The Platform Model
The One sits between your source systems and everything that consumes data from them. Sources feed in, the platform resolves and governs, and three surfaces serve the result.
Consumers
- Applications
- AI Agents
- People
The Platform
Governance
Applied on every request
- Isolation
- Permissions
- Policy
- Masking
Evidence
- Audit
Surfaces
- REST API
- Governed Agent Endpoint
- Console
- In-app Assistant
- Event Stream
Resolution
- Standardize
- Validate
- Match
- Merge
- Relate
Record
- Golden Records
- Lineage
- History
- Relationships
- Reference Data
Source Systems
- CRM Platform
- ERP System
- Core Operational Database
- Data Warehouse
- Data Lake
- SaaS Apps
- HR System
- Core Banking
- Policy Administration
- Claims
What it does
Source systems each hold their own version of the same customer, supplier, or patient. The platform resolves those into one governed golden record per real-world thing, and serves it over an API, through a governed endpoint for agents, in a console, and as an outbound event stream when something changes.
The word doing the work is governed. The same access control applies on every one of those surfaces, evaluated per request against the caller's own identity. An agent cannot see what the person operating it cannot see.
The pipeline
Data moves through a fixed sequence:
| Stage | Does |
|---|---|
| Ingest | Records arrive from a source system, each linked by a crosswalk |
| Standardize | Values are transformed into canonical form |
| Validate | Constraints are checked; violations recorded with a severity |
| Match | Records that describe the same thing are identified |
| Merge | Matched records consolidate; survivorship picks winning values |
| Serve | Golden records are read, with access control and masking applied |
Each stage depends on the one before. Matching unstandardized values produces false negatives; merging unmatched records does nothing.
What is governed
Four things are enforced on every request, independently:
- Row-level isolation, in the database itself.
- Role-based access control, at the request boundary.
- Attribute-based policy, on each record as it is loaded.
- Field masking, when the response is generated.
Because masking happens on response generation rather than per endpoint, there is no surface that returns an unmasked value to a caller who lacks permission — including the AI layer.
What is remembered
- Lineage: every value traces to the source system that supplied it.
- History: every record keeps its full version history.
- Audit: every change records who made it, appended immutably.
- Reversibility: merges can be undone, restoring what was absorbed.