The decision that comes back every quarter

Why does a settled decision keep coming back?

Because reopening is free and closing is not. Anyone in a leadership group can ask for an item to be discussed again at the cost of a slide deck. Closing it costs the person who does it: they must overrule colleagues, absorb their displeasure and own the outcome if it goes badly. In a group of peers nobody is paid to do that, so nobody does, and the decision's status stays at pending whatever the minutes say.

A decision taken by consensus can be closed only by consensus, and consensus can be rebuilt on any day by a new argument or a new member. Each person holds an equal right to reopen it, and nobody holds the right to refuse. That is why the return is so regular. The reasons for the decision may be excellent. What it lacks is a person whose job is to say that it stands.

Some returns are legitimate, because a fact the decision assumed has changed. Those are the system working. The question is what to do with the returns that have no new fact behind them, and the answer starts with who is able to close them.

What does it look like when a leadership team cannot close one?

Consider a company whose leadership team decides that its business units will stop buying separate AI tools and move onto one shared platform. The reasons are sound: one set of controls, one supplier to oversee, one place for the data to sit. The decision is taken in a good meeting and recorded in a line.

The next quarter, a unit whose tool was set aside returns with results from its own trial and asks for an exception. The chair, a peer of the unit's head, thinks it fair to hear them out. The group debates the original options and concludes that more evidence is needed. The following quarter a new head of another unit joins and asks, reasonably, why the decision was made. Nobody has the reasons in front of them. The supplier then changes its terms, which may or may not matter, since the decision never stated what it assumed. Each return is polite, and each is reasonable on its own. Exhibit 1 shows the cycle.

What does a returning decision cost?

The visible cost is the leaders' hour each quarter. The larger cost is behavior. The units told to migrate wait, because experience says the decision may be reversed. The platform team cannot commit to a plan. The unit with the set-aside tool keeps spending. A decision that returns teaches the organization to treat decisions as proposals, and the next one, which may matter more, is waited out in the same way.

Who should hold the authority to close it?

The person who will carry the consequences and has the standing to carry them: usually the executive accountable for the outcome the decision serves. In the platform example that is whoever answers for the cost and the controls of the AI estate, not the chair of the meeting and not the unit that lost. Others with a stake, such as risk, technology and the units, hold a different right: to be heard before the ruling and to have their objection written down. They do not hold a vote, because a vote is what lets the question be rebuilt.

The decider owes the group something in return. The ruling comes with reasons that can be read, a statement of what would change it, and a date on which the decision is looked at again whether or not anyone asks. A decider who rules without reasons is no better than a room that cannot decide, because the next doubt will have nothing to answer it. In the platform example, a closing note issued after the first meeting would have answered the second and third returns in a sentence each.

What does closing a decision look like?

Closing has a person in it. For each class of decision the group names, in advance, who decides after hearing the others: a role and a name, not the group. The group advises and the decider rules. A chair who is a peer of the others cannot do this, and a decider who can be overruled by the next meeting is not one, so the decider can be overridden only by someone senior to them, in writing.

The ruling is then recorded in a form that a person joining the group can read in one sitting. Michael Nygard's note on documenting architecture decisions, published in 2011 for software teams, describes the form that works: a title, the context and the forces at play, the decision in full sentences, its status, and its consequences, all of them and not only the favorable ones. When a decision is reversed, the old record is kept and marked as superseded, because it is still relevant to know that it was the decision. His reason for keeping records applies to any group: a person arriving later can otherwise only accept a decision blindly or change it blindly.

A leadership group needs additions. The dissenting views are written in their holders' own terms, so that the losing side can see they were heard. And the record states what would reopen the decision: a named assumption changing, a stated result, a date for review. Closing is then a state that can be checked, and the list below is the check.

Is this decision closed?

• One named person with authority ruled, and the record says who
• The options that were set aside are listed, each with the reason
• People who disagreed are recorded in their own words
• The facts the decision assumed are written down, and someone is named to watch each
• The conditions that would reopen it are stated, with a date for review
• The route for reopening it is written: a request to the decider that names the changed fact
• A person new to the group can read all of it in one sitting

How do you tell a good return from a bad one?

When a decision returns, ask the questions of Exhibit 2 in order. If a recorded condition has changed, the return is legitimate and the right response is to decide again on the new facts. If none has, ask whether a record exists. Without one, no doubt can be answered, so write it before anyone debates, and close the item with the writing. With one and no changed condition, the question is whether anyone can rule. If the decider exists, the decider closes it in writing, without a meeting. If nobody can, the finding of the quarter is not the original question. It is that no one holds the authority, and the useful decision is to name who does.

Does closing a decision suppress dissent?

No. A decision nobody can reopen is as unsound as one anyone can reopen at will. The point is that dissent has a channel with a rule: a request to the decider that names what has changed, answered in writing, and a review date that arrives whether or not anyone is pressing for it. The NIST AI Risk Management Framework 1.0 reaches for the same discipline when it asks that ongoing monitoring and periodic review of the risk management process are planned, with roles defined and the frequency of review determined (GOVERN 1.5).

This comes from advisory work and has not been measured. It is offered as a habit to try: take the decision that keeps returning, write its record before the next meeting, and name who rules.