Skip to main content

Introduction to AIPS

The Artificial Intelligence Product Specification (AIPS) is an open specification for defining, qualifying, realizing, governing, assuring and evolving AI Products.

AIPS specializes the Base Product Specification (BPS) for Products whose value-bearing essence is artificial intelligence. BPS establishes Productness and general Product Intrinsic Nature (PIN); AIPS establishes whether a Product version qualifies for the AI Product Family and, if it does, how its intelligence is described and governed.

Using AI does not automatically make something an AI Product.

A governed Model Product, feature platform, AI-enabled service, or AI-generated creative Product can all participate in the AI ecosystem without automatically belonging to the AIPS AI Product Family.


Why AIPS?​

AI ecosystems contain models, prompts, rules, knowledge bases, agents, feature stores, planners, reasoners, tools, services and applications. Calling all of these “AI Products” destroys useful semantic boundaries.

AIPS separates the questions:

  1. Is this thing a Product? — determined under BPS.
  2. Does its intrinsic nature qualify it as an AI Product? — determined through AIPS AI Qualification.
  3. What type of AI Product is it? — described through APT and the Product's intelligence provenance.
  4. How does it behave, act, interact, embody risk and evolve? — described through orthogonal ontology and governance dimensions.

A qualifying AI Product is supported by current, version-specific AI Qualification Evidence (AQE) and a direct or transitive path to at least one qualifying Artificial Intelligence Primitive.


AI Product ontology at a glance​

AIPS describes AI Products through independent dimensions rather than one overloaded “capability type”:

  • APT — AI Product Type: Primary or Derived;
  • IC — Intelligence Capability: e.g. Detection, Generation, Reasoning, Planning, Control;
  • CS — Consumption Scope;
  • PU — Purpose;
  • MO — Interaction / Execution Mode;
  • IE — Intelligence Effect;
  • AR — Agency / Autonomy;
  • EM — Embodiment;
  • RT — Effective Risk Tier (R0–R4).

A Product may therefore be, for example, Primary + Reasoning + Interactive + Assist + Non-embodied + R2 without confusing any of those characteristics with Product Type.

See AI Product Ontology.


Scope of AIPS​

AIPS is broader than two documents. Its principal layers are:

  • AIPROD — AI Product Semantic Specification: versioned AI qualification, meaning, ontology, lineage, governance and Product intent;
  • AIPDS — AI Product Deployment Specification: mechanism-neutral execution, deployment, observability, controls and evolution;
  • AIPS-DPP — AI Product Trust Passport: qualification, ontology, risk, governance, currentness and assurance projection;
  • PDEP / HDIP Spec C: Product-development and evolution lifecycle from intent through definitive qualification, realization, publication and Continuous Assurance;
  • Marketplace / Discovery / Consumption: verified Product projection, acquisition/entitlement and governed consumption ports;
  • COVP — Consumption Observation & Value Plane: usage, feedback, fitness and value-realization evidence;
  • PMDD / AIPCH: maturity assessment for Products that already qualify as AI Products;
  • Lifecycle Closure: cross-artifact consistency, change impact, supersession and reclassification.

Mechanism neutrality​

AIPS does not require every AI Product to own a trained ML model, retraining pipeline, or inference API.

Valid intelligence mechanisms can include:

  • learned models;
  • symbolic reasoning;
  • knowledge reasoning;
  • planning and intelligent search;
  • optimization where it materially constitutes intelligent behavior;
  • hybrid mechanisms;
  • upstream AI Product dependencies in a Derived AI Product.

Model training and retraining are conditional specializations. A symbolic AI Product may legitimately declare learningMode = NONE while still being fully conformant and mature.


Relationship to BPS​

BPS provides universal Product concepts such as identity, ownership, lifecycle, discoverability, governance, observability, Product Intrinsic Nature and contextual Business Product role.

AIPS adds AI-specific semantics including:

  • Product-family qualification and AQE;
  • AI Product Type and derivation topology;
  • Artificial Intelligence Primitive provenance;
  • intelligence capabilities, effects, agency and embodiment;
  • evidence-backed risk and Governance Obligation Profiles;
  • trust/currentness and assurance;
  • AI-specific lifecycle impact and downstream requalification.

Who should use AIPS?​

AIPS is intended for:

  • AI Product Owners and domain owners;
  • AI/ML, reasoning, knowledge, agent and robotics engineers;
  • Enterprise, solution and platform architects;
  • governance, risk, compliance and assurance teams;
  • marketplace/catalog and platform builders;
  • researchers and standards communities.

AIPS is an open specification under continuous evolution. Its goal is not to label everything AI-related as an AI Product, but to make AI Product claims precise, evidence-backed, interoperable and governable.