Episode 117: Claude Cowork is INSANE (Try These Use Cases Out)
Co-Host
Aytekin Tank
Founder & CEO, Jotform
Co-Host
Demetri Panici
Founder, Rise Productive
About the Episode
Claude Cowork is transforming the way we interact with AI in our daily workflows. In this episode, Demetri Panici and Aytekin Tank explore this groundbreaking new tool from Anthropic, showcasing powerful real-world use cases where Claude Cowork goes beyond basic chat capabilities to actually execute tasks on your desktop. From organizing files to building presentations, automating browser actions, and generating PDF reports, Claude Cowork functions as a true digital assistant, helping users with productivity-enhancing automation without needing to write code. The hosts break down how Claude Cowork leverages access to your computer and browser to perform intelligent tasks with context awareness and user permissions. Whether it's streamlining client meeting recaps, building forms with integrations like Jotform, or cleaning up cluttered folders, this AI agent offers an unparalleled level of autonomy and precision. Tune in to learn how you can start using Claude Cowork to turn ideas into action and boost your daily efficiency with minimal input.
It has many connectors like there's lots of products that you can add, so I'm going to use the Jotform connector. It's available.
I can just say create a contact form using Jotform. Basically, whatever kind of form you want to create, you can do it and it will just use the MCP connection with Jotform to create a form for us.
I'm going to show this contact form it created. Here it is. It basically created a simple contact form with all these questions on it and it's now making submissions to it.
Hi, my name is Demetri Panici and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jotform and featuring our very own CEO and founder Aytekin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show.
Hello and welcome back to another episode of the AIHS podcast. In this episode, Ida and I are back and we're talking about the hottest new tool. I was going to say in 2025, but it's 2026. It is Claude Co. That's such a fast start to the new year.
This is our first episode of the new year and we have a huge update, a new AI agent that's so powerful it's going to blow your mind. So it's going to be some exciting demos.
It showcases this new product that just came out two days ago called Cloud Co-work. Basically, what they did was they already had this amazing product called Cloud Code and lots of developers have been using this product.
They discovered that developers weren't just using this product for writing code but also for all kinds of automation on their busy work and they decided to do something about it.
You're going to tell us that story? Yeah. I was somebody who was using Cloud Code for a while for a lot of different use cases. I was making landing pages, doing keyword research, and a bunch of different things that can be done with Cloud Code effectively.
The story that came out was that Anthropic claims their buzzy new Claude Co-work tool was mostly built by AI, their Claude Code tool, in less than two weeks. So what does that mean?
Claude AI wrote product manager Felix, a project manager product manager at Claude, wrote that Co-work. Us humans meet in person to discuss foundational architecture and product decisions, but all of the devs manage anywhere between 3-day cloud instances implementing features, fixing bugs, or researching potential solutions.
As a result, they were able to basically instacode a project to turn the Cloud Code tool into this Co-work tool in two weeks. We're going to go through the whole episode giving use cases but here's a very brief minute showcase if you're trying to get enticed.
I'm just messing around because they always have this dang music on every single demo. It's cool though. So, as you can see, it's a very engaging product that's able to do a bunch of stuff. It really did cook.
We're excited to show you different ways that it has cooked. Just to kick it off, I know that you have some use cases. I have some use cases. We're going to show some demos.
What makes this powerful is that you give access to a folder or some folders in your computer and also to your browser. Basically, you install the Cloud plug-in to give access to that and it can do many things for you.
We have done episodes about browser agents, AI agents that can use your browser to get things done, and we've shown products like Mayus, general AI agents that can actually do things for you.
But this is taking it to a whole new level where basically you can do so much more, which means everyday tasks you do can be done automatically.
I wanted to compare and share screens. Are you sharing screen? You can do it. Do you want to go first or should I?
I'm going to go first. I use Cloud Co-work to describe Cloud Co-work. I asked it to do some research and create a table for me so we can explain to our viewers or listeners what is different about it.
When you're using Cloud as a regular user, you're getting answers to your questions and doing research. When you use Cloud Code, you're building software, writing code.
With Cloud Co-work, you're asking questions but basically asking it to do things for you. What you're seeing on my screen is the Cloud Mac app; they also have a Windows app available and a mobile app, but Cloud Co-work is only available on desktop right now.
Because it's a desktop app, it can do many things. It can access files on your computer, use your browser, and basically get things done for you.
I asked it to do research and create a table. Once it created this table, I also asked it to create a presentation about it, and it created a keynote presentation.
It's comparing Cloud to Cloud Co-work. It can do many things; it has skills and you can actually teach it new skills.
In regular Cloud, you get text answers. In Cloud Code, you get code. In Cloud Co-work, you get work outputs. If you give it data or ask it to do something, it will give you back data or turn that into a webpage or a slide.
