Blueprint / 16 harnesses

Intelligence needs an operating system.

The blueprint separates enterprise multi-agent capability into sixteen accountable boundaries. Each has a mandate, an owner, interfaces, risks, and observable proof that it works.

Concern plane

Runtime plane

Infrastructure & Runtime

Where does it run?

Phase 0 02

Model & Inference

Which intelligence executes?

Phase 1 03

AI Gateway

How is model traffic controlled?

Phase 1 04

Experience & Interaction

How does a human participate?

Phase 1

Concern plane

Knowledge plane

Domain & Semantic

What does the domain mean?

Phase 2 06

Data Integration & Provenance

Where does enterprise truth come from?

Phase 2 07

Retrieval & Context Engineering

What information matters now?

Phase 1 08

Memory & State

What should the system remember?

Phase 2

Concern plane

Execution plane

Protocol & Interoperability

How do agents, tools and peers talk?

Phase 1 10

Orchestration & Durable Execution

Who acts, when and in what order?

Phase 1 11

Tool, Skill & Sandbox

What can the system actually do?

Phase 2 12

ML & Decision Intelligence

When is a classical model the better tool?

Phase 3

Concern plane

Trust & lifecycle plane

Security, Safety & Guardrails

Who or what is allowed to do it?

Phase 1 14

Governance, Oversight & AgentOps

What autonomy is allowed, and how does it evolve?

Phase 2 15

Observability & FinOps

What happened — and what did it cost?

Phase 1 16

Evaluation & Assurance

Was it good, safe and compliant?

Phase 1

THE BOUNDARY PRINCIPLE

Separate the jobs that fail differently, change at different speeds, or answer to different owners.

The harnesses are not products and the planes are not deployment layers. They are a vocabulary for deciding where a responsibility belongs—and where evidence must cross a boundary.

ILLUSTRATION / SWITCH A HARNESS OFF

Absence is easier to see than architecture.

Choose a boundary to remove from the reference task. This is a scripted illustration, not a live model simulation.

Step 14 / BOUNDARY REMOVED

A retrieved document contains an instruction. The agent follows it and exposes restricted data.

Security, Safety & Guardrails would screen retrieved content again inside the loop.

Inspect harness 13 ↗