Your business proposal has one author but needs many critics. The Council of Agents method uses Claude to simulate a boardroom of adversarial experts — a CEO focused on ROI and scalability, a Lawyer scanning for liability and compliance gaps, and a Creative Director pushing for boldness and market differentiation — who debate, challenge, and pressure-test your idea from angles you'd never consider alone.
When you write a business proposal, you see it through your own lens — usually the lens of the builder. You know the product intimately. You believe in the vision. And that belief creates blind spots. A CEO would ask about unit economics and competitive moats you haven't considered. A lawyer would identify regulatory exposure you've glossed over. A creative director would challenge whether your positioning is distinctive enough to survive a crowded market.
Hiring a real advisory board costs tens of thousands of dollars. Assembling three domain experts for a single review session is logistically difficult. But Claude can embody each of these perspectives with remarkable fidelity when given the right system prompts, role constraints, and interaction structure.
The Council of Agents method doesn't replace real advisors — it prepares you for them. By running your proposal through a simulated adversarial review, you identify weaknesses before stakeholders do. You arrive at the real boardroom having already survived the hard questions.
The method uses a single Claude conversation (or Claude Project) with carefully sequenced role-switching. Each "agent" is activated by a system-level persona prompt that constrains Claude's reasoning to that character's professional priorities, vocabulary, and risk tolerance. A fourth role — the Moderator — synthesizes the debate and identifies where the agents agree, where they clash, and what remains unresolved.
You can run this in two ways: sequential mode (one agent at a time, each responding to the previous agent's critique) or town hall mode (all agents respond to the same prompt, then debate each other's positions). Sequential mode is simpler. Town hall mode produces richer, more adversarial output.
Start by establishing the three core agents and one moderator. Each agent needs a persona prompt that defines their professional identity, what they prioritize, what they're skeptical of, and how they communicate. The more specific the persona, the more distinct and useful the critique. Avoid generic descriptions — give each agent a decision-making framework.
Paste your full business proposal into the conversation. Then activate each agent one at a time: ask the CEO to review first, then the Lawyer, then the Creative Director. Each agent should provide their critique independently before seeing the others' responses. This prevents anchoring bias — the first opinion shaping all subsequent ones.
Now that each agent has given their independent critique, run the town hall. Ask Claude to have the three agents debate the most contentious points. The CEO might dismiss the Lawyer's regulatory concerns as overly cautious. The Creative might argue the CEO's cost-cutting will kill differentiation. The Lawyer might warn both that the proposed timeline creates contract exposure. Let them argue.
Activate the Moderator agent to analyze the full debate. The Moderator's job is to produce a structured summary: points of consensus, points of disagreement, unresolved risks, and a prioritized list of revisions. This synthesis is the actionable output — the document you use to revise your proposal before presenting it to real stakeholders.
Update your proposal based on the Moderator's synthesis, then run the council again. The second pass is faster and sharper — agents focus on whether you addressed their concerns adequately. Two passes typically surface 80–90% of the weaknesses a real advisory board would find.
| Agent | Role in the council | What they scrutinize |
|---|---|---|
| CEO The Strategist | Business viability + scalability | Unit economics, competitive moats, market timing, team gaps |
| LAWYER The Guardian | Risk assessment + compliance | Regulatory exposure, IP risks, contract gaps, liability |
| CREATIVE The Provocateur | Market positioning + differentiation | Brand distinctiveness, audience resonance, messaging clarity |
| MOD The Synthesizer | Debate resolution + action items | Consensus mapping, conflict resolution, revision priorities |
Copy any prompt below. Replace bracketed placeholders with your own details.
Every prompt in this guide plus all prompts across the full category — advanced workflows, specialized use cases, and production-grade templates.
Category Bundle — one-time access
Unlock Category Prompts — $19.99ONE-TIME · 30-DAY GUARANTEE · INSTANT ACCESS
The Council of Agents method exploits a well-documented cognitive phenomenon: we cannot effectively argue against our own ideas. The planning fallacy, confirmation bias, and the IKEA effect (overvaluing things we build) conspire to make us blind to flaws in our own proposals. By externalizing critique into distinct personas, Claude bypasses these biases and generates the kind of pushback you'd only get from people who don't share your emotional investment.
The key design choice is adversarial independence. Each agent reviews the proposal without seeing the others' critiques first. This mirrors best practices in structured analytic techniques used by intelligence agencies — independent analysis before group discussion prevents groupthink and anchoring. When the agents then debate each other in the town hall phase, genuine tensions emerge because their initial positions were formed independently.
"You are a CEO" produces bland output. "You are a CEO who has scaled two B2B SaaS companies past $50M ARR and was burned by a pivot that came too late" produces critique shaped by specific experience. The more biographical detail you include, the more distinctive each agent's perspective becomes.
If your proposal is for a healthtech startup, tell the Lawyer agent to specialize in HIPAA and FDA regulations. Tell the CEO agent they've run healthcare companies before. Generic expertise produces generic critique. Domain-specific expertise finds the risks that actually matter.
The most valuable output from a council session is disagreement. If the CEO says "cut the R&D budget" and the Creative says "the R&D differentiation is the entire value proposition," that tension reveals a genuine strategic decision you need to make. Don't ask Claude to resolve every disagreement — some tensions are real and should stay unresolved until you make a judgment call.
The first council pass finds the obvious problems. The second pass — after you've revised the proposal — finds the subtle ones. It also tests whether your revisions introduced new issues. Two passes is the minimum for a serious proposal. Three passes is appropriate for fundraising decks and board presentations.
Claude is simulating expertise, not providing it. The CEO agent doesn't have access to real market data. The Lawyer agent's analysis is not legal advice and should never replace consultation with a licensed attorney. The Creative Director's brand assessment isn't backed by consumer research. The council identifies questions you should be asking — it doesn't provide definitive answers.
This method works best with Claude Pro or the API. The free tier's usage limits may not sustain a full council session with debate and synthesis. Claude's 200K-token context window comfortably holds a proposal, three agent critiques, a debate transcript, and a moderator synthesis in a single conversation.
The quality of the council depends entirely on the quality of your persona prompts. Vague personas produce vague critique. Specific, constrained, opinionated personas produce the kind of sharp feedback that actually changes your proposal for the better.