Full Tutorial: How To Automate 99% Of Your Work With Claude Cowork transcript

Hello legends. In this video, I'm going to show you how to use your Claude co-work to be a virtual assistant that can automate 99% of your tasks. We're going to start by downloading and installing the app and then choosing the correct paid plan. I'll then take you through step-by-step exactly what the app can do. I'll explain the differences between the Claude chat, Claude co-work, and the Claude code apps. I'll then explain how to use it, including how to adjust your conversations and how to think about the tasks that you're raising so that you can turn Claude co-work into a personal assistant.

And then finally, we'll go over the three most important features. I'll show you my exact workflow of how I create new projects and how I manage tasks. So, the first thing you want to do is go across to claude.com/download and then download the Claude desktop app. So, the reason we need the Claude desktop app is because this is the only way that you can use Claude co-work. Claude co-work has features that give it access to local files on your computer.

It can work with Excel, Word documents. It can take control of your browser. It can actually browse the web, click through buttons and icons. And it can also connect into different tools that you're using like Gmail or ClickUp. And all that level of accessibility is only possible through the desktop app. Now, the next thing you need to do is get a paid Claude plan. I recommend going with the monthly pro plan for $20 a month.

Now, my honest opinion is that you're going to get the pro plan and it might last you for a couple of months. But, the better that you get at Claude co-work and maybe even using Claude code, you're going to be upgrading your plan. So, I started off on a pro plan when I was using the Claude chat app. Once I got Claude co-work and I started using Claude code, I had to upgrade to the max five plan and it is totally worth it.

So, once you install and sign into the Claude desktop app, this is what you see. The first thing you want to focus on is this panel over here because over here we've got access to the three different Claude products. The very first tab is Claude chat. If this is one of the first times that you're using Claude co-work, you've likely spent 99% of your time here. Claude chat is an incredible tool. You can ask your question into here.

You can add additional context like web search capabilities. You can drop in PDFs or Excel sheets. You can even plug it into your ClickUp for task management or Gmail, which means it can go out and actually read your tasks or read your emails. So, one of the things we'll try and learn in this video is how to change the way that we're having conversations with Claude because the way that we communicate in the Claude chat app, where we're just asking questions and getting responses back, is actually totally different to how we should be asking questions and communicating with Claude co-work.

My best analogy is you would use Claude chat for things like shower thoughts, the random thoughts that you have, random questions that you have that are totally unrelated to any other work that you're doing, and the intention is just to ask a question and get an answer, kind of like a chatbot. Now, when it comes to Claude co-work, we don't want to think of it as a chatbot. Instead, we want to think of it as a virtual assistant.

Imagine you hired a human to come and help you with managing your work. You wouldn't just use your virtual assistant to help you check the weather or get a daily email summary. You'd likely want your virtual assistant to actually help you draft emails or help push tasks along in your task manager, which means there's a couple of different things we need to do in Claude co-work to expose that kind of functionality. And then the final product that we have is Claude code.

So, the way I like to think about Claude code is if Claude co-work is my personal virtual assistant, Claude co-work is a virtual assistant for my business. So, a great example here is that I'm currently working on a finance automation. My finance team's come to me and they want me to help them build out a pipeline where it listens to our Zendesk email tickets. If we get any emails that contain any bills or invoices that we want to parse them and we want to add them into our Odoo accounting tool.

Instead of me setting up my Claude co-work and using AI and spending my tokens and it's not very private, I'm actually building out a custom script and a custom portal that'll be for internal team use. And that's an example of how you use these tools differently. Claude co-work for personal stuff and then Claude code for, you know, bigger project stuff for your business. So, now we're back in Claude co-work and we're ready to give our first task to our virtual assistant.

And there's actually two ways that we can do this. One is slightly better than the other. The first way is to click on new a new task and then enter in our question or our task. We can add PDFs, Excel sheets. We can give our virtual assistant access to Gmail, to ClickUp. And then finally, over here, we have to choose a project for this piece of task that we're working on. Now, a project is just a folder on our computer that we dedicate to this work.

