Leveraging AI

260 | Claude Cowork Just Changed the Game: How You Can Automate Everything (Without Coding) with Isar Meitis

• Isar Meitis • Season 1 • Episode 260

Share episode

📢 Want to thrive in 2026?
Join Isar’s flagship AI Business Transformation course and save $100 with code LEVERAGINGAI100.
🔗 Sign up here and secure your competitive advantage for 2026.  https://multiplai.ai/ai-course/

Learn more about Advance Course (Master the Art of End-to-End AI Automation): https://multiplai.ai/advance-course/


Ever wish your AI tools could actually do the work — not just talk about it?

Most business leaders are stuck in the loop of experimenting with AI, but not seeing meaningful returns. The problem? Most tools aren’t practical, integrated, or business-focused. That changes today.

In this episode, Isar Metis walks you through a full hands-on demo of Claude Cowork — Anthropic’s latest update that quietly redefines AI in business. No jargon. No code. Just real results, from summarizing invoices to building full-blown applications on the fly.

You’ll see Claude Cowork in action across three real use cases, each one designed to save time, cut costs, and actually do the work your team is bogged down by.

In this session, you'll discover:

  • How Claude Cowork works (and why it's a game-changer for non-technical professionals)
  • The surprising capabilities hidden behind Claude Code
  • How to set up Claude Cowork on your Mac (with no dev team required)
  • A live walkthrough: automating invoice summaries into structured Excel files
  • Building a Mac application to manage business expenses — in under 3 minutes
  • Creating a branded PowerPoint presentation from podcast transcripts
  • The future of connected, agentic AI tools — and what it means for 2026

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

GMT20260119-171735_Recording_avo_1280x720:

Hello and welcome to the Leveraging AI Podcast, a podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Metis, your host, and my original plan was to record an episode showing you how to use Claude code for stuff that is not coding for day-to-day business tasks. But then Anthropic came out with Claude Cowork, which is basically geared to do exactly that. It uses the infrastructure of Claude Code. It makes it easy for anyone to use it in order to do things that you can use in your day-to-day in business. But before we dive into using Claude Cowork, I want to explain what the hell is Claude Code, why the word code is really confusing and not really what it probably should have been called. Why I think Claude Cowork is brilliant. And then we're gonna dive into how can you get a. Access to it and how you can test it yourself. And I'll show you two quick use cases as well. So in the end of this episode, you will understand the power of cloud code and cloud cowork, which are very similar things, as well as how to install it and different kind of use cases. You can use it. So first of all, what the hell is cloud code? Cloud code is a variant if you want, of Claude that was geared initially for writing code, but what it actually knows how to do very well. Is it is a very good agent environment that can understand a task and then get access to multiple resources and then build a plan on how to execute the goal that you give it together with the resources that it has access to, and it can get access to a lot of other resources through Claude skills and through MCP servers and through other connections that you can give access to it and it can. Achieve different things. Now, the reason they called it cloud code is because in the beginning they built it for coders, and that was the main usage. But very quickly it became very obvious that it is really good at doing other things as well. Because what it's really doing is it understands your need, your goal that you set to it. It understands. The files or other data that you give it, and then you can build a plan and execute and test the plan until it gets to the output that you want it to get too, which is the ultimate agent if you want. Now, the way Claude. Code ran initially was in a terminal on your computer. That's a really geeky thing. It's actually really easy to run. If you're on a Mac, you just go to the, spotlight, search any type terminal, and it will pop up like a dos kind of like interface. It's just a black box with a little triangle on top where you can enter commands and you can run cloud code in there. The benefit of running something in a terminal is that it has access to everything in your computer and at least access to what you want it to have access to. So you can tell it what it can access and what it cannot, and they'll worry about it. Before Cloud code does anything, it actually asks you for permission. It says, are you allowing me to do one, two, or three? And you can say yes or no, and you say yes or no by. Clicking number one or the number two. It's basically gonna walk you through how to give it permissions. And if you give it permission to do something, it can then do it, uh, one time. Or you can say, I'm allow it to do anything it wants in that session. That was the original way to run cloud code, and it was the only way to do it was in the terminal, but then they added the ability to actually run it on cloud code on the internet, and then inside of a tab inside of the cloud application. So if you're running the desktop application, you had two tabs. One was called chat, the other was called code. Since it became obvious to a lot of geeky people like me that you can use cloud code to a lot more than just coding. And they understood that the name pushes a lot of people away from using it. They've actually just announced this past week a new variant that is called Claude Cowork. What is Claude Cowork? It's basically putting a slightly more user-friendly interface on top of cloud code and taking away the word code so it doesn't scare away people who don't want to code and just want to use it for other things. So how does Cloud cowork. Works. I'm glad you asked. This is exactly what I'm going to show you. So first of all. How do you access Claude Cowork? So the way to access Claude Cowork right now, the only option is on the desktop application on a Mac. Now, in the beginning, when they launched it early last week, so long ago, right? The only way to access it was only with a max plan, which I have, but they've changed that at the end of the week, and now can actually access it with a regular. Paid plan, which is called Pro. So if you're paying the$20 a month for Claude and you have a Mac, you can now use Claude Cowork. How do you do that? You need to install the latest version of Claude Desktop on your Mac. And now I'm going to share my screen and I just did the install. And you can see right now on the top left corner, there are now three tabs. There's a chat tab that was before. There's the code tab that I talked to you about, earlier, and now there's a middle section, Claude. Cowork. When you click on cowork, it opens this new section that has several different components. If you are watching this on YouTube, you can see it. If you are driving and you cannot watch it, then you can just listen and I'll explain what's on the screen. So on the left you have the list of tasks that you gave it. Similar to the regular conversations, in any other chat platform. In the middle, there's a section that tells you different things you can do, such as create a file. Crunch data, make a prototype, organize files, prep a meeting, draft a message, and these are little buttons that if you click will help you in a specific direction, but the reality, you don't really need that. You can just tell it what you want it to do in simple English, and it will do it on the right side. There are several different boxes. One says progress. The second says artifacts. The third says context. Uh, and then you can add different connectors to it on the bottom. Those of you who don't know. Most of these tools, Claude included, has multiple connectors that you can connect to it that helps it do its work. In this particular case, you can do it on your own or it will do it by itself as needed, and we'll ask you for permission. The

