Name: Claude Cowork in an Hour: Where do I start?
Uploaded: 2026-04-01T18:32:02.238Z
Duration: 59 min 48 s
Description: Claude Cowork in an Hour: Where do I start?

Transcript for "Claude Cowork in an Hour: Where do I start?": Hello, everyone. My name is Matt. I am on the product team here at Anthropic, and I'm really I'm I'm joined by my colleague, Amelia. Amelia, you wanna say hey? Hey, guys. I'm on the customer success team. Yeah. We are both heavily, heavily involved in building out Claude Cowork, a product that we launched, honestly, just a few months ago, which is kinda crazy to think about. And today, we're really, really excited to help you get started on Claude Cowork. So we're we're gonna go we're gonna do it. A few housekeeping items before we get started. A recording of this session will be distributed via email within twenty four hours. So if you have friends, colleagues, you know, who couldn't make it, you should definitely share the recording with them. Hopefully, we'll have a lot of good discussion and conversation. If you go to the Q&A tab in Goldcast, you can submit questions at any time. Throughout the presentation, we'll be talking a little bit. We'll be riffing a little bit. We'll show some stuff. But throughout all that, we're gonna be taking lots questions, and I think the session will be a lot better if you all are kind of asking the things that you're most excited about, and we can hopefully help answer them. And finally, give us feedback. Like, Cowork is a very new product. These webinars are very new. We wanna do better for the next one and and the next one after that, and we're probably gonna run a lot more of these as Cowork gets more mature, more people start using it. So, yeah, we would love any feedback about this presentation specifically if you have it. Just quickly, Amelia and I can do quick intros. So my name is Matt Piccolella. I'm on the product team here at Anthropic. Specifically, I work on our enterprise team, and it basically means making Cowork into a product that can work in large enterprise contexts. We think of Cowork as a product that can really help every enterprise worker in the world. Claude Code is really tailored to that kind of engineer persona, and we want Cowork to become a product that can be used by anybody in the entire world, whether you're a lawyer, a a salesperson, a marketer, or a finance person. And so that's kind of my job here at Anthropic is make sure that large organizations who employ many tens of thousands of people in, you know, hundreds or sometimes even thousands of different job titles and functions, everyone can have that AI agent that helps them do their work, and we hope Cowork can help enable that. That's me. Amelia, do wanna you wanna also give an intro? Yeah. Totally. I'm Amelia. I'm on the customer success team, and I have been focused specifically on Cowork enablement. So I'm very, very keen to learn, like, how you guys, learn how to get started. This webinar is one of our first, scaled tests of that. And, yeah, we want Cowork to help you succeed at the first task you try in it, and it's my job to make that happen. Great. Well, welcome, Amelia. Yeah. Amelia and I work really close together. So, hopefully, we will, hopefully, be able to have some good good stuff to show you. First question is, who is this even for? Like, you know, lot of different people in the room. I'm seeing tons of emojis and reactions from, you know, the thousands of people who are on this call right now. These are just some of the questions that we've been hearing from customer calls, from our users on Twitter, our community, nonusers of co work. These are some of the questions we've been hearing. One, I think that really stood out to me at least as a person who's interested to make sure Cowork works for everyone is we heard one admin from a large company say, I asked several different people who Cowork was for, and everyone had a different answer, which I think in some ways is a good thing, but we wanna help kind of, show exactly who Cowork is for today. Another is why can I why should I use Cowork when most of what I can do is available in Claude Chat? You know, Claude Chat has been on the market for a couple of years. It does a lot of really good stuff. Why is Cowork any better? You know, some people really like this projects feature we've had in Claude Chat for a while. Isn't this just kind of advanced projects? And just a few other things, and I I hope some of the stuff we're gonna show today, will go into a little bit more detail. So probably if you're here, you probably at least have some idea of what Claude Coburg is. But just to ground us today, the high level way I think about it is Anthropic launched Claude Code in 2025, and, really, 2025 became the year of agentic coding. Claude Code kinda took off. Engineers all over the world are using it to to build all sorts of amazing things. And toward the end of 2025, we started to get a lot of attention from nonengineers, actually. So, you know, folks like Amelia, myself, product managers, customer success managers, we were using Claude Code for all sorts of different things. And, really, it was the same technology that was powering coding could also be used to power other things. The only problem was that it was in a terminal. And so, really, what we did in the beginning of 2026 was we said to ourselves, how can we take that same power, that same architecture, and the same kind of functionality that has helped explode AI coding and bring it to everyone else with no terminal required? So, really, we think that's three things. It's meeting you where your work happens. So, you know, the files on your desktop, the things that you're sending in Slack, the files you have in your Google Drive. Claude