Claude Cowork 4 Everyone

🎯 What is Claude Cowork?

You've probably used a chatbot before. ChatGPT, the regular Claude in your browser. You walk up, ask a question, get an answer, close the tab. Useful. But limited. It's like stopping a knowledgeable stranger on the street — smart, helpful, gone the moment you walk away.

Cowork is different. Cowork is an intern at the desk next to yours. You give that intern a folder; they read everything in it. You give them a task; they work on it across multiple steps. You come back tomorrow; they remember you. Same brain as chatbot Claude. Completely different setup.

The three tabs. When you open the Claude desktop app, you'll see three at the top:

• Chat — the knowledgeable stranger. Quick questions, no memory.
• Cowork — the intern with your folder. What we're setting up today.
• Code — for software developers. Ignore for now.

By the end of the next four sections you'll have:

• Claude installed with access to a folder you control
• An assistant that remembers you between conversations
• A daily 9am check-in to build the habit

🔌 Install Claude Cowork

Four steps. Do them in order — the order matters.

📁 Assign Your Folder

This is the most important step in the course. The thing that makes Cowork actually Cowork instead of just another chatbot.

Why a folder? Chat assistants forget you the moment you close the window. Cowork remembers — but only because it can read and write files in a folder on your computer that you give it access to. Your assistant's memory lives on your hard drive, in plain English files you can open in Word or read on your phone.

How to assign one:

1. In the Cowork tab, look for Settings (gear icon, or under a menu in the corner).
2. Find the option called Workspace folder (sometimes "Working folder" or just "Folder").
3. Click it. You'll get a normal Mac or Windows folder-picker.
4. Recommended: Don't pick an existing folder full of work documents. Make a fresh folder on your Desktop called My Claude. Easy to find, doesn't mix with anything else.

Confirm it worked. Drop a Word doc or text file into the folder you just picked. Then go to Cowork and type:

If Claude reads it back, you're connected. If not, double-check the folder you picked is actually the one Cowork is using.

🎤 Bootstrap Your Cowork

Now you paste one prompt into your Cowork and it sets itself up. Names itself, interviews you, saves your context, and schedules a daily check-in.

Hi! I'm setting you up as my work assistant for the first time. I'm not particularly technical. I use Word and Excel for most things. Please go slowly, ask me one thing at a time, and use plain language. No jargon.

I'd like you to set yourself up by walking me through four things today. You'll know we're done when you confirm all of them.

PART A — Name and personality

First: ask me what I want to call you. That's your name from now on.

Then ask me three short questions, one at a time:
  1. Should you talk to me formally or casually?
  2. What language should we work in? (English by default — or tell me another language if you'd prefer that)
  3. Should you be friendly and conversational, or efficient and brief?

Write down my answers.

PART B — Interview me, one question at a time

  1. What's your name?
  2. What do you do for work? Tell me in plain words.
  3. What does a typical week look like for you?
  4. What tools do you use day-to-day? (Word, Excel, Outlook, Gmail, Google Drive, anything else)
  5. Are there a few different projects or clients you work with, or mostly one thing?
  6. What kind of tasks take up most of your time?
  7. What's the most annoying or repetitive thing in your work — the thing you'd hand off first if you could?
  8. Anything else I should know about you to be useful?

PART C — Set up my files

  - Save my answers from Part A and Part B to a file called `about-me.md` at the top of my workspace folder. Use clear headings. From now on, every time we work together, read that file first so you remember me.
  - Create a file called `todo.md` at the top of my workspace folder with exactly these starter tasks:

      # My To-Do List

      Each item:
      - [ ] **Task title** — what it is
        - Status: open
        - Last asked: never
        - Notes:

      ---

      ## Starter tasks (we'll work through these together)

      - [ ] **Show me a real piece of your work** — Drop a document, email, screenshot, or note into your workspace folder and tell me what you need help with. This is your first real win.
        - Status: open
        - Last asked: never
        - Notes:

      - [ ] **Connect your email** — In Cowork settings → Connectors, sign in with Outlook or Gmail. This lets me read your inbox to help with replies.
        - Status: open
        - Last asked: never
        - Notes:

      - [ ] **Tell me about one specific recurring task** — Pick one thing you do every week and walk me through it. I'll see if I can make it faster.
        - Status: open
        - Last asked: never
        - Notes:

      - [ ] **Add a second project folder, if you have multiple projects/clients** — Tell me you want to add another, and I'll create the folder structure with you.
        - Status: open
        - Last asked: never
        - Notes:

PART D — Schedule the daily check-in

Create a scheduled task that runs every weekday at 9:00am in my local timezone with this prompt:

---SCHEDULED PROMPT START---

Good morning. Read `about-me.md` and `todo.md` first.

