4 Ways Claude Co-Work Automates Real Business Tasks (Full Transcript)

[00:00:00] Speaker 1: If you're still using AI to Google and write emails, then this video is for you. You're gonna learn how Cloud Co-Work can literally do your work for you. I'll show you four use cases that you can deploy today and with each one, a specific reason that Co-Work beats classic AI chat. I'm David DeWinter, your guide to all things Cloud. Head to the link at the bottom of the screen to install Cloud Co-Work and let's dive in. Once you're here, click the download button in the center of the screen and launch the app. Now, before we jump in, to follow along, you will need a paid plan. So if you don't see Pro or Max here next to your name, that means you'll need to upgrade your plan. If that's you, settle in because my job is to teach you if it's worth it or not. The desktop version of Cloud has these three tabs at the top of the screen. If you've used Cloud before, you've used chat. We'll move past chat today into this new mode called Co-Work. Now, just like with chat, you can choose your Cloud model. I'm gonna stick with Sonnet for this tutorial. If you've got enough tokens to use Opus, go for it. Now, here's what's different from chat. Every conversation has access to one or more folders on your computer. You choose those folders at the beginning of every conversation by clicking this button. For this tutorial, I've created a set of sample files which you can access from the resource guide in the description. We'll walk through each of these exercises today and I'll start by pointing Co-Work at this first folder, 01 Contracts. Now, 95% of you won't read this, but this is a real warning. In the course of its job, Cloud can permanently delete files or share files with third-party tools it connects to. For this tutorial, that's okay, but be careful in real-world use cases. Let's put this to work with our first prompt and ask, what's in this folder? After a couple of seconds, you'll see a response which lists a couple files and a subfolder. Remember, you can grab every prompt I use from the resource guide in the description. Now, this might not seem like much so far, but watch what happens when we put this to work. Say you wanna use AI to write a contract for a new deal. You've got notes for that deal, a contract template, and a folder of past contracts that you've executed. If you were just using chat, you might use a prompt like this. Use these notes to generate a new contract based on the attached template. Then you'd attach your notes file and attach the template file to the prompt and then send it off. In a couple minutes, you'll get a contract back that actually looks pretty decent. And you can download it using the button up top or using the file card at the end of the chat. But a lot of work with AI requires iteration. So if you needed to edit this contract using a new template or updated notes or even lessons learned from past contracts, you need to re-upload any of those files back into the conversation window. When you use Cowork though, your approach changes. First of all, because Cowork sees all files in this folder, you can reference files by typing the at symbol and then scrolling through the list. And if I submit this prompt right now, you won't see much difference between this and the chat version. But here's where it gets interesting. Every contract you create lives in this folder, which means that Claude can learn from all of them. So you start to change the way you prompt. Here's an example of what I mean. In this prompt, we're still asking for that new contract, but now I'm telling Claude it should pull terms from past contracts that look similar to this deal. And if you find some terms that might apply, let me know what they are. Now, this is our first big Cowork task. And on the right-hand side, you see that Cowork has made a plan and then shows the progress inside of that progress box. As it completes tasks, items get checked off from that box. You can also see which files it's used and recognized in the folder itself. Clicking on this folder icon also opens it up in your Finder or Explorer. And in the context card, you can see what skills it's using. We'll talk more about skills later. And if you haven't done big tasks before, you'll also see progress inside the chat window itself. Now, this contract still looks similar, but this time it pulled three rules from existing contracts. It noticed that in outdoor catering events that all items must arrive in insulated cooler boxes. It included our coffee station setup and our complimentary dozen donut holes since this event has over 100 guests. And all of these key points appear in the new contract. Now, what if you had 50 contracts to look through? That's reason number one to use Cowork over chat. Chat limits you to 20 file attachments per conversation, but Cowork can read your entire folder up to its token limit. There's really no set limit on the number of files. Now let's do something that chat can't do at all, which is work inside your browser. Here's a scenario. Say your business has a car loan and every month your payment covers both principal and interest. Only the interest