can have access to those, and Cowork can make use of them. Second is the ability to do much more complex work kind of through multistep projects and tasks that it creates for itself. And then third is much more than a QA Q&A bot. Cowork is about real deliverables and not just drafts. So the idea is that Cowork can help you build things that you're gonna send to your boss, that you're going to share with your colleagues. Like, these are real deliverables, not just kind of first drafts or things that you might throw away eventually. So to get into it, I kind of like to think about Cowork and levels, and this is where I'm gonna welcome Amelia to kinda start chiming in as well. Almost like a video game, Claude Cowork really is a kind of about levels. You start at level zero with kind of a lot of what Claude chat shows and is possible today. You advance to maybe a basic kind of single player use case. You advance to another level, and you think about skills. Level three is maybe where you start to bundle some of these things into a plug in. You, you know, start scheduling tasks. And then level four is really how can I do this for my entire company, building plug ins and tasks and skills for everybody across the company? Before we get in, Amelia, anything you wanna add here, or what what's your perspective? Yeah. Definitely. So I agree with the the level framework here. And I think it's important to really start at level zero, like just getting into Cowork and trying something. When we first get started, I think you can have a lot of, like, how is this different than chat if you haven't tuned into, like, the key differentiators of cohort. So one of the things I want everyone to come away with is the importance of connectors. And I'm gonna use a a demo to bring this to life, that really hits on level zero, how it can still answer your questions just like chat, but also level one where once you start, like, picking a folder and having it reference connectors like Slack, it could start really giving you, like, boosts to your day. So we'll hit on level zero and level one. I'm going to share my screen and we'll get into it. All right. So here we are in Cowork. You can see on your desktop app that there are three toggles now at the top between chat, code, and Cowork. We'll be focused on Cowork today. And the first thing I wanted to do is just ask a basic question like, what is the last message sent by our CEO in Slack? Now it's gonna start up. It's going to realize that it has the tools to go look at Slack for me. So I don't need to go move screens and search for the CEO name. It's figuring out the users. And the fun thing about Cowork is that you actually it's very transparent like what it's doing while it gets started. So we can see like the context here, if it's working in a folder, and what tools that it's using. So I'm gonna give it a tip and say that in this demo, the CEO's name is Stephanie. Usually, it's able to match that itself. And, unfortunately, in this demo, it's not Dario. But I can basically see here that it's pulling from conversations that are happening in Slack, and I don't need to move there at all. Now it's done a good job giving me external links so that I can go pop into Slack and check its work if I want to. One thing I'll call out is like I always wanted to add a link to Slack whenever I ask about Slack. And so I'm going to add another set of directions that gives it the instructions that whenever you summarize Slack, make sure you do include the links because I want to be able to check your work. And it'll start saving the details of my specific prescription for how to do this task. So I'll pause here because we've covered a couple of things that are in the zero level. We are asking basic questions. But importantly, we're using the Slack connector. We're not just going blindly in basic mode without having context available to co work. And we can see that I can actually use Cowork itself without leaving the chat to make sure that it remembers what to do every time I do something. So these are kind of the building blocks of Cowork that I want us to take away. Now one of the things that I think is also really important is just having Claude help you. So one of the let's, like, kick off another task. I want help making agendas for some of the meetings on my calendar in this coming week. So I have calls scheduled next week between April. Can you help me prep for those? I want to know details about the companies and key questions I should ask. So if we pause here, we can see that these are this is just an open ended question. And I know that I've set up connectors like Gmail and my calendar as well as HubSpot and Slack. So it's gonna have a host of things to pull from. I can also give it specific project context too. So I have a customer management project that has, like, a whole series of files, one for each customer, and I'm just gonna throw that at it as well. And maybe maybe if I could just chime in here, Amelia. I feel like you just showed, like, two absolute magical moments, at least for me. Because when you when you really zoom out and think about it, the product is called Cowork. This is meant to be a product that allows you and Claude to Cowork together. And for me, what Cowork really means is really about selecting a folder that both I have access to and Claude has access to. So almost like, I can add some stuff. Claude can read it. Claude can add some stuff. I can read it. Really, I think of this folder as kind of the way through which people co work. So, yeah, I think the absolute critical bit is both selecting that folder so that you can have a kind of shared workspace for you and Claude and then also all of the connectors. Because what is really co working if Claude