Speaker 2:

And I have some really exciting news related to that, due to some personal reasons, which I'm not going to share right now. I am forced to postpone the beginning of the AI Business Transformation course that was supposed to start this Tuesday, January 20th. Now, while I'm personally sad that I cannot start the course this week, and I apologize for those of you who signed up, but don't worry, the course is starting. But for those of you who did not get the chance to sign up yet, it gives you the opportunity to still do that and join the course. The course is a complete game changer, and it has changed the trajectory of entire businesses and definitely careers of hundreds and maybe thousands of people who have taken this course in the last two and a half years. So if you want, make this your New Year's resolution to set up a solid baseline for your AI knowledge in order to drive your career. And if you are in a leadership position your business forward based on AI in 2026. Now, a few people have approached me on LinkedIn and on other channels, and ask me about the price point of a thousand dollars or 9 97 if you wanna be specific. All you need to do is come and join us on the AI Friday Hangouts. It is our weekly community AI meetup that we have every single Friday. It is completely free to join and there are incredible people there. All of them have taken the course, and there are people there from senior positions in large enterprises, all the way to really successful solopreneurs. Many of them became solopreneurs leveraging AI, following, taking the course. So completely changed the trajectory of their career because of the knowledge they gained in the course and asked them if it was worth a thousand dollars. So if your career and if your business is worth more than a thousand dollars, do not think twice. First of all, come and join us on Friday and literally ask the people, I have no problem with you being blunt and asking people in the community what do they think, but also come and join the course. It is going to be the best investment of the year for you. And if you do this right now, you are setting yourself up for a solid start in 2026 on how to leverage AI from a practical perspective across multiple aspects of the business. And because you are listeners of this podcast, you also get a hundred dollars off, so it's only going to be less than$900, and you can do that with promo code leveraging AI 100, all uppercase One word.

