AIPS Taxonomy
AIPS uses orthogonal classification axes. Terms that answer different questions MUST NOT be collapsed into a single “capability type” or archetype.
1. AI Product Qualification
A Product qualifies for the AI Product Family only when:
- it first qualifies as a Product under BPS;
- artificial intelligence is the value-bearing essence;
- it provides a distinct intelligence proposition;
- its direct/transitive constitution contains at least one qualifying AI Primitive;
- the Product boundary is governed and versioned.
AI qualification is version-specific and evidence-backed.
2. AI Product Type (APT)
APT classifies the provenance of productized intelligence:
- Primary — no material upstream AI Product dependency;
- Derived — materially depends on one or more upstream AI Products.
Composite is a Derived topology, not a peer APT.
3. Intelligence Capability (IC)
IC answers what intelligence the Product provides. It is multi-valued.
Canonical values include:
- Perception;
- Classification;
- Detection;
- Prediction;
- Generation;
- Recommendation / Ranking;
- Reasoning;
- Planning;
- Optimization;
- Control.
A Product can therefore be both Generative and Reasoning, or Perception + Planning + Control, without inventing a new Product Type.
4. Interaction / Execution Mode (MO)
MO answers how intelligence is invoked or operated:
- Batch;
- Real-time / Event-driven;
- Synchronous API;
- Interactive;
- Embedded;
- Continuous;
- Agentic / Goal-driven.
Execution mode is not APT.
5. Intelligence Effect (IE)
IE answers what the intelligence produces or causes:
- Score;
- Classification;
- Prediction;
- Detection / Alert;
- Recommendation;
- Generated Content;
- Explanation / Evidence;
- Decision;
- Plan;
- Digital Action;
- Physical Action.
6. Agency / Autonomy (AR)
AR describes delegated causal/decision authority:
- Inform;
- Recommend;
- Assist;
- Act with Approval;
- Autonomous Action.
AR is a major governance/risk driver but does not determine APT.
7. Embodiment (EM)
EM describes the relationship between intelligence and the physical world:
- Non-embodied;
- Physical-world Perceptive;
- Embodied / Actuating;
- Closed-loop Autonomous.
Physical AI can be Primary or Derived.
8. Consumption Scope (CS) and Purpose (PU)
CS examples: Local, Cross-Division, Global/Federated, External.
PU examples: Decision Support, Operational Automation, Compliance & Risk, Client Value Creation, Revenue Enablement, Innovation, Research/Exploration.
These are contextual dimensions and do not establish intrinsic AI Product Type.
9. Topology Scope
Canonical topology scope values are:
- Standalone;
- DomainCoupled;
- CrossDomainFederated;
- EnterpriseOrchestrated.
Topology is a governance/risk driver describing how far Product effects and coordination can propagate.
10. Risk & Governance Taxonomy
AIPS uses the canonical R0–R4 Effective Risk Tier:
- R0 — Minimal;
- R1 — Limited;
- R2 — Moderate / Material;
- R3 — High;
- R4 — Critical / Escalation-required.
Risk is computed from evidence-backed drivers rather than a single archetype. Material drivers may include:
- agency;
- intelligence effect;
- embodiment risk;
- topology;
- domain sensitivity;
- decision finality;
- action authority;
- impact scale;
- reversibility;
- qualification currentness;
- dependency topology.
The resulting Governance Obligation Profile records the controls that apply to the exact Product version.
11. Model Product and AI-enabled Product boundaries
A Model Product is a productized reusable learned computational capability. It can be a Business Product for its provider without automatically being an AIPS AI Product.
An AI-enabled Product has another value-bearing essence (for example Service, Physical/Engineered, Creative/Content) while AI contributes materially to its operation or construction.
Using AI does not automatically make a Product an AI Product.
12. Composition
Runtime composition/orchestration does not automatically create a new AI Product.
A new Derived AI Product exists only where the composition establishes a distinct governed intelligence proposition with its own Product boundary.
Principle: AIPS taxonomy is deliberately multi-axis: Product Type, capability, agency, embodiment, effect, execution, risk and context answer different questions and remain independently computable.
13. Marketplace listing and consumption taxonomy
Marketplace classification is not a substitute for PIN or AI Qualification. AIPS Marketplace 0.1 distinguishes:
- AIProduct listing - current/historical listing of a Product version that satisfies AI Product qualification;
- ModelProduct listing - productized learned capability, with no APT claim;
- AIEnabledProduct listing - another intrinsic Product Family with an enabling or constructional AI contribution.
Listing state is separately classified as Active, Restricted, Suspended, Historical, or Retired. Access policy and acquisition mechanism are also orthogonal to Product Family.
Consumption is resolved through governed Product ports after entitlement; marketplace discovery itself is not a consumption mode.
14. Consumption Observation & Value (COVP)
COVP introduces contextual evidence classifications, not intrinsic Product classifications:
- Usage Observation;
- Consumer Feedback;
- Fitness Assessment: FIT / CONDITIONAL / UNFIT / INSUFFICIENT_EVIDENCE;
- Value Realization with attribution confidence;
- Incident / assurance escalation signal.
COVP currentness is tracked separately for observation, fitness and value evidence. High usage, consumer ratings or realized value MUST NOT be used to infer AI Product Family, APT or R0-R4 risk directly.
Principle: Consumption evidence may trigger review or evolution, but ontology remains evidence-backed by PIN, AQE and the Product constitution graph.