Automation: fixed rules
An automation follows steps you define. "When a customer fills out the contact form, add them to the spreadsheet and send me a message." It does exactly that every time and nothing else. Tools like Zapier, Make and n8n build these.
Chatbot: a conversation
A chatbot answers questions in a chat window. Older ones follow scripts; newer ones use AI to answer in natural language, usually from your help articles. Tidio and Intercom are examples on the customer-support side, and ChatGPT is a general chat assistant.
AI agent: goal-driven work
An agent is given a goal and decides what steps to take, often using several tools along the way: reading an email, checking a calendar, drafting a reply and asking you to approve it. Lindy is one platform for building such agents, and several automation tools now let you add agent steps to workflows.
Which do you need?
- The task is always the same and the inputs are tidy: automation.
- Customers ask the same questions all day: chatbot trained on your help content, with a path to a human.
- The task needs judgment on messy input, like sorting a shared inbox: try an agent, but keep a human approving anything that leaves the building.
The rule of thumb: use the simplest thing that works. A plain automation is cheaper, faster and easier to trust than an agent.
Want help choosing? Tell us what you're trying to do.