Agents vs Chatbots (Recap)

A chatbot typically: user message → model reply → done. An agent typically: user goal → plan → tool call → observe → maybe plan again → final answer.

When you need an agent

• Multi-step workflows (search, then calculate, then file a ticket).
• Fresh data behind APIs or databases.
• Actions with side effects that need confirmation.

When a chatbot is enough

• FAQ and drafting inside a fixed context.
• Single-turn summarization with no tools.

One-line summary

Agents add loops, tools, and state; chatbots mainly generate the next reply.

Key terms

• One-shot — single generate step.
• Agent loop — repeated reason/act/observe until stop.