Enterprise
Organisations, business units, regions, environments, compliance domains and ownership.
Continuously derived enterprise model
AI Twin™See the AI enterprise as it is now.
The AI Twin is a continuously derived model of every governed AI system, agent, model, tool, dependency, policy, trust relationship, risk and runtime state.
It is generated from operational evidence—not maintained as a separate inventory—so leaders can work from a current model of the enterprise rather than a diagram of what the enterprise used to be.
DerivedEvidence-backedContinuously generatedRead-only by default
Most organisations cannot answer basic questions about their AI estate.
AI estates change faster than manual governance processes can document them. A spreadsheet may list approved systems. It rarely shows the agents now running, the tools they can reach, the policies currently in force or what changed since the last review.
01What AI systems exist?
02What agents are running?
03Who owns them?
04Which models do they use?
05What tools can they access?
06Which policies govern them?
07What has changed?
08Where are the risks?
Static documentation records intent. Governance needs operational reality.
Architecture diagrams, registers and control spreadsheets remain useful records. They fail when treated as the live model of a changing autonomous estate. The AI Twin closes that gap by deriving current state from the governed system itself.
Derived—not manually maintained.
A conventional inventory is another copy of the truth. Every copy needs an owner, a review cycle and reconciliation. The AI Twin is assembled from the evidence and governed records the operating platform already produces.
01
Collect
Approved provisioning records, governance decisions, runtime events and monitoring evidence.
02
Normalise
Systems, agents, models, tools, identities and environments resolve into a common schema.
03
Relate
Dependencies, permissions, trust boundaries, policy scope and reachability become graph relationships.
04
Derive
The current enterprise model is assembled from evidence rather than copied into a separate inventory.
05
Compare
The live model is evaluated against the governed baseline to identify meaningful change.
06
Project
Executive Workspaces present the same model through role-specific operational views.
Runtime evidence and governed records→Continuously derived AI Twin→Current operational and executive views
A living enterprise cannot be governed from a static picture.
ConcernStatic documentationContinuously derived model
CreationWritten or drawn by a personGenerated from operational evidence
FreshnessCurrent at the last reviewRe-derived as governed state changes
CoverageUsually systems and ownersSystems, agents, tools, policy, trust, risk and reachability
Change detectionDepends on someone reporting itCompared with the approved baseline
AuditabilityExplains what was documentedLinks current state to source evidence
Executive useSeparate reports for each functionRole-specific views of one shared model
The result is better operational awareness, stronger evidence and faster governance decisions—not because documentation disappears, but because it is no longer asked to behave like a runtime system.
Not a list of assets. A model of relationships and consequence.
The Twin connects what the enterprise owns, what each autonomous system can reach, who may authorise it, which policies apply and where operational risk can propagate.
Continuously derivedAI TwinOne enterprise model
Assets
AI systems, models, tools, MCP servers, APIs, data services and protected resources.
Agents
Deployed agents, specialist roles, autonomy levels, assigned capabilities and operating context.
Dependencies
The systems, tools, services and workflows each component relies on or can affect.
Trust relationships
Identity providers, operators, approvers, boundaries and delegated authority.
Policies
Applicable controls, approval requirements, protected actions and evidence obligations.
Risks
Open incidents, risk zones, privileged capability, policy gaps and exposed critical assets.
Reachability
Which assets and outcomes are reachable from an agent’s current permissions and tool access.
Runtime state
Current governance state, active environments, recent decisions and observed change.
Reachability view
Agentcan callToolcan accessSystemcan affectProtected outcome
Visibility follows the path of possible action, not just the inventory of components.
Every operational view traces back to evidence.
The Twin is not an executive illustration detached from the control plane. Its state is derived from provisioning, governance, approval, execution and monitoring records produced throughout the operating lifecycle.
Observed state→
Source identity→
Policy context→
Governance verdict→
Approval state→
Execution outcome→
Retained evidence
Control boundary
The AI Twin is read-only by default. Any write capability or source-system mutation remains a separate governed integration surface requiring explicit policy, authority, testing and evidence.
Know when reality moves away from the approved baseline.
Drift is operationally important because a small change in policy, permission, tool access, autonomy or trust can alter what an autonomous system can reach. Current monitoring surfaces the governed divergences below; additional estate signals are scoped to the connected evidence sources.
New AI system
A system appears outside the approved estate or expected provisioning path.
New MCP server
A server is added outside the governed baseline.
New tool
A new capability changes what an agent or AI system can reach.
Permission change
An existing tool is elevated to privileged capability.
Removed control
A protected asset, critical-system control, risk zone or governance department is removed.
Disabled policy
A policy required by the governed baseline is no longer active.
Unexpected autonomy
The operating autonomy level rises above the approved baseline.
Trust-boundary violation
A new system operates in an environment outside its declared trust boundary.
Change detected→Impact understood→Evidence attached→Authority notified→Human decision
Detection does not silently rewrite production. The platform identifies the change, relates it to the governed baseline and routes the evidence to the appropriate operational authority.
One Twin. Different questions. Consistent answers.
Each stakeholder sees a scoped projection of the same enterprise model. Security, architecture, compliance and operations no longer begin from separate inventories.
Business overview
See where AI is operating, who owns it and which decisions or risks require executive attention.
Architecture
Understand systems, agents, models, dependencies, environments and the impact of change.
Security posture
Identify privileged capability, trust-boundary exposure, permission drift and reachable assets.
Evidence
Connect obligations, policies, decisions, approvals and retained evidence to the current estate.
Live estate
Track active systems, monitoring state, open drift and operational intervention queues.
Dependencies
See tool access, integration paths, upstream dependencies and likely blast radius before change.
Five responsibilities. One enterprise operating platform.
AI Twin
What exists?Continuously derives the enterprise, its relationships and current state.
Runtime Governance
What is allowed?Evaluates proposed behaviour against policy, authority and reachability.
Guardian OS
How is the enterprise operated?Coordinates provisioning, governance, evidence, monitoring and executive operations.
Industry Intelligence Packs
How is governance adapted to the sector?Contribute domain policies, mappings, evidence requirements and workflows.
Executive Workspaces
How do leaders interact with the platform?Present the same governed state through role-specific decisions and briefings.
The AI Twin is generated as part of Guardian OS enterprise provisioning and remains the shared model used by governance, monitoring and Executive Workspaces. See the complete Guardian OS operating model →
Current state. Connected evidence. Accountable decisions.
See the AI enterprise you actually operate.
Replace periodic reconstruction with a continuously derived model built into the operating platform.