Agentic AI Chapter 3 / 16 · Chatbots vs Agents

03 Chatbots vs Agents

Stop asking questions.
Start giving goals.

A chatbot answers and stops. You ask, it replies — reactive, one shot. An agent takes a goal and goes and does the work: it plans, uses tools, checks its own results, and adapts — with barely any hand-holding. This is the shift the whole field is built on, and the keyword is autonomy.

Reactive → autonomous Perception Reasoning Planning Action Adaptation

The shift

One goal in. A sequence of actions out.

Here's the moment it clicks. You don't hand an agent a question — you hand it a goal, and it figures out the steps itself. Give it "research my top 3 competitors, compare their pricing and features, write a report, and email it to my team," and it just… does all of it. Press play.

🎯 The goal you give it
"Research my top 3 competitors, compare their pricing & features, write a report, and email it to my team."
🔍

Research the competitors

Web search + read their sites

waiting
📊

Compare pricing & features

Pull it into a structured table

waiting
✍️

Write the report

Draft the comparison & positioning

waiting
✉️

Email it to the team

Send via the email tool

waiting
idle
Four steps, zero follow-up prompts. A chatbot would have answered the first question and stopped.

Side by side

Reactive vs. autonomous.

Same underlying LLM — completely different behavior. A chatbot is a single round trip: your message in, one answer out. An agent wraps that same model in a loop with goals and tools, so it can take many steps before it's done.

🤖 Chatbot

Reactive · one shot

You ask LLM Answer

It responds to exactly what you said, then waits. Nothing happens until you type again.

🚀 Agent

Autonomous · many steps

Goal LLM + tools Result
think → act → observe, on a loop

It keeps looping — planning, calling tools, reading results — until the goal is actually met.

The difference isn't a smarter model. It's the loop, the goal, and the tools wrapped around it — which is exactly Chapters 4 and 5.

The test

Five properties make something agentic.

So when is a system truly an agent, not just a chatbot with extra steps? Look for five things. A plain chatbot has maybe the first two, and only partly. A real agent has all five — and that gap is the whole difference in what it can accomplish. Toggle between them.

5 / 5
has all five — it can own a goal end to end
👁️

Perception

Reads files, browses the web, sees images & API results

🧠

Reasoning

Thinks through a problem and breaks it into pieces

🗺️

Planning

Makes multi-step plans and adjusts as things change

Action

Actually does things — calls APIs, runs code, sends email

🔄

Adaptation

Learns from feedback and its own mistakes

A chatbot perceives your text and reasons a little. An agent perceives, reasons, plans, acts, and adapts.
The keyword

Autonomy — the capacity to act independently and make its own choices toward a goal. That's the line between a tool you operate and a teammate you delegate to.

Chapter 3 in one breath

A chatbot answers. An agent achieves.

Hand a chatbot a question and it replies. Hand an agent a goal and it perceives, reasons, plans, acts, and adapts — looping through steps until it's done. Same model underneath; autonomy on top. Next: the loop that makes that autonomy actually run.

🖥 Present this chapter

A matching slide deck with speaker notes — press S for notes, F for fullscreen.

Open the slides →