Skip to main content
Section 4

Building with AI

From chat to something that acts.

3 lessons25-question quiz
4.1

Chat, assistant, and agent

6 min read

You will hear "chatbot," "assistant," and "agent" used loosely. They roughly describe a ladder of autonomy: how much the system does on its own before handing back to you.

TermWhat it doesExample
ChatReplies to what you type, one turn at a timeA question-and-answer chatbot
AssistantHelps inside a task or app, with some contextAn in-editor coding helper
AgentPlans and takes multiple steps toward a goal, using toolsA system that edits files and runs commands to finish a task
A rough ladder of autonomy

The jump from assistant to agent is the important one. An agent does not just answer; it can take actions, check the result, and decide what to do next, looping until the goal is met. That power is useful, and it is also why permissions and review matter, which the next lessons and later courses cover.

Key idea
Chat replies, an assistant helps in context, an agent takes multi-step actions with tools. More autonomy means more usefulness and more need for guardrails.
Key terms
Agent
An AI system that can plan and take multiple actions toward a goal, not just reply.
More

An agent typically uses tools, checks results, and loops until it finishes or hands back to you. Its ability to act is why permissioning matters.

Autonomy
How much a system does on its own before returning control to you.
4.2

How tools extend a model

6 min read

On its own, a model only produces text from frozen training patterns. Tools break that boundary. A tool is a function the model is allowed to call, such as a web search, a calculator, code execution, or reading a file. The model decides when a tool would help, calls it, reads the result, and continues.

This is how a model can answer with today's news (a search tool), do exact arithmetic (a calculator or code tool), or work with your documents (a file tool). The intelligence is still the model; the tools give it hands and eyes.

  • A search or browse tool fetches live information past the knowledge cutoff.
  • A code or calculator tool does exact computation the model would otherwise guess at.
  • File and app tools let it read and change real things on your system.
Key idea
Tools are functions a model can call to act beyond text: search, compute, read and write files. They are what turn a chat model into something that can actually do things.
Because tools can touch real systems, a well-built agent asks permission or works inside limits, which is the next lesson.
Key terms
Tool
A function a model can call to do something beyond generating text.
More

Common tools include web search, code execution, calculators, and file access. The model chooses when to call a tool and uses the result to continue.

4.3

Your first tiny project

7 min read

You do not need to be a programmer to build something small with AI, and the same loop scales up later. The idea is to work in small, checkable steps rather than asking for everything at once.

  1. Describe the goal in plain language, including who it is for and what "done" looks like.
  2. Ask for the smallest first version, one piece you can actually try.
  3. Test that piece. Does it do what you expected? Note what is off.
  4. Give specific feedback and iterate: change one thing at a time.
  5. Keep a human in the loop for anything that touches real data, money, or people.

This loop works whether you are drafting a document, automating a small task, or writing your first script. The wins come from small steps and honest testing, not from one perfect prompt.

Where to go next

You now have the foundations: what a model is, how it works, its limits, how to use it well, and how people build with it. The other Academy courses go deeper: Prompting, Building with AI, Running Local AI, and Shipping Web & Apps.

Key idea
Build in small, testable steps and keep a human in the loop where it counts. Iteration beats one big ask.

Section 4 quiz

25 questions. Pass at 75% to master this section. Retakes are unlimited, and the quiz is where the learning sticks.

Section 4 quiz · Building with AIQuestion 1 of 25

What do "chat," "assistant," and "agent" roughly describe?