Skip to main content

Capability Type

Every AI Product must declare its capability type — the kind of intelligence, reasoning, or action it provides.
Capability type classification ensures products are comparable, composable, and discoverable.

Why Capability Type Matters​

  • Discoverability → Catalogs can group and filter products by capability.
  • Composability → Enables chaining of different capability types into workflows.
  • Governance → Risk assessment differs across types (e.g., symbolic vs generative).
  • Future-Proofing → Provides a taxonomy that can adapt as new AI paradigms emerge.

Core Capability Types​

1. Predictive​

  • Description: Makes predictions based on input data.
  • Examples: Fraud detection, demand forecasting, risk scoring.

2. Descriptive​

  • Description: Analyzes and summarizes existing data.
  • Examples: Report generation, clustering, topic modeling.

3. Prescriptive​

  • Description: Recommends actions or decisions.
  • Examples: Treatment recommendations, next-best-action systems.

4. Generative​

  • Description: Creates new content (text, image, audio, video, 3D models).
  • Examples: LLMs, image generation, code synthesis.

5. Agentic​

  • Description: Acts autonomously or semi-autonomously to achieve goals.
  • Examples: Multi-agent systems, autonomous assistants, robotic controllers.

6. Symbolic / Reasoning​

  • Description: Uses symbolic rules, logic, or reasoning to derive conclusions.
  • Examples: Knowledge graphs, theorem provers, rule-based expert systems.

Hybrid Capability Types​

AI Products may combine multiple capabilities, such as:

  • Predictive + Generative → Forecasting with narrative explanations.
  • Agentic + Generative → Agents creating and executing content.
  • Symbolic + Neural → Neuro-symbolic AI for reasoning with perception.

Hybrid types must be explicitly declared.

Metadata Requirements​

Each AI Product must declare:

  • Primary capability type → The dominant role of the product.
  • Secondary capability types → Supporting or hybrid roles.
  • Capability evolution → How retraining or updates may change capability scope.

Example​

AI Product: Customer Support Agent

  • Primary Capability Type: Agentic.
  • Secondary Capability Types: Generative (LLM-powered responses), Predictive (next-best-action).
  • Capability Evolution: Initial release supports structured workflows; later releases may include autonomous task chaining.

Summary​

  • Capability type defines what kind of intelligence the AI Product provides.
  • Types include predictive, descriptive, prescriptive, generative, agentic, symbolic, and hybrids.
  • Declaring capability type ensures discoverability, composability, and governance.

Principle: An AI Product without a declared capability type risks being a black box — unintelligible to consumers and unmanageable for governance.