Sponsored by
Hi, there!
Welcome to the 24th edition of Work in Beta.
In this edition, we we settle a question we get constantly: Claude Chat or Claude Cowork - which one should you use, and how do you tell the difference for the work in front of you?
Our free AI Starter Kits are the fastest way to go from "I should use AI more" to actually using it. Pick a tool, open the first lesson, and you'll have something working in 10 minutes. Start one here.
So let's dive in!
IF YOU ONLY HAVE 2 MINUTES
Image Credits: ChatGPT / Work in Beta
THE ‘HOW TO’ PLAYBOOK
Claude Chat or Cowork? How to Tell Which One Your Work Needs
We get this question a lot. You sit down to do something in Claude, and you pause: Chat or Cowork? You know both exist. You might even have a Claude Project set up. But faced with a real task, you can't tell which one it belongs in.
Here's a scene you'll recognize. You're building the monthly report. You open Chat, upload the spreadsheet, then last month's report, then the deck. Claude summarizes the numbers, drafts the sections, cleans up the wording - all of it well. But you're the one carrying each file in, copying each answer out, and pasting it back into the real document. Next month, you do the whole thing over again.
Notice what happened. Claude didn't get anything wrong. It answered everything you asked. The problem wasn't the thinking - it's that you spent an hour being the courier between Claude and your files. That's the signal you're on the wrong surface.
So the real question was never "Chat or Cowork?" It's simpler than that: are you asking Claude a question, or asking it to do a job across your files? Most people reach for Chat either way - it's the familiar one - and then do all the file-shuffling themselves.
The One Question That Decides It
Forget the tool names for a second. Ask this instead:
When Claude gives you the answer, are you finished or is that where the work begins?
Sometimes the answer is the whole job. "Draft a reply to this email." "Summarize this meeting." "What's weak about this plan?" Claude responds, you read it, you're done. The answer was the finish line. That's Chat.
Other times, the answer is only step one. Take the monthly report again. Claude can tell you what changed - but you still have to pull the new numbers from the spreadsheet, rewrite each section, refresh the charts, and save the file in the right folder. That pulling, rewriting, and saving across a stack of files is the actual work. That's Cowork.
That's the whole test in one line: if reading Claude's reply ends the task, use Chat. If the real work starts after the reply - gathering, editing, and saving across files - use Cowork.
Here's the same test run against real tasks:
What you're doing | Surface | Why |
|---|---|---|
Draft a reply to this email | Chat | The answer is the deliverable |
Summarize one meeting | Chat | One input, one answer |
Review 7 client docs and build a proposal folder | Cowork | The deliverable is built across files |
Scan this month's meetings and update the account brief | Cowork | Many inputs, a saved file as the output |
One more thing: Chat caps you at around 20 files per conversation - fine for a quick task, useless for a folder of 60. Cowork doesn't deal in uploads at all. You point it at a folder and it reads what's inside, in place - no attaching files one at a time, no ceiling to hit. (New to Cowork? We covered what it can actually do in Edition 8.)
Where a Claude Project Comes In
Sometimes the right answer is neither - it's a claude.ai Project. Reach for one when the same kind of work keeps coming back: the same client, the same brand voice, the same monthly report. A Project keeps your instructions and key files loaded, so you're not re-pasting the background every time you start. It's the middle layer - still a conversation like Chat, but one that remembers your context so each new chat starts warm.
Why This Works
The signal was never how hard or important the task is - it's how much of it lives in files. It's the same line we drew in CLAUDE.md and AGENTS.md Are Everywhere: when AI is answering your questions, a chat is enough; when it's doing work inside your files, you're in different territory. That's what Cowork was built for: it takes a multi-step job and carries it through to finished files - reading your folders, editing documents, saving the results - instead of answering one message at a time. The moment your work is mostly moving and assembling files, that's the tool for it.
The Mistakes We See People Make
Reaching for Cowork because the task feels big. A board presentation feels important, so you grab the "serious" tool. But if what you need is help thinking through what to say, that's pure thinking - it belongs in Chat. How big the task feels isn't the signal. Whether there are files to gather and build is.
Staying in Chat because the first answer was great. Claude's reply impresses you, so you keep going and then spend the next half hour copying, reformatting, and saving by hand. The answer was good. The workflow was still broken. A strong first answer doesn't mean you're on the right surface.
Loading a Project with files and assuming Claude reads them all. It doesn't. Once a Project holds a lot of files, Claude starts searching them and answers from the few pieces it finds most relevant - so it can miss something sitting right there. When every file needs to actually be read, that's Cowork's job.
Treating Cowork like the team's shared room. It lives on your computer - there's no version a teammate can open. When the whole team needs the same context, that's a claude.ai Project. Cowork is your private desk, not the group's room.
Final Thought
Most people aren't choosing the wrong tool. They're choosing by its name instead of by the work in front of them. So look at the task, not the label: do you want an answer, or do you want a finished thing made out of your files? Name the work first, and the surface is obvious.
WORK WITH US
Build With Us
Most professionals know AI can do more for them. The gap isn't awareness - it's knowing where to start, what to change, and how to make it stick.
That's what we work on through Work in Beta.
For individuals, we run working sessions - not teaching sessions, not agency engagements. You bring a real work problem and we figure out the AI solution together. You build, we guide. Think of it as borrowing our learning curve instead of building your own from scratch. You walk out with something working and the skills to keep going. When you get stuck, we're a message away.
For organisations, AI adoption is a people problem, not a technology problem. Your teams have the tools - what's missing is the translation layer between AI capability and daily work. Which processes to redesign, which habits to break, how to build genuine fluency - not just awareness. We help close that gap through hands-on training, process redesign, and deep adoption engagements. Not advisory, forward-deployed.
If any of this resonates, email us at [email protected] / [email protected] and we will figure out how to work together.