AI Board of Directors
A multi-AI discussion workflow where Claude, GPT-4, and Gemini deliberate on your questions and deliver a synthesized recommendation.
What It Does
Submit any question through a web form. Three AI models independently analyze it, then a neutral moderator synthesizes their perspectives into one actionable recommendation. You get the full transcript showing each AI's reasoning plus the final consensus.
Use cases: Business strategy decisions, technical architecture choices, marketing direction, career advice, product pivots — any question where multiple perspectives lead to better answers.
How It Works
A 4-step pipeline from question to consensus recommendation.
Claude (Anthropic)
- Sonnet 4 model
- Nuanced reasoning
- Risk-aware analysis
- ~$0.015 per query
GPT-4o (OpenAI)
- Broad knowledge base
- Structured responses
- Also serves as moderator
- ~$0.02 per query
Gemini Flash (Google)
- Fast inference
- Multi-modal capable
- Alternative perspective
- ~$0.001 per query
Key insight: Each AI responds independently — they don't see each other's answers. The moderator then finds common ground, highlights disagreements, and produces a balanced recommendation.
n8n Workflow Architecture
The complete workflow runs in n8n with parallel API calls and a synthesis step.
Workflow Nodes
n8n-provided web form collects the user's question. Generates a public URL anyone can access.
Three HTTP request nodes fire simultaneously — Claude API, OpenAI API, and Gemini API — each with the same question.
All three responses are collected and formatted into a structured prompt for the moderator.
GPT-4o receives all three perspectives and produces a consensus recommendation with areas of agreement and disagreement.
A Code node formats the full transcript — each AI's response + the final recommendation — into clean HTML for display.
Credential Setup
- Anthropic: HTTP Header Auth — header
x-api-keywith your API key - OpenAI: HTTP Header Auth — header
AuthorizationwithBearer YOUR_KEY - Gemini: HTTP Query Auth — parameter
keywith your API key
Setup Guide
Get the AI Board of Directors running in under 10 minutes.
Prerequisites
- An n8n instance (cloud or self-hosted)
- Anthropic API key — console.anthropic.com
- OpenAI API key — platform.openai.com
- Google AI API key — aistudio.google.com
Step-by-Step
Upload AI-Board-Simple.json to your n8n instance via Workflows → Add Workflow → Import from File.
In n8n Settings → Credentials, create 3 new credentials (HTTP Header Auth for Anthropic + OpenAI, HTTP Query Auth for Gemini).
Open each API node in the workflow and select the corresponding credential you just created.
Toggle the workflow Active, copy the Form Trigger URL, and open it in your browser.
Tip: Test with a simple question first — "What's the best programming language for beginners?" — to verify all three APIs respond correctly.
Customization Ideas
Extend the board with additional rounds, models, and capabilities.
Multi-Round Debate
Duplicate Round 1 nodes so AIs can respond to each other's initial takes. Creates richer, more nuanced discussions.
Specialized Boards
Create topic-specific boards with tailored system prompts — Marketing Board, Tech Board, Finance Board — each with domain expertise.
More Models
Add DeepSeek, Llama, Mistral, or any model with an API. More perspectives = richer synthesis.
Other Extensions
- Memory: Add a database to track previous discussions and reference them
- Voting: Include a consensus score (0-100%) showing how aligned the AIs were
- Translation: Add a translation step for multi-language support
- Persistence: Save discussions to Google Sheets or Postgres for later review
- Webhook mode: Replace the form with a webhook for programmatic access from your own frontend
Troubleshooting & Cost
Common issues and cost breakdown.
Common Issues
- "Invalid API Key" errors: Verify credentials match the exact format (Header Auth for Anthropic/OpenAI, Query Auth for Gemini)
- Responses look malformed: Check the "Format Output" Code node — Gemini's response structure may differ from expectations
- Workflow times out: Increase n8n's execution timeout in Settings, or reduce
max_tokensin each API call - Form URL not working: Make sure the workflow is toggled Active (top-right toggle in n8n)
Security Notes
- Never commit API keys to git — use n8n's built-in credential system
- The form URL is public by default — add authentication if deploying externally
- Consider rate limiting to prevent abuse if sharing the URL widely
Project files: form.html (standalone web form), n8n/ (workflow JSON), setup-guide.md (detailed setup instructions), testing-guide.md (QA checklist).