An AI design review cut an unnecessary charting integration

Internal and client teams were trying to design real agent systems (support triage, internal copilots, data workflows) with no shared idea of what an agent is.

Conversations kept drifting into hype and vague wishlists. One person talked about tools, another about prompts, another about UX. Projects stalled because there was no disciplined way to scope what mattered first or how deep to go.

I formalized the Agent Redwood assessment: twelve components (Personality, Planning, Mission, Constraints, Memory, Evaluation, Tools, Awareness, Reward Model, Metadata, Strategy, Integrations). In a working session the mission gets named first, then each component is decided against it: whether the first version needs it, and how far to go. In one session a team's agent had one job, a daily shortlist. The TradingView integration they'd asked for did nothing for that job, so it was cut before anything was built.

• Agent Redwood worksheets and diagrams
• Orchestration stack (LangGraph, Pydantic-based agents, custom tools)
• Standard version control and review

Teams aligned on a concrete v1, killed low-leverage ideas early, and shipped simpler agents that held up in production and matched the constraints they actually had.

Teams that said "let's experiment with AI" got a design session they could repeat, and founders and engineers agreed on what they were building.

The AI projects I've seen fail started as scoping failures. Deciding the twelve parts up front catches a part like that TradingView integration before anyone builds it.