GMT20260119-171735_Recording_avo_1280x720:

The last component is, the way it works is it works by connecting to either a single or multiple folders on your computer. So I created a folder on my Mac that's called Cloud coworks, so it will be easy for me to identify, but you can use any other. Folder that you want. When you connect the folder and you do this just by little plus sign, down here, it will tell you that it will have access to anything in this folder, and it will be able to change, manipulate, delete, anything in that folder, just to give you a warning that that's what it will be able to do. So I wanna test two simple use cases with you. The first one is we'll take random invoices that I downloaded from the internet, and we are gonna ask Claude Cowork to create a summary Excel file that will summarize the information from all these different invoices in a structured way, and literally all you have to do because the folder is already connected and I've uploaded all the invoices to the folder. All I have to do is now give it a simple instruction, and so I will do this. I would like you to review all the invoices in the folder that you're connected to, and I would like you to create a Excel that summarizes the information in all these invoices. You need to add columns for obvious. Information such as the vendor name, the type of thing that we are purchasing, whether it's a service or a product. If it is a product. Then also include a quantity, and then the price per unit, and then the total price. I also would like you to include a total summary at the bottom that will tell us the entire cost of all the invoices, and I want you to sort it by the date that the invoices were issued. That's it. And now I click on Let's Go. And once I do that, it will start working. So the first thing that you will see is that it's sending the request and now it is going to start reading the information and it start looking for skills. So it is adding skills on its own. Those of you who don't know what skills are, skills are little capabilities that Claude can. Bring into its usage as it needs them, and there are multiple of them that you'll see it's using as it's going forward. So right now in the context, uh, box, we have the folder that we've added, and one scale that is just a generic scale that will allow it to use other in the artifacts. It now uploaded all the invoices and it created a plan. The plan was read and extract data from all the invoices. Number two, creating Excel spreadsheet with invoice summary number three, add formulas for totals and sort by date. Number four, verify Excel file and share with user. So it created the plan and now it is executing the plan in the first step. It already read all the invoices. And now it is saying that it is going to create the Excel. So now it's created the Excel. It now created the formulas in the Excel, and I assume the next step it will verify the information. So now what it's saying right now, let me verify by reading it and uh, open something. And so now it's gonna add a new skill for it to be able to read the file. And now it's saying verifying Excel file content. And basically it is walking me through the processes. It's telling me exactly what it's doing in simple English, step by step. And if I look at the folder right now, and I'm gonna share, the entire screen in ch instead of just the application. So if I share the entire screen, you can see the folder down here, you can see the invoices, and now you can see that there is an invoice summary Excel that did not exist before. It just created it. It also in Claude itself tells me exactly what it found. So it found the different invoices. It tells me the grand total is 33,860, dollars, uh, broken down into services of 20,900 into products of 12,960. And if double click on the Excel. I will drag it into view so you can see that as well. Uh, it did exactly what I requested, so you can see we have different columns that says invoice state. Then it has a vendor name, invoice number, the description of each of the line items. Then it says type, whether it's a product or a service, which I assume it's guesstimating. Then it has the quantity, the unit price and the total price, and it did a grand total of 33,860. And it even did a total by type because this is what I requested. So pretty well structured Excel, nothing crazy, but definitely did the effort that I requested and it would've taken me. I don't know, probably 15, 20 minutes to do this. And instead of that, I give it a task. And the thing was ready. When it was ready during that time I was talking to you guys. But the reality is what I could have done is I could have done another email or two, or do whatever it is that I needed to at the same time and would've completed the process as well. So now let's look at a different kind of use case, and that's the ability of it to summarize information. But now let's try something a little more advanced before we dive into a completely different use case. Let's ask it to create an application that will allow us to upload new invoices that it will then include in the Excel as well. That will allow me to create a simple application for me to track my invoices and my expenses. So I go back here, uh, I don't actually have to do anything. I'm just gonna go into the prompt line and I'm gonna say. This is great work. I would like you to create a application for me that I will be able to launch from my Mac what the application needs to do. All it, all it needs to have is a simple upload interface that will allow me to drag and drop invoices into that interface. And when I do. It will automatically do the process you just did, and it will add it to the Excel Every time there is a new month, meaning there's a new date of a new month in the Excel, I