Boardroom is an opinionated Pi workflow pack for executive deliberation. It runs board sessions from structured briefs and produces ranked recommendations, tradeoffs, and final decision memos. It is a Justin Clark Network product.

Overview

Boardroom treats strategic decisions as a structured process, not a chatbot conversation. You write a brief. A board of six specialist agents deliberates. The CEO synthesizes their positions into a decision memo. Every step produces versioned, auditable artifacts. In the broader stack, Pi is the substrate and Boardroom is one packaged workflow that runs on top of it.

Core concepts

Briefs - structured inputs with situation, stakes, constraints, and a key question
Board members - six specialist agents, each optimizing for a different dimension
Deliberation rounds - parallel or sequential prompts with synthesis between rounds
Votes - each member votes accept, reject, or defer with rationale
Memos - final CEO decision with recommendations, conditions, and next moves
Artifacts - deterministic, versioned outputs at every stage

Why structured disagreement matters

A single LLM response to a strategic question will be coherent but narrow. It optimizes for one perspective. Boardroom deliberately introduces six competing lenses, then forces synthesis across them.

The Contrarian exists to break false consensus. The Moonshot exists to surface under-explored upside. Revenue exists to ground everything in economics. When these perspectives collide, the resulting decision is sharper than any single viewpoint.

A 5-accept / 1-defer vote tells you something a unanimous "yes" does not. The defer is a quality gate, not a failure.

Outputs

Every Boardroom run produces a complete artifact set:

The original brief
Full deliberation transcript (JSONL event log)
Individual board member positions and votes
Persistent run state (costs, timing, vote summary)
Final CEO decision memo

Artifacts use relative paths only. Costs are tracked in microUSD precision. Durations use canonical elapsed_ms. State files are versioned and written atomically.

Operator control

Boardroom does not pretend to be autonomous. The operator authors the brief, configures the board, reviews intermediate outputs, and decides whether to accept the recommendation.

The source repository is published under Apache License 2.0. That keeps adoption simple, permits commercial use, and gives downstream users an explicit patent grant.

It is built to be human-operable first and agent-operable by design. That means the runtime exposes both an interactive slash-command kickoff and a deterministic text kickoff, so real users and their LLMs can drive the same workflow without depending on fragile UI parsing.

Budget and time constraints are enforced at runtime. When the budget or time limit is reached, the CEO is prompted to close the deliberation. There is no runaway spending or unbounded context exhaustion.

Board composition is configurable. You can add, remove, or modify specialist roles. Each member has a persona file, a color, and a specific optimization directive.

What Boardroom is not

It is not a chatbot or conversational AI product
It is not an "autonomous AI executive" - the operator is always in control
It is not a framework - it is a specific, opinionated workflow
It is not cloud-hosted - it runs locally inside Pi
It does not make decisions for you - it produces structured decision support