doesn't have access to your Slacks, your docs, all that stuff? So, yeah, I just wanna. It might seem small what Amelia just did there, but, yeah, absolutely game changing. And I think one of the biggest differences between chat and co work. Yeah. So thanks for calling that out, Matt. You will get this this pop up that says, can I change files in my customer management folder? And this is a warning because Claude is getting access to a folder on your computer, and it can add and write to those files. And so this is extremely helpful, but also something to just keep in mind in terms of how you set up the folder structure that Claude is referencing. You want it to be something branched from, like, anything extremely, like, critical that you don't want edited, and you wanna isolate it to, like, your Coworking role folder that will start building on, like, your day to day tasks and host all of the information that you need. So I'll build on that point as we continue to show demos. But for now, I'm gonna allow it access to this customer management folder, and let's go. Let's see if it can create some prep for me for my meetings next week. I think I have, like, eight demos. And I don't know if anyone else feels this, but being back to back on calls, like, you're often kind of just going in a little bit colder than you might want to your next call. And having something like this helps me get ahead of that feeling. Yeah. And, man, it's just this always, like, I don't know, scratches my brain in the right way. Like, when Claude's cooking on a task like this, I love kind of, oh, there's it hit my calendar. It hit my HubSpot. Like, here's the progress it's doing. Like, it pulled in the doc x skills so it can make a proposal for me. Like, I just love this context window, the progress. Like, think a big thing that used to scare me about AI was I didn't really have any idea what it was doing. I didn't know whether it was staying on track. It was just like, oh, am I just gonna wait ten minutes, and it will have done, like, all the wrong things? And so, yeah, I think one of the better parts of Cowork is that it kinda shows all of this stuff that is happening. So, yeah, I it just yeah. Really love when this happens, and it it makes my work just so much more joyful to, like, feel like I'm in the loop as Claude is working. Yes. And you can offload, like, the things that just take just enough more time to do, like prep really well for every single call that might slip as you get busier with more strategic things. So offloading to Claude is, like, definitely the way to go. And I agree. You can either collapse these elements, but you can see Claude makes its own checklist. Importantly, each of these items are also steerable. So if I wanted to, as it's, like, reviewing and verifying the document, I wanted to, like, make sure you double check the HubSpot info is pulled in. I can, like, queue that up on top of the existing tasks. So you're not, like, locked in in full waiting for Coworker to complete the task. You can always use one of these, like, kinda intervene on any of these steps and give it further instructions. Maybe while we let Claude cook, Amelia, is there maybe a question that we should start to take? Yeah. I think one of the things that is getting asked is about like, existing Claude chat projects. And I know that projects are are beloved in the chat surface. So maybe we hit on, like, how you can also create projects in Cowork and hit that topic. Oh, yeah. Of course. So if you hit create new project, Amelia, Yeah. you can actually import projects from Claude chat and bring them over to Claude Cowork. And so if you have existing projects like this test project from Amelia, this is a Claude chat project. This is basically allowing you to migrate your chat projects to Cowork. There are a lot of benefits to Cowork projects over chat projects, which if you hit create, you can see some of those things. So for example, like, there's much more granular memory. You can set scheduled tasks associated with projects. You get kind of the benefits of local folders. The only problem today is that like co like, Cowork Cowork projects are only on your desktop. The team is hard at work, both bringing a Cowork remote version so that you can access it on the web. You can share things with your colleagues. But for now, these projects are just totally on your desktop, which is kind of the shortcoming. So it's almost like today, there's a lot more power in Coworker projects. You get scheduled tasks. You get complete folder access. You get each of the sessions kind of producing things in the way Coworker would. But then the downside is that you can't collaborate yet. But we expect probably in the next, you know, maybe four or six weeks or so, we'll have a much better solution there. So, yeah, that's projects. But I I absolutely love projects, and I'll show a demo later of of some of the stuff I got cooking on on my own projects. Okay. Nice. So we see the notorious blue dot that tells us that this task is done. Claude has done a lot of work. It's completed the its whole checklist and has created this doc for me that's like a call prep document that it'll load up in this sidebar. And I can see for each of my calls and their time, like, what sector is this, what is the key context it's brought in from my CRM, and also loaded up some of the, like, key questions to ask. And I'm gonna be able to, like, fly through all of my meetings next week using this, like, pre prepared doc. Now one of the other questions in the chat was around, like, skills. And so right now, all I've done is had Claude build me this beautiful call prep doc. But I want to encode what we've done so far into a repeatable skill so I don't need to ask the same question