Intro to Claude Cowork: The Power-Up with wordPower

追加: 2026-05-11

[00:00:00]Thank you so much for coming everyone.

[00:00:01]It's really nice to see you. Um huge thanks to our sponsors. Oh hello.

[00:00:07]If you don't know Oh hello, they are a fantastic professional mentoring platform with world-class mentors across marketing, tech, leadership, and many more sectors as well all through live one-to-one video sessions.

[00:00:26]Um and here's the fun part. They're giving away free membership for one team. So one team can get up to five of their staff free membership.

[00:00:38]If you already opted in during registration, that's great. There's also another chance to opt in in the post survey.

[00:00:48]And yeah, you might well be the winner of that mentorship.

[00:00:54]Um for those of you that don't know us, it's really nice to meet you.

[00:00:59]I've gone with a bit of a like a construction work style bar for our avatars today.

[00:01:06]Um but my name is Daff. I'm the chief learning officer at WordUp.

[00:01:11]And I'm joined here by Tom. Tom, do you want to introduce yourself?

[00:01:15]Hey there PowerUp crew. My name is Tom Rudden. I founded WordUp about 3 years ago to help businesses go from zero to one on their AI journey. We're increasingly helping businesses go from one to two on their AI journey at this stage as well. Um so as Daff mentioned, I'm excited to to jam out with you all.

[00:01:31]We've been doing these PowerUp sessions once a month. This is our second series. So excited to have another good audience with us today.

[00:01:40]Thank you. Yes, and that's us too, but you're definitely involved as well.

[00:01:46]Please do ask questions as we're going.

[00:01:49]You can use the microphone, you can use the chat, you can raise your hand. It'd be lovely to hear from you. And um, definitely get hands-on if you can.

[00:01:59]We'll be doing some work in, um, Claude Co-work. Um, so if you have access to the desktop app, do fire that up cuz we'll be doing some prompting all together, um, shortly.

[00:02:11]Uh Tom, what is the Power Up? What's going on?

[00:02:14]Yeah, so I hinted at it a minute ago.

[00:02:15]We're at We're at two of two so far. But we basically at Word Power, we talked, um, all day long to different businesses about how to implement AI in our day-to-day. And it felt like we had, um, extra We had more to say than maybe we were distributing to the world. We had a lot of individuals saying, "Hey, I'd love to get a peek at what you guys are doing and start learning alongside of you." So we started doing the Power Up as a monthly session to just kind of include more in our working development of AI knowledge. So we'll be picking a topic each month. Um, our commitment to you is it's only going to take 30 minutes. And as long as you kind of show up with the right tool and the right mindset, you will leave with something that you can use that day. So we're going to try to stay super practical. Uh you'll always see a slide or two from us. We have We have concepts to share.

[00:02:57]We'll stay super practical, grounded in actually doing work. So, um, you'll see you can come off mute and chat. You can raise your hand. This is meant to be an interactive work session. This is not us just lecturing to you. So, uh if you dig this, uh invite more folks for next month. I will do a little preview at the end of this part as what we're going to going to cover.

[00:03:15]Cool. And what are we up to today? Well, today is all about Claude Co-work. Um, so we've used multiple AI tools, um, in the past. Um, but we've started getting huge value out of the Co-work uh Co-work. Um, so we wanted to share that with you today. Share what we've learned so far. Um, so there's two quick, uh, parts to the session. Firstly, we're just going to talk about what makes uh Claude Co-work different from all other platforms. What makes it special. Um, and then secondly, we're going to talk through how to get the most out of the platform. Uh and that's where we'll get hands-on and do some prompting together, as well.

[00:03:52]Um, but Tom, you know, why do we love Claude CoWork so much? Why Why is it so good?

[00:03:57]Yeah, I think it's a pretty well-named uh chat platform. Um a lot of times when we're talking about AI, we're talking about chats, prompts, conversations, and the exchange of text. Claude CoWork is very much about completing actual work.

[00:04:09]Um and so one of the first things that you'll notice when you use um Claude CoWork is it can create plans and execute what's called long-horizon tasks or um work that takes a while. Maybe it has different um steps along the way. So when you're using CoWork, um a lot of times it'll create a nice robust plan.

