From Harvard Business Review · · 2m
AI Signals Psychological Safety
AI-generated concerns reveal hidden team tensions without attribution.
In brief
AI-generated concerns reveal hidden team tensions without attribution. Leaders using AI to surface concerns without naming them can detect psychological safety gaps before they become critical. Originally reported by Harvard Business Review.
Safety by proxy
Marcus called his team to present a regional expansion plan two days before an anonymous email arrived with a document listing concerns about the expansion.
“Someone had asked an AI tool to make the strongest case for the expansion and the strongest case against it.”
AI as anonymous survey
The anonymous email contained three pages of concerns about market saturation, customer base size, and unsecured partnerships that were previously unspoken.
“The concerns were not necessarily new. What was new was that they were now in writing, circulating, without a name attached.”
Safety by proxy definition
Safety by proxy describes using AI to surface concerns when people feel unable to raise them directly in person.
“The term describes a substitution, not a solution.”
AI’s perceived neutrality
AI outputs arrive with greater status and legitimacy than anonymous surveys, making them a powerful signal of psychological safety gaps.
“Safety by proxy describes the specific phenomenon that occurs when people use AI to surface a genuine, substantive concern when they do not feel safe enough to raise it directly.”
Why it matters
Research shows team psychological safety directly impacts how teams learn, adapt, and perform under pressure.
“One of us (Amy) conducted research on team psychological safety, which is a shared belief that the team is safe for interpersonal risk-taking in service of team goals.”
Leadership’s role
Leaders must model fallibility and respond non-defensively to dissent to build psychological safety.
“While recognizing that speaking up looks different for different people, and is shaped by personality, communication style, and comfort with direct challenge.”
Four patterns
AI’s perceived neutrality interacts with psychological safety in four distinct ways: workaround, equalizer, testing space, and mirror.
“If safety is absent, AI can serve as a workaround. If safety is unequal, AI can work as an equalizer. If safety is developing, AI can serve as a testing space. Because even the best leaders are separated by authority, AI can be used as a mirror.”
Workaround pattern
People using AI to surface concerns without owning them when speaking up carries personal cost.
“The concern enters the room. The person stays safe. AI takes the attribution, at least, for now.”
Equalizer pattern
AI helps underrepresented team members bypass communication barriers to voice concerns without attribution.
“An underrepresented team member can route a concern through AI and have it arrive without a name attached.”
Testing space pattern
Teams use AI to test arguments before raising them directly, signaling growing psychological safety.
“Whether this should be seen as accelerating psychological safety or as a substitution remains an open question.”
Mirror pattern
AI reveals hidden concerns by bypassing power dynamics that suppress honest input.
“The higher a leader’s authority, the more people modulate what they say to reflect what they believe the leader wants to hear.”
3M case study
3M uses AI to help scientists challenge marketing assertions without direct confrontation.
“Established scientists, those with deep technical expertise and years of credibility in their domain, are among the most enthusiastic users of AI-assisted analysis when the conversation moves outside that domain.”
AI stress-testing
Leaders should run AI stress-tests before decisions to identify concerns and build psychological safety.
“What are the strongest arguments against this direction? What risks haven’t been addressed? What assumptions haven’t been tested?”
Leadership response
Leaders must respond to AI-surfaced concerns with curiosity, not defensiveness, to build psychological safety.
“Interpret AI-surfaced concerns as a signal, not a threat.”
What happens next
Leaders who build psychological safety now can create durable cultural change as AI’s perceived neutrality erodes.
“Leaders who build genuine psychological safety now, while AI is still perceived as neutral, can create something durable.”
The main idea
Leaders using AI to surface concerns without naming them can detect psychological safety gaps before they become critical.
