Decision formation and action
Decision Workspace
Fast explanation of the decision workspace contract: signal, context, evidence, policy, agent formation, human judgment, action, replay, and learning.
Latest Runtime Capability Update
This page is evaluated against the current runtime baseline: governed chat, WIIF assistant portability, Token Economy controls, fail-closed policy enforcement, deterministic replay, signed evidence, runtime adoption audit, skill learning promotion, autonomy demotion enforcement, case-state propagation, and observability readiness.
Evidence-backed answersPolicy before actionReplay-ready decisionsValue attributionAssistant portabilityRuntime observability
Compact-SOR proof anchor for BP validation pages
INV-LIVE-DUP-1778594510 · OceanFloor Subsea Services Ltd · £2.8M
Policy FIN-P2P-DUP-04 · Evidence EP-P2P-OCEANFLOOR-LIVE · PO PO-100421 · $2.3M protected.
Latest Runtime Capability Update: governed chat, WIIF assistant portability, Token Economy controls, fail-closed policy, deterministic replay, signed evidence, runtime adoption audit, case-state propagation, and observability readiness.
Governed action controls
Operator ACT mode exposes only bounded, approval-aware actions. Approve records an in-authority human decision; Escalate routes the case to the supervisor or control owner when authority, policy, or evidence is insufficient.
Approve
Escalate
Request evidence
Create replay record
Decision crates
Every crate must show situation, impact, confidence, recommendation, alternatives, and status.
ImpactConfidenceOwnerStatus
Agentic Fabric trace
Shows which agents contributed and what remains deterministic or human-controlled.
SignalContextEvidencePolicyImpactExecution
Human accountability
Records who approved, rejected, modified, or escalated the recommendation.
AnalystSupervisorControllerCFO
Action execution
Actions remain governed: draft, approve, execute through API or workflow, then replay.
SAPAribaServiceNowEmail