Generative AI may be creating more work than it saves

Abstraction: Wharton professor argues LLM backends create more labor than they save

Key points:

• Wharton management professor Peter Cappelli argues that on a cumulative basis generative AI creates more work than it saves, due to heavy backend labor for building and sustaining LLMs
• Many LLM use cases are overkill: existing form letters and rote automation already handle most business correspondence, and AI-generated text still requires legal review
• LLM infrastructure costs (compute, electricity, database management, guardrails) will rise as adoption increases
• Validating AI outputs requires domain experts — using other LLMs for validation is a reliability issue, not a validity solution
• Reliability problem: the same prompt can yield different LLM responses, creating "dueling reports" that organizations must adjudicate
• People tend to ignore algorithmic recommendations in favor of personal judgment, undermining investment in decision-support AI

Connections: Wharton School · Mit · Generative AI · Large Language Models · AI Productivity

Source: https://www.zdnet.com/article/generative-ai-may-be-creating-more-work-than-it-saves/