Ai Decisioning

Operator enables AI Decisioning across the workspace. A subset of users have previously requested data-processing opt-out under GDPR. Their events still enter the training corpus.

Filter opt-out users out of the AI Decisioning training corpus at source. Per docs, the training corpus is derived from warehouse data the operator exposes — the operator must enforce opt-out via the Model query. Document the opt-out filter as a contract.

AI Decisioning runs treatment (model-picked action) vs control (operator-specified baseline) at 80/20. Operator pulls treatment users into other campaigns; control gets cross-contaminated.

Treat the control group as a protected hold-out: no other AI Decisioning campaign and no overlapping personalization touches control users during the experiment window. Document the hold-out contract and audit overlaps. Per docs, lift measurement requires clean control.

AI Decisioning picks send-time and channel per user. A user in a consent-revoked SMS state still receives an SMS recommendation because the model wasn't told about channel eligibility.

Per docs, AI Decisioning supports per-user action eligibility constraints — encode consent state (sms_opted_in, email_opted_in, push_opted_in) as inputs or hard constraints. The model must respect eligibility regardless of predicted lift. Audit suggestions against consent state.