is tax deductible, but your bank shows one payment. So when you log this in QuickBooks, you need to split it into two lines, principal and interest. Now you could read the loan statement and type these numbers in by hand, which is fine for one loan. But what if you had 50 of those sitting in your bank feed? Here's what you need to do. First, install the Clawed Chrome extension. You'll find a link to it inside the resource guide in the description. Then head back to Cowork and click the plus button in the left sidebar to start a new task. From the exercises folder that's also in the resources guide, choose number two, accounting. This folder has five loan statements which correspond to the five transactions we see in QuickBooks. Before we get into the prompt, make sure you've selected only the accounting folder. Cowork likes to select multiple folders by default. Now here's the prompt. We tell it to use the loan statements in this folder to categorize transactions at a specific link. We then give it rules on how to split the transaction between principal and interest. Let's start this off and set up the browser and Clawed side by side. Clawed starts by opening a new tab and it's attempting to open QuickBooks in the browser. I'll allow it just on this website so it doesn't try to go somewhere else. Let's switch to the new tab and see what Clawed does. It's already made it into QuickBooks and notice that I logged into it ahead of time so I didn't have to give Clawed my username and password. It takes its time reading the page and trying to figure out how to interact with it. About two minutes in, it makes it to the split page and then it fills out the amounts and the categories. Now the good thing about Clawed is it keeps checking itself to make sure it got the task right. It's not infallible, but it gets pretty close most of the time. This is the type of task you can leave on autopilot and get back to other work or do something else. That's case number two where co-work beats chat. It can actually do the work in the browser for just about any web app. While categorizing transactions isn't the most glamorous part of running a business, the money you save in finding tax deductions sure makes up for it. That's why for my real business, I rely on the sponsor of today's video, QuickBooks Online. With it, I can keep track of our products, invoice our customers, monitor sales, and track our growth. Plus they've got AI agents to save me even more time by auto-categorizing transactions and sending custom payment reminders. With all my financial data in one place, QuickBooks helps me make smarter decisions and gives me the confidence that I'm on track. Check out the link in the description below to get a discount off your first three months of QuickBooks Online. This next one is where co-work starts pulling away from other AI tools. We're going to build a PowerPoint presentation, but we're gonna do it in two steps so you can see something important. To get prepared, start a new task inside of co-work and then set your folder context to 03 Slides. Now this presentation is to teach my team how to use Cloud on the web. So step one is research. And instead of doing the research ourselves, we're going to outsource this to Cloud. We'll just point it to two sources. One is a previous YouTube video that I've done on Cloud, which you can find in the top right, and the other is the official Cloud help. While I get this started, notice that I'm not asking for the PowerPoint presentation right away. I'm asking for an outline, which I want written in a file called outline.md. This takes a few minutes to run, and here's where it gets good. The outline should appear on the right side of the screen. This is called a markdown file. Now just like Microsoft Word and Google Docs have headings, bold text, links, et cetera, markdown is a machine and human-readable text file that also has this formatting. In fact, you can see it's not magic. Go to the folder, right-click it, and open it with Notepad. You can read it and edit it before Cloud does anything else with it. Now let's say you're looking through this outline and you see something that you want to cut from the presentation. You'll see different slides in your outline, but I'm gonna cut this slide on artifacts. When I save this, Cloud Co-Work now sees this updated file. In chat, you'd have to download it, edit it, and then re-upload it. It's time to make our slides. Copy and paste this prompt from the resource guide. This is not a prompting tutorial, but there are four things I want to call out to you. This time we're clear that we're creating a PowerPoint presentation based on the outline. Number two, we explicitly do not want Cloud to research more for each slide. Number three, I've included a template that I've downloaded from the internet inside of this folder that I want it to use. And number four, we want more pictures on the slides and fewer words. Now, if you're doing this live with me, this is a good time to take a coffee break because this could take about 15 minutes. We'll see you when it comes back. In case you're just enjoying the show, this is the template that we gave it, and this is