SCENARIO ATLAS
Five paths through the Control Plane
Turn an early product idea into a structured foundation.
Define the user, problem, intended outcome, scope, assumptions, evidence and first useful release. Preserve those decisions in a Project Model that can later support implementation and continued development.
- Starting condition
- An idea without a durable product definition
- Primary decision
- What is the first useful product?
- Desired outcome
- A structured foundation for a bounded first release
Reconstruct the project before adding more change.
Collect the available sources, separate observed behaviour from claims, identify risks and missing context, and decide the next responsible action. The aim is to continue from project truth rather than from the previous tool’s chat history.
- Starting condition
- Existing or AI-built software with fragmented context
- Primary decision
- What is true, uncertain and responsible to do next?
- Desired outcome
- A reliable continuation path based on project truth
Make the decision from evidence, not frustration.
Evaluate product fit, architecture, implementation, operations, ownership and evidence quality. Compare repair, stabilisation, rebuild and defer options before scope is committed.
- Starting condition
- Software that is difficult to trust or continue
- Primary decision
- Should the project clarify, stabilise, repair or rebuild?
- Desired outcome
- An evidence-based decision before scope is committed
Give the next builder a durable project contract.
Bring product intent, decisions, architecture, scope, acceptance criteria, ownership, access boundaries and verification expectations into a reviewable Build Specification and continuation record.
- Starting condition
- A project moving to another developer or delivery stage
- Primary decision
- What must the next builder understand and be allowed to do?
- Desired outcome
- A reviewable project contract with explicit boundaries
Keep AI agents, people and external systems in one visible workflow.
The active Control Plane programme introduces Work Items, Agent Tasks, human-action coordination and progressive technical evidence. Bounded execution, independent verification and Tested Preview follow through the roadmap.
- Starting condition
- Work split across AI agents, people and external systems
- Primary decision
- How should work, evidence and approvals stay visible?
- Desired outcome
- Coordinated delivery with future verification gates kept distinct
In active developmentWork Items, Agent Tasks and human-action coordination
On the roadmapbounded execution, independent verification and Tested Preview
