liveBook · Manning

Planning

As explained in Retrieval Augmented Generation, The Foundational Ideas

Planning is an agentic workflow pattern in which a system constructs an action sequence before or during task execution.

Planning is an agentic workflow pattern that decomposes a complex query before retrieval, retrieves for each sub-query, and synthesizes the results. section 15.2.1

Use planning when a single retrieval pass cannot assemble the required evidence.

  • Treat planning as a runtime decision: determine how to decompose the question and how many sub-queries it requires. section 15.2.1
  • Retrieve independently for each sub-query, then combine the results into one answer. section 15.2.1
  • Prefer planning over fixed query expansion when the question's information needs vary at runtime. section 15.2.1
  • Balance quality against latency because additional model calls increase response time. section 15.9

Comparing claims processing with industry benchmarks

A planning agent separates the question into internal claims-process documentation and external auto-insurance industry data, retrieves each independently, and synthesizes the comparison. section 15.2.1

How Retrieval Augmented Generation, The Foundational Ideas covers it

How deep it goes
Explains decomposition, independent retrieval, synthesis, and contrast with fixed query expansion, but not implementation internals.
Examples
Uses claims-processing documentation, external auto-insurance benchmarks, and independently retrieved sub-queries as the central example.
What it assumes
Assumes familiarity with retrieval and query expansion while treating runtime decomposition as the distinguishing capability.
What you can do after
Apply planning to split complex questions, retrieve evidence per sub-query, and synthesize a multi-source answer.
How it connects
Connects planning with agentic RAG, query expansion, retrieval, synthesis, reflection, tool use, and multi-agent collaboration.

Also explained in

Compare how 11 books explain planning →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Retrieval Augmented Generation, The Foundational Ideas. The section links point to the parts of the book it draws on.