10 Levers for Shaping Generative AI That Truly Improves Worker Performance
Draws on Humans in the Loop, from the MIT Working Group on Generative AI and the Work of the Future, built on interviews at more than 20 companies across healthcare, retail, finance and manufacturing. It sets out three operating principles — gather evidence before scaling, one size does not fit all, and learn when to trust — alongside seven outcomes including minimising drudgery, promoting learning, preserving teamwork and continuing to invest in domain expertise. The researchers warn specifically against “mental offloading”, where workers bypass the learning a task would otherwise have produced.
Deployment decisions are also learning-design decisions: a rollout that strips out the effortful part of a task removes the mechanism that built the judgment needed to tell when the output is wrong. Capability builders should name which work deliberately stays manual in the areas where domain expertise is the differentiator, and treat evidence-before-scaling as a precondition rather than a post-hoc review. Note the framing is September coverage; the underlying report dates from April 2026.
Save this story, pin it to the top of your feed, or keep a private note on it.
Sign in