Guide · Claude Cowork

Claude Cowork isn’t a chat. It’s where we run the company.

A chat answers you and forgets. Claude Cowork works on your files, remembers the decisions, and runs the recurring tasks. This is the guide we actually work from: what it is, the method, the use cases, and the free kit to get started, even if you don’t write code.

Raffaele Zarrelli·Founder, Yempik·June 23, 2026·7 min read
In summary
  • Claude Cowork works on your files and processes, not just in words: it reads, writes, organizes, and runs tasks.
  • The leap isn’t a smarter AI, it’s giving it a memory: the context in tidy files it rereads and updates.
  • A single rule keeps it alive, the Memory Update: at the end of a task you write down what changed and where.
  • Start for free with cowork-os, the open-source kit that builds a custom workspace, even without technical skills.
  • It doesn’t solve everything: when a process has to run in production at high volume, you need real automation.
The starting point

What Claude Cowork is

Claude Cowork is the mode of Claude, Anthropic’s AI, built to work on your files and your processes instead of only in words. You open a project, give it access to a folder on your computer, and the assistant reads the documents, writes and organizes the files, runs tasks, and can even run on a schedule.

At Yempik we use it to run an entire company: marketing, decisions, the website, content. This page is the index of everything we’ve learned, the method, the real use cases, and the kit to get started, and from here you reach every deep dive.

In one line

A chat answers you; Claude Cowork works inside your material and remembers where you left off.

The difference

A chat forgets. A workspace remembers.

Generative AI is brilliant at the single question. In real work, though, it trips over continuity: what you decided in March has to count in June. Meanwhile the company’s context stays scattered, some in your head, some in ten chats, some in folders no one can find. It’s not a detail: according to McKinsey a person spends nearly a full workday a week just searching for internal information[1], and on average we switch applications more than a thousand times a day[2].

Throwaway chat

Starts from scratch

who are you?
let me explain again…
who are you?

Every conversation starts over: you re-explain everything each time.

Workspace with a memory

Remembers and grows

context/
decisions/
marketing/
linkedin/

The context stays in the files: every week the AI knows a little more.

Two ways to use the same AI: on the left you always start over, on the right you build a memory that compounds.

An AI with no memory isn’t a collaborator. It’s a very fast answer engine.

The method

Memory, structure, and a single rule

The method doesn’t depend on Cowork: it’s portable to any agentic tool. Three pieces hold it up. Memory: the context doesn’t live in your head or in ten chats, but in files the assistant rereads before it answers. Structure: everything has one place, a single source of truth. And a single rule, the Memory Update, which at the end of every task notes what changed and where, the one almost everyone skips.

The map
  • context/

    Who you are, how you position yourself, your tone of voice.

  • decisions/

    The decisions made and the questions still open.

  • marketing/

    Strategy, campaigns, and content ideas.

  • website/

    Notes and tasks for the site: copy, SEO, conversion.

  • reviews/

    The recurring reviews and the briefs already done.

Every piece of information has one place. When it’s needed, the AI knows exactly where to read and where to write.

The full method, with the Memory Update rule explained step by step, is in organizing work with Claude Cowork.

In practice

What you actually do

Three concrete uses to start with right away, and how they tie into the kit and our services.

Hub · pillarWorking with Claude Cowork
The cluster guides
Organizing workGiving the AI a memory
Data entryFrom documents to an Excel
Automatic routinesReviews, briefs, checks
From free to work done
cowork-os
The free kit: start on your own
Training
AI Adoption Sprint: with your team
Services
Real automation in production
One hub, three hands-on guides, and a path: the free kit opens the door, training brings it into the team, services put it in production.

Each card leads to the matching hands-on guide.

Apply it Monday

The free kit: cowork-os

You don’t have to build everything from scratch. We packaged our system and made it open source. It’s called cowork-os: the folder structure, the instructions, the automations, and a real, already filled-in example.

You copy it into your Cowork project, answer six questions, and Claude builds the workspace to fit you. Even if you’ve never written a line of code, and even if you’re starting from a messy folder.

How you start · ~30 min
1
You copy cowork-os
into your Cowork project
2
You answer 6 questions
in plain language
3
Claude generates the workspace
ready and tailored to you
What’s inside
context/
decisions/
marketing/
website/
reviews/

Plus the instructions, the routine automations, and a real, already filled-in example to copy from.

You don’t start from a blank page: you paste in a file, answer six questions, and the workspace builds itself.
With your team

Bringing it into the company, not just trying it out

The kit gets you started on your own. If instead you want the whole team to use it well, on your real process, not on an example, the path is training: you start from your flow and come out with something that runs on your data.

Want to learn it with your team?

In an AI Adoption Sprint we start from your process and come out with an automation that runs on your data, in 2–4 weeks.

Explore the course
The honest take

When a tidy workspace isn’t enough

This method organizes the work and makes it repeatable, but it doesn’t solve everything. When a process has to run in production, at high volume, connected to your back-office systems and without a person hitting “send” every time, you don’t need a tidy chat: you need real automation, built and put in place, with code you own and fixed timelines.

It’s the same principle as our method for AI agents: process first, then the tool. If you want the full picture, start from the pillar how to integrate AI agents into your company.

FAQs

The questions we get asked most

What is Claude Cowork?

Claude Cowork is the mode of Claude, Anthropic’s AI, built to work on your files and your processes instead of only in words. You open a project, give it access to a folder on your computer, and the assistant reads the documents, writes and organizes the files, runs tasks, and can even run on a schedule. The difference from a chat is simple: a chat answers you, Cowork works inside your material.

Is Claude Cowork a good fit for an SME or for non-technical people?

Yes. You don’t need to know how to code: you give it access to a folder, upload what you have, and give instructions in plain language. The value isn’t technical, it’s organizational: you keep the company’s context in one place and stop repeating yourself. With the free cowork-os kit, the workspace builds itself by answering a few questions.

What’s the difference between Claude Cowork and a regular AI chat or a chatbot?

A chat (or a chatbot) answers the single question and then forgets: every time you start over. Claude Cowork works inside your files, keeps a persistent memory of the context, and can run tasks, not just converse. For high-volume processes to put into production, though, even Cowork isn’t enough: you need real, custom-built automation.

Does my data stay private with Claude Cowork?

The workspace lives in your folder, on your computer. Our public cowork-os kit contains only the folder structure and sanitized examples, with no real client data. For the company-wide rules on using AI, from risk classification to policy, see our guide on AI governance for SMEs.

How do I get started? What is cowork-os?

cowork-os is the free, open-source kit in which we package our method: the folder structure, the instructions, the automations, and a real, already filled-in example. You copy it into your Cowork project, answer six questions, and Claude builds the workspace to fit you, even starting from a messy folder. The first setup takes about thirty minutes.

Transparency note

I wrote this guide myself. The method and the opinions are the result of my work and of the way we use Claude Cowork every day at Yempik. For the writing, I had Claude help me with editing, clarity, and layout. The substance is mine; the tool shown is Claude Cowork and it’s disclosed.

Transparency

Sources

  1. [1]McKinsey & Company: knowledge workers spend about a fifth of the week, nearly a full day, searching for internal information. www.mckinsey.com
  2. [2]Harvard Business Review: on average we switch from one application to another about 1,200 times a day. hbr.org