Notice: This page requires JavaScript to function properly.
Please enable JavaScript in your browser settings or update your browser.
Learn Your First Task | Understanding Cowork
Claude Cowork

Your First Task

Swipe to show menu

A live end-to-end task run through Cowork. You give it a goal - drafting an email to a colleague - and Cowork handles connecting to Gmail, drafting the message, and leaving it in your drafts folder ready to review. The chapter's real lesson isn't the email. It's the pattern: describe the outcome, let Cowork handle the tools.

Core Concept

I want you to feel this in your hands as fast as possible. Everything else in this course is theory until you've watched Cowork take a natural-language request, figure out which connectors it needs, ask permission where it should, and deliver a finished result.

The task itself is intentionally small - a friendly email to a colleague. What matters is the shape of the interaction. You describe an outcome in the way you'd describe it to a smart teammate. Cowork figures out the tools.

How It Works (Step-by-Step)

Step 1 - Open a new Cowork task

New task, blank chat. If you have an existing project you want the task tied to, open it inside that project. Otherwise a standalone task is fine.

Step 2 - Describe the outcome in natural language

Type or dictate a full description of what you want. Not a step-by-step instruction - a description. Something like:

"Hey Claude, I want to draft an email to a colleague of mine. The email is jordan@example.com. I want to invite them to my birthday, which is on June 13th 2026, at a bar in the Gold Coast, Australia. I'll let them know the exact address closer to the date. Please make it casual but semi-formal - we talk to each other formally but we're friends. I use Gmail. If we need to connect anything, please help me do so."

Notice what's in there - the person, the topic, the tone, the date, the constraint about the address, and the meta-instruction about connecting Gmail if needed. Give Cowork the same context you'd give a human assistant.

Step 3 - Let Cowork handle the connector setup inline

If Gmail isn't already connected, Cowork will surface a "Connect Gmail" button right in the chat. Click Connect. Sign into Gmail. Grant read and draft access. Return to the chat. You are connected.

Note
Note

This is the pattern for every connector in the course - Calendar, Notion, Slack, Drive. Don't hunt for a Settings menu. Ask Cowork to connect the tool, and the flow shows up inline.

Step 4 - Read the draft Cowork produces

Cowork will draft the email in the chat window first. Read it. If it hits the tone and the details right, tell Cowork to create it as a draft in your Gmail account. If it needs adjustment ("change 'colleague' to Jordan's actual name," "make it warmer," "shorter"), say so. Cowork edits, then drafts.

Step 5 - Approve the draft action

When Cowork moves to create the actual Gmail draft, "Ask before acting" will surface an approval request. Allow it. Cowork uses the Gmail tool, creates the draft, and confirms.

Step 6 - Open Gmail and check the draft

The email is in your Drafts folder. Ready to send. You didn't open Gmail. You didn't copy-paste. You described what you wanted, and it was done.

Why This Matters

This chapter is where the whole idea of Cowork clicks for most people. The interaction feels like talking to a competent colleague, not typing prompts into a machine. That shift - from prompting to briefing - is the muscle you'll use for the rest of the course. Every workflow we build later is a variation on the same pattern: describe the outcome, let Cowork handle the tools, verify the result.

What Should Be Possible After This

  • Open a Cowork task and describe an outcome in natural language.
  • Let Cowork connect Gmail (or any other native connector) inline, without opening the Customize panel.
  • Read a draft Cowork produces and give it a natural-language edit.
  • Approve a draft action with the "Ask before acting" checkpoint.
  • Recognize the pattern - describe outcome, let Cowork handle tools, verify - and apply it to your own tasks.
question mark

What is the recommended way to describe a task to Cowork so it can execute it effectively?

Select the correct answer

Everything was clear?

How can we improve it?

Thanks for your feedback!

Section 1. Chapter 3

Ask AI

expand

Ask AI

ChatGPT

Ask anything or try one of the suggested questions to begin our chat

Your First Task

A live end-to-end task run through Cowork. You give it a goal - drafting an email to a colleague - and Cowork handles connecting to Gmail, drafting the message, and leaving it in your drafts folder ready to review. The chapter's real lesson isn't the email. It's the pattern: describe the outcome, let Cowork handle the tools.

Core Concept

I want you to feel this in your hands as fast as possible. Everything else in this course is theory until you've watched Cowork take a natural-language request, figure out which connectors it needs, ask permission where it should, and deliver a finished result.

The task itself is intentionally small - a friendly email to a colleague. What matters is the shape of the interaction. You describe an outcome in the way you'd describe it to a smart teammate. Cowork figures out the tools.

How It Works (Step-by-Step)

Step 1 - Open a new Cowork task

New task, blank chat. If you have an existing project you want the task tied to, open it inside that project. Otherwise a standalone task is fine.

Step 2 - Describe the outcome in natural language

Type or dictate a full description of what you want. Not a step-by-step instruction - a description. Something like:

"Hey Claude, I want to draft an email to a colleague of mine. The email is jordan@example.com. I want to invite them to my birthday, which is on June 13th 2026, at a bar in the Gold Coast, Australia. I'll let them know the exact address closer to the date. Please make it casual but semi-formal - we talk to each other formally but we're friends. I use Gmail. If we need to connect anything, please help me do so."

Notice what's in there - the person, the topic, the tone, the date, the constraint about the address, and the meta-instruction about connecting Gmail if needed. Give Cowork the same context you'd give a human assistant.

Step 3 - Let Cowork handle the connector setup inline

If Gmail isn't already connected, Cowork will surface a "Connect Gmail" button right in the chat. Click Connect. Sign into Gmail. Grant read and draft access. Return to the chat. You are connected.

Note
Note

This is the pattern for every connector in the course - Calendar, Notion, Slack, Drive. Don't hunt for a Settings menu. Ask Cowork to connect the tool, and the flow shows up inline.

Step 4 - Read the draft Cowork produces

Cowork will draft the email in the chat window first. Read it. If it hits the tone and the details right, tell Cowork to create it as a draft in your Gmail account. If it needs adjustment ("change 'colleague' to Jordan's actual name," "make it warmer," "shorter"), say so. Cowork edits, then drafts.

Step 5 - Approve the draft action

When Cowork moves to create the actual Gmail draft, "Ask before acting" will surface an approval request. Allow it. Cowork uses the Gmail tool, creates the draft, and confirms.

Step 6 - Open Gmail and check the draft

The email is in your Drafts folder. Ready to send. You didn't open Gmail. You didn't copy-paste. You described what you wanted, and it was done.

Why This Matters

This chapter is where the whole idea of Cowork clicks for most people. The interaction feels like talking to a competent colleague, not typing prompts into a machine. That shift - from prompting to briefing - is the muscle you'll use for the rest of the course. Every workflow we build later is a variation on the same pattern: describe the outcome, let Cowork handle the tools, verify the result.

What Should Be Possible After This

  • Open a Cowork task and describe an outcome in natural language.
  • Let Cowork connect Gmail (or any other native connector) inline, without opening the Customize panel.
  • Read a draft Cowork produces and give it a natural-language edit.
  • Approve a draft action with the "Ask before acting" checkpoint.
  • Recognize the pattern - describe outcome, let Cowork handle tools, verify - and apply it to your own tasks.
Everything was clear?

How can we improve it?

Thanks for your feedback!

Section 1. Chapter 3
some-alt