Automation & Organizational Memory
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.
Kristal is designed to capitalize work done by both people and AI.
Without persistent structured memory
AI researches
↓
answer is produced
↓
next task starts again
With Kristal
AI or human work
↓
assertions + evidence + applicability + valuations
↓
Kristal State
↓
validation / corrections / new observations
↓
updated state
↓
next agent inherits previous work
Progressive reduction of friction
The first occurrence of an unusual case may require a human decision. If the organization later decides that the same decision should apply whenever explicit conditions X/Y/Z hold, that judgment can be represented as an organizational rule.
Future cases matching X/Y/Z may then become automatically actionable.
This is not “the AI silently learned what to do”. It is explicit institutional learning with provenance, scope and authority.
Knowledge / policy / state / decision / action
A healthy architecture keeps these separate:
KNOWLEDGE what is represented
POLICY how the organization wants to act
STATE what is currently observed
DECISION application of knowledge + policy to state
ACTION execution by an authorized owner
Kristal can preserve all five as traceable records without turning them into one undifferentiated workflow state.
The same principle applies to external constraints, observed state and organizational rules: they can contribute to one decision while retaining different authority and change boundaries. See Information-Layers-and-Operational-Boundaries.
Because Kristal semantics are portable across storage forms, this accumulated memory does not need to live in one monolithic database. Graph, relational, document and analytical projections may coexist when they preserve the canonical semantic boundaries. See Universal-Structure-and-Transposition.