[00:04:26]It'll work and check items off along the plan. It will review its own work before deciding it's ready to move on to the next step. So you can really offload task. Um it's one of my favorite tool to um fire something off and then go get a coffee or like go make lunch and come back 15 minutes later and have actual work produced. So very different from chat. Uh local file access. So CoWork is built um it's it's an iteration of Claude Code. If you have heard of Claude Code, it's been uh probably the single most um powerful AI tool for power users over the last year. Um and one of the things about Claude Code is that it actually works on your local computer with your local files, which gives you just way more scale um and power than what you would get from working in the cloud. CoWork uh being built on that same architecture does the same thing and it's a lot easier. It's built for non-coders specifically. So we get that sort of firepower um without having to be be technical. And then thirdly, connectors and user prone. Um so CoWork is very much about interacting with the outside world. Uh by connectors, um we mean uh a connection to an external application either to get data and use information and knowledge in your uh chat session or to actually complete actions um and and um uh create files or make changes to applications, which is a big uh paradigm shift from normal chat. A lot of times when you're using third-party apps or connectors um in a normal chat, you get you're getting uh information. Maybe you're getting um um chunks of documents and things like that. Using Kwork, you can do things like update your website or create new project tickets in your ticket management queue, um create and ship files, things that really um um are are real-life work. Um and also uh the native Chrome browser, which is one of the the connectors that's built into Kwork or or the native Chrome extension, you can actually pull up Chrome and start doing things um that you would have previously had to do using your mouse and keyboard right in front of you. So, you've got a major paradigm shift towards completing actions and actually doing work over chat.

[00:06:23]And I just wanted to give you a quick flavor of that. Um so, for those of you that haven't seen it before, this is the Claude desktop app. Um there's three tabs along the top. There is chat, which we still use quite a lot for kind of brainstorming and having a kind of general conversation with Claude.

[00:06:42]Uh there is uh Claude Code, which is for the nerds who want to code stuff. Don't tend to go in there very much. Um but there's the fantastic tool in the middle, which is called Kwork, which is built for knowledge workers like us, which harnesses a lot of the power from um Claude Code.

[00:07:01]And just to give a flavor of the type of work we're doing here, um this is a uh you know, example workflow where I've asked Claude for um building some session slides. I've provided it with some documents in the folder and an outline of the session that we're building, a deck template, and an example. And I'm asking it to pull in relevant information from Outlook and Slack in case that influences the content. Uh I'm then asking it to draft all of the slides. And then afterwards, I want it to mark the relevant task complete in Asana and message the Slack channel to say that it's been done.

[00:07:42]And what it does is it write it writes itself a progress um some tasks to do. So, as you can see, six tasks there that it's going to go through. First, it goes through all the materials that I've provided for it. It goes through Outlook. It goes through Slack looking for information. It asks a few clarifying questions about the deck that I want built. And then it goes away and it, you know, understands the Word Power Deck. It's using a Word Power PowerPoint skill, which we'll talk about a little later.

[00:08:16]Creates lots of sub-agents that do some of this work at the same time. And then it goes through and QAs what it's building as well. It spots issues. It wants to use the demo bursts and spotlights that we use in our slides quite a lot. It fixes them. And eventually, maybe it takes 30 minutes or so, it creates the deck for me for me to look at. And it goes ahead and updates Asana and Slack accordingly.

[00:08:48]As you can see, it's gone and done a a lot of work all on its own with a few bits of input from from me, but it's it goes ahead and does a a huge amount of work on its on its own.

[00:09:03]Um if I pop back into the presentation quickly, um so just want to touch on what makes um the Cohort so so powerful. So, there are four key elements to a really great Cohort setup.

[00:09:23]One is the connection, so using other apps as I demonstrated in that demo to really get the most out of your AI tool.

[00:09:33]Second is the account level instructions, so telling Claude exactly how you like to work with it.

