The way in
The 90-Day Causal Growth Control Pilot
Begin observe-only. Advance to human-approved actions only after evidence quality, economics, and policy boundaries are validated. Autonomy is earned, never assumed — every advance in authority is a deliberate, human-made decision backed by the pilot's own evidence. Pilot-ready
Deliberately narrow
Pilot scope
A pilot that tries to govern everything proves nothing. This one scopes tightly so its evidence is unambiguous.
- One brand or business unit
- Limited media and commerce sources
- One performance-incident decision class
- One budget-allocation or profit decision class
- Explicit evidence boundaries
- Explicit human authority
- Observe-only starting mode
- Measured predicted-versus-realized outcomes
- Clear expansion criteria
Three phases
The structured plan
Days 1–30 · Connect and observe
The system earns the right to an opinion before it offers one.
- Connect the scoped systems and confirm source access
- Validate identities, timestamps, and economics
- Establish canonical event definitions
- Map the critical entities and constraints
- Run in observe-only mode — no recommendations, no actions
Governance checkpoint: evidence-quality review with your team. The pilot does not advance until the evidence base is trusted.
Days 31–60 · Diagnose and recommend
Recommendations appear — every one of them for human review, none of them acting.
- Generate causal recommendations for the scoped decision classes
- Evaluate counterfactuals, including the no-action baseline
- Measure and report confidence on every recommendation
- Validate every candidate against configured policy
- Require human approval on everything — no exceptions in this phase
Governance checkpoint: recommendation-quality review. Your team scores the diagnoses against ground truth before any action authority is discussed.
Days 61–90 · Approve, act, and learn
Narrowly scoped actions — human-approved, bounded, reversible — and honest scoring of the results.
- Enable limited governed actions within explicit bounds
- Preserve rollback paths on every action
- Measure predicted versus realized outcomes
- Produce the executive scorecard
- Recommend an expansion roadmap — or recommend stopping
Governance checkpoint: exit review against the criteria below. Expansion is a decision you make on the pilot's record, not a renewal that happens by default.
How the pilot is judged
Success metrics and exit criteria
Success metrics
Agreed before day one, measured throughout, reported without editing:
- evidence completeness and quality
- time-to-detection on scoped incidents
- diagnostic accuracy vs. ground truth
- recommendation acceptance rate
- policy-check coverage and veto record
- prediction accuracy on acted decisions
- decision latency, sense to authority
Targets for each metric are set with your team during scoping — the pilot commits to measuring honestly, not to hitting a number we invented before seeing your data.
Exit criteria
The pilot ends in one of three explicit states:
The pilot's product is evidence about your own decisions — whichever exit you choose, you keep it.
Next step
Scope it around your most expensive decision
Tell us the decision that costs you the most when it goes ungoverned. We'll shape the two pilot decision classes around it.