Before
Teams read scattered card comments, searched research notes, and manually translated each idea into a hypothesis and pilot proposal. They also had to check for duplicate review handoffs and reformat unresolved questions for an innovation lead.
How it works
- Collect up to three open ideas, their discussions, approved research, and pilot constraints.
- Assess the evidence and determine whether the idea supports a bounded test with measurable success and stopping criteria.
- Verify a pilot comment on the original idea, or reuse or create a review card when human judgment is required.
Workflow details
- Intake: When started, the workflow reads up to three open ideas from the configured experiment intake list and preserves each idea’s source key, description, and problem statement.
- Evidence assembly: It loads the latest approved research documents and pilot rules, including the two-week duration, one opt-in depot scope, and supervisor approval requirement for wider rollout.
- Assessment: For each idea, it reads the discussion and supporting study, separates observations from assumptions, and determines whether the evidence supports a bounded test with a target group, falsifiable hypothesis, comparison, metric, guardrail, and stopping rule.
- Decision routing: If the evidence and permission are sufficient, the workflow adds a pilot proposal to the original idea and verifies the returned comment action. It does not start the pilot.
- Controlled handoff: If evidence or permission is inadequate, it searches for an existing review card before creating one, then records the evidence, completed work, unresolved issue, and recommended decision for an innovation lead.
What changed
Each idea is automatically evaluated against its discussion, research, and pilot guardrails; eligible proposals are attached and verified on the original card, while incomplete cases are routed to an innovation lead with the open decision clearly documented.
Work removed
Copying research excerpts, separating observations from assumptions, reformating ideas into pilot proposals, checking for existing review cards, and manually creating evidence-based handoff cards.
Expected outcome
A single run can assess up to three open ideas and route each to a verified pilot-comment outcome or a review decision card, without starting a pilot automatically.


