Unclassified

20 · Projection & Context Assembly

One business reality → different participant projections. Projection Engine, contracts, Active Context Bundle — how context assembly and relevance work.

One Business Reality → Different Participant Projections

                 BUSINESS CONTEXT SPINE
                      │
          ┌───────────┴───────────┐
          ▼                       ▼
    HUMAN PROJECTION          AI PROJECTION
          │                       │
    Human Workbench          AI Workbench
          │                       │
  Personal Contextual      AI Continuity
      Memory (serves BC)     Memory (serves BC)
          │                       │
          └───────────┬───────────┘
                      ▼
                Shared Work
          (same reality, scoped view)

Humans and AI share the same Business Context Spine. They do not share the same projection. Twin never needs to re-explain the business because it hydrates the same Work + Spine slice as the human.

Active Context Bundle

The Workbench and the AI Workbench hydrate from the same Active Context Bundle:

  • Business Context slice — the relevant entities, relationships, and state for this Work and participant
  • Work — objective, state, participants, and history
  • Evidence — what has already been proven to have happened
  • Authority — what this participant may do in this context
  • Continuity — personal or AI working memory relevant to this Work

How Assembly Works

Human opens Work or Twin receives Work — both request the Active Context Bundle for that Work and identity.


The engine selects the governed slice of the Spine for this participant, tenant, and Work context — per projection contracts and authority.


Additional contextual context is assembled (related Work, prior decisions, related entities) and ranked by relevance — not dumped as raw payloads.


Both workbenches hydrate from the resulting bundle. Handoffs carry Work state and evidence, not transcripts.

Projection Contracts

Projection is governed. Contracts define:

  • Which entities and fields are visible to which participant in which Work type
  • Which memory (Personal Contextual vs AI Continuity) contributes to the bundle and how
  • How provenance and evidence are surfaced vs kept internal
  • When projections invalidate and workbenches must rehydrate (after verify → reconcile → Spine)

Where to Go Next

<CardGroup cols={2}> 
Why two working memories serve one organizational memory — ownership, privacy, durability.


What the human actually experiences of this projection.


How the AI Workbench uses the same bundle.
</CardGroup>