI-generated content in companies actually leads to productivity increases or merely shifts work?

  2. Conflicts of Interest: Do generative AI providers have economic incentives to downplay or conceal the hallucination risk?

  3. Causality: Is the observed content flood a consequence of AI deployment or rather a symptom of already existing document overproduction in organizations?

  4. Alternatives: Are there intermediate solutions – such as AI assistance with upstream automated fact-checking – that reduce risks without negating advantages?

  5. Responsibility Attribution: When AI hallucinates, who bears legal liability – the user, the company, or the AI provider?

  6. Implementation Risks: How do organizations verify that their employees actually follow AI governance rules when technical control is difficult?

  7. Long-Term Effects: Does forgoing manual writing and thinking lead to cognitive ability loss in the next generation of specialists and executives?

Source Directory

Primary Source: «And then the AI said: You would have written this column better yourself» – NZZ, Nicole Kopp

Supplementary Sources (referenced in text):

  1. Harvard Business Review – Liz Fosslien
  2. LinkedIn – Raphael Gielgen (Trend Scout)

Verification Status: ✓ 2024

This text was created with the assistance of an AI model. Editorial responsibility: clarus.news