Decision Model

The 7th Layer: Decision & Governance

virtauto models decision-making itself as a dedicated architectural layer — between enterprise planning, AI proposals and operational execution. This layer does not replace ERP, MES or domain systems. It orchestrates, validates, governs and records decisions across them.

virtauto is a dedicated decision layer that orchestrates and governs actions across agents and enterprise systems.

7th Layer Decision-first architecture Governance by evidence Transformation beyond technology

Why this layer exists Decision model Decision graph backbone Core functions Decision flow Live implementation In practice Operating principles

Why this layer exists

Digital transformation does not fail only because systems are fragmented. It also fails because decision-making remains structurally unresolved. ERP systems define constraints. MES systems expose operational reality. AI generates insights and proposals. But without an explicit layer for governed decisions, responsibility gets blurred, trade-offs stay implicit and execution becomes inconsistent.

ERP / Enterprise Planning

Defines constraints, costs, orders, resources, capacities and policy boundaries.

Strong on planning. Weak on situational orchestration.

MES / Operations

Reflects real execution conditions, disturbances, quality signals and plant reality.

Strong on operational truth. Weak on cross-domain governance.

AI / Agent Systems

Generate proposals, forecasts, optimizations and alternative action paths.

Strong on analysis. Weak on accountable decision ownership.

The missing link: who turns constraints, signals and proposals into a governed decision — and who records why this decision was acceptable?

The Decision Model

virtauto treats decision-making not as a hidden side effect inside applications, but as an explicit model with inputs, governance, output and evidence.

1. Inputs

enterprise constraints
operational signals
AI proposals
system state and context

→

2. Decision Layer

orchestrates agents
evaluates options
applies governance rules
selects an allowed action path

→

3. Outputs

validated decision
execution intent
decision trace
audit-grade evidence

Decision = architectural object

In virtauto, a decision is no longer a hidden consequence inside ERP logic, a prompt output, or a workflow side effect. It becomes an explicit system object with identity, context, validation and evidence.

Governance = runtime capability

Governance is not only documentation. It becomes a runtime capability that decides whether a proposal is allowed, blocked, escalated or routed into review.

Decision Graph as the governing backbone

Context alone is not enough. The system also needs an explicit graph that governs what can happen next, who is allowed to decide, which constraints apply and where escalation is required.

Decision Nodes

Represent decision objects with identity, state, context and intended outcome.

State Nodes

Represent the current condition of systems, processes, resources or governance status.

Constraint Edges

Define what is allowed, forbidden, conditional or rate-limited across the graph.

Authority Edges

Define who is allowed to decide at a given point — agent, human or governance gate.

Escalation Paths

Route uncertain, conflicting or blocked decisions into review instead of silent execution.

Outcomes

Every decision path resolves into approved, blocked or escalated.

The Decision Graph is therefore not a visualization gimmick. It is the governing backbone that connects decision objects, authority, constraints, escalation and evidence into one reviewable runtime structure.

Context Graphs vs. Decision Graphs

Context Graphs	Decision Graphs
describe reality	govern what is allowed to happen next
connect facts, entities and relationships	connect decision objects, authority, constraints and outcomes
support understanding and navigation	support decision control and escalation
can improve context quality	can improve decision integrity
inform agents and humans	bound agents and humans
do not by themselves authorize action	determine whether action is approved, blocked or escalated

Key distinction: context may explain reality, but it does not yet govern action. The Decision Graph is the structure that converts context into bounded, reviewable and accountable decisions.

The four core functions of the Decision Layer

Orchestrate agents

Coordinate specialized agents and information sources instead of letting them act independently.

Validate decisions

Check proposals against enterprise constraints, governance rules and operational boundaries.

Enforce constraints

Ensure that decisions remain inside allowed business, operational and compliance limits.

Record evidence

Create decision traces, validation artifacts and audit records for reviewable execution.

Not just more automation — better decision design

Traditional digital transformation	Decision-layer transformation
adds systems, dashboards and automation tools	adds an explicit decision architecture between systems
focuses on process digitization	focuses on governed action selection
assumes better data automatically yields better decisions	models how decisions are actually made, validated and owned
treats governance as ex-post documentation	treats governance as part of runtime decision execution
optimizes separate functions	coordinates trade-offs across the enterprise

Decision flow in virtauto

Conceptual flow

Signals, constraints and proposals enter the system
GEORGE creates a decision intent
GUARDIAN validates the decision
Status aggregates the system state
Self-audit verifies governance completeness
The website renders the visible governance surface

Runtime governance loop

Decision → Validation → Status → Audit → Reviewable system state

This closes the loop between architecture, automation and accountability.

Approved

Decision is allowed to proceed because constraints and authority checks passed.

Blocked

Decision is stopped because it violates policy, authority, state or governance rules.

Escalated

Decision is routed to human or higher-level governance because uncertainty remains unresolved.

How the model is implemented in the current repository

The current virtauto repository already contains the first operational proof objects for the 7th Layer.

Decision status/latest_decision.json

Current decision output generated by GEORGE.

Decision Trace governance/decision_traces/GEORGE-example.json

Evidence that the decision exists as a traceable object.

Validation governance/validation/GUARDIAN-example.json

Governance proof that the decision has been checked.

Status Surface status/system_health.json
status/dashboard_summary.json

Aggregated operational view of the governance state.

Audit audit/self_audit_report.json

Self-verification that the governance chain is present and intact.

Live Surface /status/

The current website renders governance artifacts as visible operational status.

The Decision Graph in practice

The website already contains the main building blocks of a governed decision system. Together, these pages form the practical surface of the Decision Graph.

Decision Contracts

Define what can be decided and which fields, classes and constraints belong to a decision object.

decisioncontracts.html

Authority Graph

Define who is allowed to decide, where governance boundaries are placed and where escalation paths begin.

authoritygraph.html

Decision Trace

Make the decision path visible over time: intent, validation, state transition and evidence.

decision-trace.html

Governance Status

Expose the currently visible governance artifacts of the system as a live operational surface.

/status/

In other words: Decision Model explains the architecture, Decision Contracts define decision objects, Authority Graph defines permission structure, Decision Trace records history, and Status exposes runtime visibility.

What makes this the 7th Layer

Beyond data ERP, MES and AI already produce data, context and proposals. The missing problem is not data access alone — it is governed action selection.

Beyond orchestration Pure orchestration is not enough. The layer must also decide what is allowed, what is blocked and what must be escalated.

Beyond reporting Evidence is not only historical. It becomes a runtime condition for trusted industrial action.

Operating principles of the Decision Layer

No change without governance evidence

Every meaningful action path must leave behind decision, validation and audit artifacts.

No decision without explicit ownership

Decisions must be attributable to a decision actor, governance step and reviewable context.

No autonomy without boundaries

Agents may propose and orchestrate, but only inside constraints that can be validated and enforced.

No uncertainty without escalation

If confidence, authority or policy alignment are insufficient, the system must escalate rather than silently act.

No transformation without cultural reset

Digital transformation is also a redesign of who decides, how decisions are governed and how accountability is preserved.

No architecture without observable state

The system must be able to expose its own current condition, recent decision and governance chain transparently.

Current maturity

virtauto is currently a Decision Governance Proof of Concept. The existing implementation already demonstrates the structural core of the 7th Layer: decision, validation, status, audit and repository-governed evidence. The next phases extend this toward richer agent orchestration, knowledge integration, enterprise connectors and industrial execution paths.

In one sentence

virtauto is not just another AI layer. It is the missing decision and governance layer that connects enterprise intent, operational reality and agentic action.