Information Layers & Operational Boundaries
v9 compatibility note: This page documents the retained v6 portable-state semantics. V9 preserves the contract as a frozen substrate and may wrap/project it through Logical Artifacts without redefining it. See Version-Stack-v6-v7-v8-v9.
A useful Kristal often combines information with very different relationships to authority and change.
The most important distinction is not “structured vs unstructured”. It is what kind of thing the record represents and who is allowed to change its meaning.
1. Authoritative external constraints
Examples:
- laws and regulations;
- official government procedures;
- standards;
- manufacturer requirements;
- contractual constraints;
- regulated schedules or thresholds.
Kristal does not improve or rewrite these. It records:
- the authoritative source;
- the exact represented rule;
- its applicability;
- effective period;
- succession / supersession;
- recognition context.
Typical record_role: authoritative_constraint.
2. Neutral observed state
Examples:
- document received;
- payment posted;
- temperature measured;
- symptom observed;
- product label captured;
- inventory count recorded;
- workflow event occurred.
These records primarily capture what happened. They should not silently become policy or conclusions.
Typical record_role: observed_state.
3. Organizational interpretation and policy
Examples:
- escalation thresholds;
- preferred diagnostic sequence;
- internal definition of “ready”;
- review policy;
- risk tolerance;
- exception handling;
- operational prioritization.
These records reflect how an organization chooses to operate. They may be highly important without being universal facts.
Typical record_role: organizational_rule.
4. Derived knowledge and decisions
AI, rules or humans may derive conclusions from the three layers above.
Examples:
- this permit is applicable;
- this case is missing one required document;
- this automotive cause is strongly supported;
- this product formulation conflicts with a dietary requirement;
- this task can proceed automatically.
Typical roles include derived_state, decision, or action.
Why the separation matters
EXTERNAL CONSTRAINTS
+
OBSERVED STATE
+
ORGANIZATIONAL POLICY
↓
DERIVED DECISION / ACTIONABILITY
If these layers are collapsed, automation becomes dangerous because the system can no longer tell the difference between:
- what the law requires;
- what was observed;
- what the organization prefers;
- what a model inferred.
Do not redesign the process by accident
Kristal's goal is not to replace an organization's existing process with a new abstract workflow.
The goal is to make the existing process explicit enough to become action-ready:
- deterministic checks can run automatically;
- missing information can be detected automatically;
- routing can happen automatically;
- humans receive only the ambiguity, authority, exception or judgment that actually needs them.
Human expertise as a deliberate boundary
A healthy system asks humans for work where human expertise creates value:
- interpretation;
- ambiguity resolution;
- policy choice;
- novel cases;
- legal or organizational authority;
- negotiation;
- safety-sensitive judgment;
- exception approval.
Everything else can be progressively reduced to lower-friction machine work when policy explicitly allows it.
Institutional learning
When a human resolves a novel case, the decision can remain a one-off decision or be promoted — through an explicit governance step — into a reusable organizational rule.
novel case
↓
human decision
↓
traceable decision record
↓
optional policy generalization
↓
future matching cases may become deterministic
This is how Kristal improves over time without silently inventing new organizational policy.