would like you to open a new tab in the Excel for that month and then keep adding the invoices into that tab before you begin the development. Let me know if you have any questions about the application. That's it. So now it's going to ask me a few questions. You will see it sending the request. And then it's gonna come up with a few questions that will allow it to develop a better application For me, whenever I create applications, I usually do a long brainstorming process, and this is just a very simple, short example of how I do this. So now the first question you'd ask, uh, what type of file format should the app support? It's saying images, p and g, jpeg, web, uh, p, images plus pdf, all common formats or type something else. I will go with all common formats, so I click on that. It's actually making it like a multiple choice that I can actually click, which is actually pretty cool, but it also gives me numbers one, two, and three. Which is kind of like the old school user interface inside of cloud code when you're running it in terminal, which doesn't mean anything to you probably if you haven't used it. But here I can just click on the screen, uh, and ask it to do what it needs to do. Uh, the second question is how should the app extract data from the invoices, uh, AI powered manual entry or something else, and it's recommending AI powered, so I will use that. Uh, where should the Excel file be saved? Same folder as the app. Choose, uh, in the first run, fixed location or type something else. Let's use same as the app, should months be named by date or let you customize. So I will use, uh, the first suggestion which says, auto name 2025 dash zero one. That makes sense to me. And that's it. It doesn't have any more questions, and I'll see what it's going to do. So now it is going to create a new plan. It will show us a new process on the left. It will pull up new skills. Okay, so what it's done now. It says this is a great project. Before I start, I would like, uh, clarifying questions. Do you have a cloud API key or should I include the instructions for getting one? So this is actually really interesting. I wonder what will happen if I click skip because in cloud code. It actually allows you to run, not necessarily by using Claude tokens through the API key, but actually by using tokens from the regular license that you have, in my case, the MAX program. But there's gotta be on the pro program as well. So I click Skip and let's see what it does. So what it's telling me, it's telling me. Got it. I will build the app prompt for the API key, uh, on first run, and we'll include a link. To get it if needed. Let me create this Mac application for you. So despite the fact I didn't give it a key, it seems to be doing something and now it is created a new plan, creating the Python invoice processing application. Number two, build a drag and drop graphic user interface. Number three, implement AI powered invoice data instruction. Number four, add Excel file management with monthly tabs. And then there's two more package the Mac application test and verify the application. So again, it's gonna do all the steps, including testing it and running it. And now it's just working on the progress. In this particular case, it is basically running just like Claude Code used to run, and it identified what I'm trying to develop and it's just going to develop it for me. So it is now done. It took about two and a half to three minutes. It completed the process and it tells me exactly how. To install it. Uh, it says run the tracker content and then, uh, what are the steps that I need to do? And it created a full documentation for it so I can understand how to run. And there is a setup. File that will then create the application that I can then run locally, uh, and even give it to other people as a package if I wanted to. So it did all these things in about two minutes, and now I have an application that will actually do everything that I requested. It says that the features are supporting PNG jpeg, PDF, and Web P, uh, and other image formats. AI powered extraction for vendor date items, quantities and prices, automatic monthly tab creation and summary dashboard with totals and breakdown. Everything that I requested and I can now run the application and keep using it from now till forever, uh, basically so it can create any kind of code, even really sophisticated things because again, in the backend, this is actually cloud code, which is currently the most powerful coding platform out there as far as a raw model, and really all they did is they just created a really simple and easy to use user interface for it so you don't have to run it in terminal and don't understand what to install and so on. You just run it right here on the desktop application. So now let's go to a third use case. On the third use case, I will start a new task. And what I've done is I've created a new sub folder in the main folder it has access to, and it has access to everything in the folder, including the sub folders and what I've dropped in that subfolder is transcripts of the news episodes of this podcast. And what I'm gonna ask it to do is I'm gonna ask it to create a PowerPoint presentation that will summarize the most important news from these three different episodes and make the PowerPoint align with my brand guidelines. So let's do it. I would like you to look at the subfolder called transcripts. In that folder, you will find three transcripts of previous episodes of this podcast that includes different news about AI from the past several different weeks. What I would like you to do is to go thoroughly through the transcripts, decide what are the five most critical pieces of AI news that is most impactful for business people, and I would like you to create a