Designing a Personal Board of Directors Using ChatGPT Projects

The Problem: Advice Arrives Too Slowly

Executives rely on mentors, peers, and trusted advisors, but human counsel is episodic. Decisions often move faster than advice can arrive. Leadership today requires continuous, multi-perspective reasoning, not scattered one-offs.

Previously, I used OpenCode and Ollama to simulate advisory agents in separate local workspaces. Today, ChatGPT Projects allow all agents to exist in a single persistent environment, creating a cohesive Personal Board of Directors.

One Project, All Advisors

Instead of creating separate workspaces per agent, I now create one ChatGPT Project that contains:

AGENTS.md
challenger.md
sponsor.md
mentor.md
peer.md
outsider.md
personal-anchor.md
operator.md

How it works

  • AGENTS.md — defines the board structure, inter-agent relationships, and rules for cross-perspective reasoning.
  • Role files (challenger.md, operator.md, etc.) — define each agent’s perspective, priorities, heuristics, and how they integrate insights from the other advisors.

By loading all agents into the same Project, every advisory perspective is always available for any query. There is no need to switch workspaces or re-load context.

Custom Instructions for Multi-Agent Awareness

The Project’s custom instructions tell ChatGPT how to use the agents effectively:

1. Primary Guidance: Use the content in AGENTS.md as the foundation for understanding the board’s structure and inter-agent relationships.

2. Role Perspective: Consult each role file (challenger.md, operator.md, etc.) to apply the specific advisory lens.

3. Cross-Agent Awareness: When reasoning or evaluating tradeoffs, integrate relevant input from all advisors.

4. Consistency: Maintain continuity with prior decisions and guidance stored in the Project’s memory.

This ensures each response considers all perspectives, while still respecting each agent’s role.

Using the Personal Board in Practice

Quick Queries

You can ask a single question and request multi-agent analysis:

“What risks does this plan overlook?”  
- Challenger: tests assumptions  
- Operator: identifies execution bottlenecks  
- Peer: evaluates realism  
- Outsider: introduces fresh perspective

Major Decisions

For complex decisions, the Project allows you to run a full “board session”:

  1. Prompt all agents sequentially or simultaneously
  2. Collect and compare viewpoints
  3. Aggregate insights into a coherent decision plan

Because all agents are in one Project, their reasoning is aware of each other, producing responses that reflect interdependencies and nuanced tradeoffs.

Benefits of This Approach

  • Continuous, consistent reasoning — memory is persistent, so advisory context accumulates over time.
  • Cross-referenced perspectives — agents consult each other via the base AGENTS.md guidance.
  • Simplified workflow — no multiple Projects; all advisors live in one space.
  • Actionable insights — output can feed directly into strategy, execution planning, and communication.

The Emerging Pattern

The real leverage is not prompt engineering. It’s workspace design:

  • Identify key advisory perspectives
  • Capture them in structured files (AGENTS.md + role files)
  • Load them into a single Project with custom instructions
  • Leverage persistent memory and multi-agent awareness

The AI becomes more than a chat assistant — it becomes a standing personal board of directors, continuously available, aligned across perspectives, and capable of compounding executive insight.

Written on March 3, 2026