How to delegate decision-making authority so it stays delegated
You told your commercial director that pricing is hers. You meant it. Three weeks later she is at your desk with a spreadsheet, asking whether twelve percent off the enterprise tier is acceptable.
Nothing went wrong in that conversation. She was not testing you and she was not dodging responsibility. She came to you because asking is still cheaper than deciding, and nothing in the structure has changed that.
Most advice on how to delegate decision-making authority treats this as a coaching problem: pick the right person, set clear boundaries, resist the urge to step in. None of that is wrong, and none of it works on its own, because it describes the behaviour you want without touching what produces the behaviour you get. You moved the right to decide. You kept everything that determines who decides.
- Delegation fails structurally rather than personally. You can hand over the right to decide and keep the information, the audience and the review, and the decision keeps arriving at your desk because that is where deciding still happens.
- The Delegation Envelope: three lines written before the handover. The class of decisions being moved (not one instance), the limit at which they return to you (in a unit both of you can read), and the channel that carries the information and the reporting.
- Philippe Aghion and Jean Tirole separated formal authority, the right to decide, from real authority, effective control over the decision. In their model real authority follows the structure of information, so a superior who stays informed and available keeps real authority no matter what the org chart says.
- A limit that is never stated defaults to zero. Faced with an unwritten boundary, a competent person escalates, because escalating has a known cost and guessing wrong does not.
- Overturning one decision inside the envelope without changing the envelope teaches the team that the envelope is decorative. Either the decision stands, or the boundary changes in writing.
- Aghion, Bloom, Lucking, Sadun and Van Reenen found that firms in ten OECD countries and US establishments that had delegated more power to local managers before the Great Recession outperformed their centralized counterparts in the sectors the crisis hit hardest.
Why do delegated decisions come back to you?
Because you transferred the right to decide and kept the conditions under which deciding actually happens. The org chart moved. The information, the review and the availability did not, and those are what determine who really holds the decision.
Philippe Aghion and Jean Tirole made this precise almost thirty years ago. They separated formal authority, the right to decide, from real authority, effective control over what gets decided, and showed in a formal model that the two come apart routinely inside the same structure. Real authority, in their account, follows the structure of information: whoever holds the knowledge a decision needs ends up controlling it, whatever the reporting line says (Aghion & Tirole, 1997). This is an economic model rather than a study of firms, so treat it as a way of seeing the problem, not as measured evidence about your company.
What their model predicts is uncomfortable for anyone who has tried to delegate and watched it bounce. A subordinate's real authority rises when the superior is overloaded, when rules are lenient, when decisions are urgent, and when the subordinate's track record is strong. Read that backwards. If you are available, if you review, if you remain the person who knows the customer history and the board's mood, then you hold real authority permanently, and no announcement transfers it.
This is also why delegated decisions are the ones people rate worst. In a McKinsey Global Survey of 1,259 executives across 91 countries, most of them in Europe (36 percent) and North America (25 percent), 46 percent of respondents agreed their organizations' delegated decisions were high quality, against 54 percent for cross-cutting decisions and 65 percent for big bets (Aminov, De Smet, Jost & Mendelsohn, 2019). These are self-reported perceptions, not audited outcomes, and the sample skews senior and Western. Still, note the direction. The decisions organizations think least of are the ones they claim to have handed over, which is what you would expect if handing over is mostly announced rather than built. The same figure appears in the one-hour decision rights audit, for the same reason.
One finding in that survey cuts against my own argument, and it deserves saying out loud. Looking specifically at delegated decisions, the researchers found coaching people and giving them room to fail safely more strongly associated with winning organizations than defining clear roles or setting escalation guidelines. Structure alone did not carry it. My reading is that a role definition without a stated boundary and without an information change is not much of a structure, which is exactly the failure this article is about. That is my interpretation, and the survey did not test it.
What are you actually delegating when you delegate decision-making authority?
A class of decisions, a boundary, and the information that decisions in that class require. Not a single instance, and not a task.
In my experience the confusion between task delegation and authority delegation is behind most failed handovers. “Run the pricing analysis and bring me a recommendation” is task delegation. The work moves, the decision does not, and the person doing the work knows it. “Discounting on the enterprise tier is yours below fifteen percent” is authority delegation. Different sentence, different structure, and most leaders believe they said the second one when they said the first.
Three things habitually stay behind.
