We hand decisions down in the calm and pull them back in the storm, then call the result delegation.
Most of us have signed the document that pushes authority closer to the work. The delegation matrix is drawn, the thresholds are agreed, the field is told it can decide within a defined range. For a while it holds. Then something goes wrong somewhere in the system, an audit finding lands, a public misstep happens, a single decision turns out badly. And without anyone announcing it, the authority quietly climbs back up the chain. A new sign-off appears. A threshold drops. A choice that used to be made in a week now needs a committee that meets monthly. Each step is reasonable on its own. Together they undo the delegation we were proud of, and the people closest to the work end up with less room than they had a year ago.
This is not a failure of will. It is a ratchet built into the structure, and it moves in one direction unless we deliberately stop it.
Why authority climbs
The pull upward is not appetite for control for its own sake. It is the predictable response to how risk is felt. When a decision made low in the system goes wrong, the discomfort travels high, to the people who carry the formal accountability and the public exposure. The fastest way to feel safe again is to move the decision closer to that exposure, where it can be watched. So authority migrates toward whoever feels most at risk, not toward whoever has the best information. Over time the organization optimizes for the comfort of the center rather than the judgment of the edge, and it does so sincerely, one defensible tightening at a time.
The cost lands where we can least afford it. Every threshold that drifts down is a decision that now travels further and arrives later. The people we serve do not experience that as caution. They experience it as the help that came after the window closed. We bought reassurance at the top and paid for it at the door.
There is a quieter cost too. When authority keeps sliding back up, the people we asked to lead learn not to use the room they are given, because they have watched it disappear before. Delegation that can be revoked at the first hard moment trains everyone to wait for permission, which is the opposite of what we said we wanted.
Pinning authority in place
The fix is not to delegate harder and hope. It is to make delegation sticky, so that taking it back costs something visible and deliberate. A few moves do this.
Write the floor, not only the grant. Most matrices say what authority is delegated. Few say what may not be quietly removed without an explicit, recorded decision at a named level. A delegation that can be eroded silently is not a delegation. It is a loan.
Make every tightening name its trigger and its cost. When a new approval is added or a threshold is lowered, require a one-line record of what prompted it and what it will slow down. A control that has to state the delay it creates is a control we can weigh, keep if it earns its place, and remove when it does not.
Review the controls, not only the decisions. Once a year, walk back through the sign-offs that accumulated since the last review and ask of each: would we add this today, knowing what it costs downstream. The ones that fail are candidates for removal. Subtraction has to be as routine as addition, or addition always wins.
Separate the rare failure from the standing rule. One decision going wrong is not proof the level was wrong. Treat a bad outcome as a case to learn from, not an automatic reason to move all such decisions upward. Otherwise a single hard event permanently shrinks the judgment of everyone who had nothing to do with it.
None of this asks us to be careless with risk. It asks us to stop letting a sincere fear of error rewire our structure by stealth. Authority placed close to the work is authority better used, but only if it stays there long enough to be relied on.
Delegation that vanishes at the first hard moment was never delegation. It was a loan.
So here is the standard worth holding. A delegation is real only when it survives a bad day. Building the discipline that lets it survive, rather than drift, is something we can start at the next review of who is allowed to decide what.