[00:09:42]Thirdly, our projects, so getting yourself set up with a project which has consistent memory and context that will will use to build for you. And then fourthly, our skills, so scaling your repeating tasks using prompts and assets to really get the most out of Claude.

[00:10:02]And so we're going to go through quickly each of these and demonstrate kind of how to get the the best setup to get the most from your from your tool.

[00:10:12]Tom, is your Claude behaving itself and ready to go?

[00:10:14]>> Yeah, so I'm I'm back. We've got a shared a lot. Do any folks have any quick questions before we we hop into the exercise for today?

[00:10:22]Just a random question. And so for you said it would create a deck for you. So did you have it like connected to a deck creation tool like Canva or like what did you use?

[00:10:33]Yeah, it No, it was using its built-in PowerPoint tool.

[00:10:38]Yeah, it can do that natively. It can read PowerPoints and it can create PowerPoints for you.

[00:10:44]Pretty cool. Thank you.

[00:10:46]Same with Word documents, same with PDFs.

[00:10:49]Yeah, so major major differentiator versus chat platforms.

[00:10:57]All right. Should we get into an exercise?

[00:11:00]Let's do it.

[00:11:01]Let's do it. So I'm going to pull up my co-work now and we're going to walk you through a little exercise we built to try and show off a few of the different elements of co-work and really emphasize how it's different again from from traditional chat.

[00:11:16]Everything sharing okay?

[00:11:19]So this is my co-work account. Like you saw with Daffy, you have chat here, co-work and code. This is only available on the desktop app. So again, if you're not seeing this in your Claude web account, it's because you need to be in the the app. And a couple things I'll show you real quick and then we'll get into our our exercise. So, because CoWork is very much about doing work, you want to be mindful of your connections to the outside world. When I click this customize button, I have those skills that we kind of just hinted at a minute ago, um, but I also have these connectors, which are my external applications that I want to allow CoWork to interact with in different ways.

[00:11:55]And so, I'd encourage you all to navigate over there. That was customize connectors.

[00:12:01]And then if you hit this plus button, you can browse off-the-shelf connectors.

[00:12:05]And you'll see at the top some really familiar ones, um, that you can imagine are quite useful. Plugging into your Google Drive and allowing CoWork to peruse through all your information, um, create Google documents for you, um, really powerful. Gmail, um, drafting emails for you, things like that. Um, all sorts of third-party applications.

[00:12:23]We use WordPress, um, as one to control WordPress and build things into our website. So, we'll talk more about connectors. We're actually going to hit on it um, next month as our focus. Um, but if you're new to CoWork, suggest getting something plugged in so you can start to get used to interacting with, um, sort of the the outside world beyond your chat.

[00:12:41]Okay, the next thing I want to show you all is your account level instructions.

[00:12:44]So, in chat in, um, traditional AI chat, you've likely experienced how over several conversations there's a building memory of you. And that can be useful in some regards, kind of annoying in other regards, um, where you don't want context to bleed between projects. And CoWork has really firm rules about what information is shared in a given conversation. And if you start from scratch, um, in a fresh task or fresh project, CoWork doesn't know much about you the way it does in traditional chat.

[00:13:10]So, you have to provide that information ahead of time if you think it's going to benefit you. So, I'm clicking in my profile in the bottom left and clicking settings.

[00:13:18]And here I have instructions for my Claude account. Um, and these up here in the general tab actually apply to all of my Claude accounts. So, this is both chat and and co-work. And here I'm just going to provide some universal information like who I am, maybe where I live, um my teammates right now, a little bit of information about what I'm uh generally working on over like a long time period.

[00:13:41]And then, importantly, some direction on how to work with me. So, I just have some like work style nuances that I like. Um I want to get alignment from Claude before it goes away and waste too much time and burn through a bunch of tokens. Um I want it to always suggest a proposed solutions. So, based on how you work, there's all sorts of little notes you can put in here. Um I also explain how I want it to use connectors anytime it can use connectors.

[00:14:04]I suggest if you're new to um to Claude, um start simple here and then just kind of consider um going back here um adding to it over time.

