Unclassified

Twin — AI Workspace

The user's AI work partner — one Twin per user, shared task assignment layer, operating within bounded context.

Twin is the user's AI work partner. Each user has one primary Twin that operates as their work coordinator, digital execution coordinator, and orchestration boundary.

Twin is not a panel, layer, or sidebar destination. Twin is a capability of the Workbench that surfaces contextually via the Four Buttons.

Twin Operating Model

UNDERSTAND
  ↓
CONTEXTUALIZE
  ↓
TRIAGE
  ↓
ALLOCATE
  ↓
PLAN / PROPOSE
  ↓
DELEGATE
  ↓
EXECUTE
  ↓
VERIFY
  ↓
REPORT
  ↓
ADAPT

Internal reasoning follows OODA:

OBSERVE → ORIENT → DECIDE → ACT

One Twin Per User

Each user has exactly one Twin. The Twin is:

Scoped to the user's role domain
Aware of the user's work context
Connected to the user's organizational memory
Bounded by the user's permissions and authority

The Twin is NOT:

Unrestricted enterprise authority
A second source of truth
A replacement for governance
A standalone product surface

Shared Task Assignment Layer

Twin and humans share a task assignment layer. This layer:

Receives work from multiple sources (human assignment, system events, agent proposals)
Routes work based on capability, availability, and policy
Tracks assignment state across sessions
Supports handoff between human and AI

WORK SOURCES
    ├── Human assigns to self
    ├── Human assigns to Twin
    ├── System event generates work
    ├── Agent proposes work
    └── Workflow generates work
          ↓
SHARED TASK ASSIGNMENT LAYER
          ↓
CAPABILITY-BASED ROUTING
          ↓
EXECUTION

Context Assembly + AI Sandbox

Twin operates within bounded context assembled from Spine Fabric:

CANONICAL SPINE
      ↓
CONTEXT ASSEMBLY
      ├── Permission (what Twin may see)
      ├── Purpose (why Twin needs this context)
      ├── Relevance (what is related to current work)
      └── Freshness (how current is this context)
      ↓
AI OPERATIONAL SANDBOX
      ↓
TWIN REASONING

Rules:

Canonical Spine is never directly exposed to AI
Context assembly is bounded by permission, purpose, relevance, freshness
AI operates only within the operational sandbox
Twin consumes projected context, not raw Spine

Twin Modes

Passive Mode — Always Observe, Stay Lightweight

Twin continuously observes without continuously performing expensive reasoning:

Pipeline events
Spine changes
Current Workbench surface
Current entity
User actions
OODA state
Approvals and outcomes

Only material signals produce lightweight surfacing:

"Renewal risk detected."
"Meeting starts in 10 minutes."
"Three approvals pending."

Active Mode — Ask your Twin

When the user clicks Ask your Twin, Twin escalates from observation to grounded reasoning:

Ask your Twin
    ↓
ORIENT
    ├── Current Workbench surface
    ├── Current entity and related entities
    ├── Spine projection, timeline, relationships, provenance
    ├── Pipeline state, freshness, ingestion status
    ├── OODA state
    ├── Evidence and connected-system context
    └── Organizational memory
    ↓
THINK
    ↓
Evidence-backed recommendation
    ↓
User decides

Four Buttons (OODA Grammar)

The Four Buttons are the explicit human-to-platform control grammar:

Button	OODA Phase	User Intent	Platform Behavior
Store in Spine	Observe	Capture durable observation	Governed write through Pipeline
Ask your Twin	Orient	Request understanding	Twin reads bounded context, reasons
Assign your Twin	Decide	Delegate bounded task	Twin creates draft/proposal from context
Approve Twin's Action	Act	Authorize proposal	Governance validates, Pipeline executes

Grounded Twin Inputs

Every Twin response is grounded in bounded context:

Current Surface
    + Current Entity
    + Pipeline State
    + OODA State
    + Adaptive Spine
    + Evidence
    + Connected Systems
    + Organizational Memory
    ↓
Twin Reasoning

Twin must distinguish facts, evidence, freshness, uncertainty, and recommendations. It never invents missing operational truth.

Twin Lifecycle

INITIALIZE → OBSERVE → ORIENT → PROPOSE → WAIT → EXECUTE → VERIFY → REPORT
     ↓          ↓         ↓         ↓        ↓        ↓         ↓        ↓
  Load      Passive   Context   Draft    Human    Governed  Check    Update
  context   mode      assembly  proposal decides  execution outcome  Spine

Workbench Canonical Model — the surface Twin operates on
Agent Fabric — the agent fleet Twin orchestrates
Governance — the policy boundary Twin respects
Work Graph — the work Twin allocates