Source: Noahpinion
As AI systems become capable enough to handle complex work but unreliable enough to need constant correction, a new class of labor emerges—people whose primary function is prompt engineering, output validation, and behavioral steering rather than the underlying work itself. This mirrors previous tech transitions where new tools created new overhead roles (quality assurance after manufacturing automation, content moderation after social platforms), except this time the overhead might rival the productive capacity of the systems themselves, potentially offsetting efficiency gains. Organizations that win will be those that either build reliable AI systems that need less supervision, or develop efficient workflows that don't treat human-AI collaboration as a one-to-one babysitting arrangement.