Frequently asked questions

What is an AI agent in simple terms?

A program with an AI model inside that you give a goal, not a question. It decides on its own which steps to take and with which tools (opening websites, reading spreadsheets, writing drafts) until it gets a result.

How is an AI agent different from a chatbot?

A chatbot replies to a message. A workflow runs the same predefined steps. An agent plans its steps and acts with tools on its own. OpenAI states plainly that a simple chatbot is not an agent.

How is an AI agent different from an AI assistant?

An assistant helps; an agent acts. Under the hood, an assistant is often the same agent, but it leaves decisions (send, pay, delete) to you. The test comes down to two questions: who chooses the next step, and who presses “send.”

What types of AI agents are there?

By design, there are five classic types: simple reflex, model-based, goal-based, option-comparing, and learning. By task, there are agents for the browser, coding, research, customer support, personal affairs, and teams of several agents.

Are AI agents dangerous?

The main risks are errors that compound from step to step, going in circles, and commands from strangers hidden in emails and on websites. That's why agents get minimal permissions, a step limit, and human approval before buying, sending, or deleting anything.

How do I build my own AI agent?

A simple one without code, in n8n or Yandex AI Studio; a complex one on a framework such as the OpenAI Agents SDK or LangGraph. Step by step: see how to build an AI agent.

Read next

Try it on your own task

Message our trial bot in Telegram and describe what you’d like to hand over. It will do the task right in the chat, and you’ll see how the assistant will work for you.

Your first task is free