BOA — Boundary-Oriented Architecture

Separate Fact, Meaning, and Responsibility.

BOA is one of the four entry points of the VCDesign ecosystem. It is a construction method that separates Fact, Meaning, and Responsibility across boundaries by default-async. BOA can be used standalone.

VCDesign IDG RP VCD Library

Start here

BOA prevents silent meaning drift and responsibility leakage by making boundaries explicit and operable.

For boundary safety when governance is incomplete, a lightweight interlock can be deployed independently: → Interface Determinability Gate (IDG)

BOA is not a philosophy. It is a method you can apply immediately.

→ boa_core.yaml (raw) GitHub VCDesign 
Fact: [Immutable, Context-free]
Meaning: [Revisable, Context-bound] # AI operates here
Responsibility: [Human-owned, Explicit]

What BOA is

  • A boundary-first construction method Declare boundaries where meaning changes and responsibility cannot pass silently.
  • Default-async by design Assume delays, retries, and partial failure as the normal operating condition.
  • Sharper separation of concerns Separate immutable facts from contextual interpretations and human-owned responsibility.

What BOA is not

  • Not a reference architecture BOA does not mandate topology, products, or layers.
  • Not an autonomy license for AI AI may suggest hypotheses, but never owns decisions or responsibility.
  • Not a way to eliminate boundaries Boundaries are inevitable. BOA makes them visible and operable.

The three artifacts

BOA treats these as different artifacts with different owners. Mixing them is the most common failure mode.

  1. Fact Immutable observations. No meaning. No overwrites.
  2. Meaning Contextual interpretations. Revisable. Must be labeled and scoped.
  3. Responsibility Human-owned decisions. Explicit owner. Auditable rationale.

AI positioning

Deferral is a valid outcome

In BOA, the inability to produce bounded Meaning is not treated as an error.

Forcing an interpretation when uncertainty remains risks contaminating Fact or silently promoting Meaning into Responsibility.

Therefore, BOA explicitly allows systems to defer judgment. Not deciding is a normal and correct state.

This deferral represents a structured handoff from automation back to human responsibility — not a failure of the system.

BOA is AI-aware, not AI-delegated.

AI may generate candidate meanings, compare interpretations, and expose trade-offs. AI must never overwrite facts, make final decisions, or silently promote meaning.

Promotion requires explicit human ownership. → Resolution Protocol (RP)

For boundary-level determinability checks, see Interface Determinability Gate (IDG).

When to use BOA

  • You have multiple worlds OT/IT, teams, vendors, systems, or time horizons with different meanings.
  • Meaning drift is expensive Small interpretation changes cause large operational risk.
  • You need AI assistance without responsibility leakage LLMs can help, but decisions must remain owned.

Frequently misunderstood

  • Do I need VCDesign to use BOA? No. BOA is designed to be applied standalone.
  • Is BOA only for factories? No. It is domain-agnostic. “Worlds” can be teams, products, systems, or time.
  • Does BOA replace good engineering? No. It makes meaning and responsibility explicit so engineering can stay honest.

BOA can be applied on its own. VCDesign defines the broader constraints BOA operates within.