You just switched from ChatGPT to Claude. But you’re still not using Cowork.

When Claude released it, software stocks lost $830 billion in 6 days because of it.

And since I published my guide ‘Claude’ (my most-shared newsletter I’ve ever written, over +2,000 shares), people keep asking me:

“I installed Claude. But how do I actually start using Claude Cowork?”

If you don’t code, you must master Claude Cowork now.

First, you might be lost between different Claude modes. So here’s a quick recap:

And that’s not it.

Here’s the entire Claude product line, in short:

1. Claude “Chat” → it’s like ChatGPT. Probably the only one you know.

2. Claude “Project” → it’s still Claude Chat, but separated as individual Projects.

3. Claude “Code” → massive revolution for developers to code (much) faster.

4. Claude “Cowork” → like Claude Code but for us, knowledge workers.

5. Claude “Skills” → teach Claude repeatable workflows. Like better Projects.

6. Claude “Connectors” → plug Claude directly into apps like Slack, Google Calendar, Gmail, etc. It reads, writes, and acts inside the tools you already use.

7. Claude “Plugins” → like Connectors, but you do it (you don’t want that). Connectors = download from the App Store, Plugins = you upload your app.

This newsletter is the full playbook of Cowork. How to set it up. How I use it every single week to write this newsletter, deliver consulting work, and research faster than I ever could alone. Also, where it falls short (I keep promising honesty).

Two things before we start:

1. Save this guide and spend 30 minutes this weekend to explore Cowork.

2. Send it to anyone who still hasn’t tried Cowork (or Claude).

Share

You must forget about prompts.

You read my Claude guide. So you installed it correctly.

Just a quick reminder for those who didn’t read it:

1. Go to claude.com/download. Download the app.

2. You need a Pro account ($20/month). It’s very much worth it.

3. Open the app. Click on the Cowork tab at the top between Chat & Code.

4. Select a folder from your computer. More about it right after this set up.

5. Make sure to always select “Opus 4.6” and “Extended thinking”.

ChatGPT trained you to write better prompts. Longer prompts. Forget that.

With Cowork, the game is text files. Take everything you know (your writing style, your company’s rules, your best examples, your past work) and put it in text files. Drop them in a folder. Point Claude to that folder.

It’s kind of like having an SOP for an employee. But here, Claude is your employee. And the SOP is your “Claude Cowork” folder.

The more context you give it as files, the less prompting you need. The output goes from “generic AI” to “this actually sounds like a full-time employee.”

Here’s how to create your folder:

Step 1: Build your folder.

Create a dedicated folder for Cowork on your computer.

It needs a clean, intentional space. Here’s mine:

ABOUT ME → a folder with 1/ “about me”, 2/ “anti AI writing style”.

PROJECTS → live work. The projects you're building right now. The brief, the drafts, the reference material for that specific job. one subfolder per project.

TEMPLATES → finished work so good you want to reuse it as a pattern. It’s not the content itself, but the “perfect” structure, as templates.

CLAUDE OUTPUTS → where Claude Cowork delivers finished work for you.

This keeps things organized and limits what Claude can see. Cowork has real read/write access to whatever folder you share. If something goes wrong, you want the damage contained. You must keep it tight.

Now here’s how to create your own core files:

Step 2: Create your core files.

You now have 4 folders within your mega “Claude Cowork” folder.

Let’s create the “About me” files now:

File 1: about-me.md — Who you are. What you do. Your current priorities. What matters to you right now. I wrote an entire newsletter on how to do it.

File 2: anti-ai-writing-style.md — Because I hate the AI style. So I made a text file on how to NEVER write like an AI. I share how to download it below.

Download my 3x md files by subscribing to my newsletter for free. Do not pay for anything. Open my welcome email & you will access my Notion doc with md files.

If you are free subscribers already, and can’t find this email, leave a comment.

Quick note: a markdown file is just a plain text file with an .md extension. Open any text editor (even a Google Doc works), write your content, and save it as about-me.md instead of about-me.txt. Claude reads them better.

One great markdown file is worth more than 50 random uploads. Don’t dump everything into the folder. Be intentional about what you include.

Step 3: Set Global Instructions.

Go to: Settings → Cowork → Edit Global Instructions.

Global Instructions handle how Claude must (always) behave.

