Psychological Safety by Design
Safety is a structural outcome. Organizations keep funding it as a culture initiative, and then read the quiet that follows as a people problem.
Silence has two causes, and they need opposite repairs.
Safety is not the reward for competence. It is the precondition for building it.
Most silence lives in the space between roles, not inside them.
Safety that depends on one person in the room is a mood, and moods do not scale.
A specialist sits in a monthly review and says nothing. She knows the number on the slide is wrong. She built her own version of that model alone, because no standard existed and the five colleagues she asked gave her five different answers. She has no language for the objection and no standing to make it. The number moves forward into planning, and into an investment review that prices the next five years. She stays quiet. The room reads it as temperament.
Silence is not a feeling. It is an output. Something produced it.
Values get posted. Leaders are asked to model vulnerability. Workshops run on speaking up. Then people return to an operating model where speaking costs more than it returns.
Organizational research has documented one producer for decades. People stay quiet when they expect that speaking will be ineffective, risky, or unwelcome, and those expectations are learned from the system rather than carried in from personality. That is the silence of cost, and it is the one most organizations are trying to solve.
There is a second one, and I have never seen it named. A person cannot object to something they do not know reaches them. The topic in the room was never connected to anything they own, so there is nothing for them to say. Not fear. Not indifference. The connection was never drawn.
They cannot predict what happens next
Quiet in the room, articulate in the hallway. The person has the objection and withholds it, because they cannot predict what speaking will do to them. The room set the price, not the person.
They do not see how it touches their work
Quiet everywhere, and it reads as disinterest. A few people map those connections on their own. Most people see what they have been shown, and no one showed them.
The two look identical in a room and need opposite repairs. One requires changing what speaking costs. The other requires building the connection first. Diagnosing the wrong one wastes the intervention.
The first thing I change is never the culture. It is the conversation nobody is having about what is actually expected, regardless of role, regardless of where the org chart moves.
The first fracture is never dramatic. It is the gap nobody names.
Everyone senses the distance between what people believe is expected and what anyone is willing to say out loud. Nobody names it. So the organization lives in limbo, repeating the same patterns under different leaders and different reorganizations, holding the same unspoken agreement under a new name each time.
Then the reorganization arrives and the reset lands. The person who invested in a relationship learns the lesson again: trust here does not compound. They do not leave because the work stopped mattering. They leave because the system taught them that investment does not carry forward.
Both silences concentrate in four places. Each one is visible before anyone reports it, and none of them are cultural.
Undefined roles
Ask someone to describe their work and they list tasks instead of naming what they owe. They feel the misfit. They have nothing precise to say about it, so they absorb whatever the nearest business asks for.
The unclaimed middle
Roles are defined. The space between them is not. Nobody is sure whose problem the handoff is, so nobody claims it, and the issue sits there until it becomes expensive. Most silence lives here.
Standards without authors
Standards exist and nobody can say who wrote them. Nobody defends them, corrects them, or objects when they are wrong. Compliance produces silence. Authorship produces objection.
No line of sight
Ask how their work affects the team next to them and the answer is thin. Not because they do not care. Because nobody ever drew the line. Without it, most people will never form the objection at all.
The second signal is the one organizations miss, and it is the one that hides the damage. When a handoff has no owner, someone eventually absorbs it to keep the work moving. Compensating for a gap conceals it. The schedule holds, the capability never builds, and the system loses the signal that anything was missing. It looks like collaboration. It is debt.
Safety is not a state you cultivate. It is a set of conditions you either build or leave absent.
The same four moves apply here that apply to any system I build, and each one answers a signal directly.
Diagnose
Talk to every person individually, then separate the two silences before designing anything. Build the line of sight first, because you cannot tell disengagement from blindness until you have. Build it, and the people who still say nothing are the ones who actually chose to.
Design
Name what the function owes and publish it, so a person can point at a document and say this is mine, that is not. Then make the standard something to understand rather than something to obey.
Orchestrate
Resize the rooms to fit the risk, then build the path that carries what surfaces there back across the function. Give authorship of each standard to the people who will live inside it. A safe room with no outlet is a vent. The outlet is what makes it architecture.
Build Conditions
Assign the handoffs the way you assign the roles, then transfer ownership to named people who teach it forward. Safety that depends on one person's presence in the room is not architecture. It is a mood, and moods do not scale.