The first is the class. You delegate one decision and the person reasonably treats it as one decision. Next month a similar case arrives, no rule covers it, and they ask. You experience this as a lack of initiative. From where they sit, you granted permission once and never said it generalised.
The second is the boundary. Nobody wrote where the authority stops, so the person has to guess, and guessing high has an obvious downside while guessing low has almost none. Escalating costs them a few minutes of your time. Deciding wrong costs them standing. A competent person picks the cheap option, and you read caution as timidity.
The third is the channel, and it is the one I most often find missing. If the customer still emails you, if the finance report still lands on your desk, if the board still asks you about pricing, then the person who formally owns pricing has to come through you to do the job. Aghion and Tirole's point is structural, not motivational: leave the information where it was and you have moved a label.
The Delegation Envelope: class, limit, channel
Write three lines before the handover. They are the difference between announcing a delegation and building one.
Name the set, not the case. “All discounting on the enterprise tier”, not “this discount for this customer”. A class tells the person what to do next month without asking, which is the entire point. If you cannot write the set, you have not decided what you are giving away, and neither has anyone else.
State where the authority stops, in a unit you can both read on a Tuesday afternoon. Money is the easiest, but it is rarely the only one that matters: reversibility, headcount, a public commitment, or a precedent that would bind other accounts. Write the threshold, not the sentiment. “Use your judgement on the big ones” is not a limit, it is a request that someone else invent one and then be responsible for having guessed.
Redirect the information and the reporting. Say who the person now hears from, who now hears from them, and what you get instead of approval rights. Usually what you get is notice: a line in a weekly summary, after the fact. This is the field that actually moves real authority, and it is the one that takes real work, because it means telling a customer, a colleague, or a board member to stop calling you about it.
Class, limit, channel. The shape deliberately echoes the Reopening Threshold, which sets what it costs to reverse a decision after it closes. That one works backwards in time. This one works downwards through the structure. A leadership team can have both and still be slow, but a team with neither is not slow because of its people.
The pattern I meet most often, and I am describing a recurring shape across engagements rather than one company, is a founder who has delegated the same decision three or four times. Each handover was sincere. None was written. When we go back through the last month of approvals and ask what boundary would have let each one proceed without the founder, the answer is nearly always available, uncontroversial, and takes one sentence. Nobody wrote it, because at the moment of handing over, “pricing is yours” felt complete.
What do you do when you disagree with a decision inside the envelope?
You let it stand, or you change the envelope in writing. What you cannot do is overturn the instance and leave the envelope untouched, because that is the move that empties it.
Sort your objection into one of three cases before you say anything.
The decision sat outside the envelope. Then the problem is the boundary, not the judgement. Say so plainly, restate the limit, and move on. This is the easy case and it is the rarest.
The decision sat inside the envelope and you would have chosen differently. Then it stands. You bought this outcome when you drew the line, and reversing it now converts the envelope into a suggestion. If you cannot live with the class of outcomes the limit permits, the limit was wrong, which is the third case, not this one.
The envelope itself is wrong. Then change it, out loud, with a reason, and let it apply from the next decision rather than retroactively to this one. Changing a boundary is legitimate. Changing it silently, in the specific case where it produced an answer you disliked, is not a boundary at all.
Say the cost of this honestly, because the trade is real. Aghion and Tirole's model has no free lunch in it: giving a subordinate real authority raises their initiative and lowers the principal's control, at the same time, by the same mechanism. Anyone selling delegation as pure upside has not read the argument. What you are buying is speed and ownership at the price of some decisions you would have made differently.
There is evidence that the trade is often worth taking, though it comes from a very different setting than your leadership team. Philippe Aghion, Nicholas Bloom, Brian Lucking, Raffaella Sadun and John Van Reenen used two large datasets covering firms in ten OECD countries and US establishments, and found that firms which had delegated more power from headquarters to local plant managers before the Great Recession outperformed their centralized counterparts in the sectors the crisis hit hardest (Aghion et al., 2021). Their explanation is that turbulence raises the value of local information. The evidence is about manufacturing plants rather than executive teams, and about crisis conditions rather than normal ones, so read it as support for the mechanism rather than as a promise about your quarter.
How do you audit whether a delegation actually held?
Count, rather than ask. People will tell you they feel empowered while routing everything through you, and both statements are sincere.
- List every decision you approved in the last month that someone else formally owns. Use your calendar and your sent mail, not your memory. Memory undercounts this badly, because each individual approval is small and feels like helping.
