board-meeting

Board Meeting Protocol

Structured multi-agent deliberation that prevents groupthink, captures minority views, and produces clean, actionable decisions. Every phase has a purpose, a format, and rules that cannot be skipped.

Keywords

board meeting, executive deliberation, strategic decision, C-suite, multi-agent, founder review, decision extraction, independent perspectives, groupthink prevention, synthesis, critic analysis, structured deliberation

---

Clarify First

Before generating, confirm these inputs. If any is unknown or vague, ASK — do not assume:

• The exact decision to be made (stated in one sentence) — the whole protocol synthesizes toward a single decision; a fuzzy topic produces a fuzzy synthesis
• Which roles to activate — the role-activation matrix decides whose perspectives appear in Phase 2; the wrong roster means a missing voice or noise
• What data and context is available vs. assumed — Phase 2 contributions tag confidence and source, so known facts must be separated from guesses
• Reversibility and stakes of the decision — drives whether the full 6-phase protocol or an abbreviated advisory meeting is appropriate

Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.

---

The 6-Phase Protocol

PHASE 1: Context Gathering
    |
PHASE 2: Independent Contributions (ISOLATED)
    |
PHASE 3: Critic Analysis (Executive Mentor)
    |
PHASE 4: Synthesis (Chief of Staff)
    |
PHASE 5: Founder Review (FULL STOP -- human decides)
    |
PHASE 6: Decision Extraction and Logging

---

Phase 1: Context Gathering

Purpose: Load all relevant context before anyone contributes.

Step 1: Load company context (if exists)
Step 2: Load decision history (Layer 2 ONLY -- NEVER raw transcripts)
Step 3: Reset session state -- no bleed from previous conversations
Step 4: Present agenda and activated roles
Step 5: Wait for founder confirmation before proceeding

Role Activation Matrix

Not all roles attend every meeting. Select based on topic:

Topic Domain	Activate	Exclude
Market expansion	CEO, CMO, CFO, CRO, COO	CTO (unless tech expansion)
Product direction	CEO, CPO, CTO, CMO	CFO (unless budget question)
Hiring / org	CEO, CHRO, CFO, COO	CMO, CTO (unless their teams)
Pricing	CMO, CFO, CRO, CPO	CTO, CHRO
Technology	CTO, CPO, CFO, CISO	CMO, CRO
Fundraising	CEO, CFO, CRO	CISO, CHRO
Security incident	CEO, CTO, CISO, COO	CMO, CRO
M&A	CEO, CFO, CTO, CHRO, COO	-- (all relevant)

Maximum attendees: 6 roles per meeting. More than 6 creates noise, not insight.

---

Phase 2: Independent Contributions (ISOLATED)

Critical Rule: No cross-pollination. Each advisor contributes without seeing others' outputs. This is the primary groupthink prevention mechanism.

Contribution Order

1. Research/data gathering (if needed)
2. CMO  -- market perspective
3. CFO  -- financial perspective
4. CEO  -- strategic perspective
5. CTO  -- technical perspective
6. COO  -- operational perspective
7. CHRO -- people perspective
8. CRO  -- revenue perspective
9. CISO -- security/risk perspective
10. CPO -- product perspective

Contribution Format (Strict)

Each advisor's contribution must follow this exact format:

## [ROLE] -- [DATE]

Key Points (maximum 5):
  1. [Finding] -- Confidence: [High/Medium/Low] -- Source: [data source]
  2. [Finding] -- Confidence: [High/Medium/Low] -- Source: [data source]
  3. [Finding] -- Confidence: [High/Medium/Low] -- Source: [data source]

Recommendation: [Clear position statement]
Confidence: [High / Medium / Low]
Key Assumption: [The one assumption this recommendation depends on]
What Would Change My Mind: [Specific condition or data point]

Reasoning Techniques by Role

Role	Technique	How It Works
CEO	Tree of Thought	Explore 3 possible futures, evaluate each
CFO	Chain of Thought	Show the math, step by step
CMO	Recursion of Thought	Draft -> self-critique -> refine
CPO	First Principles	Decompose to fundamental user needs
CRO	Chain of Thought	Pipeline math must be explicit
COO	Step by Step	Map the operational process
CTO	Analyze then Act	Research -> analyze -> recommend
CISO	Risk-Based	Probability x Impact for every option
CHRO	Empathy + Data	Human impact first, then validate with metrics

---

Phase 3: Critic Analysis

Purpose: The Executive Mentor receives ALL Phase 2 outputs simultaneously and performs adversarial review.