Find the OLDEST task with status "open". Just one. Ignore the rest.

Greet me warmly in one short sentence using my name. Then ask me about that one task in plain language. Something like:

"Morning! Today's open one is still: [task title]. Where are you with it — done, not yet, or want to skip it?"

Wait for my reply. Then:

- If I say "done" or anything that confirms completion → mark it [x], change status to "done", note today's date, write one warm one-liner, and tell me what's next on the list (preview only — don't ask about it yet).

- If I say "not yet" or describe a blocker → keep it open, update "Last asked" to today, save any notes to "Notes:". Ask if I want to do it RIGHT NOW with your help. If yes, walk me through it. If no, say "ok, tomorrow then" and end.

- If I say "skip" or "not relevant" → change status to "skipped", note the date, ask briefly why (one short question), save my answer to Notes. Don't argue.

- If my reply is vague, ask ONE clarifying question. Don't pile on.

Tone: warm, brief, peer-level. Like a calm friend who knows I'm capable.

End every check-in by saving the updated `todo.md`. That's how tomorrow's check-in knows where we are.

---SCHEDULED PROMPT END---

After all four parts are done, confirm out loud:
  1. The name I picked for you
  2. The three personality settings
  3. That `about-me.md` and `todo.md` both exist in my folder
  4. That the daily 9am check-in is scheduled

Then say: "We're set up. Your first real task is in your to-do list — drop something you're working on into the folder and tell me what you need. I'll be here. And I'll see you tomorrow morning at 9 for our first check-in."

Start now with Part A — ask me what to call you.

🧠 How It Remembers

Here's the strangest thing about your new assistant: it has perfect memory inside a single conversation, and goldfish-level memory between them. Close the chat, open a new one — gone. Doesn't remember yesterday. Doesn't remember the article you mentioned last week.

I know. Sounds bad. There's a fix, and it's elegant.

Your folder is your assistant's filing cabinet. Anything you want it to remember tomorrow, save it in there today.

You've already seen this in action. The about-me.md file you made during bootstrap is the intern's onboarding manual — the first thing they read every morning, so they know who you are. The todo.md file holds your tasks across days. The folder holds everything.

The rule: anything important, ask your assistant to save it. A draft you liked? Save it. A useful insight? Save it. The conversation is just where you talk; the folder is the memory.

Smaller version of this inside a single conversation: if a chat goes really long — hundreds of messages — your assistant starts to forget things from earlier. Same fix. Start a new chat, let it re-read the folder, you're back to full strength.

Mental shift: stop asking "what does my assistant remember?" Start asking "what's in the filing cabinet?" That's the answer.

🗣️ Talking to Your Assistant

This is the highest-leverage chapter in the course. Master these habits and your assistant goes from "okay" to genuinely useful.

Here's the pattern most people fall into. They open the chat and bark: "write me an email about X." Hit send. Get something mediocre. Blame the AI. It's like handing a new hire one sentence and expecting them to read your mind.

The fix isn't more AI knowledge. It's three small habits.

Habit 1 — Brief before you ask. Don't lead with the request — lead with the situation. Imagine briefing a competent freelancer who's never met your business. Bad: "write a pitch." Good: "I work at a credit ratings company. We just released a survey on small business credit trends. I need to pitch fintech journalists. Tone should feel newsworthy, not promotional. Now draft me a pitch." Same task. Dramatically better output.

Habit 2 — Make your assistant interview YOU first. Counterintuitive but it works beautifully. Just say: "Before you draft anything, ask me three questions to make sure you understand." Answering its questions forces both of you to anchor on what actually matters. What you get after the interview is twice as good as cold output.

Habit 3 — Push back. If a draft feels too corporate, tell it: "Shorter. Less formal. Punchier opening. Drop the second paragraph." It's not insulted. It just redoes it. The more you correct, the more it learns your voice. Treat it like a junior who's eager to please, not a sacred oracle.

A few other moves that consistently work:

• Ask for two options instead of one. Comparing is easier than judging.
• Tell it to act as a specific role. "Act as an editor and tighten this paragraph."
• Tell it what good looks like. "A good pitch is under 120 words, has a number in the subject line, ends with one clear ask." Now it has something to hit.
• When it nails something — tell it. "That's the right tone. Remember this voice for future work." Feedback compounds.

