Can AI Actually Break Down Every Job Into Learnable Tasks?
Source: JOSH BERSIN
Job-Task Analysis—a decades-old industrial engineering discipline—is experiencing renewed vendor interest because it allows framing AI as capable of replacing specific work functions rather than entire roles, making the threat appear more surgical and less sweeping. The friction isn't methodological; task decomposition has been standard since the 1960s. The actual problem is that most jobs contain tacit knowledge, judgment calls, and context-dependent decisions that resist clean categorization. Vendors and executives have stronger incentives to ignore this friction—cleaner ROI narratives, simpler automation pitches—than to acknowledge what remains stubbornly human. This creates real consequences. Executives betting on task-level automation tend to miscalculate implementation costs and underestimate retraining needs. Workers in roles that *are* easily decomposable face genuine displacement risk regardless of whether the AI performs as promised.