Topic 7

Do you need an agent? And wrap-up

Before building an agent, ask whether you need one. Anthropic's "Building effective agents" (2024) recommends finding the simplest solution possible and only increasing complexity when needed, and notes that agentic systems often trade latency and cost for better task performance. For many applications, a single well-prompted call with retrieval is enough.

Wrap-up. From this lesson, you can now:

• define an agent by its environment and its actions, and recognise RAG systems as simple agents;
• sort tools into knowledge augmentation, capability extension and write actions, and say why writes need the most care;
• describe the agent loop: think, act, observe, repeat, until the model answers or the harness stops it;
• read a trace and count model calls, tool callsTool callingA model emits a structured request (a tool name and JSON arguments) instead of text; the application checks and runs the function, then returns the result into the model's context.Where this is taught and write actions;
• separate what the model decides from what the harness enforces;
• explain compound errors, and the three levers against them.