The project starts off as a completely empty folder and then as we give resources to Claude or Claude generates resources for us, all that stuff will be added into this project. And these are all just different folders of work that I've been working on. And at the very bottom, you have another tab which is called projects. Now, projects are essentially the same thing as assigning a folder to this task that you're working on.

But, whenever you want to kick off a brand new piece of work with your Claude co-work assistant, you actually want to do that by going into this projects panel. Now, before I do that, I want to spend some time explaining to you what's the difference between a task and a project. Like, why couldn't I just ask a question here, assign a folder, and then just get my task complete? Like, why why isn't that good enough? And the answer is that it actually is good enough for a beginner level because you can give access to your Gmail and ask Claude to summarize your emails and you get a response back.

You can give it an Excel report and ask it to analyze that report and give you findings in a summary and it'll do that. Or, you can ask it to create a PowerPoint from an Excel report that you give it and then email it to yourself or add it to ClickUp as a task for everyone to view and it'll actually do that. But, the real unlock is when you stop thinking about tasks, which are once-off events, and you start thinking about projects or goals.

So, let's say we just break down this analyze this report request. Now, if I was to ask you, why do you want to analyze this report? You might say, "Oh, every Monday I get an export of the previous week's sales and I just want to see how we're doing." So, then we'll add that over here. And then if I ask you, "Okay, what kind of insights are you looking for from this report?" You might say that I want to compare the results of this week from this year to the results of the same week from last year.

And we can add that over here as well. And then I'll ask you, "Okay, so what are you going to do with this report once you have it?" You might say that on a weekly level, you just monitor it yourself. On a monthly level, you want to provide a summary to your team. And on a quarterly level, you actually want to review the strategy. You want to see if your sales are actually trending up against your goals and then if they're not, you need to make some decisions.

So, even by asking just very basic questions, you can see that actually the point of this task isn't to just get an analysis of a report that you give it. What you really want is an autonomous virtual assistant that will automatically go off and download that report on a weekly basis. It's going to know that it has to compare to last year's results. Or, maybe you'll give it a set of benchmark results where if any of the numbers drop below a certain level, it'll automatically flag you and and either tell you directly within the app or it'll raise a task for your team to work on.

And then you want your virtual assistant to complete this task every Monday, but then every four weeks to create a summary aggregating the results of the previous four weeks. And since we're working within a folder or a project in our computer, all the reports are going to live there. And then finally, for the quarterly summary, you want to define a five-step process so your virtual assistant goes across to your ClickUp, pulls your team goals and team benchmarks, it then aggregates the last 12 weeks of your results, it generates some findings, and then it creates an informative PowerPoint that you can share with your leadership team.

Now, the thing is that you don't want to blindly trust whatever Claude is telling you. Claude is a very smart AI model, but it doesn't replace your own nuanced understanding of your own business or your division, your department. So, really, whenever it generates these reports, the main thing that you want to be giving it is also additional context around decisions that you're making around your business. For example, imagine you had a virtual assistant that had no idea about what products you were selling, about what your goals were.

You never had a meeting with them to keep them up to date about key decisions or changes that are happening within the business. Of course, the analysis of the report is not going to be very accurate. So, now the second most important thing that you can do when you're creating a project is to drop in a bunch of context. Just like when you're hiring a brand new person for the first time, they might be very smart and capable.

They can click the keyboard. They can click around on the mouse. They can navigate the computer. But, you really want to install a bunch of knowledge into their brain. And the only way to do that is by transferring it from either your brain or your tools that you're using and then front-loading it into the project. Giving Claude context is easier than you think it is. So, whenever I start a brand new project in my very first conversation, the very first session that I start, I give a basic set of instructions similar to this.