Paste this:

# GLOBAL INSTRUCTIONS

## BEFORE EVERY TASK

1. Read `ABOUT ME/`. No task starts without reading both.
2. If the task relates to a project, read everything in the matching `PROJECTS/` subfolder before proceeding.
3. If the task involves a content type that has a matching pattern in `TEMPLATES/`, study that template's structure first. Use the structure. Don't copy the content.

## FOLDER PROTOCOL

You have three read-only folders and one write folder.

### Read-only — never create, edit, or delete anything here:
- `ABOUT ME/` → My identity and writing rules.
- `TEMPLATES/` → Proven structures to reuse as patterns.
- `PROJECTS/` → My briefs, references, and finished work organized by project.

### Write folder — the only place you deliver work:
- `CLAUDE OUTPUTS/` → Everything you create goes here. Organize with one subfolder per project, mirroring the structure of `PROJECTS/`. Create the subfolder if it doesn't exist yet.

## NAMING CONVENTION

All files you create must follow this format:

`project_content-type_v1.ext`

Content types: Newsletter, LinkedIn Post, Brief, Deck, Report.

Examples:
- `How-To-AI_Newsletter_v1.md`
- `EasyGen_Deck_v1.pptx`
- `GPC_Report_v2.docx`

Increment the version number if a file with the same name already exists.

## OPERATING RULES

- If the brief is unclear or incomplete, use the `AskUserQuestion` tool. Don't fill gaps with generic filler.
- Don't over-explain. Deliver the work. Save the commentary unless I ask for it.
- Never delete files anywhere.

You set this once. It runs every time. You never type it again.

Combined with your context files, this means your prompts can be 10 words long and still produce work that sounds like you. With the same level of quality.

Here’s a prompt example:

I want to write a cold DM on Linkedin to a Chief of Staff who needs an AI workshop.

I will DM her on Linkedin, and I need to have a clear strategy to make her open the message and book a call. 

Start by using AskUserQuestion. Develop 5x completely different strategies.

This level of output is literally what a good employee would do.

Now extrapolate to legal work, creative agencies, academic research.

You might think it was “too simple of a benchmark”.

Alright, let’s do another example. A heavy Excel file together:

I want to create a spreadsheet that maps out a potential exit for my socials (if I could sell them). Start by using AskUserQuestion to be up to date with my numbers.

Then and only then, create a spreadsheet with Wall Street financial model style.

If you’re curious about the one-shot results (I didn’t change anything from its first answer), go here: https://docs.google.com/spreadsheets/d/1WkUTtE41zdHT616a2zLxLtoC79Zvv_wZDLJGi5tr_Fk/edit?usp=sharing.

Don’t prompt Claude. Let it prompt you.

This is the feature that changed how I work. And I haven’t seen a single Cowork guide explain it properly.

Always add this to your prompt:

Start by using the AskUserQuestion tool before answering to gather enough context.

When you do this, Cowork generates an interactive form. Actual buttons. Clickable options. Multi-select choices. Rankings you can drag and reorder.

AI is finally prompt you.

This is called AskUserQuestion. It’s a tool built into Cowork. Claude forces you to be clear. It asks the right questions so it can give you the right output.

The one prompt I use for everything.

80% of my chats with Cowork starts like this:

I want to [TASK] to [SUCCESS CRITERIA].

First, explore my CLAUDE COWORK folder. Then, ask me questions using the AskUserQuestion tool. I want to refine the approach with you before you execute.

What happens: Claude reads your context files, generates a clickable form asking about your audience, your goals, your preferences. You click through in under a minute. Claude shows a plan. You approve. It executes — creating real files in your folder. If something’s off midway, you interrupt. Claude recalibrates with a new form. Picks up where it left off.

The entire process feels like directing someone smart instead of wrestling with a text box. I’m obsessed with this feature.

(Yes, I stopped writing prompts. My prompt folder is collecting dust.)

I even added a “Text Replacement” on my Mac so that whenever I am typing the command “/prompt” it pastes this one prompt:

If you use a Mac, search for “Text replacement”. Click on the “+” and simply add a command (I like /prompt) and paste my prompt template.

How to use Plugins inside Cowork.

Remember that $830 billion stock crash I opened with?

