liveBook · Manning

Agentic application

As explained in Agent Design Patterns

A software application whose operation relies on AI capabilities for essential functions while combining model-driven behavior with traditional software.

An agentic application is software that relies on AI for an essential part of its operation. The AI need not control the entire process: traditional code can orchestrate when and how the model is used. section 1.1.1

The design choice is how much control to give the model:

  • Code orchestration specifies a rigid, tractable process while retaining AI capability where it is needed. section 1.1.1
  • Greater model agency lets the model make more high-level decisions, but that freedom may be unwise when hallucinations could affect customers. section 1.1.1
  • Composable agent design patterns can help make the resulting application transparent, standardized, and automatically testable. section 1.0

Reliability-critical applications

Legal tech and medtech agents illustrate the preference for code orchestration when reliability matters. A firm process does not rule out situational attentiveness from the AI components within it. section 1.1.1

How Agent Design Patterns covers it

How deep it goes
Defines the agent boundary and compares orchestration choices without specifying an implementation or testing procedure.
Examples
Points to legal tech and medtech agents as reliability-critical cases favoring code orchestration.
What you can do after
Evaluate/understand when AI dependence makes an application agentic and why reliability may favor code orchestration.
How it connects
Connects agent design patterns, code orchestration, model agency, LLMs, and vision-language-action models.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Agent Design Patterns. The section links point to the parts of the book it draws on.