Pattern to tool mapping

Find the relevant passages first, then answer from them with the source cited.

RAG over a knowledge graph of entities and relationships, so an answer can follow connections across documents.

Find documents by meaning rather than keywords: the retrieval half of RAG, with no generated answer.

Models that write, summarise and converse. The models and model platforms every generative pattern runs on.

Delivered by: Any generative pattern: RAG, AI workflow, AI agent

Read documents into structured fields, check them against rules, and send them to a person when unsure.

Assistants and bots that hold a dialogue across channels, handing over to a person when needed.

Delivered by: RAG for answers; AI agent when the bot acts

A known process run by a workflow engine, with AI doing one or two steps such as classifying or drafting.

A model in a loop with tools, memory and a goal, deciding its own steps.

Several agents, each with a narrow job, coordinated by one of them or by a workflow.

Classic machine learning that scores or forecasts from history: churn, demand, faults.

Delivered by: Predictive analytics pipeline

Flag what departs from the normal pattern in metrics, logs, usage or network behaviour.

Delivered by: Predictive analytics pipeline, or built into an observability product

Solvers that choose the best plan under constraints: routes, rosters, capacity, network planning.

Delivered by: Predictive analytics pipeline, with a solver in place of the model