"Hey, before we do any work, I'm just going to spend a lot of time speaking with you, giving you a bunch of context about this project that we're working on. I don't want us to do any work at all, no tasks, nothing. I just want to be giving you a bunch of information and if anything is interesting or unclear, just ask me some follow-up questions." And as we go on, so this is like persistent persistent instructions for future chats as well.

As we go on, just document key learnings or key findings or any nuanced bits of information that seem interesting and just document them within a file in our folder. And then from there, you can use your favorite transcription tool and you can just spend 5, 10, 15 minutes just speaking into your session with Claude. If you don't have a transcription tool, you can just backspace all this and then just use the built-in transcription tool from Claude.

And now you might be asking, "Well, how much context do I need to give to a project? Like, how do I know when I've given enough information and I can start?" Well, some of my projects might take me like 15 minutes, literally a 15-minute sit-down on a computer where I have a couple of back-and-forth conversations with Claude where I'm actually speaking. I don't type. Typing is really slow. I kind of get into a flow state when I'm just speaking and transcribing directly.

And it's actually been times where I set up projects like my longest-standing project is um literally the lifetime of me using this app, so over 30 days. And for that project, I spent three days adding context into my session. And it wasn't just all at once. Like, I just didn't sit at my computer for 24/7 giving context. I came into the project. I gave it high-level instructions like what I just showed you and I started speaking to it.

And then I started asking it, "Hey, hey, go across to this ClickUp task and like read this and tell me what you think." And then over the next few days, I would like have another idea as I'm going for my walk or as I'm doing different piece of work and then I come back into a session and I continue just dropping some notes. And then after three days, I literally had like a smart version of myself, smart version, capable version that could go into my Gmail, go into my ClickUp, control the browser, do literally everything that I could do, and it had all the memory of this specific project that we were working on.

And then it became really cool, even from the point of view of just like imagine you could speak to yourself about problems that you have sometimes, and ask yourself to like give you different opinions based on your own knowledge. Like it's really difficult as a human to kind of have that introspection, but when you give all this knowledge into your co-work, into this like one specific folder, you can just have really high-quality conversations with it.

So then different types of context that you could give is for example in our reporting example, we wanted to compare the reports for the same time previous year. So you could give the results from the same time previous year. As you go through, you also accumulate different results. So let's say I give a report every single week for 12 weeks. Well, all those reports are staying within that folder and Claude can access them at any stage as well.

And like I said, you also want to drop voice notes on a weekly basis. Like just to give it update on like, "Hey, the company is actually picking up a brand new product. We're rolling it out next week. Or we're dropping this specific product. Or we're we're creating a brand new service for this type of customer. Or we're off-boarding this customer." All these little nuances, if they're related to the report that it's pulling and what what it's working on, definitely drop that into there as well.

Other ways are they can read your emails every day, generate an internal summary for it, also read different tasks that you're working on, read decision logs. And like I mentioned before, this is a really good way for yourself to just like create a really smart version of yourself that has access to all different tools that you can just have a brainstorming session with. Like it's no longer a generic one-off Claude session that you're trying to like never close that window because you lose all the memory.

When you're using projects, that memory is retained and that context stays there. Now you're not just limited to using Claude co-work for professional tasks like helping you automate stuff that you do at work. You can also help it to automate your private tasks like things that you want to do outside of work. For example, if you were to go to Claude and ask, "Find me the cheapest flights to Europe." And I was going to say, "Well, why do you want to find the flights to Europe?" Your answer might be, "Hey, I want to go to Europe for summer and I just want to plan my trip." Well, in that case, you might have a bunch of different tasks that you need to do.

Your very first stage of planning the trip might just be deciding on which countries do you want to go to, which dates, what your budget is, what the itinerary would be. Once you've locked this down, you then know roughly which dates you can go, roughly what dates you want to find your flights for. And as you go through this process, you might be uploading all the different hotels that you book, all the budget that you have, your itinerary, your flights.