If you take one line from this whole course: a sharp prompt beats a smart AI. Your job is to set it up well.

🔗 Connectors

A connector is a wire between your assistant and an app you already use. Outlook, Gmail, Google Drive, Slack, your calendar. Plug it in once, and your assistant can reach into that app the way you'd reach into a file cabinet.

To plug one in: Cowork settings → Connectors. You'll see what's available. Click sign-in for the app you want. Microsoft for Outlook. Google for Gmail, Drive, Calendar.

Important default: most connectors start in read-only mode. Your assistant can see your inbox but can't send anything on its own. Drafts go to your folder. You send manually. Think of it as letting your intern read your email pile but not letting them mail anything out yet. Once you trust them, you can expand permissions.

Real example. You sent five pitches this week. Three days later you want to know who replied. You ask:

Your assistant runs the search, lists the replies, asks which one to draft a response to first. You pick. It drafts in your folder. You polish and send.

That's the loop. Read in the connected app. Draft in your folder. Send manually. Clean and safe.

Don't plug everything in on day one. Each connector adds capability, but also adds surface area for things to go sideways. Add one at a time, when you have a real reason. Start with where you spend most of your time — for most people, that's email.

✅ Trust, but Verify

Your assistant is right most of the time. Not always. Sometimes it'll be confidently wrong about a fact — a statistic, a date, a person's title, a contact detail. It's not common. But it happens.

The technical word is "hallucination." The practical word: confident-but-wrong. Picture the eager intern who really wants to be useful, and occasionally invents a small detail rather than admit they don't know. Same energy. Same fix.

Two categories to hold in your head. Facts that came FROM you — your files, your notes, what you typed into this chat — your assistant handles those fine. They're right in front of it. But facts that come from its own knowledge — a stat it remembers, a person it claims to know — those are drafts until you've eyeballed them.

The trick is making your assistant tell you which is which.

That last one is the most important. A good assistant says "I don't know." A weak one fakes it. You can train yours toward the former just by asking for it.

One more move worth knowing: "Tell me what you're least sure about in this draft." Your assistant will literally point at its own weak spots. Use it before sending anything important.

For any fact you'll act on — a contact email, a quote you'll attribute, a date in a doc — verify it. Same instinct you already use when reading a junior's draft before it goes out the door. New junior. Same instinct.

🚫 What It Can't Do

A short, honest list, so you can plan around them.

• Can't make phone calls. Voice is still your job.
• Can't predict the future. It can draft, analyze, suggest — but it can't tell you whether a pitch will land, whether a strategy will work, whether a person will say yes. That's the unknowable part. Same as any junior — they can prep the deck, they can't make the client buy.
• Can't replace your judgment. It drafts; you decide. The same way a senior overrides a junior's work, you override your assistant's. That dynamic is normal, healthy, and the right way to work.
• Doesn't know recent events. Its knowledge has a cutoff date — it stopped reading the news at some point. For anything time-sensitive, check or paste the latest source in.
• Can be wrong about facts (see the previous chapter). Verify what matters.

The frame to leave with: a great assistant makes you faster. The best ones also know when to step back. Yours is trying to be the second kind. So if it ever tries to make a call that's really yours — push back. Make it work on the inputs, not the verdicts. That's the divide between a useful intern and a presumptuous one.

☕ Building the Habit

Most people who try AI tools quit in week two. Not because the tools are bad — because they never built the habit. The tools sit there gathering dust like a gym membership in February. Don't be that person.

Here's the daily rhythm that works.

Morning, five minutes. Open Cowork. Name your project, if you have more than one: "I'm working on this project today." Ask your assistant to walk you through yesterday's tracker — what moved, what's left. Set today's top three priorities.

End of day, two minutes. Log what happened. Who replied. What got finished. What's owed tomorrow.

That's the whole ritual. Five minutes in, two minutes out. The daily 9am check-in you set up earlier will start the morning ritual for you. Don't skip it.

Two weeks of this and something shifts. Your assistant stops being a tool you use and becomes a colleague you work with. You'll know it's working when two things happen: first, you stop having that "wait, what was I doing again?" moment in the middle of the day — the tracker holds it for you. Second, you stop thinking about it as "AI" at all. You just ask your assistant for the thing and keep moving.

That's the goal. The course ends here. Your work begins.