1
Intake
Connectors receive events from Workday, Greenhouse, Lever, DocuSign, Slack, and Gmail via API polling and webhooks.
2
Load
Raw events are ingested into Spine Fabric with source timestamps and metadata.
3
Normalize
Events are mapped to a common schema. A candidate update from Greenhouse and a calendar event from Gmail become comparable entities.
4
Schema Mapping
Entities are linked across sources. The same person exists as Candidate in Greenhouse and Employee in Workday once hired.
5
Context Assembly
Related entities, events, and evidence are gathered on demand into a context bundle for the requested view.
6
Projection
The context bundle is rendered into a domain view: Open Roles, Interview Loops, People Directory, or another projected surface.
7
OODA
Twin and agents observe the updated state, orient against policy, and propose actions.
8
Governance
Proposed actions are checked against capabilities, authorization, and audit rules before execution.
9
Execution
Approved actions write back to source systems through connectors.
10
State Update
Source system responses generate new events, which re-enter intake and keep the loop continuous.