Critic Checklist

Check	Question
Suspicious consensus	Where did agents agree too easily?
Shared assumptions	What assumptions are shared but unvalidated?
Missing voice	Who is not in the room? (customer voice? front-line ops?)
Unmentioned risk	What risk has nobody mentioned?
Domain bleed	Did any agent operate outside their domain?
Data quality	Which claims are backed by data vs. assumption?
Reversibility	Has anyone assessed if this decision can be undone?

Critic Output Format

## CRITIC ANALYSIS

Consensus Assessment:
  [Genuine agreement / Suspicious alignment / Split decision]

Unvalidated Assumptions:
  1. [Assumption shared by multiple advisors but not verified]
  2. [Assumption]

Missing Perspectives:
  - [Voice or data point not represented]

Unmentioned Risks:
  - [Risk nobody raised]

Domain Violations:
  - [If any agent operated outside their domain]

The Uncomfortable Truth:
  [The one thing nobody wants to say but needs to be said]

---

Phase 4: Synthesis

Purpose: Chief of Staff combines all inputs into a decision-ready format.

Synthesis Structure

## BOARD MEETING SYNTHESIS
Topic: [topic]
Date: [date]
Attendees: [roles]

### Decision Required
[One sentence: what must be decided]

### Perspectives Summary
| Role | Position | Confidence | Key Concern |
|------|----------|-----------|-------------|
| [Role] | [1-line summary] | [H/M/L] | [Top concern] |
| [Role] | [1-line summary] | [H/M/L] | [Top concern] |

### Where They Agree
[2-3 consensus points]

### Where They Disagree
[Named conflicts with each side's reasoning]
[What the disagreement is really about]

### Critic's View
[The uncomfortable truth from Phase 3]

### Recommended Decision
[Clear recommendation with rationale]

### Action Items (if approved)
1. [Action] -- Owner: [role] -- Deadline: [date]
2. [Action] -- Owner: [role] -- Deadline: [date]
3. [Action] -- Owner: [role] -- Deadline: [date]

### Your Call
[If you disagree with the recommendation, here are alternatives:]
Option A: [description] -- Trade-off: [what you gain/lose]
Option B: [description] -- Trade-off: [what you gain/lose]

---

Phase 5: Founder Review

FULL STOP. Wait for the founder. No agent acts beyond this point.

FOUNDER REVIEW

[Paste synthesis above]

Options:
  [A] Approve as recommended
  [M] Modify (specify changes)
  [R] Reject (specify reason)
  [Q] Ask follow-up question to specific role
  [D] Defer decision (specify timeline)

Phase 5 Rules

Rule	Rationale
Founder corrections override all agent proposals	Human judgment is final
No pushback on founder decisions	Agents advise, founder decides
30-minute inactivity auto-closes as "pending review"	Prevents zombie meetings
Founder can reopen any time	Decisions are not time-locked
Follow-up questions go to specific role	Keeps discussion focused

---

Phase 6: Decision Extraction

Purpose: After founder approval, extract and log all decisions.

Step 1: Write full transcript to Layer 1
  --> memory/board-meetings/YYYY-MM-DD-raw.md

Step 2: Run conflict detection against existing decisions
  --> Check for DO_NOT_RESURFACE violations
  --> Check for topic contradictions
  --> Check for owner conflicts

Step 3: Surface any conflicts to founder for resolution

Step 4: Append approved decisions to Layer 2
  --> memory/board-meetings/decisions.md

Step 5: Mark rejected proposals with DO_NOT_RESURFACE

Step 6: Confirm to founder:
  "Meeting concluded. Logged: [N] decisions, [M] action items,
   [K] DO_NOT_RESURFACE flags."

---

Failure Mode Reference

Failure	Detection	Fix
Groupthink	All advisors agree without tension	Re-run Phase 2 isolated; force "strongest argument against"
Analysis paralysis	Discussion exceeds 5 points per advisor	Cap at 5; force recommendation even with Low confidence
Bikeshedding	Discussion on minor points, major decisions deferred	Log as async action; return to main agenda
Role bleed	CFO making product calls, CTO pricing	Critic flags in Phase 3; exclude from synthesis
Layer contamination	Raw transcripts loaded in Phase 1	Hard rule: decisions.md only. Never raw.
Founder absence	Phase 5 timeout	Auto-close as pending. No decisions without founder.
Stale context	Company context not loaded	Phase 1 mandatory context check
Missing role	Key perspective not activated	Chief of Staff reviews topic against routing matrix

---

Meeting Cadence

Trigger	Meeting Type	Typical Duration
Scheduled quarterly	Full strategic review	2-3 hours
Complexity score >= 8	On-demand strategic	1-2 hours
Cross-functional conflict	Resolution meeting	1 hour
Crisis or urgent decision	Emergency session	30-60 minutes
Founder request	Any topic	Varies