You could upload that into the folder that you're working on with Claude. And then while you're on your trip, you could just use it as a chatbot. Because the Claude session can access all the files within that folder, can access previous conversations that you've had within that folder. And you can ask it, "Hey, today's the 4th of April. What's, you know, what are we doing tomorrow? What's the hotel contact's phone number?" "Hey, does the next place allow us to have a parking on the street?" So you can literally use it as a very helpful assistant that has all the context of your entire trip.

And then when you're finished with your trip, you can do a post-trip review. Maybe you've been uploading all your different receipts and expenses directly into that folder. At the end of your trip, you can say, "Hey, here was the budget that we had. Here is what we spent. Let us know how it went." As you can see now, it literally starts from just looking at what kind of question you have for Claude. And 9 times out of 10, if you've got a question for Claude, it's usually with a deeper purpose.

Like if you just take a step back, that question is part of a bigger task that you're working on. So it's all about taking a step back, asking why, and just understanding what are the different moving pieces. So at this stage, I think we're ready to actually create our first piece of work for our virtual assistant. There's a couple different ways that we can do this. We can either start from scratch, which is what I would recommend 9 times out of 10.

We can import a project. So if you were in the Claude chat mode and you had a project inside here, you can actually just import it directly into co-work. And then you can use an existing folder. This just means if you've got a folder with a bunch of files already in it, you can just directly tap into that folder. For me, I'm just going to go with start from scratch. Let's actually use our Europe example. I'm just going to call this Europe summer trip.

I'm going to copy each of these four different stages. I'm going to paste it into my instructions. I'm going to say, "You are helping me manage my summer trip to Europe across all the below stages. You're going to help me first IDA and brainstorm, decide which places I want to visit, what kind of budget I have. We're then going to find the cheapest flights. You'll then be helping me by being a contextually aware chatbot, and then help me with a post-trip expense review." Now I might spend some more time actually going through what each of these stages are just to give more context to Claude.

And then you can add some resources into here if you have them. For example, if you're doing that reporting trip, maybe right now you import the last year's results reports into here. Maybe you'll upload your benchmark Excel file. So that if any of the numbers drop below a certain mark, it already knows to escalate that to you. For our summer trip, we don't have any additional files. And then over here, the most important feature we want is that the memory is turned on.

If your memory is turned off, I'll show you how to get that turned on in just a second. Let us create this trip. And now what we see is our main high-level workspace for the Europe summer trip. Just like we saw before, this is exactly where we start new tasks and ask questions. We can attach additional resources. We can make sure that we're plugged into things like Gmail and ClickUp. All the same features we can either use the transcription service.

Now on the right-hand side, those instructions that we just added are actually project-level instructions. They will apply to every new session that you start with Claude. So I've just added some really quick initial context. I'm going to fire off this chat. And now we've just switched windows to actually go into this session view. But I'm going to go back one level, back to our parent folder. As you can see, that's the current session that is running.

I can go off and create a brand new session here with a different task that I want to work on. And then all the sessions are going to be stacked over here. So if I have three, four, five, or 10 sessions, they're all going to be available. Now the interesting thing is that the co-work project feature actually has persistent memory across different sessions. So if I have five or 10 different conversations across different sessions, my next session that I create will be able to call on the memories from my previous conversations.

It looks like we're needed. And now you can see that our virtual assistant is asking us some follow-up questions to try be even more helpful. I've clearly not given enough context letting it know where I was going to be flying out from, so let's just say New York. I'm going to be traveling with one more person. I actually like all these places, but I must also go to Poland, and I want a super relaxed trip. And as you can see, Claude's already started building different files for us.

The project Europe trip MD. MD is just markdown format. It's just like a Word doc or PDF. It's just a type of format. And then you also have a user profile, which is oh, down here now based on my preferences. I can click on these and I can actually open up whatever Claude's written down. So I have the Europe trip preferences, which is now going to be a hard memory file, which is going to be shared across all different sessions that I create.