Anthropic released 11 official plugins. Sales. Marketing. Legal. Finance. Data analysis. Product management. Customer support. Each one gives Claude specific skills, workflows, and slash commands for that function.

Legal software lost so much value because Claude can do the work. And the plugins sitting in your Cowork sidebar are a big part of what triggered that reaction.

You don’t need to be technical to install it. I promise.

How to install:

1. Open Cowork.

2. Click Customize in the left sidebar → Browse plugins.

3. Pick one that matches your work. Install.

4. Type / in the chat to see available slash commands.

What this looks like in practice:

Marketing plugin

The prompt: /marketing:draft-content → “Write a LinkedIn post about [topic]. Use my voice profile. Target [audience].”

What happens: Claude reads your about-me.md, drafts a post that actually sounds like you, and suggests hook variations. You pick one, edit, post. Five minutes instead of thirty.

Data plugin

The prompt: /data:explore → Drop a CSV into the folder.

What happens: Claude summarizes every column, flags anomalies, suggests analyses, and can build an interactive dashboard. It writes SQL in plain English. You don’t touch a formula.

Legal plugin

The prompt: “Review the NDA in this folder. Flag anything unusual or one-sided.”

What happens: Claude reads the contract, highlights risky clauses, explains each one in plain English, and suggests alternative language. This is what wiped $285 billion off the stock market.

Connectors: Claude, inside your tools.

Cowork can also connect to your existing tools. Slack, Google Drive, Notion, Figma, and 50+ others. They’re called Connectors.

Go to Settings → Connectors. Browse the directory. Click “Add.” Done.

Once connected, it reads your tools. Claude can search your Slack messages, pull from your Google Docs, or reference your Notion pages mid-conversation.

Here’s an example with Gamma.

Oh, and I forgot - connectors are free.

How I actually use Cowork every day.

I don’t use anything else these days.

Use case 1: Writing this newsletter.

The setup: My folder has my about-me.md, my anti-AI-writing-guide.md, past newsletters that performed well, reference guides from other creators, and official documentation from companies (like OpenAI/Anthropic/Google).

The prompt:

I want to write my next newsletter on using Gemini to grow on Linkedin with infographics using the new Nano Banana 2.

First, explore my CLAUDE COWORK folder. Then, ask me questions using the AskUserQuestion tool. I want to refine the approach with you before you execute.

What happens: Claude reads every file. Generates a form asking about my audience, my tone, what length, what angles the other guides missed. I click answers. It produces an outline. I push back on weak sections. It adjusts. Then it writes — and because it has my voice profile and anti-AI writing guide, the output actually sounds like me.*

I edit. But the heavy lifting is done. This newsletter you’re reading right now was outlined by Cowork and written by me. But you don’t care and you keep reading :)

*look, he used an em-dash, it must be AI!!! no i did it on purpose.

Use case 2: Consulting deliverables.

The setup: Client sends a brief. I drop it in the folder next to my templates and past deliverables.

The prompt:

A client just sent a brief for a 2026 AI adoption strategy. The brief is in /projects/client-x/.

Read the brief, my deliverable template, and my past examples. 
Create a first draft as a .docx. Ask me questions first (AskUserQuestion).

What happens: Claude reads the brief. Compares it to my template format. Then it asks me things I didn't think of — "Should this include a timeline or just recommendations?" and "Do you want competitor examples or keep it internal?" I click answers. It creates a .docx file directly in my folder.

Use case 3: Research and competitive analysis.

The setup: I drop 3-5 competitor articles or reports into a subfolder.

The prompt:

I uploaded 4 newsletters from other creators covering Claude Cowork. Read all of them.
Create a comparison table: what each one covered, what they missed, and where I can be the only one saying something new. Ask me questions first.

What happens: That used to be a junior job in my company. Now it’s a prompt.

Use case 4: Automated weekly briefing.

This one is different. Cowork works without you even being there.

The setup: You create a folder called /weekly-briefings/.

The prompt (with /schedule, it’s a plugin):

Every Monday at 7am, research [competitor names] for news, product updates, or pricing changes. Save a summary to /weekly-briefings/ as a markdown file. Only include items from the past 7 days.

What happens: Cowork runs automatically every Monday, as long as your computer is awake and the app is open. You wake up to a briefing doc ready to read. That’s the endgame.

