Skip to main content

Architecture Alignment

AI Products participate in a broader Product and enterprise architecture. AIPS therefore separates intrinsic Product semantics from implementation platforms, runtime technology and contextual Business Product roles.


Relationship to BPS and PIN​

  • BPS defines universal Product concepts, Product Intrinsic Nature (PIN), Product identity, contextual Business Product role and general Product governance.
  • AIPS specializes BPS for Product versions whose value-bearing essence qualifies for the AI Product Family.
  • AI Product qualification, APT and AI Primitive provenance are version-specific and evidence-backed.

AIPS does not assume that every AI-domain Product is an AI Product. A governed downloadable model can remain a Model Product; a Service or Physical Product can be AI-enabled; a Compound Product can have AI as one of several co-essential family contributions.


Relationship to Data and Information Products​

AI Products can consume Data Products, Information Products, Model Products, knowledge and other AI Products.

Typical relationships include:

  • Data Products provide governed business truth and training/inference inputs;
  • Information Products can provide interpreted context or consume AI effects;
  • AI Products provide inference, reasoning, generation, planning or action capabilities;
  • Derived AI Products may materially depend on upstream AI Products;
  • AI outputs may later be productized separately as Data or Information Products.

Dependency does not imply Product-family inheritance.


Enterprise architecture integration​

AI Products are first-class Product/version entities in enterprise architecture and Product registries. They integrate with:

  • Data / Information Platforms for governed inputs and outputs;
  • AI Runtime / Execution Platforms for learned, symbolic, agentic and hybrid intelligence;
  • Application / Experience Platforms for consumer-facing experiences;
  • Physical / Edge Platforms for embodied and closed-loop AI;
  • Governance Fabric for risk, policy, obligations and prohibited-use enforcement;
  • Observability / Assurance for behavior, currentness, incidents and fitness evidence;
  • Marketplace / Catalog for discovery and entitlement;
  • COVP for usage, feedback, fitness and value-realization evidence.

Deployment architecture is not Product ontology​

An AI Product can be realized as:

  • cloud service;
  • on-premise runtime;
  • edge/embedded execution;
  • synchronous API;
  • event-driven service;
  • continuous/agentic runtime;
  • physical-control deployment.

These choices belong to AIPDS and MO/EM semantics. They do not determine APT or Product Family.


Compound and Physical AI​

A physical system containing AI does not automatically become exclusively an AI Product.

For example, an autonomous humanoid can be modeled as a Compound Product with co-essential Physical/Engineered and AI contributions, while a separately governed Humanoid Intelligence capability is an AI Product with its own APT, AQE, AIPROD, AIPDS and DPP.

This enables architecture governance to preserve both system-level and intelligence-level Product boundaries.


Version and change impact​

Architecture repositories should preserve:

  • logical Product identity;
  • immutable Product-version semantics;
  • current AQE and APT;
  • versioned dependency bindings;
  • Governance Obligation Profiles;
  • deployment and trust projections;
  • downstream impact relationships.

A new upstream version that removes the only qualifying AI Primitive does not rewrite history. Downstream impact is computed only for Product versions that adopt the changed dependency.


Summary​

AIPS aligns architecture at four levels:

  1. BPS / PIN — what kind of Product is this?
  2. AIPS ontology — why does this version qualify as AI and where does its intelligence come from?
  3. Enterprise realization — where and how is it executed, governed, observed and consumed?
  4. Lifecycle graph — how do version changes propagate through dependent Products and trust projections?

This keeps ontology stable while allowing implementation technology and ecosystem architecture to evolve independently.