A new paper argues that the real bottleneck on AI adoption isn't the models. It's whether people and organizations can absorb the disruption without losing the plot.
That's the case made in "Human Resilience in the AI Era, What Machines Can't Replace" (arXiv:2510.25218), which treats resilience as a trainable skill built on three layers: psychological (staying goal-directed under stress), social (getting trusted feedback and support from a group), and organizational (turning a caught problem into an actual fix rather than a postmortem slide). The authors connect older resilience and technostress research with newer experiments on people working alongside AI tools. Those experiments found AI assistance can raise productivity and spread expertise to less-skilled workers, and in some cases improve expressed empathy and more calibrated trust in the tool. The paper turns those findings into recommendations for AI training, workplace design, governance, and evaluation.
Most AI coverage obsesses over capability and who's ahead on benchmarks. This paper's point is that the actual constraint on deploying AI well is often the humans and institutions around it, not the model itself. Companies that skip training and support and just bolt on a chatbot are usually the ones whose rollouts backfire hardest.
It's a useful reframe, but "resilience" is doing a lot of work here. The paper itself admits the AI-specific evidence is thinner than the general resilience research it leans on.