Across all four use cases, the pattern is the same. I never write a long prompt. I write a short task, point to my folder, and say “ask me questions.”

The workflow is always the same. Only the context changes.

Where Cowork falls short (I promised honesty).

It eats your usage fast. A single Cowork session can burn through what would normally be dozens of regular chat conversations. On the Pro plan ($20/month), you’ll feel it within a week if you use it daily. If Cowork becomes your main workflow, consider Max ($100/month). I’m being direct about this because I don’t want you surprised.

It’s still a research preview. Anthropic says this explicitly. It can make mistakes. It can misread files. It sometimes takes an odd approach to a task when a simpler one would work. You need to review what it produces. Don’t send a client deliverable without reading it first.

It needs the app open. Close the Claude Desktop app, and the session dies. There’s no mobile version. No web version. Cowork only runs inside the desktop app on macOS or Windows.

It’s not for quick questions. If you want to ask “what’s the capital of France,” use Chat. Cowork is for tasks, not trivia. It’s designed for multi-step work.

Agents can be hit or miss. For complex tasks, Cowork breaks the work into parts and runs them in parallel using multiple agents. Most of the time, it’s fast and accurate. Sometimes one agent goes in a weird direction, and the final output has a section that doesn’t match the rest. Keep an eye on it. Happens 10% of the time.

Cowork is not the best at everything. But it’s getting better every week.

And if this is Claude Opus 4.6 + Cowork, I can’t imagine Claude 5, Claude 6…

Your first 30 minutes with Cowork.

Open your calendar. Book 30 minutes with yourself, this newsletter, and Claude.

Minutes 0–5: Install and open Cowork.

→ Go to claude.com/download. Download the desktop app.

→ Sign in with your Pro account ($20/month. Or $17/month if you pay annually).

→ Open the app. Click the Cowork tab at the top.

→ You’re in.

Minutes 5–10: Build your context folder.

→ Create a folder on your computer called “Claude-Cowork.” Inside it, create four subfolders: ABOUT ME, TEMPLATES, PROJECTS, CLAUDE OUTPUTS.

→ In the context folder, create your first file: about-me.md. Write three things: (1) What you do for work. (2) How you like to communicate. (3) One example of writing you’re proud of. Paste it in.

→ No time to write it yourself? Skip to minute 10 and let Cowork create it for you.

Pro tip: Instead of typing, use Wispr Flow to talk. It’s 4x faster.

Minutes 10–15: Start your first Cowork conversation.

→ In Cowork, click Add Folder and select your Claude-Cowork folder.

→ Make sure to select Opus 4.6 + Extended thinking for the smartest AI.

→ Type: “I want [task] for [success criteria]. Go through my folder first, and use AskUserQuestion tool so you gather enough content before executing.”

→ Watch what happens. A form appears. Click answers. Let it create your context files.

Minutes 15–20: Install a plugin.

→ Click Customize in the sidebar → Browse plugins.

→ Pick one that fits your work. Good starters: Productivity (tasks and workflows), Marketing (content drafting), or Sales (prospect research).

→ Click to install. Start a new conversation with the plugin active.

→ Type / to see what slash commands are available. Try one.

Minutes 20-30: Create a tough deliverable.

→ Give Cowork a task you’d actually need for work this week.

→ Type: “Create a [report / deck / document] based on the files in this folder. Ask me questions first.”

→ Watch it create a real file. Open it. Edit it. Use it.

→ Optional: be amazed.

I don’t care about Claude.

I don’t care about Claude, ChatGPT, Grok, Gemini, or any other model.

I don’t pick sides. I’m not paid to make this newsletter.

I’m sharing, twice a week, how my work is transforming (very fast) with AI.

As I’m trying to keep up, I want you to keep up. So we move just as fast.

I want to be the greatest filter to the AI noise. And 338,000+ people read this twice a week to focus on the How. Some came because of my LinkedIn. But most readers subscribed because someone they trusted sent one of my articles to them.

If this article helped you, be that person for someone else (and share it):

Share

It does not cost you anything to share. And sharing is caring :)

If someone did send you this, thank them and subscribe for free here:

archive of my past blogs: https://docs.google.com/document/d/1pWuMCBVQo1zKcgKltX_BZxAr31KgxmOlp3Vzvmc5Hxc/edit?usp=sharing