the presentation that we got. Obviously, it got the template, the colors, and fonts right, but let's see what else it did right. We do see a lot of the talk track is in the presenter notes and not the slides, just like we asked. It's also chosen some interesting images for a few slides. It also stuck to our 13-slide outline, including removing that slide on artifacts. It doesn't get the layout right on every slide, as you can see here, but man, this is a great start. So through this, you've seen reason number three that cowork outshines chat, when your workflow has multiple steps and you want to check the work in between them. If we were in the chatbot interface and needed to revise the outline, we'd have to download it, then edit it, then re-upload it for this to work, and cowork just streamlines that away. We've created documents, controlled a browser, and built a presentation. Now let's tackle data, and these data sets have a twist. In exercise folder number four, we have three reports extracted from three different tools, account data from our CRM, sales data from our e-commerce engine, and survey data. Our goal is to join them all together and see how sales and customer satisfaction trend across where we do business. Obviously, you could do this by hand, but that's not why you're watching this video. So let's head back to cowork and then start a new task. Now change your folder context to 04 data, and then grab this prompt from the resource guide. This is a hefty prompt, so let's submit it, and then while it's running, we'll go through it together. First up, we are creating a new Excel file called dashboard that has four tabs. The first three tabs have the data from the original three files I just showed you, and then the fourth tab joins all of it into a master table based on the customer email. That tab also needs a horizontal bar chart that shows the average NPS by state. And finally, we should be able to filter that data by month so that we can see changes over time. After a few minutes, you'll see a preview of the file that you just created on the right-hand side of the screen. You'll see all four tabs created, including the three from our original files, as well as the dashboard tab. This gives you a sense of what you're creating, but you probably just wanna go into the Excel file directly. In this view, it's clear that it's met the goal. It's got the bar chart of the average NPS by state, and we can filter by month using this dropdown. So why use cowork over chat here? Well, next month, you'll get three new files, and you don't want another dashboard Excel file. You wanna update this one. In chat, you'd have to re-upload this dashboard and re-explain the whole setup. So that's reason number four. When you're building something you'll come back to, like this dashboard page, Cowork keeps your files where they are and lets you build on them. Here's the twist though. What if you never had to write that prompt again? Go back to the exercises folder and open 05 more data. Pull both the April and the May folder inside the 04 data folder. Then I'll head back to Cowork. Make sure you are still pointed at that data folder. And this time, instead of using a prompt to create the dashboard folder, I'm using one to update it based on the April folder only. You can grab this prompt from the resource guide. No surprise, you'll get the updated file in a few minutes, but imagine doing this every month. It's very easy to lose track of prompts like this. Well, what if you could just say update the report and Claude knew what to do? That's called a skill. So after you've completed a task, literally tell Cowork to turn the update into a skill. Then Cowork reviews your conversation history to make sure it understands the task. If it doesn't, it may ask you some follow-up questions. Then it creates the skill and makes sure that it works. It does its own checking, but it also wants your help. And that's what this eval review screen is for. An eval is just a prompt and a result. So for all of the evals that Claude presents to you, it would like you to verify that it actually created the right output. I'll leave that verification step as an exercise for you at home. When you're done, you can either click submit all reviews or you can tell Claude to package up the skill. Then when you have a skill card in the chat that looks like this, click the button save skill. That installs your skill. So now you can click on new task, leave it as the same folder, and then type slash your skill name to invoke your skill. Now, instead of typing that whole prompt again, I can just say dashboard updater and then do this with the May folder. And voila, your dashboard updates are done in less than two minutes. Once you install a skill, it's available on any computer with Co-Work installed and Claude on the web. Now, you've seen Claude Co-Work create documents, navigate your browser, build presentations, and analyze your data, all from a folder on your desktop. This literally is a game changer. So post your questions down in the comments below and check out QuickBooks online with the link in the description. I'm David DeWinter, and I'll see you in the next video.