PowerPoint presentation with five slides each. That will include one piece of news and make it easy to consume for busy business people. In addition, I would like you to include a final slide with additional things that business people can benefit from or learn that we reviewed in those episodes, even if it's not directly related to the other five pieces of news. The PowerPoint presentation needs to be aligned with the multiply AI brand guidelines. That's it. I click go and now I'll explain a few things. Uh, it's going to do very similar to what we've done before, so it's gonna go and add new skills, which it's doing already. Uh, one of the skills that it is going to use afterwards is a skill that I created that will allow it to. I create things based on my brand guideline, but it's also now adding skills for HTML to PPTX, which will allow it to create code that will create an HTML file. It, uh, also added skills for CSS, uh, capabilities, and it's also now looking at the slides, examples that I have in the scale for the brand guidelines. And it created a plan. Reading and analyzing transcript files. Number two, identify the five most impactful AI news stories for business people. Number three, create PowerPoint presentation with multiply AI brand guidelines. Number four, add final slide with additional business learnings. Oh, and number five, validate presentation visually and deliver to user. And now it's gonna run through the plan just like we've seen it doing before. So it's gonna explain to me along the way exactly what it's doing and it is just going to run and hopefully will give us a great PowerPoint presentation with five slides that are summarizing the five most important news from three different episodes that has. 45 minutes to an hour worth of content in it and it will pick on its own what is the most important. So we'll see, uh, shortly what it's doing. So now it's already now telling what are the top five news it's going to include. So number one, open AI apps, ecosystem launch. Definitely a big one. Number two, Claude Code, cowork Revolution, which is exactly what we're talking about right now. So if I decided to choose to do an episode about it, that obviously makes sense. Number three, enterprise AI adoption. He inflection point sounds definitely relevant. Uh, and it's gonna bring reports from Deloitte, Google Cloud, Melo Ventures, and Ernest and Young. So definitely sounds, uh, relevant. Number four, AI workforce impact becoming real. And number five, AI shopping. And. Personalization era begins. So all sounds really relevant. And then it added bonus learning about Nvidia meta acquiring Manus, uh, data centers in space being planned by SpaceX, uh, new Nvidia Reen, GPUs and healthcare in ai. So all of them are definitely relevant. So this is kinda like the bonus slide that I requested. And now it is already in section three in its plan, which is creating a branded PowerPoint presentation. That is what it's working on right now. And as it's doing it, it's adding more and more capabilities and skills to its context, so it will be able to do the work. And it's also even telling me which slide it is working on at every single step. So now it's writing slide number five, and it's telling me how many seconds it's working on each one. Again, those of you who have used cloud code before will find the user interface a significantly nicer, but from a flow perspective, it works almost identical to cloud code because in practicality, this is cloud code just with a nicer wrapper. The process has finished and something interesting actually happened along the way. Obviously in the editing, we skipped that, but it actually got stuck in the process. The HTML to PPTX library is having issues with the browser in the sandbox environment. Let me use the skills tool. So basically it got stuck with creating the presentation on its own, replaced the implementation of that to a different kind of implementation, was able to create the presentation. And verify the presentation, so let's see how this looks. Like. You can see now in the folder, there's now a presentation called AI News business update and if I open that, then it will open a presentation. Those of you who know my brand guidelines, you can see it is perfectly aligned with my brand guidelines and it actually looks pretty good. So the first slide says chat's app store moment, uh, 900 million plus weekly active users. Open. OpenAI launched a full app ecosystem inside of Chachi pt, the biggest platform shift since smartphone app stores, et cetera, et cetera. And you can see the rest of the slides. And they actually have a really cool kind of like, uh, vibe to them. They have the same branding as my brand guidelines, and they even have. Really nice graphics and these boxes in similar colors across all the different slides with the main aspects of the different news and what we covered. Uh, there is some little overlap between the different components. Uh, nothing that I can't fix very easily because it is a PowerPoint file, uh, that I can edit. And there's even the. More key development final slide that we discussed that has multiple news in smaller boxes, so it definitely followed the process. Now, if you generalize what I just did, it means it knows how to take lots and lots and lots of information. This could be notes from meetings. These could be files from an RFP. This could be information from resources that could be your clients open sources or internal sources, and turn them into any kind of a report while you are doing something else. Now, if I would've used a specific format as a template, it would've followed that format. So if you need to create a weekly report of any sort. All you have to do now is actually drag and drop the source file into a folder that Claude has