MIT researchers found that LLM-assisted writers showed lower brain engagement and performed worse once the tool was removed. Dr. Vishal Kapoor argues organizations need to measure not just what AI automates, but whether their people can still reason without it.