Why AI Leads to More Work, Not Less

What It Covers

Berkeley Haas researchers studying a 200-employee tech company from April to December 2024 found AI power users work more intensely, not less. The study identifies three forms of work intensification: task expansion into others' roles, blurred work-life boundaries, and increased multitasking, challenging assumptions about AI-driven productivity gains.

Key Questions Answered

• •Task Expansion Effect: AI enables workers to perform tasks previously requiring specialists. Product managers write code, researchers handle engineering work, and employees tackle responsibilities they would have outsourced before. This creates intrinsic rewards through new capability mastery but generates spillover effects, requiring engineers to spend more time reviewing and correcting AI-assisted work from colleagues who are vibe coding without full technical expertise.
• •Ambient Work Phenomenon: Workers prompt AI during lunch breaks, meetings, and file loading waits, sending final prompts before leaving desks so agents work during absence. This habitual behavior eliminates recovery time, making work feel unbounded and ambient rather than discrete. The AI rhythm raises speed expectations through normalized visibility of what becomes possible, not through explicit demands from management.
• •Multi-Agent Management Burden: Single agents evolve into coordinated teams, creating new pressure to constantly deploy agents. Users report losing sleep trying to push one more feature with one more prompt, feeling exhausted after one to two hours of parallel work across multiple projects. The shift from assisted to agentic AI intensifies feelings of underutilizing a capable team when agents are not running.
• •Organizational Expansion Opportunity: Companies using AI to expand capabilities rather than cut costs will win long-term. Winners view AI as opportunity-creating technology enabling new product lines, revenue streams, and market categories rather than efficiency technology for doing the same with less. Databricks demonstrates this with 5.4 billion dollar revenue run rate, up 65 percent year over year, with 25 percent from AI products launched during their agentic transformation.
• •Democratized Coding Impact: Nontechnical use cases expand across organizations as coding capabilities democratize beyond engineering departments. Domain experts implement solutions directly, extending productivity gains organization-wide. This trend suggests companies will hire dedicated vibe coders for non-engineering issues, internally deployed specialists helping different departments use software to solve problems without traditional technical skills or engineering department dependency.