Executive Summary

A study by Berkeley Haas professors shows: AI does not reduce workload, but intensifies it. Employees take on new tasks, work during breaks, and engage in multitasking. The effect is ambivalent – organizations become more productive, but employees experience greater sustained stress. The research refutes the assumption that AI time savings automatically lead to less work.

People

Topics

  • Artificial Intelligence & Work Environment
  • Productivity and Burnout
  • Organizational Management
  • Agentic AI

Clarus Lead

Work intensity increases despite AI tools. A study by Berkeley Haas professors Aruna Ranganathan and Ching-Chi Maggie Yi demonstrates: AI tools do not lead to shorter working hours, but rather to expanded task scopes and constant availability. The research observed 200 employees of a technology company from April to December 2025. Decision-makers must develop new management strategies to avoid sustained stress – a critical issue for personnel development and employee retention.

Detailed Summary

The study identifies three main forms of work intensification. First, task expansion: AI tools reduce knowledge gaps, enabling employees to assume responsibilities that were previously held by other specialists. Product managers write code, researchers take on engineering tasks. This boundary crossing initially feels empowering – new skills, immediate feedback – but accumulates into significantly larger job descriptions.

Second, blurring of work and leisure time: Because AI prompts are so simple, employees work during breaks, in meetings, or quickly send a prompt before leaving the desk. These moments accumulate into permanent availability.

Third, intensive multitasking: Multiple AI agents run in parallel while employees simultaneously work manually. The expectation that agents "never run empty" becomes a psychological burden.

The side effect is substantial: Engineers spend more time on code review and coaching colleagues who engage in "vibe coding." A "boiling frog" effect occurs – employees only realize in hindsight that downtime has disappeared and performance expectations have risen, without explicit instruction. Sleep deprivation and mental exhaustion are reported.

Key Messages

  • Work scope grows: Employees take on tasks outside their original domain; the job becomes broader, not shorter.
  • Permanent availability: AI usage during breaks and meetings leads to blurred boundaries between work and leisure.
  • Psychological stress despite productivity: Although people accomplish more, stress levels rise – the feeling of not doing enough persists.
  • Organizational consequences: Other employees must invest more time in review and quality control.

Critical Questions

  1. Evidence (Data Quality): The study is based on ethnographic observation of a single 200-person firm. How representative are these findings for larger enterprises or other industries? What control group without AI usage was compared?

  2. Conflicts of Interest (Independence): Ranganathan and Yi published in the Harvard Business Review. Are there conflicts of interest with AI providers or investors? What funding financed the research?

  3. Causality (Alternative Hypotheses): Could the observed work intensification also result from understaffing or raised expectations independently of AI? Was it measured whether companies deliberately assigned more work because AI was available?

  4. Feasibility (Solutions): The authors mention "intentional breaks" and "sequencing" as management strategies. How concretely do these measures work? Were they tested and what results do they show?

  5. Long-term Effects: Is burnout risk elevated? Are there staff turnover or resignations? The study ended in December 2025 – what longer-term data are available?

  6. Industry Specificity: The study was conducted in a "technology company" with positive AI attitudes. How do results differ in conservative or regulated industries (finance, healthcare)?

  7. Measurement Accuracy: How was "work intensity" operationalized? Is it based on self-report, system tracking, or objective metrics such as working hours?

Additional News

  • ByteDance Seed Dance 2.0: Chinese video generator with native audio synchronization outperforms Sora 2 and OpenAI models; first Chinese AI with genuine lip-sync.
  • OpenAI Launches Ads: ChatGPT now shows ads for free and $8 Go users; Pro/Enterprise/Education users remain ad-free; users can control data usage.
  • Databricks $7 Billion Funding: Revenue run rate at $5.4 billion/year (+65% YoY); one quarter from AI products; CEO positions this as refutation of the "SaaS is dead" narrative.

Source Directory

Primary Source: The AI Daily Brief – Podcast Episode on Berkeley Haas Study (Harvard Business Review; Transcript: 2026-02-11)

Study Authors:

  • Aruna Ranganathan, UC Berkeley Haas School of Business
  • Ching-Chi Maggie Yi, UC Berkeley Haas School of Business
  • Publication: Harvard Business Review (2026)

Supplementary Sources:

  1. Anthropic Agentic Coding Trends Report (2026)
  2. CNBC: OpenAI Model Announcement (Feb. 2026)
  3. Politico: White House AI Data Center Pact (Feb. 2026)
  4. Simon Willison Blog Post on Haas Study (2026)

Verification Status: ✓ 2026-02-11

This text was created with the support of an AI model. Editorial responsibility: clarus.news | Fact-checking: 2026-02-11