AI is eliminating the friction that holds teams together. Product designers skip researchers, PMs skip designers, engineers skip accessibility teams. The workflow gets faster. The team gets thinner. A 2025 study from Harvard, Columbia, and Yeshiva University found that AI-driven automation decreased overall team performance and increased coordination failures, with the largest effects hitting low- and medium-skilled teams first. Automation also decreased team trust. This is not a soft finding. It is a measurable collapse in the connective tissue of how work actually gets done.
The research trail behind this argument is worth your time. MIT's Human Dynamics Lab found in 2012 that informal communication, not meetings, was the top predictor of team productivity, with the highest-interaction teams showing 35% more successful outcomes. Google's Project Aristotle studied 180 teams and identified psychological safety built through low-stakes micro-interactions as the single strongest predictor of high performance. A 2024 Korean study on private-sector innovation found that weak ties, the occasional cross-team conversations, sustained innovative output more reliably than core team collaboration. McKinsey puts the cost of disengagement and attrition at between $228 million and $355 million annually for a median S&P 500 company. The 'bug-free' workflow has a price tag.
The article does not argue against AI. It argues for precision. Use AI to eliminate repetitive toil. Protect the interactions that build belonging, surface misalignment, and create the conditions for breakthroughs. The piece also introduces a finding from a March 2026 HBR study of 1,488 workers that deserves attention on its own: a concept called 'AI Brain Fry,' where 34% of workers experiencing cognitive exhaustion from AI use intended to quit. The efficiency gains are real. So is what you are trading away to get them.
[READ ORIGINAL →]