What's the difference between a chatbot and an AI agent?

The old-school chatbot most people have experience with is a decision tree wearing a chat interface: "Press 1 for support, 2 for sales" translated into buttons or keyword matching. Type something the tree didn't anticipate and it either fails silently, loops back to the menu, or gives a generic "I don't understand" — because there's no actual comprehension happening, just pattern matching against a fixed script written in advance.

An AI agent replaces that scripted core with a language model that reads the user's actual words, holds context across the conversation, and reasons about intent even when the phrasing is nothing like anything the builder anticipated. Someone can ask a question sideways, change their mind mid-conversation, or combine two requests in one message, and a well-built agent handles it because it's reasoning about meaning, not matching a regex.

The bigger difference is what happens after the conversation. A chatbot's job usually ends at the reply — it tells you the store hours or shows you an FAQ answer. An AI agent is typically wired to tools: it can look up a real calendar and check actual availability, write a real lead into a real CRM, query an order status against a live database, or trigger a handoff to a human when the situation calls for it. That's the "agent" part of the name — it acts on the world, not just on the conversation.

This distinction matters practically because it changes what can go wrong and what needs guarding against. A scripted chatbot fails predictably (it just doesn't understand). An agent with tool access needs real safeguards: it shouldn't book a slot that's actually taken, shouldn't write bad data into a CRM, and should know when to stop and hand off to a human rather than guessing on something high-stakes — a booking confirmation, a price commitment, a legal or medical question. Good agent design includes explicit boundaries on what it's allowed to do autonomously versus what needs a human sign-off.

In practice the line blurs — plenty of products marketed as "chatbots" are LLM-powered agents underneath, and the terms get used loosely in sales conversations. The useful question to ask a vendor isn't which word they use, but: does it actually understand free-form input, and does it take real actions in real systems, or does it just generate text back at you. That one question separates a real evaluation from a naming debate.