Agent vs. chatbot vs. automation
The honest definition — and where the line is genuinely fuzzy.
"AI agent" is the phrase of the moment, and it's used loosely enough to be confusing. Before you build one, it's worth getting the honest definition straight — including where the line genuinely blurs.
The three things, distinguished.
- A chatbot answers. You ask, it responds — one turn at a time. ChatGPT in its basic form is a chatbot: it produces text, but it doesn't go do anything in the world or pursue a goal across multiple steps.
- A fixed automation follows a script a human wrote. "When a new email arrives, save the attachment to this folder and send a confirmation." The steps are decided in advance by a person; the automation just runs them. (This is what tools like Zapier have done for years.)
- An agent figures out the steps itself. You give it a goal, and it plans a sequence of actions, uses tools to carry them out, checks the results, and adjusts — looping until the goal is met. The useful one-liner: **you define the what; the agent figures out the how.** A chatbot with the ability to search the web, book something, and email you a summary — deciding for itself how to get there — is acting as an agent.
The simplest test. If a human designed every step in advance, it's an automation (a workflow). If the system decides its own sequence of actions based on the goal, it's an agent. Put differently: automations follow instructions; agents interpret them.
Where the line is genuinely fuzzy (and why honesty matters). "Agentic" is a spectrum, not an on/off switch. A chatbot that can call a single tool (like web search) is slightly agentic. An automation with one AI decision-step in the middle is a hybrid. In fact, most useful real-world "agents" today are tightly-scoped hybrids — mostly-predictable workflows with a bit of AI judgment inserted where it's needed — not fully autonomous systems roaming free. And because "agent" sells, companies slap the label on lots of things. So take the marketing with a grain of salt, and judge any tool by what it actually does, not what it's called.
Why this matters for you. Getting this straight saves you from two mistakes: being underwhelmed ("this 'agent' is just a chatbot with one tool") and being overwhelmed (thinking you need to build a sci-fi autonomous system). What you'll actually build in this course is the useful, grounded kind — an assistant with clear instructions, some knowledge, and a few tools, kept on a short leash. That's where agents genuinely work today.
The mindset: a chatbot answers, a fixed automation follows human-designed steps, and an agent figures out the steps itself to reach a goal you set — you define the what, it works out the how. The simple test: if a human designed every step it's an automation; if the system chooses its own actions it's an agent. But "agentic" is a spectrum, not a binary — most real-world agents are tightly-scoped hybrids of predictable workflow plus a little AI judgment, and vendors over-use the "agent" label, so judge tools by what they do. Understanding this keeps you from being underwhelmed by glorified chatbots or overwhelmed by sci-fi expectations; what actually works today is a well-scoped assistant with instructions, knowledge, and a few tools.
Classify three things you've heard called "AI": for each, decide — is it a *chatbot* (just answers), an *automation* (fixed human-designed steps), or an *agent* (decides its own steps toward a goal)? Then think of one task you'd like AI to handle for you and describe it as a goal (the *what*) rather than a fixed procedure. That goal-framing is how you'll brief your first agent.
Enjoying the free lessons? Get an email when we publish new courses and updates — no spam, unsubscribe anytime.
Discussion (0)
Ask a question or share what worked for you. Comments are reviewed before they appear.
No comments yet. Be the first to start the discussion!