You can use any kind of data to do your work. Cloud uses browser chat, Cloud Code works in a terminal, and Cloud Co-work is the desktop app I'm using right now.
Basically, it took them two weeks to turn Cloud Code into Co-work. They've been investing in Cloud Code for a long time, so it's a really good product with a huge community of users.
Now you have this major product that can be used for other purposes. It can access your computer and browser, do things for you, and when accessing your computer, it asks for permissions.
It creates a virtual machine limited to one folder so it cannot go crazy and destroy things on your computer. You only give access to one folder.
The presentations it creates, even this table, are HTML files created within that folder. You can open them separately. It's good for everyone.
While it's working, it shows you all the artifacts and context. It has all the power of Cloud Code but doesn't require you to be a developer. You just tell it what you need and it does it.
I find the three levels of explanation apt. Cloud Co-work is a clear worker and researcher side for chat, Cloud Code is for coding.
If you're using this tool, make sure you use it for its use case. It might not hit 100% for all uses, but keeping it in those buckets is a fair delineation.
Let me share my screen and show a good example of something I've done. I told it to clean up a folder, scan the content, propose an organizational plan, show me the plan before making changes, and execute only after approval.
I took my downloads folder, which was insanely disorganized with many file types like audio files, images, MP3s, HTMLs, zip files, and Office docs.
I asked it to organize it, and it put everything into 14 different folders. I approved it and it took 45 steps, about 4 minutes in real time.
The result is these 14 folders created within four minutes by just saying, 'Hey, this is disorganized, can you organize it?'
This is a common use case people showcase. It's incredible because many of us are disorganized and now you have a product that can help you with your desktop.
If you think about it, a lot of people do their work on their computer full of files, making it hard to find, organize, and process stuff.
You can ask it to find all different documents about clients and organize them by client in a few minutes, making it easier to find stuff.
I just asked a question: 'Continue to organize my files,' and it organized all remaining files in 30 seconds.
One powerful thing about Cloud Co-work is that it's not one shot; it creates a plan and then executes it.
While it's working, you can interrupt it and change its plan without canceling the whole process. It won't time out.
This is similar to Mayus, but Mayus doesn't allow interruptions easily. Cloud Co-work is continuous and agentic with many powers and connections.
I'm going to show a demo about connectors. It has many connectors for different products. Jotform is one of them.
I'm going to use the Jotform connector. Because I'm connected to my Jotform account, I can say create a contact form using Jotform.
It decides what questions to have on the form and sends a MCP request to Jotform to create the form.
It created the form. If I open it, it's a regular contact form. I'll ask it to do some test submissions on this form, for example, three test submissions.
It can fill out the forms using MCP so we can see how it works. While it's working, I'll show the contact form it created.
We created a simple contact form with all these questions and it's now making submissions to it.
It created these test submissions for us. We can ask it to get the data into an Excel spreadsheet.
I'm making a secondary request to create a report about the data. It's using the same connection to get the data and list submissions.
If you go to settings, you can see capabilities and skills. You can teach it new skills, like a Slack gift creator skill that creates gift files.
It created the Excel file and placed it in the folder. We can see the submissions in the Excel file.
You can ask it to create a report about the data, including charts and other things.
I love the cloud animation. It has some of the best aesthetics. The styles it creates are beautiful compared to other AI tools.
The report is still working. I see a doc file. It looks like a great report showing total submissions, phone numbers, active status, and detailed submission documents.
I'm going to ask it to turn this docx file to PDF. Because it's a desktop app, it won't time out and will keep going.
This morning, I was playing with it. It included competitive information and required browsing many pages. I started it and it kept running, creating amazing reports with excellent suggestions.
Sometimes it creates code, like Python code. I'm not sure what it's doing right now but it's looking for files to convert to PDF.
I have some stuff to show too. It has been improving its ability to have MCCP servers work with Co-work. Some connectors had issues but were working in chat.
I'm going to show a cool use case with a tool called Grain, a meeting recording tool. MCP is available even for plans not using the API.
I had a meeting with a client. I can ask, 'What did we talk about in my meeting with Arcade yesterday?' It references Grain and other connectors like Google Calendar.
It found my Arcade meeting and listed attendees. We discussed meeting cadence, contracts, billing, and many things.
I asked it to put together a meeting action plan for our three meeting system so I can have agendas and write a recap message for the team.
This is basic but something you can consistently do and have your team members do well connected work.
It asks me questions to be specific if it's missing context. The meetings are Monday paid call, Monday organic call, and the Wednesday holistic call.
I clarified the deliverable format as a visually good-looking PDF and asked it to draft the team recap message for Slack.
This guided me where chat doesn't have this capability. It attempts to do tasks correctly by creating files, running prompts, and asking clarifying questions.
It reduces hallucinations by giving options and trying to understand what you're trying to achieve, which is a huge improvement.
