What: A KPMG and University of Texas study shows that AI performance depends less on AI fluency than on human judgment inside AI workflows.

Why it is important: Human judgment, workflow design and process-based assessment are becoming decisive factors in whether AI improves performance.

A Harvard Business Review article based on a KPMG and University of Texas field study finds that early-career employees create value with AI not through AI fluency alone, but through how they direct, evaluate and refine AI output. The study involved 523 U.S.-based KPMG professionals using an AI agent on business-specific tasks benchmarked against an AI-only baseline.
The research identified three profiles. AI amplifiers, representing 50.1% of participants, outperformed the AI baseline by framing problems clearly, applying domain frameworks, challenging assumptions and refining results. AI delegators, at 25.8%, produced work comparable to AI alone by accepting outputs with limited scrutiny. AI apprentices, at 24.1%, performed below the baseline despite strong foundational skills, because their critiques often failed to improve the work. The findings suggest that domain knowledge, critical thinking and AI literacy are necessary but insufficient. Organisations should redesign early-career development around task-based training, visible judgment and workflow-based assessment, evaluating how employees interact with AI rather than only the final deliverable.

IADS Notes: The KPMG and University of Texas study sharpens a growing workforce lesson: AI fluency only creates value when employees know how to direct, challenge and improve machine output. In June 2026, BCG argued that AI upskilling must be embedded in real workflows to translate capability into performance, with judgment, collaboration and problem-solving treated as core skills. Another BCG analysis in June 2026 warned that widespread AI use can erode critical capabilities such as judgment and problem framing if organisations do not design deliberate practice into work. Harvard Business Review’s March 2026 work on expertise similarly showed that generative AI does not make employees experts without training and repeated application, while its March 2026 analysis of entry-level jobs warned that automating junior work can weaken the experiences through which future leaders develop. Seramount’s June 2026 framework reinforced the same boundary: AI can automate and augment many tasks, but decisions involving people, ethics, capability building and accountability still require human judgment.

Why some junior employees work well with AI—and others don’t

Shaping early-career success in the age of AI - KPMG study