In this article
“Agent” is the most over-used word in AI right now. Strip the hype away and an agent is something simple: a language model running in a loop, with tools it can choose to use, until a task is done.
Advertisement
The loop
- You give the model a goal and a list of tools (functions it may call).
- The model either answers, or asks to call a tool with some inputs.
- Your code runs the tool and sends the result back.
- Repeat until the model answers without asking for a tool.
That is the whole idea. The “intelligence” is in step 2: the model decides which tool to use and what to pass it, based on what it has seen so far.
Chatbot, workflow or agent?
| Type | Who decides the steps | Example |
|---|---|---|
| Single call | Nobody — one question, one answer | Summarise this email |
| Workflow | Your code, in a fixed order | Extract invoice fields → validate → save to Excel |
| Agent | The model, step by step | “Find why March sales dropped” — it decides what data to look at next |
⚠️ Most business tasks are workflows, not agents. A fixed sequence is cheaper, faster and easier to test. Reach for an agent only when the steps genuinely can’t be known in advance.
A four-question checklist
- Is the task open-ended? If you can write the steps down, write a workflow.
- Is it worth the cost? Agents make several model calls per task.
- Can the model do it? Try the task by hand in a chat first.
- Can mistakes be caught? Prefer tasks where a wrong step is visible and reversible.
What we will build in this course
We start with a tiny agent that uses two tools, then teach it to search your own documents (RAG), make it refuse when it doesn’t know, add guardrails and cost limits, and finish with a policy Q&A assistant you could give to colleagues.
Next: Lesson 2 — your first tool-using agent in about 40 lines of Python.
✨ Ask AI about this article
Stuck on a step? Ask a question and the AI answers using this article.
Free · AI can be wrong