By VADM David Lewis (USN, Ret.), President of the American Society of Naval Engineers

The MIT report, "Humans in the Loop: The evolution of work in early experiments with Generative AI," cuts through the hype around generative AI and lands on a more consequential reality: this is not (yet) a story of job replacement, it is a story of job redesign. Across industries, the dominant pattern is a shift from execution to supervision. Workers are no longer primarily doing the task; they are overseeing, validating, and troubleshooting AI-generated outputs.

That sounds incremental, but it is not. It represents the spread of “supervisory control” work, long familiar in aviation and industrial systems, into white-collar domains like law, medicine, finance, and engineering. The real transformation is not fewer workers, but different work: judgment, context, and accountability becoming the scarce and valuable skills.

The report identifies three core use cases: relieving bottlenecks, integrating expertise, and accelerating learning and shows that outcomes depend heavily on how organizations deploy AI against them. In some cases, AI is genuinely skill-leveling; in others, it increases the premium on deep domain expertise to validate outputs. There is no “AI autopilot” for organizations: the most successful applications solve pre-existing problems, combine multiple technologies, and retain humans firmly in the loop.

The takeaway for leaders is clear: generative AI is not a plug-and-play productivity tool. It is a management and design challenge. Those who treat it as such, focusing on job design, training, and integration will capture its value. Those who don’t will get noise, not advantage.

Main image credit: Photo by Maj. Lindsay Roman