In partnership with
TL;DR:
What it is: Cowork is Anthropic's new desktop AI assistant. You give it access to a folder on your Mac, tell it what you need, and it gets to work—organizing messy files, turning receipt screenshots into expense spreadsheets, building presentations from scattered notes, or drafting reports from multiple documents. You don't babysit it. You check back when it's done.
What's really interesting: Anthropic's engineers built the whole thing in about a week and a half by prompting their existing AI tool, Claude Code, to write most of it.
Why try it: If you spend hours on repetitive file work—renaming downloads, pulling together decks, organizing client folders—Cowork handles that in the background while you do something else. It's Mac only for now and a way to see what "delegating to AI" actually feels like in practice.
But here's the catch: The tools are ready. Most organizations aren't. More on that below.
INTRODUCTION
💡 What Cowork Actually Does
If you've used ChatGPT or Claude over the past two years, you know the drill: type something, AI responds, you refine, it responds again. Useful, but it requires your constant attention. You're driving.
Cowork works differently. You point it at a folder on your Mac, tell it what you need, and walk away. It makes a plan, executes the steps, and checks back in when it hits a decision point or finishes. Anthropic describes the experience as "less like a back-and-forth and more like leaving messages for a coworker."
Claude Cowork
Here's what it can handle:
File organization — Drop a messy Downloads folder on it and ask it to sort by type, date, or project. It'll create the folder structure and move everything into place.
Document creation — Give it a collection of notes, meeting transcripts, or research and ask for a summary report, presentation outline, or client brief. It reads through everything and produces a draft.
Data transformation — Screenshot a bunch of receipts? It can extract the info and build an expense spreadsheet. Have a PDF table you need in Excel? Done.
Batch processing — Need to rename 200 files with a consistent naming convention, resize a folder of images, or convert documents between formats? This is where it saves real time.
Research synthesis — Point it at multiple documents and ask a question. It reads them all, pulls the relevant pieces, and writes up a coherent answer with context.
The key difference from regular chat-based AI: you're not prompting back and forth. You describe the outcome, Cowork figures out the steps, and you check back later. It asks for clarification when it genuinely needs input, but otherwise it just works.
10 DAYS TO BUILD
🔧 How It Got Built
Here's something worth knowing: Anthropic's engineers didn't build Cowork in the traditional sense. They built it by prompting their existing AI coding tool, Claude Code, to write most of it.
According to VentureBeat, the entire thing shipped in about a week and a half. An AI built a product. That product is now available to paying customers.
That's a preview of how software development is changing—and how quickly new tools can emerge.
COWORK VS. THE OTHERS
🎯 How Cowork Compares to the Others
Three major AI assistants now operate with some level of autonomy. They're built for different things.
ChatGPT Agent (OpenAI, July 2025) is web-first. It browses the internet, fills out forms, books travel, does research across multiple sites. Think of it as an assistant that can navigate Chrome for you.
Gemini Agent (Google, rolling out now) is ecosystem-first. It connects to Gmail, Calendar, Drive, and the rest of Google Workspace. It can draft emails, schedule meetings, search your Drive, and handle multi-step tasks across Google's apps. Still experimental, but deeply integrated with tools most businesses already use.
Cowork (Anthropic, January 2026) is files-first. It lives on your desktop, works inside folders you designate, and handles document creation, file organization, and data transformation. No browser, no cloud apps—just your local files.
Tool | What It's Built For | Strongest Use Case |
|---|---|---|
ChatGPT Agent | Web browsing & online tasks | Research, bookings, form filling |
Gemini Agent | Google Workspace | Email, calendar, Drive management |
Claude Cowork | Local files & documents | File organization, document creation, batch processing |
Different tools, different jobs.
This issue is supported by Attio. If you're exploring AI-native tools for your team, worth a look below.
FROM OUR PARTNERS
Introducing the first AI-native CRM
Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.
With AI at the core, Attio lets you:
Prospect and route leads with research agents
Get real-time insights during customer calls
Build powerful automations for your complex workflows
Join industry leaders like Granola, Taskrabbit, Flatfile and more.
THE CATCH
🔍 So What's the Catch?
Getting access to these tools is easy and getting easier. How to get value out of them and make sure you don't create an organizational security disaster is hard.
A November 2025 study from MIT Sloan and BCG found that 76% of executives now view agentic AI as "more like a coworker than a tool." But here's the gap: 35% of organizations are already using these tools, another 44% plan to deploy soon—and almost none have updated their governance, workflows, or talent strategies to match.
We know how to manage tools. We know how to manage people. We don't know how to manage something that sits in between.
You don't write HR policies for a hammer. You don't depreciate a human employee.
Agentic AI asks you to think about both for the same system. Who's accountable when it makes a mistake? What data can it access? How do you measure its output? These aren't technical questions—they're organizational ones.
The companies moving fastest right now aren't necessarily the ones with the best AI tools. They're the ones building the internal frameworks to actually use them: clear policies on data access, defined decision rights, accountability structures that make sense. Everyone else is buying subscriptions and hoping it works out.
That's the real gap. Not the technology. The management layer around it.
FINAL THOUGHTS
AI just built its own coworker in ten days. That's the headline. But the real story is simpler: these tools are becoming normal, and most organizations haven't figured out how to work with them yet or how to gain meaningful ROI from these tools.
Just be cautious about having a garage full off tools when you aren’t a mechanic or do your own maintenance. That can lead to an expensive pile of rubbish.
How helpful was this week's email?
We are out of tokens for this week's context window!✋
Keep reading and learning and, LEAD the AI Revolution 💪
Hashi & The Context Window Team!
Follow Hashi:
X at @hashisiva | LinkedIn
