Liturgy, Doctrine & Formation
Tier 4 touches sacred text and teaching — high doctrinal sensitivity. The system retrieves and cites from sealed, verified sources, and is bound by one rule above all: cite, but never explain. It may show what the text says and where; it may never invent commentary, speak for God, or teach on its own authority.
The Inverse Risk-to-Autonomy Scaling Law. As human vulnerability and doctrinal impact rise, the autonomy ceiling falls — Tier 4 operates at Level 4 · Augmented Autonomy.
The five-level AI value maturity scale.
Every Sanctum feature is assessed against this scale.
Software only records what humans already decided. Every step — noticing, deciding, acting, confirming — is done by a person.
Human role: Does everythingAI surfaces information and suggestions, but a person must still carry out every action by hand.
Human role: Executes every actionAI triages and prepares drafts, but nothing happens until a human approves each individual action. Used for soul shepherding and crisis care.
Human role: Approves every actionAI executes routine work on its own, but anything leaving the building — a message, a payment, a dispatch — waits for a human seal.
Human role: Approves external dispatchAI senses, decides, acts, and audits itself on deterministic rules. People supervise by exception: they only appear when the system halts and asks.
Human role: Supervises by exceptionChoose a domain for its assessment & walkthrough.
Each domain opens a worked example with a step-by-step maturity commentary.
Demonstration content — all scenarios are local samples. Remaining tiers will follow one tier at a time, each with its own autonomy ceiling.