BlueHourThe eight-layer Operating Architecture
Select any layer to examine its purpose, inputs, outputs, dependencies, control boundary, IP, CIM functions and MOMs. Then switch to failure propagation and see where the architecture stops a failure.
Layers are shown bottom to top. Select a layer to inspect it.
Layer responsibilities are the established eight-layer spine. Inputs, outputs, boundaries and the IP, CIM and MOM associations are proposed architectural mappings pending BlueHour approval.
All eight layers at a glance
The same information as the explorer, as a table.
| # | Layer | Inputs | Outputs | Control boundary | Depends on |
|---|---|---|---|---|---|
| 1 | Data / Source of Truth Trusted information, provenance, data integrity | System-of-record transactions; External and partner data; Event streams from workflows and agents | Authoritative records with provenance; Source-authority registry (which system is authoritative for which fact); Integrity exceptions | Write authority per record class. Only the registered authoritative source may assert a fact; every other copy is a reference. | None |
| 2 | Integration / Workflow Applications, APIs, dependencies, workflow execution | Authoritative records; Business events; Approved workflow designs | Executed workflow steps; API calls and integration events; Dependency map updates | Integration contracts and the dependency map. A workflow may call only the systems its contract names. | 1 |
| 3 | Agents / Skills AI agents, specialized capabilities, bounded autonomy | Tasks and work allocation; Delegated authority grants; Authoritative data; Model access through orchestration | Agent actions and proposals; Action receipts (what was done, by whom, under what authority); Escalations | Bounded autonomy. An agent acts only within its registered scope and limits; anything beyond them escalates to a person. | 1, 2, 4, 5 |
| 4 | Model Orchestration Model selection, coordination, execution policy | Requests from agents and workflows; Execution policy; Cost and performance measurements | Model selection and routing; Usage and cost records; Policy exceptions | Approved model list and execution policy. Switching providers must not require rebuilding what depends on them. | 1, 5 |
| 5 | Security / Governance Authority, policy enforcement, security, accountability | Human and machine identities; Policies and delegated authority; Threat and control signals from cybersecurity | Allow, deny and escalate decisions; Approvals with named owners; Audit and accountability records | Policy enforcement points at run time, for people and agents. This layer is where failure propagation is meant to stop. | 1 |
| 6 | CX / EX Value Flows Customer and employee experience, value creation | Workflow outcomes; Agent and human work; Customer and employee signal | Customer and employee experiences; Value delivered; Recourse and explanation requests | Customer-facing authority. What an agent may promise or commit to a customer, and how a customer reaches a person. | 2, 3, 5 |
| 7 | Measurement / BigBoard Economics, performance, complexity, enterprise outcomes | Telemetry and costs from all layers; Outcome measurements; Complexity signals | Economic and operating measures; Conversion Yield and BlueHour Units; Complexity position against the Operating Complexity Ceiling | Measurement integrity. A measure carries its source and confidence; modeled figures are never presented as realized results. | 1, 2, 3, 4, 5, 6 |
| 8 | Operating Model Design & Governance Enterprise architecture, decision rights, lifecycle control | Strategy and economic objectives; Measurement from BigBoard; Risk and complexity position; Stress-test results | MOM designs and activation decisions; Decision rights; Complexity thresholds and operating constraints | Lifecycle control. No MOM expands without evidence; any MOM can be held or stopped. | 7 |
How failure propagation is calculated
A failure in a layer reaches every layer that depends on it, directly or through others. When Security / Governance boundaries are enforced, a failure that originates outside layer 5 does not pass into a layer whose actions execute through layer 5's enforcement points (Agents / Skills and CX / EX Value Flows): it is held at the boundary. Measurement still observes the failure, and design still responds to it. This is a structural illustration of where boundaries sit, not a prediction for any specific enterprise.