Skip to main content

AIPCH11 — Addressable

“Provides Well-Defined Output Ports”

What AIPCH11 is really asserting​

AIPCH11 is not asserting that:

“The AI Product exposes an API or interface.”

It is asserting that:

The AI Product exposes explicit, stable, and governed output ports that represent its capability, such that consumers interact with the product through these ports — independent of any user interface, application, or experience layer.

Addressability is about:

defining the product boundary and how it is consumed

The Essence (HDIP + AIPS Interpretation)​

An AI Product is addressable if and only if:

  1. It exposes well-defined output ports
  2. These ports represent the capability of the product
  3. Consumers interact with the product only through these ports

If the AI capability is only accessible through:

  • applications
  • dashboards
  • UIs
  • embedded workflows

then AIPCH11 is not met, even if APIs exist internally.

What Are Output Ports?​

Output ports are:

The formal access points through which the AI Product delivers its capability

These may include:

  • APIs (REST, gRPC)
  • event streams
  • batch interfaces
  • agent interfaces
  • query endpoints

👉 Important:

Ports are not UI elements.
Ports are product-level access mechanisms.

Positive Criteria — When AIPCH11 is met​

AIPCH11 is met when all of the following are true:

1. Output ports are explicitly defined​

The AI Product defines:

  • how it is accessed
  • what inputs it accepts
  • what outputs it produces
  • what contracts govern usage

These are:

  • documented
  • versioned
  • stable

2. Ports represent the capability, not implementation​

Ports expose:

  • decision outcomes
  • predictions
  • recommendations
  • actions

Not:

  • model internals
  • pipeline steps
  • intermediate states

3. Experience layers are separate from the product​

Applications, dashboards, or UIs:

  • consume the AI Product via its ports
  • are not part of the product itself

This ensures:

product ≠ application

4. Ports are governed and controlled​

Access to ports is:

  • controlled via policies (AIPCH10)
  • monitored and observable (AIPCH09)
  • tied to trust and usage signals (AIPCH07)

5. Ports are stable across consumers​

Consumers:

  • rely on consistent contracts
  • are not impacted by internal implementation changes

This enables:

reuse and composability

Negative Criteria — When AIPCH11 is not met​

AIPCH11 is not met if any of the following are true:

❌ AI capability is embedded in applications​

Examples:

  • logic exists only in a UI
  • capability accessible only via dashboard
  • no independent access point

This hides the product.

❌ APIs exist but are not productized​

Examples:

  • internal APIs without documentation
  • unstable interfaces
  • tightly coupled to implementation

This is not true addressability.

❌ Ports expose implementation details​

Examples:

  • endpoints expose model internals
  • pipeline steps visible to consumers
  • consumers must understand internal architecture

This breaks abstraction.

❌ Consumers bypass product boundary​

Examples:

  • direct access to underlying models
  • direct database queries
  • use of internal components instead of product

This violates product encapsulation.

Edge Cases (Important Guidance for Agents)​

Case 1: “AI capability accessible only via UI”​

❌ Not met

Rationale:

  • no independent product interface
  • tightly coupled to experience

Case 2: “API exists but undocumented and unstable”​

⚠️ Partial

Rationale:

  • interface exists
  • but not product-grade

Case 3: “Well-defined API with stable contract and governance”​

✅ Met

Rationale:

  • clear product boundary
  • reusable across consumers

Case 4: “Agent interface exposed for composition”​

✅ Met

Rationale:

  • aligns with composability (AIPCH20)
  • supports agent ecosystems

Evidence Signals an Agent Should Look For​

Authoritative evidence:​

  • defined API/port contracts
  • versioned interface definitions
  • access control policies on ports

Supporting evidence:​

  • usage metrics per port
  • multiple consumers using same port
  • separation between product and UI

Red flags:​

  • UI as primary access mechanism
  • undocumented APIs
  • direct access to internal components
  • frequent breaking changes in interfaces

How an Agent Should Decide​

Decision rule (simplified):

If consumers cannot access the AI Product through explicit, stable, and governed output ports independent of any application or UI, AIPCH11 is not met.

Why AIPCH11 Is Non-Negotiable​

Without AIPCH11:

  • products collapse into applications
  • reuse becomes difficult
  • composition becomes fragile
  • governance cannot be enforced consistently

AIPCH11 enables:

  • clear product boundaries
  • independent consumption
  • scalable reuse and composition
  • separation of concerns (product vs experience)

Canonical Statement (for AIPS)​

AIPCH11 is satisfied only when an AI Product exposes explicit, stable, and governed output ports that represent its capability, enabling consumers to interact with the product independently of any application or experience layer, with clear contracts and enforced access controls.