A few AI basics

Chat, skills and agents: what’s the difference?

By the end of this reading, you should be able to choose when to use a chat, a reusable skill or an agent for a teaching task.

Imagine you are preparing a worksheet for a project on food waste. You can work with AI in different ways, depending on what you need.

Tools use these terms differently. The definitions below are the working distinctions we use in this course—not three levels you must progress through.

Chat: work through a request

In a chat, you ask AI for help, inspect its response and give feedback. You can type or dictate your requests.

For example:

“Draft a worksheet that helps pupils compare ways to reduce food waste.”

After reviewing the draft, you might add:

“Include a question that asks pupils to justify their choice using evidence.”

This back-and-forth lets you develop the resource through specific requests and revisions.

Use a chat when: you want to explore an idea, ask questions or improve a particular draft.

Skill: reuse a process

A skill packages instructions for a task you expect to repeat. Depending on the tool, it can also include examples, reference materials or other resources.

For example, a worksheet skill could specify:

• The structure your worksheets should follow.
• Your expectations for language and accessibility.
• Checks for learning-goal alignment and pupil reasoning.

Next time, you provide the new topic, pupil context and learning goal rather than rebuilding the whole brief.

A skill reuses a process—not a guarantee of quality. You still check each result.

Use a skill when: you want to apply a repeatable approach across similar tasks.

Agent: carry out actions towards a goal

An agent can take a sequence of actions towards a goal, using the tools and permissions available. It can check results and decide what to do next within those boundaries.

For example, an agent might:

1. Read an approved project brief.
2. Draft a worksheet using your worksheet skill.
3. Check the draft against your criteria and revise it.
4. Save it for your review.

You might start the agent yourself, or a system might start it on a schedule or after an event. A trigger starts the task; it does not define the agent. A scheduled task is not necessarily an agent.

Before using one, decide:

• Access: which information and tools may it use?
• Actions: what may it create or change?
• Approval: when must it stop and ask you?

For example: “Save a draft, but ask before sharing it with pupils.”

Use an agent when: you want a task carried out through several actions, with appropriate limits and review.

Apply it to your work

Think of one teaching task you expect to do soon. Complete this sentence:

“For ___, I would start with a ___ because ___. Before using the result, I would check ___.”

There may be more than one reasonable approach. The important part is explaining your choice and identifying what needs your judgement.

The focus of the first week of the course is exactly on how to assemble a high-quality skill. By week 3, we'll cover how and when to implement agents properly.

Remember: chat to work through a request; use a skill to reuse a process; use an agent to carry out actions within clear boundaries. An agent may use a skill, and a chat may help you develop one.