Coming back down to here, we have our user profile with my email address and different preferences. And just to continue exploring, now we've got a really rich task list. So like I mentioned, this is actually going to be our virtual assistant. And actually at this stage, it still seems like a bit of a chatbot because I have to come into here. I've got to ask some questions. I'm going to get some responses back. But let's just figure out how to take this a step further and actually give an ongoing task to our assistant.

Now this is already pretty cool because it looks like we're making some strategic decisions around the cheapest route for us to go. New York to Athens to Krakow and then back to New York City. We got some high-level prices. But now what I want to do is actually go set up a task because I don't need to book my tickets just yet. And actually my second task is to find the cheapest tickets. And my favorite place to buy tickets is Skyscanner.

I want you to create a scheduled task where you go across to Skyscanner using the Chrome extension. And for the next 2 weeks, I want you to be searching for the best flights based on the trip plan that we have. I want you to be logging all the flight prices to an Excel file. And I want you to be checking every single day. Every time you check, log it into the Excel sheet. After 7 days, determine the lowest price of that last 7 days.

And then in the next week, the next 7-day period, as soon as we go lower price, book the ticket. I'm just going to make sure that I have my Chrome extension installed. Let's just go down to here. I have Claude in Chrome. Now in order to use that, you have to download the Chrome extension. Once you download the Chrome extension, you can't really see here, but there is a little Claude icon. If this is your first time, you'll just be prompted to sign in.

But after you've signed in, you you now have Claude access in your Chrome browser. So once this is ready, then this is fully enabled because it's going to be using it. So let's actually kickstart this process. As we can see, Claude had to check our trip preferences, which is going across to that project Europe trip MD file that that we created in that previous session. Coming down here, we have a warning about booking automatically, which is yeah, fair enough.

I wouldn't get it to book it automatically, but it just flagged it just in case. I've just answered a few more questions to give Claude some more context. And over here, we can see it's creating a scheduled task. All the scheduled tasks will live in this panel over here. And if we click on details, you can see the prompt that Claude created for this specific scheduled task. So it gives all the instructions about going off and finding the best prices, then creating an Excel sheet if it's not already created for the first time, or updating the Excel sheet if it's already there.

Literally all the instructions already created. So I'm just going to hit schedule. And one thing that you should do every single time you create a new scheduled task, is just go ahead and click run now on the task itself. So I'm going to go into the schedule tab. I'm going to find that task, which is over here. And I'm just going to hit run now. We can see the history says today cuz this is the first time it's running.

But then the accumulated history will just populate down over here. Let's jump into this task. As you can see, this task is its own brand new session. So I'm just going to go back into our parent folder. And you can see that that task just sits nicely in the session list. Now Claude's just opened up Skyscanner on my computer. And this is now Claude controlling the Skyscanner browser. As you can see, it's already input the details from New York to Warsaw.

And it looks like that first run is completed and we've got that data logged into our Excel sheet. Let's just spread this open a little bit more. We have the run date, the time, the lowest price $816, and then some more information about the flights. Now one final thing that I want to speak about is this live artifact section. And this just came out like within the last 24 hours, so I'm still playing around with this.

But let's say we have this Excel data over here, and instead of me going directly into a chat or into my folder and trying to open this Excel file to read some of this data, I'm just going to go back to my parent folder, and I'm going to ask, "Can you create me a live artifact that shows me the data from my Excel sheet in this folder?" I'm going to see if Claude can create me an artifact, which is just a visual representation of this data, so that it's just a little bit more engaging and interactive for me.

And here is that artifact that Claude created. As you can see, it's just a little dashboard, a portal, as a bit of a progress tracker on our goal, which is just trying to find the cheapest flights. The currently low price that we have, all the runs that we've done, and the URLs, like the key information from that spreadsheet. And as you can see here, it's already pinned, which means if I go into my pin tab and I drop it down, I can always just click on this button and then see that dashboard in a really easy way.

All right, guys, thank you very much for watching. If you've watched this far, I'd appreciate it if you could subscribe to my channel. If you want to see a follow-up of how to automate business tasks using Claude Code, drop a comment below. All right, see you in the next one.