Seven specialist agents deliberate under time and budget constraints to stress-test a client's growth thesis from every angle. The output is a 90-day roadmap with ranked initiatives, revenue projections, and execution assignments.
Seven advisors with deliberately conflicting biases. Each optimizes for a different dimension of growth. The orchestrator — Growth Strategist — runs on Opus 4.6. All board members run on Sonnet 4.6.
“I don't care about the funnel until someone shows me the conversion rate at each step. Optimizing a leaky bucket is cheaper than filling it faster.”
“Before we spend $5K on ads, can we test a 15% price increase on new signups? If conversion holds, that's pure margin with zero acquisition cost.”
“The current market is fine, but there's a bigger one next door. 30% of their signups are freelance designers — that's not a bug, that's a market.”
“Revenue Architect wants annual contracts. But 60% of customers are month-to-month freelancers. This will increase churn, not revenue.”
“Before we charge in, let me show you who's already there and why they haven't been beaten.”
“This is a 5-person team already behind on their current roadmap. Pick one. Do it well. Then we talk about the next one.”
“The client has 3,000 email subscribers they haven't emailed in 4 months. That's the fastest revenue opportunity in the room.”
Four deliberate tension pairs force genuine disagreement. The value comes from forced conflict, not consensus. Each pair embodies a real strategic trade-off that every SMB faces.
The Growth Strategist is the integrator. They hold the client's full picture while forcing the board to stress-test every assumption. Runs on Opus 4.6 for maximum reasoning depth.
Every recommendation must pass: “Can a 5-person team actually do this in 90 days?”
Under $50K MRR, revenue-generating moves beat infrastructure moves. Always.
One initiative that moves the needle meaningfully, plus two fast wins that build momentum and fund the big bet.
Never recommend something requiring resources the client doesn't have. Stretch is OK. Fantasy is not.
| OBJECTIVE_FUNCTION | Highest-leverage growth moves within actual constraints |
| TIME_HORIZON_PRIMARY | 90 days |
| TIME_HORIZON_SECONDARY | 6–12 months |
| CORE_BIAS | Leverage and feasibility — big impact, realistic execution |
| RISK_TOLERANCE | Moderate — SMBs can't absorb large failures |
| DEFAULT_STANCE | “What is the single move that creates the most revenue with the least resource drain?” |
Before the board meets, researcher agents gather live market data via web search. Their findings are written to the brief directory and injected into the deliberation as context.
Market size, growth trends, customer segments, adjacent opportunities, geographic opportunities.
Direct competitors, indirect alternatives, pricing landscape, strengths/weaknesses, positioning gaps.
Web presence, SEO landscape, social media, industry benchmarks, channel opportunities.
All three domains in one pass: 35% competitive, 35% market, 30% digital. Used when time or budget is tight.
The orchestrator drives the session through a structured loop. All targeted board members respond in parallel. The session ends when the orchestrator is satisfied or constraints force a wrap-up.
converse(to, message)
Send a prompt to one member, an array of members, or "all". All targeted members respond in parallel.
end_deliberation(message)
Wrap the meeting. Collects final statements from all members. Operations Realist speaks last.
write + edit
Produce the final memo after deliberation ends. Written to the client's memos/ directory.
Constraints prevent infinite deliberation and force prioritization. The system tracks elapsed time and token cost in real-time, displayed as gauges in the TUI.
Deliberation must continue. Not enough depth yet.
Orchestrator may end when satisfied with the debate quality.
System signals wrap-up. Orchestrator starts converging.
Forced call to end_deliberation. No more rounds.
Every deliberation transforms a structured brief into a comprehensive memo. The brief is the input contract; the memo is the deliverable.
Four mandatory sections. Researchers may add context files alongside the brief.
Comprehensive deliverable. Becomes the agency's proposal to the client.
Artisan ceramics studio in Gothenburg. The board diagnosed a pricing problem disguised as a growth problem and recommended a 25–30% price increase as the critical first move.
Each board member has a personal scratch pad — a markdown file that accumulates patterns, observations, and risk notes across deliberations. The expertise block is injected into the agent's prompt at session start and updated after sessions.
Patterns observed across clients. Which strategies work for which business types. Argument strategies that proved effective.
Expertise files live in teams/growth-board/expertise/. Injected via {{EXPERTISE_BLOCK}} at session start. Agents update their own scratch pads.
Deliberation quality improves over time without manual tuning. Each engagement makes the next one sharper.
Key pattern: handcrafted artisan e-commerce
Artisan brands routinely underprice relative to their positioning. Nordic Ceramics at 350 SEK/mug is barely above Kähler (mass production, 100+ employees) at 299–399 SEK. The market signal is not “you're expensive” — it's “you're confused about your category.”
Silent price increases (no announcement) are the right default for small-catalog artisan stores. 30-day conversion monitoring is the test — if conversion holds within 5pp, the increase is absorbed.
Risk worth tracking: Founder psychology is the real execution risk. Founders who have underpriced for years have internalized the low price as part of their identity. This is not a pricing problem — it's a coaching problem.
Team configurations are shared across all clients. Client data is isolated per engagement. Names are kebab-case, briefs are date-prefixed.
teams/growth-board/ config.yaml ← board config, constraints, brief sections agents/ orchestrator.md ← Growth Strategist (Opus 4.6) revenue-architect.md ← board members (Sonnet 4.6) pricing-analyst.md market-scout.md customer-voice.md competitive-intel.md operations-realist.md digital-strategist.md researchers/ market-researcher.md competitive-researcher.md digital-researcher.md generalist-researcher.md expertise/ *.md ← per-member scratch pads clients/<client-name>/ context/ ← business snapshot briefs/YYYY-MM-DD-<slug>/ brief.md ← input research-*.md ← researcher outputs deliberations/<session>/ conversation.jsonl ← full transcript tool-use.jsonl ← all tool calls memos/<session>/ memo.md ← final output templates/ growth-board-brief.md client-context.md