Skip to main content

Capability Types of AI Products

AI Products are defined not only by their packaging and governance, but also by the capability they provide.
The capability type is a required attribute of every AI Product in AIPS.

Core Capability Types​

1. Model​

A trained algorithm that transforms inputs into outputs.

  • Examples: classifiers, regressors, embeddings generators, LLMs, vision models.
  • Often delivered as APIs or deployable services.

2. Agent​

An autonomous or semi-autonomous system that perceives, reasons, and acts to achieve goals.

  • Examples: task-oriented chatbots, RPA bots, autonomous vehicles, personal assistants.
  • May combine multiple models with orchestration logic.

3. Multi-Agent System​

A coordinated collection of agents that interact with each other and with environments.

  • Examples: agent swarms for problem-solving, trading bots, distributed simulations.
  • Requires additional governance for coordination, safety, and emergent behaviors.

4. Hybrid AI System​

An AI Product that combines multiple paradigms:

  • Symbolic + Statistical AI (e.g., neuro-symbolic reasoning).
  • Human-in-the-loop AI where human judgment is embedded.
  • Cross-modal systems (text + vision + speech).

Future-Oriented Capability Types​

5. Artificial General Intelligence (AGI)​

A system with broad, human-comparable cognitive abilities.

  • Can generalize across domains.
  • Requires governance and accountability frameworks beyond narrow AI.

6. Artificial Superintelligence (ASI)​

A system surpassing human intelligence across most dimensions.

  • Still hypothetical, but AIPS must remain future-proof.
  • Requires explicit treatment of existential risk, alignment, and global governance.

Why Capability Types Matter​

  • Discoverability: Consumers can search for AI Products by capability type.
  • Governance: Risk, explainability, and prohibited uses differ by type.
  • Composability: Different types can be orchestrated (e.g., a Model inside an Agent).
  • Future-Proofing: Ensures AIPS remains valid as AI evolves.

Summary​

Every AI Product must declare its capability type (see Glossary). .
This anchors its description, governance, and lifecycle in AIPS, and ensures products are:

  • Findable in catalogs.
  • Governed according to risk level.
  • Composable into larger systems.