Product demo · No login
Change a rule. Watch the decision move.
Pick a sample case, run it through the pipeline, then edit the policy and run it again.
This runs in your browser on sample data and sends nothing. The rules mirror how the policy engine decides; your real policy runs on our servers.
Sandbox
Case RC-1042
Arrived cracked, photo attached · $38.00
Press Run to send this case through the pipeline.
1. Intake
Not yet2. Signals
Not yet3. Policy
Not yet4. Outcome
Not yet5. Execution
Not yet6. Audit trail
Not yet
Sample policy
Edit a value, then rerun. Each edit publishes a new version, and a run pins the version it started with.
P10Payment on hold
The processor reports a hold on the payment
High riskP20Return window
Order is older than the window
Auto-denyP30Abuse risk escalation
Abuse risk at or above the level
EscalateP40Minimum proof of purchase
Proof level below the floor
Manual reviewP50Returnless ceiling
Damaged or wrong item, amount at or under the ceiling
Auto-approveP60Auto-approve threshold
Amount at or under the threshold
Auto-approveDefault outcome
When no rule matches · Manual review
The next run pins version 7.
What is real
The same rules the engine follows.
Same rule semantics
Rules run by priority, the first match wins, and the default outcome covers the rest. The five outcomes here are the engine's five.
AI advises, policy decides
Classifier and sentiment readings arrive first and only advise. The router sends them to one of three cleared model vendors.
Pinned versions, full audit
Published versions never change; an edit makes a new one. Every step writes to an append-only audit trail.
Get started
Run it on your refunds.
Start on Free or the 14-day Growth trial. No card needed.