1. Decision-support, not autonomous authority
RestoreCheck IQ may use deterministic rules, parsing, calculation, OCR, workflow automation, classification, or forecasting to organize information and identify items for review. These tools support human decisions; they do not approve claims, determine insurance coverage, establish legal compliance, replace source documents, or make employment decisions on RestoreCheck IQ’s behalf.
A confidence indicator is not proof of correctness. Material requirements, calculations, matches, exceptions, risk scores, and exports must remain traceable to source data and reviewable by an authorized person.
2. Accuracy and improvement controls
- Preserve the source document beside extracted requirements so reviewers can compare the output.
- Use explainable rules, versioned requirement sets, confidence, validation failures, and human approval states.
- Separate training, test, demo, and production data; never use one customer’s identifiable data to improve another customer’s output without lawful authority.
- Measure extraction and workflow performance against a representative regression corpus before release.
- Record corrections in a tenant-scoped feedback history and promote rule changes only after review and regression testing.
- Provide an override, correction, escalation, or appeal path for consequential outputs.
3. Customer-connected AI and agents
A customer may connect its own AI or automation only through approved tenant-scoped credentials and documented API permissions. It must not receive platform-admin access, reuse credentials across organizations, train on data beyond the customer’s authority, or execute destructive or high-impact actions without an explicit human control.
RestoreCheck IQ does not enable platform-wide AI processing merely because one customer connects an AI tool. The connecting customer is responsible for that provider’s notices, consent, retention, security, output review, and lawful use.