[00:14:12]If you have this on your general Claude account, you actually have a co-work specific version of this as well. One thing you'll pick up on today is there's all sorts of different places you can drop instructions for how you want co-work to behave, and these layers are by design. Don't worry about getting too confused in them. The point is um start using them. You'll kind of get a sense of of where the value comes from.

[00:14:33]So, now we've connected our co-work account to external applications. We've given it some high-level information about uh who we are.

[00:14:40]And uh so, it seems right. One good way of getting that information is to talk to your AI tool as well, maybe in chat mode, and ask it to interview you. Ask it, you know, ask me 10 questions on how I like to work with AI, uh and then summarize that conversation into instructions. You can then, you know, copy and paste that straight into uh those instruction field.

[00:15:01]Yep. Thank you, Dov.

[00:15:05]Okay, great. So, now let's kick things off and actually do some work. When you're working in co-work, if you make an individual task because of um what I was just talking about, it kind of exists in isolation. Um so, anytime you're going to do sort of meaningful work, like come back to might have several chats. Strongly recommend uh creating a project. You can either click here in projects in the top left or when you're making a chat, you can scroll down and click create a new project down here on the work in a project button.

[00:15:35]So when we create a a new project, we have a few options. We can actually grab an existing project that we had in chat and bring it into co-work. If we have a folder where we've done a lot of work, we can tell it to use that existing folder. Or for today, let's start from scratch and we'll set up a new folder.

[00:15:50]So definitely if you if you can.

[00:15:54]Yeah, definitely we're dropping prompts in for folks in the chat.

[00:15:58]Yeah. Okay, awesome. You'll have the exact exact text that I'm dropping in and you can drop it as well.

[00:16:05]Great. So I'm going to name this Nike demo 3.

[00:16:10]And then I'm going to put some high-level instructions here and you don't need to waste too much time kicking this off. You can always adjust these later, but for this project, I'm going to guide co-work on general sort of work I'll always be doing in this project.

[00:16:22]So in this example, I'm going to say that there's a client Nike and generally speaking, we're going to be doing Google Analytics reporting here.

[00:16:29]I then can importantly pick the project location and you probably want one place where you where you save all of your projects.

[00:16:35]Um but you can grab You want to put this somewhere that you know how to access on your local computer. So Claude's going to work out of a folder on your computer and then you're going to want to go back to it over time, grab files out of it. So basically, just going to pick a location that is familiar and easy to access.

[00:16:56]Okay, so I have created my new project. You can see up here I have those instructions in my top right and then here's that folder that I'm working out of.

[00:17:06]So let's go ahead and start a conversation. For today's demo, I'm going to switch to Haiku so that everything goes nice and fast.

[00:17:13]This model selector down here allows you to pick the fastest, most efficient model, mid-grade model in Sonic, or slow thinking model in Opus.

[00:17:22]A lot of power users spend most of their time in Opus, but for today's demo, we're going to do Haiku.

[00:17:28]Yeah, and I'd recommend you um do the same um for the speed. Obviously, if you're doing real work, you'll probably want to use the uh most capable model, but uh for today use Haiku. Um Sam, there's a great question from Thomas Hartman in the chat. He says, "Can the folder be in the cloud like uh Google Drive?" Yeah, I don't know if you noticed, the one that I actually selected is in OneDrive. Um I We've kind of experienced mixed results with setting them up in Google Drive. Um as of our last testing, it was working just fine. So, I would encourage you on your when if you're trying to use a shared drive, you have to mount it like a local drive on your computer so that you can select it um like you would be able to through any file menu.

[00:18:07]So, you're not really using the Google Drive connector so much as you're you've mounted your shared drive as a local drive and have uh Co-work treat it as such.

[00:18:21]Cool. Okay, I'm going to kick off this prompt. Um so, I'm asking him to make a couple weeks of mock data. So, one thing you can do in Co-work that's really powerful is create files. So, instead of sending you out files today, we're just going to go ahead and make some right here in Co-work.

[00:18:39]We're asking it to include some basic GA columns, and we're going to um essentially build out a system here so that we can do Google Analytics analysis for our client at Nike.

[00:18:53]Now, as I'm working in Co-work, I have that progress bar here on the top right.