---

Red Flags

• Board meetings consistently produce no decisions -- meeting is theater
• Same topic discussed in 3+ meetings -- decision avoidance
• Phase 2 contributions all align perfectly -- isolation was breached or topic is trivial
• No Phase 3 (critic) conducted -- groupthink risk
• Founder skipping Phase 5 -- decisions without accountability
• Decisions logged but never reviewed -- decision logger not functioning
• Meeting attendees always include all roles -- topic selection not working

---

Output Artifacts

Request	Deliverable
"Convene the board on [topic]"	Full 6-phase protocol execution
"Quick advisory meeting"	Abbreviated: Phase 1-2-4-5 (skip critic)
"Review a past meeting"	Load Layer 1 raw transcript (explicit request only)
"What did we decide about [topic]?"	Search Layer 2 decision history
"Resume a pending meeting"	Reload Phase 5 with pending synthesis

---

Tool Reference

meeting_simulator.py

Validates role activation, contribution completeness, and phase sequencing.

# Simulate with defaults
python scripts/meeting_simulator.py

# Specify topic and complexity
python scripts/meeting_simulator.py --topic "Series B timing" --type fundraising --complexity 9

# Specify activated roles
python scripts/meeting_simulator.py --type m_and_a --roles CEO CFO CTO CHRO

# List all topic types and required roles
python scripts/meeting_simulator.py --list-topics

# JSON output
python scripts/meeting_simulator.py --type strategy --json

decision_tracker.py

Tracks board decisions, detects conflicts, flags overdue reviews and actions.

# Track demo decisions
python scripts/decision_tracker.py

# From decision log file
python scripts/decision_tracker.py --input decisions.json

# JSON output
python scripts/decision_tracker.py --json

complexity_scorer.py

Scores decision complexity to determine single/dual/multi-advisor or board routing.

# Score with CLI flags
python scripts/complexity_scorer.py --topic "Market expansion" --domains 2 --reversibility 2 --financial 1 --team 2 --urgency 0

# Add modifiers
python scripts/complexity_scorer.py --topic "Acquisition" --domains 2 --reversibility 2 --financial 2 --team 2 --urgency 1 --modifiers cross_functional external_stakeholders sets_precedent

# JSON output
python scripts/complexity_scorer.py --topic "Pricing change" --json

---

Troubleshooting

Problem	Likely Cause	Fix
All advisors agree without any tension in Phase 2	Groupthink or trivial topic; isolation may have been breached	Re-run Phase 2 with forced "strongest argument against" from each role
Discussion exceeds 5 points per advisor	Analysis paralysis; no cap enforced	Hard cap at 5 key points; force a recommendation even with Low confidence
Phase 5 times out with no founder response	Founder absence or decision avoidance	Auto-close as "pending review" at 30 min; no decisions without founder
Same topic discussed in 3+ meetings	Decision avoidance or new data not surfaced	Escalate: force decision or formally defer with stated timeline
Decisions logged but never reviewed	Decision logger not integrated into meeting cadence	Add "previous decisions review" to Phase 1 context loading
Roles operating outside their domain	No critic analysis conducted or critic missed it	Enforce Phase 3 critic checklist; flag domain violations explicitly

---

Success Criteria

• Every board meeting produces at least 1 logged decision with owner, deadline, and review date
• Phase 2 contributions are independently generated (zero cross-pollination incidents per quarter)
• Phase 3 critic analysis identifies at least 1 unvalidated assumption per meeting
• Founder approval/modification/rejection captured within 30 minutes of synthesis presentation
• Decision history has zero conflicting active decisions (conflicts detected and resolved)
• Meeting duration stays within 2 hours for standard strategic reviews, 1 hour for resolution meetings
• 90%+ of logged decisions have action items completed by their stated deadlines

---

Scope & Limitations

In Scope: Multi-agent deliberation protocol, role activation matrix, contribution formats, critic analysis, synthesis, decision extraction, decision conflict detection, meeting simulation.

Out of Scope: Actual AI agent orchestration (this is a protocol specification, not runtime code), real-time meeting facilitation, video/audio recording, external board member management.

Limitations: The protocol assumes all advisor contributions are available in text format. Complexity scoring provides routing guidance but cannot account for political dynamics. Decision conflict detection works on exact topic matching -- semantic conflicts across different topics require human judgment.

---

Integration Points

Skill	Integration
chief-of-staff	Routes questions that score 9-10 complexity into the board meeting protocol
decision-logger	Phase 6 feeds decisions directly into the two-layer decision memory
board-deck-builder	Board deck sections provide pre-read context for Phase 1
executive-mentor	Phase 3 critic analysis can be performed by the Executive Mentor skill
ceo-advisor through ciso-advisor	All C-suite advisors contribute independently in Phase 2
strategic-alignment	Validates that meeting decisions align with strategic goals