What actually makes something an "agent"
A chatbot answers one message at a time and stops. An "agent" is a system that can take multiple steps on its own toward a goal - deciding what to do next, using tools, and continuing without a human approving every single step. The word gets used loosely in marketing; the real test is: does it make its own next-step decisions, or is a human choosing every action?
Agentic systems typically run a loop: observe the current state → decide an action (possibly using a tool) → take the action → observe the result → repeat, until a goal is met or a limit is hit. The "intelligence" isn't one big magic model - it's the same LLM being called repeatedly in a loop with tools and memory wired around it (this is the "orchestrator mindset" - and it's exactly correct). Real products built on this loop include Claude's computer use, ChatGPT's agent mode, and Manus - each just wires more tools and longer autonomy around the same underlying mechanism.
Take something you already automated manually in n8n/Make and ask: at which step did you make the decision vs. the AI? Relabeling your own workflow this way is the fastest way to see the human/agent line clearly.
What actually separates an "agent" from a plain chatbot?