access to. Give it access to the format output, or create it as a scale just like it has a scale for my brand guidelines, and it will create the output for you. So what does the hell does that mean on the bigger picture? It means a few, several different things. Number one, which I said multiple times on this podcast before, is that the tools, the way the AI capabilities is going to be packaged, is gonna make a bigger. Impact on our work than the actual raw capabilities of the model itself, because all the models right now are really good. What's gonna make a difference on whether you're gonna use it or what we're gonna use it for is how easy is it going to be to use it for these specific applications? And what philanthropic did with Claude Cowork is make it really easy. To do really advanced stuff with access to basically any information you want to give it access to. That is obviously a huge win for Anthropic and I urge you to go and test it out as well. Now, before you go and test it out, a few things to remember. One, this is still in testing preview, meaning this is not a final product. It may break. It may do surprising things, so make sure these folders, you give it access to. Do not have a, any information that may leak outside that you don't want. It leaked to whatever, even though they're saying they're not gonna train on this data, but it is again, a preview. Product B is that it's just a backup of these files, and it's not the original files because it may delete them from the folder and then you would lose the information, uh, that you wanted to actually analyze. Uh, but with these two things in mind, you understand that this is an incredibly powerful agentic capability that can understand what you're trying to do. Can create a plan, can add whatever skills it needs in order to execute the plan and can then execute the plan and create files, create sub folders, do whatever you want it to do within the confinement of the folder slash folders you want it to have access to, and it can do. Anything. Literally create PowerPoint presentation, create Excel files, edit excel files, uh, create new sub folders, write software that you can use, et cetera, et cetera. Extremely powerful capability within our fingertips that is connected. To our real data and our real environment, and all you have to do to activate it is speak in simple English. Now, my suggestion if you want to use it at scale from a deployment perspective is do some planning before you ask it to do the thing you wanted to do. So in the example I gave you, I just mumbled the. Instructions away, and what I probably would do if I need this for more than a demo is I will first brainstorm with it exactly what I want the output to be, create a better defined requirements document, and then use that as the definition of what I will give to Claude Cowork to do. But overall, this is an incredible capability. It builds on top of something that was a lot more geeky before, which is cloud code, and just makes it a very simple user interface. That now allows you to do magic with information that you have on your computer or that you can bring to your computer. By the way, with the connectors, it can connect to many other sources, so in the connectors section. You can connect it to tools such as Google Drive and GitHub and Gmail, and Google Calendar, and even the Claude in Chrome extension. So then it has full access to your browser and can do anything in your browser. So now it can go in. Run third party applications or third party websites on the browser. Get access to them. If you give it access to it and your login, get information from there, operate them as part of the overall process that it's doing. It has every MCP server that I connected available as well, and many other connectors that Claude provides. So it's almost unlimited with the things that it can do and the data sources, it can connect to either through MCP or through native connectors on Claude. On the bigger picture, what that tells us is what 2026 is going to look like. First of all, if this is just the beginning of the year, I can't even imagine where we're going to go, but the direction is clear. It is going to be agentic, meaning it will create its own plans and use a variety of tools to execute the plan. It will be connected. Two, the tools and systems and environments we work in right now, and the limitation is going to be our imagination and how much we trust the system and hence the level of access we're willing to give it. But other than that, more or less, the s. More or less, the sky is the limit, and I'm not even sure there is a limit even in the sky. So go and experiment with it early on. This is just a few examples for you to potentially test so you are ready to do much more sophisticated things as the tool evolves and becomes safer to use. That's it for today. We will be back on Saturday with the news and we are back to doing the lives more or less every Thursday. I know there was a big, big break or inconsistency with our live sessions on Thursdays. We have continued and we're doing the AI Friday Hangouts every single Friday. There's an incredible community, active community of people who meet every single week and talk about ai from general concepts to specific use cases, help each other solve problems, uh, share things that we were able to achieve and so on. So if you want to join us, there's a link in the show notes for you to do that. It is very easy. It's just gonna put a link on your. Calendar and you can come and join us every Friday. It's 1:00 PM Eastern time. Uh, but there's people from Europe, people from Asia that are joining every single week. So regardless of where you are in the world, feel free to join us and have an amazing rest of your week.

Head Shot

Isar Meitis

Host