👥 Team
A squad owns a customer outcome from client meeting to production.
Who does what
The shift in roles
AI writes specifications, code, tests, and documentation. A Product Engineer combines product and development responsibilities:
- No manual coders: AI agents generate 100% of code, tests, and documentation.
- Product Engineer: Owns Meet through Prove, leads client conversations, sets priorities, and maintains context and the delivery harness.
- Meet: Understand the client's problem and desired outcome.
- Think: Set the constraints that guide AI decisions.
- Client empathy: Keep the client's intent clear throughout delivery.
Shape
A typical squad has one or two Product Engineers. Architecture, DevOps, and QA and maintenance are shared across squads:
| Member | Primary Focus | Role in Meeting-Driven Development |
|---|---|---|
| Product Engineer | Meet, Triage, Think, Plan, Code, and Prove | Leads client dialogue, defines value and priorities, steers delivery through the harness, and resolves escalated failures |
| QA and maintainer | Quality and production support | Runs Learn: watches QA and production, handles incidents, and reports what monitoring finds; shared across squads |
| DevOps | Automated deployment and pipeline | Runs Release: automated deployment; shared across squads |
| Architect | The harness: Platform Rails, Persistent Context, Deterministic Controls, and the Forge | Owns system constraints, decision policy, and risk controls; agents execute architectural decisions autonomously and escalate unresolved failures |
Boundaries
A squad is defined by a user outcome, never by a technology layer.
| Good outcome-oriented boundary | Bad technology-layer boundary |
|---|---|
| "Client Onboarding & Identity" | "Frontend team" |
| "Payments, Billing & Invoicing" | "Backend API team" |
| "Catalog & Search Relevance" | "Database & SQL team" |
Technology-layer squads create hand-offs and fragment the context agents need.
Ownership of code and context
The outcome-owning squad owns the code, tests, and context. Agents can propose cross-repository changes through change requests; each repository's controls enforce its business rules and boundaries.