[00:18:57]You can see there's not a lot going on here cuz this is a very simple task, whereas what DAF showed you earlier, it had done some pretty comprehensive planning.

[00:19:04]Uh you can also see the assets of the folder here that it's working with.

[00:19:08]So, it has this um this is the folder, and then it has these two files that it just created right in front of us.

[00:19:14]It gives us a little bit of information, and that'll be our raw material that we will then use for our analysis.

[00:19:20]Okay, dropping in our next prompt here.

[00:19:23]Now, we're going to do a little organization in the folder. We're going to say, "Create an outputs folder in our project folder alongside that raw reports folder.

[00:19:31]Then, update our project Why would you do that? Why would you have a separate outputs folder?

[00:19:36]Yes. So, um as we continue to work in Cogram, we're actually going to produce files, we're going to produce work, we're going to have several conversations, we're going to come back weeks from now. Um actually pulling up the the folder that we're working in and grabbing files out of it is something that we're likely going to be doing. And so, arranging that so it's easy for you to make sense of is a good idea. And also, so it's easy for Claude to make sense of. So, as your as your project gets more complicated, your Claude's going to be more efficient if it knows where things are arranged, just like you will be uh more efficient um in actually going through that folder yourself.

[00:20:10]So, we created the outputs folder, and we also asked it to update those folder instructions.

[00:20:17]And that's what it's referring to as this claude.md file. This is an important concept that you'll see pop up the more you work with Cogram. These claude.md files are little instruction files that sit in folders that guide Claude on what to do.

[00:20:32]And so, here we have our um structure built out in kind of detail here for Claude.

[00:20:38]So, it knows that it has this raw reports folder. These contain our raw Google Analytics files.

[00:20:44]And if we're trying to produce a new report or a dashboard, start here. This is going to be your your raw material.

[00:20:50]We then have this outputs folder, and this is where we're going to produce our customer-facing deliverables.

[00:20:55]We also indicate, "Make these PDFs unless I tell you otherwise."

[00:20:59]So, I'm giving it more and more information. I'm building this system so that as I interact with this in the future, it already knows what to do, already knows how to arrange things.

[00:21:13]Okay, great.

[00:21:15]Now, one other awesome thing about Coda Work is, just like I'm able to see these files that it's working with and building in front of me, I'm able to see these instructions, I'm also able to see the memories that it's constructing. So, here I'm going to say, "Remember that our key Nike holds stakeholder is this person, and here's what she cares about." So, I'm giving it some kind of off-the-cuff a fact that will likely be relevant for other things that we do, and I'm telling Claude explicitly, "Remember this."

[00:21:43]That's a really important paradigm shift. Maybe you at times tell your AI chat to remember things, but the more you use Coda Work, you're going to be doing it much more regularly.

[00:21:52]The memory can get a lot more robust and detailed. You basically have an a index file that guides Claude on where to find all your detailed memories. So, you can really get a really comprehensive memory built about.

[00:22:06]And here you can see that right here. Here's the memory I just wrote.

[00:22:11]Nike stakeholder priorities.

[00:22:13]The client that I mentioned.

[00:22:15]And what's important for her, and what to do about it.

[00:22:19]So, imagine again, you're working with dozens of people at Nike, you can continue to add personal preferences, enhance your profile of Sarah, you know, and and store and make sense of what other what other facts you want on the account, and not lose track of those.

[00:22:37]All right.

[00:22:42]So, now we'll start a new task and kind of show off how the that context will will come with me.

[00:22:49]Here you can see that neat little folder arrangement with the outputs where our reports are going to go, the raw reports where our raw Google Analytics reports go.

[00:22:58]You can see the memory, that's that high-level index file, and then the individual memory files that it starts to create.

[00:23:05]I'm going to tell Claude to run a an analysis on week one and produce a single client-facing PDF.

[00:23:14]Actually didn't need to say PDF cuz it knows that I like PDF as we had said in the instructions.