- Sort each one into three piles. Outside the envelope, which is correct. Inside the envelope but they asked anyway, which is a channel or a limit problem. Inside the envelope and you took it back, which is an override you owe someone an explanation for.
- For the second pile, find the missing input. What did they need to know that only you knew? That answer is the channel line you have not written yet.
- Write one envelope for the class that appeared most often. Three lines, on the decision that cost you the most time this month. Not a delegation framework for the whole company.
- Count again in four weeks. Same method, same three piles. What you want to see is the second pile shrinking. Do not read sentiment as progress; count the piles.
This connects to what Leadership Architecture measures. Escalation Discipline asks whether decisions travel upward on a defined trigger or by habit. Leadership Load Balance asks whether decision weight is distributed or concentrated. And Decision Authority Dependency, which we treat as a moderator rather than a capability, asks how much of the system runs through one person. A pile of approvals on decisions you formally gave away is a cheap and visible reading of all three at once. If the same pattern shows up across many unrelated decision types, the constraint is probably decision rights rather than delegation technique, and if it concentrates on you specifically, the frame that fits is scaling a founder-led company.
Here is what I would ask you to notice. When a delegated decision comes back, the conversation is never about whether the person is capable. It is about what they were allowed to assume. That is not a question about character, and no amount of encouragement to take ownership will settle it. Somebody has to write down where the authority stops, once, before anyone needs to know.
Frequently asked questions
Why do delegated decisions keep coming back to me? Because the right to decide moved and the conditions for deciding did not. Three things usually stay behind: the class (you delegated one instance, so nothing generalises to the next case), the limit (nobody wrote where the authority stops, so escalating is the cheap and safe option), and the channel (the information the decision needs still arrives at your desk, so the person has to come through you to do the job).
What is the Delegation Envelope? The Delegation Envelope is three lines written before a handover: the class of decisions being transferred, the limit at which they return to you, and the channel that carries the information and the reporting. It exists to make the interior of a delegation real, so the person can act inside it without asking and you can tell, afterwards, whether they stayed inside it.
What is the difference between delegating a task and delegating decision-making authority? Delegating a task moves the work and keeps the decision: “run the analysis and bring me a recommendation”. Delegating authority moves the decision itself for a defined class of cases: “discounting on this tier is yours below fifteen percent”. Most handovers that feel like the second are the first, which is why the person keeps returning with recommendations rather than outcomes.
Should I overrule a decision I disagree with if it was inside the boundary I set? Let it stand, or change the boundary in writing for future decisions. Overturning the instance while leaving the boundary untouched is what converts an envelope into a suggestion, and it does more damage than the decision you disliked. If you cannot live with the range of outcomes the limit permits, the limit was drawn in the wrong place, and that is what you should change.
How do I know whether a delegation actually worked? Count approvals rather than asking about confidence. List every decision you signed off in the last month that someone else formally owns, and sort them into outside the boundary, inside but escalated anyway, and inside but reclaimed by you. The second and third piles are your work. Repeat in four weeks and watch whether the second pile shrinks.
Sources
Philippe Aghion & Jean Tirole, Formal and Real Authority in Organizations, Journal of Political Economy, 1997, 105(1), 1–29. Source of the formal/real authority distinction, the result that real authority follows the structure of information, and the conditions that raise a subordinate’s real authority (overload, lenient rules, urgency, reputation). A theoretical model rather than a study of firms.
Iskandar Aminov, Aaron De Smet, Gregor Jost & David Mendelsohn, Decision making in the age of urgency, McKinsey Global Survey, 2019. Online survey of 1,259 participants in 91 countries, 36 percent in Europe and 25 percent in North America, skewing senior. Source of the 46, 54 and 65 percent quality figures for delegated, cross-cutting and big-bet decisions, and of the finding on coaching and failing safely. Self-reported perceptions, not audited outcomes.
Philippe Aghion, Nicholas Bloom, Brian Lucking, Raffaella Sadun & John Van Reenen, Turbulence, Firm Decentralization, and Growth in Bad Times, American Economic Journal: Applied Economics, 2021, 13(1), 133–169. Two datasets covering firms in ten OECD countries (WMS) and US establishments (MOPS); source of the finding that firms which had delegated more power to local plant managers before the Great Recession outperformed centralized counterparts in the hardest-hit sectors.
Client examples are anonymized composites; no figures are invented.
How many of the decisions you gave away still come back?
Five minutes. No account. A structural read on whether your decision system, not your team, is the real constraint.