What an AI agent is, and what it is not

Three things people confuse

The words "chatbot", "automation" and "agent" get mixed up. They describe three different designs.

• A chatbot answers. You ask, a model replies, and the conversation ends there. It does not change anything in the world unless you copy its answer somewhere yourself.
• Automation follows a fixed recipe. "When a form is submitted, add a row to the sheet." Every step was decided by a person beforehand. It is fast, cheap and predictable, but it cannot cope with a situation nobody anticipated.
• An agent pursues a goal. You describe the outcome, and the agent decides which steps to take, uses tools to take them, looks at what happened, and adjusts until the job is done or it must stop.

A useful one-line test: who chooses the next step? In automation, a person chose it earlier. In a chatbot, there is no next step. In an agent, the model chooses it while running.

The ingredients of an agent

Every agent, on any platform, is built from the same parts:

1. A model, the reasoning engine that reads text and decides what to do.
2. Instructions, a description of the job, the rules and the expected output.
3. Tools, the actions the agent is allowed to take beyond writing text.
4. Memory and context, what the agent can see and remember while it works.
5. A loop, which repeats think, act and observe until finished.
6. Controls, the limits and approvals that keep it safe.

The rest of this course takes these in turn.

How this looks in PGB Agents

PGB Agents has exactly these parts on one page. An agent has Instructions, a Model (Fast, Smart or Code), Tools, Triggers, a Budget and Memory. Every time it starts, it creates a run with a step-by-step record.

Try it yourself

You will use one practice agent through this course.

1. Open agents.pgbhub.com and sign in with your PGBHub account.
2. Open AI keys and add a key from your AI provider if you have not already. The course "PGB Agents 101" shows how.
3. Choose Agents, then New agent. Under Blank agent, type the Name "Concept lab" and choose Create agent.
4. On the agent page, find Instructions, Tools, Triggers, Budget and Memory. Match each one to an ingredient from this lesson. Notice that Tools is empty: this agent cannot do anything except write text yet, which makes it a chatbot-shaped agent.

Recap

• A chatbot answers, automation follows a fixed recipe, and an agent chooses its own next steps toward a goal.
• An agent is a model plus instructions, tools, memory, a loop and controls.
• In PGB Agents, those parts are the sections of the agent page.