references/agenda_discipline.md
6.3 KB查看排版效果| 1 | # Agenda Discipline — questions, timeboxes, decisions first, an owner, a pre-read |
| 2 | |
| 3 | The second discipline of the `meetings` skill: a meeting that passed the cost gate still fails if |
| 4 | its agenda is a list of nouns. This file documents the canon behind `agenda_builder.py` — why every |
| 5 | topic must carry a desired outcome, why timeboxes are budgets rather than suggestions, why decision |
| 6 | topics go first, and why the pre-read and the closing actions-recap slot are non-negotiable. |
| 7 | |
| 8 | ## Agendas as questions (the desired-outcome rule) |
| 9 | |
| 10 | Steven Rogelberg's meeting research produced a counterintuitive result: merely *having* an agenda |
| 11 | does not predict meeting effectiveness — attendees rate agenda'd and agenda-less meetings about the |
| 12 | same. What predicts effectiveness is an agenda built from **questions to be answered**, because a |
| 13 | question forces the organizer to know what "done" looks like, makes the right attendee list |
| 14 | self-evident (invite whoever is needed to answer), and gives the meeting a natural end (the |
| 15 | question is answered) (Rogelberg, *The Surprising Science of Meetings*, Oxford University Press, |
| 16 | 2019; and his HBR guidance "Why Your Meetings Stink — and What to Do About It," *Harvard Business |
| 17 | Review*, January–February 2019). |
| 18 | |
| 19 | `agenda_builder.py` operationalizes this as a hard rule: **no desired outcome, no agenda slot.** |
| 20 | A topic supplied with an empty outcome field is refused by name (exit 2). "Roadmap" is a noun; |
| 21 | "Decide whether the roadmap slips two weeks" is a meeting. |
| 22 | |
| 23 | ## Timeboxing (Parkinson's law made operational) |
| 24 | |
| 25 | "Work expands so as to fill the time available for its completion" (C. Northcote Parkinson, |
| 26 | "Parkinson's Law," *The Economist*, 1955). Discussion is the purest case: an unbounded topic |
| 27 | consumes whatever remains. The counter-discipline is the timebox — a fixed allocation that ends |
| 28 | the topic whether or not the room feels finished — which is the load-bearing mechanism of Scrum's |
| 29 | every ceremony: the sprint, the daily scrum's fifteen minutes, the review, the retro (Jeff |
| 30 | Sutherland, *Scrum: The Art of Doing Twice the Work in Half the Time*, Crown Business, 2014; |
| 31 | Schwaber & Sutherland, *The Scrum Guide*). The builder therefore: |
| 32 | |
| 33 | - requires per-topic minutes and refuses an agenda whose timeboxes (plus the closing buffer) |
| 34 | exceed the meeting length, naming the exact overflow (exit 3) — a budget you can silently |
| 35 | overdraw is not a budget; |
| 36 | - prints start–end offsets per topic so the running meeting can see the clock, not just the list; |
| 37 | - treats slack as a feature: finish early, end early. |
| 38 | |
| 39 | ## Decisions first |
| 40 | |
| 41 | The builder sorts topics whose outcome starts with **decide / choose / approve** ahead of |
| 42 | discuss/inform topics. Three converging reasons: |
| 43 | |
| 44 | 1. **Energy and attention decay.** Decision quality degrades as a session wears on; putting the |
| 45 | decision in the last five minutes buys either a rushed call or a follow-up meeting. HBR's |
| 46 | practical agenda guidance consistently recommends sequencing the most important items first — |
| 47 | also because meetings compress from the end when they start late or run over. |
| 48 | 2. **Grove's mission-meeting logic.** If the meeting exists to produce a decision (Grove, *High |
| 49 | Output Management*, 1983), the decision *is* the meeting; everything else is garnish and should |
| 50 | be treated as such — or moved to the pre-read. |
| 51 | 3. **Inform topics are the weakest claim on synchronous time.** If an inform topic gets squeezed |
| 52 | out by the sort order, that is the system working: it likely belonged in the memo anyway. |
| 53 | |
| 54 | ## The owner role |
| 55 | |
| 56 | Every topic carries a named owner — the person who drives that timebox, states the question, and |
| 57 | is accountable for its outcome landing in the actions recap. The meeting as a whole has an owner |
| 58 | too (checked upstream by the cost gate). This is Rogelberg's "meeting leader as steward of |
| 59 | everyone's time" plus Grove's chairman-accountability rule, made explicit per-slot so |
| 60 | responsibility cannot diffuse across the invite list. |
| 61 | |
| 62 | ## The pre-read line |
| 63 | |
| 64 | Every generated agenda opens with a pre-read instruction: circulate the material ahead; the |
| 65 | meeting starts assuming it was read. This is Amazon's narrative-memo practice inverted for |
| 66 | ordinary meetings — Bezos's argument for the six-page memo (documented in Amazon shareholder |
| 67 | letters and public remarks) is that writing forces complete thoughts and reading in advance means |
| 68 | the meeting spends its synchronous minutes on the decision, not the download. A meeting that |
| 69 | re-presents the pre-read punishes exactly the people who prepared. |
| 70 | |
| 71 | ## The closing actions-recap slot |
| 72 | |
| 73 | The builder reserves a mandatory 5-minute closing slot: read every action aloud — owner and due |
| 74 | date — before anyone leaves. Two safety properties motivate making it structural rather than |
| 75 | aspirational. First, follow-through is the most commonly dropped meeting phase (Rogelberg), and a |
| 76 | slot on the clock cannot be skipped invisibly. Second, in decision meetings, dissent and confusion |
| 77 | surface late; a recap slot is the last cheap moment to catch "wait, that's not what I agreed to" — |
| 78 | which only happens if people feel safe saying it, Amy Edmondson's core argument for psychological |
| 79 | safety in decision-making teams (Amy C. Edmondson, *The Fearless Organization*, Wiley, 2018). |
| 80 | |
| 81 | ## Sources |
| 82 | |
| 83 | 1. Steven G. Rogelberg, *The Surprising Science of Meetings*, Oxford University Press, 2019 |
| 84 | (agenda-as-questions; leader as steward; follow-through gap). |
| 85 | 2. Steven G. Rogelberg, "Why Your Meetings Stink — and What to Do About It," *Harvard Business |
| 86 | Review*, January–February 2019 (question-shaped agendas; invite-list implications). |
| 87 | 3. C. Northcote Parkinson, "Parkinson's Law," *The Economist*, 1955 (work expands to fill the time |
| 88 | available — the case for timeboxes). |
| 89 | 4. Jeff Sutherland, *Scrum: The Art of Doing Twice the Work in Half the Time*, Crown Business, |
| 90 | 2014 (timeboxing as the core ceremony mechanism; also Schwaber & Sutherland, *The Scrum Guide*). |
| 91 | 5. Andrew S. Grove, *High Output Management*, Random House, 1983 (process vs mission meetings; |
| 92 | the decision is the meeting; chairman accountability). |
| 93 | 6. Jeff Bezos, Amazon shareholder letters + public remarks (narrative memo pre-read; synchronous |
| 94 | time spent on the decision, not the download). |
| 95 | 7. Amy C. Edmondson, *The Fearless Organization*, Wiley, 2018 (psychological safety in decision |
| 96 | meetings — why the recap slot must invite dissent). |
文件内容是平台检查时读出的文本，只供查看。平台不会运行包里的任何脚本或程序。
