2026-10-23
October 23, 2026
Before you fix the agenda, the format or the attendee list, answer one question: who decides? A meeting with no named decider is a rehearsal for a second meeting.
Look at your own calendar this week and sort the recurring invites by what they are for. Status meetings are easy to spot, and so are standups and reviews. The expensive ones are different. They are the meetings that exist because a decision is stuck, and nobody has said so. The invite says "sync" or "alignment" or "working session". Eight people attend, everyone speaks, and the meeting ends with "let's circle back." The meeting did not fail. It was never able to succeed, because it was asked to produce a decision from a room where nobody holds the pen.
Paul Rogers and Marcia Blenko of Bain made this point in their 2006 Harvard Business Review article "Who Has the D?". They argue that organizations stall because decision accountability is unclear, and that the fix is a single named point of accountability, the "D". Consensus, they warn, is not required, and chasing it produces stalled action or lowest-common-denominator compromises. Twenty years on, the diagnosis holds, and the remedy is cheaper than most teams think. You do not need a new tool or a new ceremony. You need one line at the top of the invite: this meeting exists to produce this decision, and this person decides.
That line also settles the question every leader is now asking about meetings, which is whether to hold them at all. If the D is named and the options are written down, much of the meeting can happen asynchronously. If the D is not named, no format will save you.
Most advice on meetings is about format: shorter, standing, camera off, async-first. Format matters, but it is the second question. The first is the decision architecture. Until you know what is being decided, by whom, and with whose input, you cannot know whether the work belongs in a room, a document or a thread.
Here is the test I would apply to every recurring meeting on a program calendar. Ask what decision this meeting produces. If the answer is "none, it shares information", it is a candidate for an async update, and the evidence from teams who have tried that is more mixed than the evangelists admit. Ben Darfler of Honeycomb has argued that asynchronous standups in a chat channel fail because people do not read them, so they never create shared context or surface problems. Honeycomb's answer was not to go async. It was to replace the short standup with a longer daily team sync with a rotating facilitator and a shared agenda doc filled in the day before. The lesson is not "meetings good, async bad." It is that async only works when someone is accountable for the reading, and a sync only works when the agenda was prepared before anyone sat down. Both are about ownership, not medium.
If the answer is "it makes a decision", the next question is whether anyone can say who the decider is, in one sentence, without hedging. In my experience the honest answer for stuck cross-team decisions is "it depends" or "the group". "The group" is not a decider. It is a way to avoid choosing one. Rogers and Blenko list the usual victims: decisions stuck between global and local, between the center and a business unit, between functions. Those are the bottlenecks TPMs live inside, and none is solved by a better agenda.
Once the D is named, the rest of the design falls out of the roles around them. Someone owns the recommendation, which means they do the pre-reading, gather input and bring a proposal rather than an open question. A small number of people must agree because of legal, security or feasibility constraints, and their input must be reflected, not merely heard. Others have input and should be asked in writing before the meeting, because the meeting is the worst place to discover an objection. The people who will perform the work should be in the loop early enough to flag what will not survive implementation.
Notice how much of that can be async. Gathering input is a document. Raising objections is a comment thread with a deadline. What is left for the live meeting is the part humans are good at and documents are bad at: resolving a genuine disagreement between the recommender and a required agreer, and letting the D ask hard questions face to face. That is a thirty-minute meeting, not a ninety-minute one, and often it is unnecessary.
The AI notetaker wave does not change this. Meeting summarizers and "agentic" tools, which file tickets and draft follow-ups from a call, are useful for capturing outcomes. They cannot capture a decision that was never made. A vendor page I read this month is candid that its category's value is turning conversation into recorded actions, and the claims are marketing, not benchmarks. Treat the output as a draft record that the D confirms. If nobody can say what was decided, the tool will faithfully summarize the confusion.
The practical upshot for a senior TPM: your highest-leverage meeting skill is not facilitation. It is noticing, early, that a meeting has no D, and saying so, gently and in writing, before the second meeting is scheduled. Doing this well is political work. You are asking a leader to own a call they may prefer to leave ambiguous. Frame it as protecting them from relitigation, because that is true.
Try this week. Pick the one recurring meeting on your calendar that most often ends in "let's circle back." Before the next occurrence, add one line to the invite: "Decision needed: . Decider: . Recommendation due: ___ (written, 24 hours before)." If you cannot fill in the blanks, message the person most likely to be the D and ask. If there is no decision, cancel the meeting or convert it to a written update with a named reader.
What it is. A decision-accountability tool that assigns five roles for any significant decision: Recommend, Agree, Perform, Input and Decide. Bain's published description stresses a single recommender and, ideally, a single decider.
When to use it. When a decision keeps resurfacing across meetings, when more than one team or function has a stake, or when you see "alignment" meetings that never end in a call. Bain advises starting with high-value or high-frequency decisions rather than applying it to everything.
How to run it:
When NOT to use it. On small, reversible decisions where the overhead of five named roles exceeds the cost of getting it wrong, or on genuinely exploratory work where nobody yet knows what the decision is.
You may know DACI, which this newsletter has covered. They overlap, but RAPID's Agree role and its explicit Perform role make it a better fit when sign-off constraints and implementation feasibility are the real sources of delay.
Will Larson on The Pragmatic Engineer: the rules for engineering managers have changed — Larson, CTO of Imprint, notes manager spans growing from roughly 4-8 to 12-14 reports. Wider spans mean fewer synchronous touchpoints per person, which raises the cost of every unowned decision.
Katrin Freihofner at LeadDev: Fast, informed, impactful decision making — From a 2025 talk drawing on interviews with 50+ tech professionals: scaling teams tend to have either over-complicated decision processes or none, and RFCs meant to speed decisions often become bottlenecks. DACI appears as one way to map stakeholder roles.
Tana: What are agentic meetings in 2026 — A vendor page (July 2026) separating notetakers from tools that act on the conversation. Useful for the framing, but its competitor and market claims are unsourced, so verify before repeating them.
"They have worked incredibly hard, but the rules have changed."
— Will Larson, on managers navigating a shifting job market, The Pragmatic Engineer, October 21, 2026
Don't miss what's next. Subscribe to Critical Path: