Enterprise Decision & Control Runtime Audit

This page audits the runtime through the enterprise operating-model lens: identity, authorization, data contracts, controls, resilience, accountability, change management, and operating model. It shows what is already proven by the runtime and what must be hardened for production.

Enterprise operating-model test

The question is not whether an agent can detect duplicate invoices. The question is whether the platform can safely run finance decisions inside a global enterprise.

Identity	Who is this agent operationally?
Authorization	What can this agent or human actually do?
Data contracts	Which governed data products and SORs are used?
Controls	Which control, policy, evidence, and attestation apply?
Resilience	How does the process continue when AI or integration fails?
Accountability	Who approved, who executed, and what proves it?
Change management	How are agents certified, piloted, promoted, and retired?
Operating model	Which roles own execution, approval, governance, cyber, audit, and operations?

Identity

Who is this agent operationally?

Authorization

What can this agent or human actually do?

Data contracts

Which governed data products and SORs are used?

Controls

Which control, policy, evidence, and attestation apply?

Resilience

How does the process continue when AI or integration fails?

Accountability

Who approved, who executed, and what proves it?

Change management

How are agents certified, piloted, promoted, and retired?

Operating model

Which roles own execution, approval, governance, cyber, audit, and operations?

Enterprise Control Plane

The control plane is the business operating layer above the agent data plane. It makes authorities, policies, controls, owners, fallback paths, and replay visible.

Identity registry	Human and agent identities, roles, owners, risk classes
Policy registry	Business, control, and AI policies
Control registry	SOX, finance, operational, and cyber controls
Agent registry	Versions, approvals, certifications, and risk levels
Runtime governance	Who did what, why, when, with which evidence
Resilience center	Fallback, business continuity, DR, and recovery status

Identity registry

Human and agent identities, roles, owners, risk classes

Policy registry

Business, control, and AI policies

Control registry

SOX, finance, operational, and cyber controls

Agent registry

Versions, approvals, certifications, and risk levels

Runtime governance

Who did what, why, when, with which evidence

Resilience center

Fallback, business continuity, DR, and recovery status

Live Enterprise Decision & Control Runtime Audit

Live runtime

Overall readiness

60%

Agents

Skills

Replayable

Layer	Score	Purpose	Runtime proof	Remaining gap
Enterprise Identity	50%	Every agent and skill has an operational identity, owner, risk class, and authority boundary.	1 agent identities linked to decision ledger; 7 skill owners resolved; Skill owner and risk class exposed in Skills + Policy Audit	No critical gap captured
Enterprise Authorization	12%	Agent action is gated by identity, authority, policy, evidence, and replay.	1 skills authorized by policy/evidence/context; 7 skills forced to review or block state; Authority and execution mode are explicit per skill	Add live Entra delegated-authority token introspection before production writes.
Enterprise Data Contracts	100%	Agents consume governed context through source lineage and data contracts, not uncontrolled SOR polling.	8/8 audited cases carry source lineage; 11/26 context queries marked used in decisions; Context completeness evaluates required entities, relationships, policy, evidence, replay, freshness, and learning	Bind each source ref to formal UDP/Databricks/OneData data-product contract IDs.
Enterprise Control Framework	12%	Controls connect policy, execution, evidence, replay, and attestation.	1/8 cases have policy linkage; 1/8 cases have evidence linkage; 1/8 cases are replayable	Add SOX/control IDs, control owner, attestation cadence, and testing schedule to each policy binding.
Operational Resilience	94%	Failure paths degrade to policy fallback and human workflow instead of stopping finance operations.	8/8 skills expose human override boundary; 7 skill decisions route to review/block instead of unsafe execute; System has live Enterprise Resilience Center and control-plane resilience APIs	Attach explicit RTO/RPO and failure playbook IDs per skill before production rollout.
Enterprise Accountability	59%	Every action identifies responsible actors, accountable owners, policy, evidence, version, and replay.	1 policy owners resolved; 8 human override paths visible; 1 replay records linked	Persist full RACI per policy/control and bind approver identity at action execution time.
Enterprise Change Management	56%	Agents and skills move through versioning, certification, pilot, parallel run, and production gates.	8/8 skill versions captured; 1/8 policy versions captured; Manifest certification service exists for lifecycle gates	Expose certification record, pilot state, parallel-run status, and retirement date per skill.
Enterprise Operating Model	94%	Analyst, supervisor, tower lead, process owner, controller, audit, AI governance, cyber, and platform teams have explicit responsibilities.	7 business skill owners; 1 policy governance owners; 8 human accountability boundaries	Add a formal operating-model registry with role-specific responsibilities and escalation SLAs.

Enterprise Control Plane Model

Identity Registry	{"agents": ["iaf-invoice-decision-agent"], "skill_owners": ["Close Controller", "Credit Risk Owner", "FP&A Lead", "P2P Control Owner", "R2R Controller", "Trading Risk", "Treasury Operations"]}
Policy Registry	{"policy_owners": ["Finance Governance"], "policies_linked": 1}
Control Registry	{"policy_evidence_replay_controls": 1, "attestation_status": "partial_until_control_ids_bound"}
Agent Registry	{"skills": ["SKILL-P2P-DUP-HOLD", "SKILL-CREDIT-EXPOSURE-ASSESS", "SKILL-CLOSE-MAINT-ACCRUAL", "SKILL-TRADING-EXPOSURE-ASSESS", "SKILL-TRSY-PAYMENT-TIMING", "SKILL-FPA-VARIANCE-EXPLAIN", "SKILL-R2R-JOURNAL-VALIDATE", "SKILL-P2P-DUP-HOLD"]}
Runtime Governance	{"context_usage": 11, "human_override_paths": 8}
Resilience Center	{"fallback_paths": 7, "manual_override_paths": 8}

Workshop position

The platform is an Enterprise Decision & Control Runtime when identity, authorization, controls, evidence, replay, resilience, lifecycle, and operating model are visible and enforced.

The audit intentionally surfaces gaps. That is more credible than claiming every enterprise operating-model capability is complete.

Credible enterprise answer

The runtime is moving from Decision Runtime to Enterprise Decision & Control Runtime by making identity, authorization, controls, accountability, resilience, certification, and operating model visible and enforceable.