Level: Supervise. You hand the AI a whole task. You watch it work, you steer it when it drifts, and you check the result before anyone else sees it. Supervise is the third rung of the Nantucket AI trust ladder, after Orient and Assist and before Delegate.

Course overview

You have probably used an AI chat window. You ask, it answers, and then you do the real work of turning that answer into a document, a spreadsheet, or an email. Claude Cowork is Anthropic’s product for the second half of that job. You describe an outcome, Cowork works across your files and connected tools, and it hands back finished work for your review.

That change is bigger than it sounds. A chat window produces text. Cowork takes actions on real files and real connected accounts. That is why this course sits at Supervise and not at Assist. The skill you are learning here is not prompting. It is supervision.

Module 1 gives you the product as its official documentation describes it. What it does, who gets it, where it runs, and how Anthropic itself distinguishes it from Claude chat and from Claude Code.

Module 2 is the working module. You learn the permission and safety model, how to set up a first task, how to watch and steer a task in progress, and how to review a deliverable before it reaches a customer, a donor, or a client.

Module 3 makes it stick. You learn which small business tasks fit this shape and which do not, the limits the official documentation names, and how to keep your own knowledge current as the product changes.

Every example comes from the kinds of businesses that operate on this island. A retail shop on Main Street. A nonprofit with three staff. A restaurant that closes for the winter. A real estate office in August.

An honesty note about sources

Every product claim in this course comes from Anthropic’s official product page, official Help Center articles, or dated official blog posts, checked in August 2026. Where official pages disagree with each other, this course says nothing rather than picking a side. Cowork is a new product that is changing quickly. Some of what you read here will age. Module 3 teaches you where to look for the current answer.

Who this is for

• Small business owners and managers who already use AI chat and want more from it
• Nonprofit executive directors and staff with more work than hands
• Office managers and operations staff who handle files, spreadsheets, and reporting
• Anyone on a paid Claude plan who has seen the Cowork option and not used it

No technical background is required. You do not need to write code. You do need an account on a paid Claude plan to follow along with the exercises.

What you will be able to do

By the end of this course you will be able to:

• Explain in plain language what Claude Cowork is and what it produces
• Say which Claude plans include Cowork and which do not
• State the difference between Claude chat, Claude Code, and Cowork using Anthropic’s own framing
• Describe where a Cowork session actually runs and what it can reach
• Name the three connector approval modes and choose the right one for a task
• Set up a first Cowork task with a clear outcome and a connected folder
• Watch a task in progress and stop it when something looks wrong
• Review a Cowork deliverable against its cited sources before you use it
• Sort your own recurring work into tasks worth handing over and tasks to keep
• Name the current documented limits of the product
• Find the official documentation and check a claim yourself

Assessment

Each module ends with a short quiz. The course ends with a final assessment covering all three modules.

Passing grade: 70 percent. Retakes are unlimited. Your score is not published or shared. The quizzes exist so you can check your own understanding before you point an agentic tool at your real files.

Modules

1. What Cowork Is (3 lessons) – What the product does, who gets it and where it runs, and how Anthropic distinguishes it from Claude chat and Claude Code.
2. Working with Cowork (3 lessons) – The permission and safety model, setting up a first task, and watching, steering, and reviewing the result.
3. Making It Stick (3 lessons) – Task patterns for a small business, the documented limits, and keeping current as the product changes.

After this course

The next rung of the trust ladder is Delegate. Before you get there, the companion course at this level is Intro to Claude Code, which covers the agentic coding tool. Cowork and Claude Code are different products for different jobs. Taking both gives you the full picture of what supervised agentic work looks like.

Frequently asked questions

Do I need a paid Claude plan to take this course?

No. You can complete the course and the quizzes without one. To follow the hands-on exercises you need a paid plan, because Cowork is available on Pro, Max, Team, and Enterprise and is not included on the Free plan.

Is this the same as your Claude Code course?

No. They are different products for different jobs. Anthropic’s own framing is chat for conversational drafting, Claude Code for coding, and Cowork for cross-app knowledge work. If your business touches code or a website, take both. Start with whichever matches the work in front of you.

Do I need to know how to code?

No. Cowork is built for knowledge work rather than coding, and it has no terminal or IDE surface. If you work with documents, spreadsheets, and presentations, you are the audience.

Is it safe to let an AI tool touch my business files?

That question is Module 2, and the course answers it with Anthropic’s documented permission model rather than reassurance. You will learn where sessions run, what Cowork can and cannot reach, the three connector approval modes, and when to insist on manual approval.

Cowork is new. Will this course be out of date?

Parts of it will be, and the course says so. Every product claim was checked against Anthropic’s official documentation in August 2026, and where official pages disagree the course says nothing rather than guessing. Module 3 teaches you to check the current answer yourself in five minutes.

Where this course sits

This is a Supervise course, the third rung of the Nantucket AI trust ladder. Take What Is an AI Agent first if the word “agent” is still unfamiliar, and see Intro to Claude Code for the coding-side companion.