It's making a PDF and going through other stuff. It has a connector for Claude and Chrome with a Chrome extension for agentic interaction.
It created a visually polished PDF explaining the standard agenda for paid ads, organic content, and holistic strategy calls.
The Slack message recap mentions a holiday for Arcade and key action items. It remembered fine details like the holiday and meeting notes.
I tell my team to come to me with suggestions rather than asking what to do. Claude Co-work can come to you with suggestions, commentary, and ideas.
They ask clarifying questions like 'Do you mean this, this, or this?' and think about it, asking for minimal input.
One manager advice is to tell what you want to achieve, not just what to do. These tools work best when you explain the end result you need.
Good managers talk about goals and end results. If you describe it well, your coworkers, including Cloud Co-work, will do a better job.
Instead of asking it to do something, explain what you're trying to achieve and see what happens. It asks clarifying questions to make conclusions better.
MCP is breaching hack levels, giving access points to many tools you use. In the time it takes to have a conversation, AI can solve problems and take action.
You can copy and paste your conversation into Claude Co-work with your MCP setup and it'll do what you asked for. If connected to your meeting recorder, it can act on meetings.
If you can get AI to tell other AI what to do, you can automate many tasks. I'll show an example of this.
I have a prompt for consistent prompts. I want to use Claude in Chrome to do a task with the lovable app.
Cloud Co-work uses your browser to manage Lovable. On the left is perfect for creating and improving an app. I'm looking to enhance the system.
The main goal is to take the basic CSS/TypeScript look of the app and make it look sleek and modern.
You can use the Cloud Chrome extension. Make sure you're logged in. I told it to do a 30-minute session.
It might have bugged because I use Edge instead of Chrome. Let me make sure I'm logged in.
It's crazy how this works. AI telling AI what to do. It's navigating the app, clicking, and debugging the browser.
It took a screenshot and clicked on the chat input directly. This blew my mind because I had a rudimentary tool and the app looked bad.
I don't want to spend mental energy going back and forth. I just want it to continue improving the design automatically.
You can build an app with forms and prompts. People enter prompts written by AI and paste them into the prompt box. Cloud Co-work does the work while you sip your tea.
I tried saving time by capturing audio input with Whisper. I ran Cloud Opus 4.5 for 3 hours straight on a $200 plan.
Sonnet is more efficient and can run infinitely. The model 4.5 Opus costs more but Sonnet is cheaper.
Cloud Co-work goes back and forth, takes screenshots, waits for progress, and follows the plan to implement improvements.
It scrolls and presses the implement plan button automatically. It does everything you'd imagine a tool would do.
I'm enamored, shocked, scared, excited, happy, and sad all at once. It just pressed the button and is building.
This is ridiculous. It takes us back to the HTML5 hard coding days. Now AI is doing it for us. All you need is the idea.
It will do this for a while. You can have multiple tasks and instances fixing one prerogative at a time.
If you have multiple apps or reports, you can do multiple things at once. Thanks for letting me set off this prompt chain so I don't have to do anything for an hour or two.
That's all I have. What about you?
That was all for me. I tested these things and was impressed. We've tried many tools like AI browsers and Mayus, but nothing has passed this one.
This is more than a demo; it's a real game changer. Give this to every employee and they have superpowers to do the job of five people.
It can do lots of things you used to do manually. Now you can let it go and do the work while you oversee as the manager.
This will improve our company's productivity. Everyone at my company will use Cloud extensively and understand AI is a requirement.
Having people who cannot use this is an insane opportunity cost. If I had 10 people who know how to use this versus 50 who don't, I'd pick the 10 every time.
It's like the PC revolution. Some people didn't want to learn, others became much more productive.
Spreadsheets improved productivity 100-fold for accountants, moving from paper to computer in a few years.
This is the future. We've started 2026 fast. I have a prompt that tells it to QA everything.
With coding tools, you have to ask if it did what you asked. It's simple but effective.
I have a second screen where I copy-paste QA commands and check if it did the job correctly, then go back to work.
Do you have ideas on how to improve the prerogative of making it look nicer? Waiting, getting distracted?
I'm not thinking; I have a prerogative in my head to make it look pretty. I just tell Co-work to do it while I check emails.
I don't want to deal with this today. Just answer my emails for me. I'm going on vacation and Co-work will handle everything else.
Thanks for hanging out and nerding out about this great new improvement. You've been amazing and supported the show.
Please try all the stuff we have going on at the Agents podcast. Try this out specifically; it's going to do wonders for you.
If you liked this episode, hit the like button and leave a positive review on Apple Reviews and Spotify. This tool is really game-changing.
Thank you so much for watching and listening. We'll see you in the next one. Peace.
Stay Ahead with the AI Agents Podcast
Get the latest insights on AI agents, their future, and developments in the AI form industry.