Three rules hold this together. Each one is where the model usually breaks.
Safety comes before the standard, not after it
People need safety to learn the standard at all. Asking questions, admitting you do not know, getting it wrong the first time, all of that is the learning, and all of it requires safety already in place. Safety is not the reward for competence. It is the precondition for building it. So the practical move is to make a new person's first observations a required deliverable rather than a welcome one. Nobody needs courage to hand in an assignment, and fresh eyes stop being fresh in about ninety days.
Crossing between roles is not the risk. Unprotected crossing is
When someone picks up another role's work by design, capability spreads and everyone learns. When they pick it up because waiting took too long, the person who owns that work never gets the reps, and no one finds out until it surfaces somewhere expensive. The test is simple. Did the person who owns that work still get to learn it?
Do not ask leaders for systems thinking. Assign the boundary
Seeing past your own role is a rare capacity, and an operating model that depends on rare capacity is a fragile one. The space between two roles is not a skill you hope someone has. It is unowned work. Name it, give it an owner, hold it to a standard, and it stops depending on who happens to be in the role.
I led a discipline that was losing its people, spread across businesses I did not run.
Each practitioner was the only person of their kind at their site, with nobody nearby who did the same work. No shared definition of the role. No standards. No community.
I owned its standards, its hiring, and how contribution was assessed. What I did not own was the reporting lines. Their daily work was directed elsewhere, which meant nothing here could be mandated into place. It had to be useful enough that people used it when no one was watching.
I started with the people, one at a time
Everyone at that scale and a structured sample at any larger one. They were being handed work they sensed did not belong to their role and could not say so, because the role had never been defined.
Five colleagues, five answers, then a sixth version built alone
Anyone learning a core method hit the same wall. Asking had become costly and useless at the same time, so people stopped asking and improvised.
Seven bodies of work named. Community staffed as the eighth
What the function owed the enterprise, published so a person could point at a document and say this is mine, that is not.
The standards were written by the people who would live inside them
Small teams, each authoring the standard its own work depended on, run as a full project lifecycle. They were learning to lead projects. They were producing what the enterprise required. The capability build and the deliverable were the same object.
Grew thirty percent in the first year and has held since
A jump is a hiring push. A level that stays is a system.
Each body of work carries a champion and co champion with named administrators beneath them, and every incoming practitioner is taught the standards by the people who wrote them. Not by me. Letting go of the teaching was the hardest part of the build and the clearest proof it worked. I stayed accountable. They became responsible.
You do not ask people to be brave inside a system that makes it costly to speak. You change what it costs.
I had authority over the standard. I had none over the rooms these people worked in every day. That constraint is the reason the conditions had to be built rather than announced, and it is the reason they survived me. Anything that resets when one leader moves was never architecture.
The underlying distinction is old and well documented. Structure is enabling when it is built to help people master their work, and coercive when it exists to compel compliance. Nobody argues with that. What I would add is where to apply it.
Define the handoff, and assign it. Role clarity is a solved conversation. The space between roles is not.
That space is where the silence lives, because nobody is certain whose problem it is. It is also where someone starts quietly covering the gap, which hides it from everyone. Draw the boundary, name its owner, and you have removed the most expensive ambiguity in the operating model.
Design led companies invest heavily in the right people and the right culture. Both are real. Both get canceled by the operating model underneath them, and then the brand hires for diversity of thought and wonders why the room sounds the same.
Diversity of thought is a hiring conversation. The one after it is structural, and it is the same question every time: what does this system make it cost to speak.
Safety is where the answer shows first. Courage is what you ask for when the architecture is missing.
Adler, P. S., and Borys, B. (1996). Two Types of Bureaucracy: Enabling and Coercive. Administrative Science Quarterly, 41(1), 61 to 89.
Morrison, E. W., and Milliken, F. J. (2000). Organizational Silence: A Barrier to Change and Development in a Pluralistic World. Academy of Management Review, 25(4), 706 to 725.
Detert, J. R., and Edmondson, A. C. (2011). Implicit Voice Theories: Taken for Granted Rules of Self Censorship at Work. Academy of Management Journal, 54(3), 461 to 488.
Clark, T. R. (2020). The 4 Stages of Psychological Safety. Berrett Koehler.
