What is an AI Product?
AIPS separates two questions that are often conflated:
- Is this thing a Product? — determined using BPS Product characteristics.
- If it is a Product, does its intrinsic nature qualify it for the AI Product Family? — determined using the AIPS AI Product ontology.
AI Product (AIPS): An AI Product is a governed Product whose value-bearing essence is a distinct artificial-intelligence capability and whose current constitution directly or transitively contains at least one qualifying Artificial Intelligence Primitive.
This means that using AI does not automatically make something an AI Product.
Qualification Conditions
A candidate must satisfy all of the following:
- Product qualification — it first qualifies as a Product under BPS.
- Intelligence primacy — artificial intelligence itself is the value-bearing essence.
- Distinct intelligence proposition — the Product exposes a governed intelligence capability of its own.
- AI Primitive dependency — its current direct or transitive constitution contains at least one qualifying AI Primitive.
- Governed Product boundary — it has ownership, lifecycle, contracts, controls, versioning, observability, evaluation, and governance appropriate to the Product.
See AI Product Ontology for the normative conceptual model.
What Counts as an Artificial Intelligence Primitive?
An AI Primitive is a computational capability that materially contributes artificial inference, learning, perception, reasoning, generation, planning, optimization, decision formation, or intelligent action.
Examples include:
- learned models such as LLMs, classifiers, and vision models;
- symbolic inference engines;
- knowledge reasoners;
- planners;
- intelligent search;
- qualifying optimization engines;
- hybrid intelligence mechanisms.
A trained ML model is therefore one form of AI Primitive, not a universal requirement for all AI Products.
Productness Is Not the Same as AI Product Qualification
AIPS does not assume that packaging an AI-related artifact as a Product automatically makes it an AI Product.
For example, a downloadable, governed, versioned LLM may become a Model Product. It can be a valuable Business Product for the organization selling it and can act as an AI Primitive dependency of downstream AI Products, while remaining ontologically distinct from those downstream AI Products.
Likewise:
- a feature store can be a productized platform or data capability without being an AI Product;
- a rules engine can be infrastructure, or may qualify as an intelligence primitive depending on its semantics and use;
- an AI-generated movie remains a Movie Product if the movie is the value-bearing essence;
- a smartphone can be AI-enabled while remaining a physical/engineered Product.
AI Product Type (APT)
Every qualified AI Product declares an AI Product Type (APT):
- Primary — intelligence is constituted without a material upstream AI Product dependency, directly from AI primitives together with data, knowledge, rules, tools, and controls.
- Derived — the Product's distinct intelligence proposition materially depends on one or more upstream AI Products.
Where a Derived AI Product materially composes multiple upstream AI Products, its derivation topology may be described as Composite. Composite is not a separate peer APT.
Orthogonal AI Characteristics
APT must not be confused with other classifications. AIPS describes these separately:
- IC — Intelligence Capability
- CS — Consumption Scope
- PU — Purpose
- MO — Interaction / Execution Mode
- IE — Intelligence Effect
- AR — Agency / Autonomy
- EM — Embodiment
- RT — Risk Tier (R0–R4)
An AI Product can therefore be, for example:
Primary + Detection + Real-time + Recommend + Non-embodied + R3
without any of those dimensions competing to be its Product Type.
Version-specific Qualification
AI Product Family membership is a version-specific ontological claim, not a permanent label attached to Product identity.
If a later Product version removes its only qualifying AI Primitive path, that new version must be re-evaluated. Historical Product versions retain their historical classification.
AIPS therefore expects AI Product qualification to be supported by current, version-specific evidence.
Examples
Qualifies as an AI Product
- a governed fraud-detection intelligence Product backed by a registered model;
- a governed symbolic reasoning Product backed by a knowledge base and inference engine;
- an autonomous navigation intelligence Product that perceives, plans, and controls physical movement;
- a Derived transaction-risk AI Product that materially composes upstream fraud and sanctions AI Products.
Does not qualify merely because it uses AI
- a smartphone with an embedded AI assistant;
- a streaming service that uses recommendation AI;
- an AI-generated movie;
- a generic model package merely because it has been productized;
- a router that only delegates requests among existing AI Products without establishing a new intelligence proposition.
Relationship to BPS
AIPS inherits the universal Product concepts defined by BPS, including identity, ownership, purpose, discoverability, governance, observability, evolvability, economic accountability, and composability.
AIPS then adds the AI-specific ontology needed to establish and govern AI Product Family membership, AI Product Type, intelligence characteristics, AI provenance, risk, agency, embodiment, and qualification currentness.