[00:23:29]You can see down here um the skills that it uses during the session. So it grabbed a native PDF skill that it has to make PDFs. Going back to the question earlier about creating documents and PowerPoints and things like that, it has off-the-shelf skills for certain file types. Um so it's going to just use those uh kind of organically.

[00:23:54]The first >> uh following along here able to uh keep up?

[00:24:02]Jeremy's got a question. What are your thoughts on importing previous chat history from ChatGPT into Claude?

[00:24:08]Yeah, great question. So at the beginning when I was doing that account level um profile of myself, that would be a great place to include pertinent details from previous chat platforms. Um if you think they're important enough to influence every conversation, you could totally do something like ask ChatGPT to make a brief profile of you and your work styles and then use that as your your profile um for your uh chat platform. So don't need to start from scratch.

[00:24:37]Okay.

[00:24:38]So looks like I'm having a little bit of a preview issue here.

[00:24:43]But it made me a report.

[00:24:47]See if I can preview that from my screen. Put it in my outputs folder.

[00:24:53]And it's pretty darn slick.

[00:24:57]It for the time that it took.

[00:25:06]So, we get a table here at the top. We get key findings. Remember organic search was really important for our key stakeholders Sarah Chen.

[00:25:15]And then we get four actually fairly crude visuals here at the bottom. So, I'm going to blame this on hype food that it's not the most robust looking report, but the fact that it did this very quickly hopefully is impressive.

[00:25:27]We're starting to run up on time, but what we would do at this point is iterate on this report with Claude directly. So, we wouldn't start messing around with the PDF. We'd keep chatting with Claude and ask for improvements.

[00:25:39]Maybe we'd ask for visual changes. And when we get to a point that we really like, then we lean in on that skill creation, and we'd create a skill where we can replicate this sort of work in the future.

[00:25:51]So, you saw generally the flow of orienting your profile or orienting your account, building a project, doing work in the project, building a system that supports that, and then as you see repeatable work building skills for that repeatable work. And Tom, can you give us a one-minute on what a skill is?

[00:26:09]Yes. So, a skill is essentially a process that Claude remembers for you.

[00:26:15]And Claude Co-work has a native skill creation tool in it or skill creation skill in it.

[00:26:22]So, you can just ask Claude at any point in time, help me make a skill for XYZ.

[00:26:27]And it will work with you on that skill.

[00:26:30]In the future when you invoke that skill and say, "Hey, I want to make a on-brand PowerPoint presentation." Claude knows that and will go grab that package of repeatable process that you created with it.

[00:26:41]So, again a good workflow is to just do work in Claude. Like have a nice conversation, complete a task, and then say, "Hey, let's package that up and make it into a skill."

[00:26:49]I love that. And we will likely do a full session on skills in particular cuz they are so powerful but yeah, take a little bit of work to to get perfect. So here's what we covered today on the four elements of a great co-work setup. We talked about how connecting your apps, we talked about your account level instructions, we talked about setting up a project and we touched on creating skills to scale your work.

[00:27:18]I've just put into the chat the form. I would love your feedback please on this session so we can uh guide future future sessions.

[00:27:29]The next session we are doing is all about connectors, Tom.

[00:27:34]Yes, so sorry to hit today at how co-work uh some of the magic comes from actually influencing apps and actually making changes to actually doing your work maybe better said. So we're going to lean in specifically on this topic not just for co-work this is actually across all AI chat tools you see it in Gemini, you see it in chat GPT.

[00:27:50]Um we're trying the AI companies are trying to allow us to use AI as like a central command center more than operating system to control software around it. So we're going to lean in on that trend it's been really powerful for us and for our customers at Word Power and we want everybody messing around with it alongside us.

[00:28:07]Thank you. And if you feel like your team could do with some training on AI either fundamentals or creating workflows or building skills or any topics like that across any AI tool, um do drop us a note we'd love to hear from you.

[00:28:23]But thanks everyone really appreciate your interactivity today. We'll be sharing the recording and the deck and the prompt guide so you can recreate this yourself in your own time and questions over LinkedIn happy to talk further.

[00:28:40]Thank you very much everyone lovely to see you all.

[00:28:42]Thanks everyone.

[00:28:47]Bye folks.