Effects of AI on Knowledge Worker Productivity and Quality

Research studied 750+ consultants using AI. It increased productivity 12%, quality 40%, and speed 25% for tasks AI could do. But for some work outside its abilities, AI decreased accuracy 19%. The "jagged frontier" of AI capabilities makes its effects unpredictable.

Effects of AI on Knowledge Worker Productivity and Quality

A recent study took an in-depth look at how artificial intelligence (AI) is affecting the work of knowledge workers. Researchers partnered with Boston Consulting Group (BCG), a large global consulting firm, to conduct randomized controlled experiments with over 750 of their employees. The goal was to evaluate real-world impacts of AI on productivity, quality, and work processes.

Boosting Productivity and Quality

The experiments included tasks that were designed to be well-suited for current AI capabilities. When consultants had access to AI for these "within frontier" tasks, the results were striking. On average, they completed 12.2% more work compared to those without AI access. Quality of their work also dramatically improved - independent reviews found it to be over 40% higher. Speed increased significantly as well, with AI users finishing tasks 25.1% faster.

Interestingly, AI seemed to help participants at all skill levels. Those below average saw quality rise 43% with AI assistance. Even top performers still benefited, with quality boosting 17%. So while AI reduced differences, it elevated standards across the board.

Challenges Outside Current AI Capabilities

To study limitations, researchers selected one task believed to be outside what AI could readily handle. For this "out of frontier" work, AI access backfired. Consultants were 19% less likely to solve it correctly versus the control group without AI. Quality also declined for incorrect answers when relying on AI rather than their own expertise.

However, AI still improved how good recommendations sounded - it enhanced qualitative aspects regardless of solution accuracy. This nuanced impact highlighted that effects are multi-faceted, not simply about right versus wrong answers.

The study introduced the concept of AI abilities forming a "jagged frontier" that varies unpredictably between similar tasks. Two main patterns emerged for successfully working with AI: "Centaurs" who strategically split tasks between human and machine strengths, and "Cyborgs" who tightly integrated their work with AI.

More research is still needed to understand which integration strategy works best in different situations. But the experiments provided useful insights on current realities of AI adoption. As capabilities continue advancing rapidly, these types of impacts can only be expected to grow in both scope and significance for knowledge-intensive work. Understanding optimal human-AI collaboration practices will become increasingly important for professionals and organizations.

In summary, this in-depth empirical research demonstrated AI's transformative potential to dramatically boost productivity, quality and speed for certain knowledge work when applied appropriately. However, discerning where current technological frontiers lie remains challenging. Continued study of effective integration models can help maximize AI's benefits as its abilities expand in the future.


Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality by Fabrizio Dell'Acqua, Edward McFowland, Ethan R. Mollick, Hila Lifshitz-Assaf, Katherine Kellogg, Saran Rajendran, Lisa Krayer, François Candelon, Karim R. Lakhani :: SSRN