There are two kinds of change, and most organizations treat them as if they're the same thing.

The first kind is mostly irreversible, or at least the cost of reversing becomes prohibitive. A merger. A platform migration. A restructure that eliminates a division. Once you've walked through, the door closes behind you. These decisions warrant the full weight of deliberation: aligned leadership, clear mandate, rigorous analysis, top-down commitment. The cost of getting it wrong is high and the path back is long.

The second kind is reversible. A new onboarding process. A pilot program. A team-level experiment in how work gets reviewed. You walk through, look around, and walk back if it isn't working. These decisions warrant something different entirely: a hypothesis, a timeframe, and permission to learn from what actually happens.

Amazon calls these one-way and two-way doors. Dave Snowden calls them fail-safe and safe–to-fail. The language may vary, but the distinction doesn't.

The failure mode nobody names

The problem isn't that organizations lack the right tools for each type of change. Most leadership teams know, at least conceptually, that some decisions warrant more deliberation than others. The problem is misdiagnosis. Leaders apply one-way-door processes to two-way-door decisions, and vice versa, because they haven't stopped to ask which kind of change they're making. The door they walk through is often not the door they thought they were walking through.

So reversible decisions get treated like permanent ones. Pilots get the governance load of enterprise rollouts. Team-level experiments go through approval chains designed for strategic commitments. A client described it to me this way: his organization was waiting for perfect information before making decisions that could be updated the following quarter. The delay felt like prudence, but it was an invisible cost — one that doesn't show up on any dashboard, but accumulates in exactly the same way organizational mass does. I wrote about this here: The Most Corrosive Force in Every Organization.

Over-processing reversible decisions makes the organization slower, heavier, and less capable of learning from the thing it was afraid to try.

Discontinuity makes this harder

In a stable environment, defaulting to the fail-safe mode is expensive but survivable. You move slower than you need to. You miss some opportunities. The organization carries weight it didn't need to carry, making it harder to turn when it needs to.

In a discontinuous environment, it becomes actively dangerous.

Discontinuity is the condition where the causal relationships that used to hold stop holding. What worked reliably before is no longer a trustworthy guide. The map stops matching the territory. In that condition, the only way to learn what works is to try things, watch what happens, and update based on what you actually observe. Safe-to-fail experimentation isn't a nice-to-have in discontinuous environments. It's the primary mechanism of organizational learning.

But discontinuity also feels threatening. And when things feel threatening, leaders reach for the most controlling approach available, because it feels like protection. What it's doing is preventing the learning that the situation requires.

Donna Haraway writes that in urgent times, we are tempted to address trouble by making imagined futures safe — by clearing away the present in order to secure what comes next. Staying with the trouble, she argues, requires something harder: learning to be truly present to what is happening, not to what we wish were happening or fear might happen. That is precisely what safe-to-fail change asks of an organization. Not certainty. Presence.

Who decides which door it is

Before you can apply the right mode to a decision, someone has to correctly identify which kind of change it is. And in discontinuous environments, that judgment is exactly the kind of judgment that individual leaders are least equipped to make alone.

The people closest to the work usually know whether something is reversible. They know whether the organization has tried a version of this before, what happened, and why it stalled. They know which assumptions are load-bearing and which ones are guesses pretending to be strategy. What they often lack is permission to say so.

This is where collective intelligence becomes structural rather than aspirational. Collaboration pools what people already know and tends to confirm what leadership already believes. Collective intelligence is something different. It's what becomes possible when people with different vantage points are in the room, with actual influence over how the question gets framed. An organization that can surface which door it's facing, before it commits to a mode, is an organization that can learn. One that can't is left with individual executive judgment operating on incomplete information, masquerading as decisive leadership.

The question leaders aren't asking

Before any significant decision, most leadership teams ask some version of: how do we make sure this goes right?

The more useful question is: what kind of door is this?

If it's a one-way door, slow down. Get aligned. Bring the full weight of the organization's deliberative capacity to bear. The cost of speed here is real.

If it's a two-way door, and most decisions in a discontinuous environment are, you should move—quickly. Design a clean experiment. Define what you're trying to learn and how you'll know if it's working. Set a timeframe. Walk through the door.

If you're not sure whether a decision is a one-way or two-way and you need a little help, just reply here. Happy to put our brains together.