Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality.

This article investigates the concept of a "jagged technology frontier" to describe the uneven effects of artificial intelligence (AI) capabilities on knowledge work, where AI assistance improves performance on some tasks but degrades it on others within the same workflow. Conducted in collaboration with the global management consulting firm Boston Consulting Group (BCG), the preregistered randomized experiment involved 758 highly skilled management consultants performing realistic consulting tasks either without AI, with GPT-4 AI assistance, or with GPT-4 plus a prompt engineering overview. Results show that for tasks within GPT-4’s capability frontier—spanning creative, analytical, writing, and persuasive activities—AI use increased task completion by 12.2%, improved quality by over 30%, and reduced completion time by about 25%. Conversely, for a complex managerial task designed to lie outside the AI frontier, AI assistance reduced correctness by 19 percentage points despite improving the persuasiveness and coherence of responses, highlighting risks of overreliance on AI in tasks beyond its current capabilities. The study underscores the importance of human judgment in discerning when AI augmentation is beneficial and calls for organizational strategies that consider the jagged nature of AI’s capabilities across knowledge workflows.