Strategic Schema
If strategy is infrastructure, the schema is the blueprint. Every building has load-bearing walls, electrical systems, plumbing — distinct systems with distinct purposes that connect at defined points.
Generic AI infrastructure stores all of this as embeddings — high-dimensional vectors that capture semantic similarity but destroy structural meaning. You can ask "what's similar to this?" but not "what depends on this?" or "what breaks if this changes?"
Stratafy's schema encodes strategic architecture as typed entities with explicit relationships. Every entity has a type, a role in the structure, and defined connections to other entities. The schema is the blueprint that makes strategy queryable, traceable, and machine-readable.
Strategic Entity Types
Foundation Layer
Identity and purpose — the root of every decision tree
Mission
Why the organisation exists. The anchor point.
Vision
The future state. Ambitious enough to be directional, concrete enough to be falsifiable.
Values
What matters most when things get hard. Not platitudes — real choices.
Beliefs
Assertions about reality. Unlike values, beliefs can be wrong.
Principles
Where values and beliefs become operational.
Strategy Layer
Choices and commitments that bridge identity to execution
Strategies
Choices about where to play and how to win. Hierarchy: corporate → functional → sub-strategies.
Initiatives
Commitments within a strategy. Three types: strategic, tactical, operational.
Measurement Layer
How you know whether strategy is working
Objectives
Bounded targets. Linked to strategies and initiatives.
Metrics
Continuous measures you always track.
Intelligence Layer
What you're learning, what could go wrong, what's changing
Assumptions
What must be true. Tracked by confidence.
Risks
What could go wrong. Scored on likelihood × impact.
Decisions
What you chose and why. Type 1 vs Type 2.
Insights
What you're learning. Feedback from execution.
Signals
What's happening outside. External events filtered through strategic context.
Relational Architecture
Typed entities alone are not enough. The power of a schema comes from the relationships between entities — the edges that make the graph queryable and traceable. Stratafy encodes four types of relationships.
Hierarchical
Parent-child relationships that create traceability. Corporate strategy → functional strategy → initiative.
Contextual
Cross-cutting links that attach intelligence to the execution stack. Assumptions, risks, and insights link to the entities they inform.
Temporal
Time-based relationships that track how entities evolve. Version history, status transitions, and confidence changes over time.
Scored
Quantified relationships that enable prioritisation. Risk scores, assumption confidence levels, OKR progress, and metric thresholds.
Why This Matters
A typed, relational schema transforms what is possible with strategy. Three capabilities emerge that are impossible with flat documents or generic embeddings.
Every AI interaction is strategy-aware
Because entities have types and relationships, AI agents can retrieve not just semantically similar content but structurally relevant context. A question about an initiative surfaces its parent strategy, linked assumptions, and associated risks.
Strategy becomes queryable
You can ask structural questions: "Which strategies have no linked metrics?" "What assumptions have low confidence?" "Which initiatives are not connected to any objective?" These queries are impossible with documents.
Structure compounds over time
Every entity added, every relationship created, every score updated makes the schema more valuable. The system gets smarter with use — the opposite of documents, which decay from the moment they are written.
Technical Details
The strategic schema is implemented in PostgreSQL via Supabase, with pgvector for semantic search across entities. This gives you the best of both worlds: relational queries for structural traversal and vector similarity for semantic discovery.
The entire schema is accessible programmatically through 169+ MCP tools, enabling AI agents to read, write, and reason about strategy with full type safety and relationship awareness. Every tool operates on typed entities with validated inputs — not free-text fields that lose structure.