twice. And the best thing about Claude and Cowork is that you can always ask it to help you do what you need to do. And this is like the most important thing to come away from this webinar with is like I can just ask, hey, can you package the call prep workflow into a skill for me? And Claude will get to work actually doing this on my behalf and come away with like a nice skill in a zip file saved that I can use over and over again. And I don't need to go through the trouble of like writing the directions from scratch. Just to, like, zoom out for a second while this is working, if I go to customize, this is where you'll find skills and connectors. And so and plug ins, which we'll talk about later. But for now, I'll investigate some of the skills that are already available. If you're in cohort, you'll see schedule and context are pre built. And you also will have this skill creator skill, which is kind of meta, but it's honestly extraordinarily important because it allows me to ask any task to get packaged into a skill for me. So I don't need to, like, write down every single step myself. All that a skill is is really a set of instructions that tells Claude how to do something. We can see in the, like, skill MD file, which is like the home for the instructions. There's always a description field and a name. And so when I'm talking to Claude in a chat and I'm saying, hey, can you make this a skill? It knows to invoke that skill creator skill and do that on my behalf and follow like all of the directions written down here. Well, I have made a call prep skill before. And so we can see some of the outputs here where this skill has this description. It's for building my meeting agendas, and it's called call prep. So we'll see what it's done. It's still working, but it's using the skill creator skill over here to get this all all documented on my behalf. Since it already preexists, I'll just show how when you invoke a skill, you can do it by writing a slash command and then invoking call prep, and it will do all of the, like, instructions that have been encoded there. Matt, anything I've missed on skills, or do you have any demos that you'd wanna show on skills that you rely on? Or Yeah. I I think it makes a lot of sense. Maybe I can take over screen share, Amelia, because I I'm seeing a couple pretty uploaded questions about the differences between all these different things. yeah. And I actually think it's it's it's pretty basic, honestly. So let let me show you. So this is my customized screen, and I think a big part of what we've done with Cowork is bring everything that you need to customize your agent into this customized menu. I think of it basically there are two concepts you really well, maybe three concepts you really need to know. Skills are what Claude does. So almost like think of them like workflows. I think in maybe 2010 or 2020, you might have built these workflows as, like, drag and drop, like, if this, then that, or, like, do this. And then when this happens, do that, like, in kind of a very, like, graphical way. Skills basically allow you to do that, but with human language. So this is probably the exact same document that I might send to a report or a colleague or something like that. Like, if I had a chief of staff that I wanted to prep my morning briefing, this is basically exactly what I would send them. So I must think of these as, like, scripts for Claude that can be written in human language exactly as you might tell a colleague. And to Amelia's point, all you need to do is slash command and run them. Second are connectors, and this is just giving Claude access to all the data that you need. So in my case, I'm a product manager. I'm giving it access to Asana, Figma, HubSpot, MS 365, sixty five, Slack. So that's basically it. Skills are what you do, and connectors are, you know, what data you're accessing or things like that. Then all you need to think about for plug ins is that they're the bundle of many of these things. So the way we think plug ins work best are kind of on the per job function basis. The idea is that, you know, a function like a product manager or a salesperson has kind of many things that they do on a day to day basis. So if I go to sales, you see that it's just the same two things always. Skills are what you do, and connectors are what data you connect to. And so I almost think of these plug ins as, like, mini apps. Like, in some ways, it it feels kinda crazy to say, but I think these are almost like the iPhone apps of the future where, like, you used to write apps with code. Now you can just kinda write them with human language and connecting to other data sources. So if you're a salesperson, these are some of the common things that you do. You know? Amelia showed call prep. But it's not just call prep. It's account research. It's competitive intelligence. It's creating assets. It's doing daily briefings. It's drafting your outreach. And I think what we were seeing from a lot of customers when we had launched skills maybe five or six months ago is that they were just getting overwhelmed by the volume of these skills. They had hundreds or thousands kind of floating around. There was no kind of way to provision these to people. Plus, there was also no way to kind of connect them with the tools that you might need. So the sales plug in here, it kind of allows you to pick and choose which tools that you access. So these are some of the most common sales tools that we've seen. So, you know, Close, Clay, ZoomInfo, Notion, MS 365. So this kind of allows each user to kinda customize what tools that they're using. To the point around slash commands, we actually just deprecated slash commands. So we heard all the feedback, and we were like, yeah. Skills and slash commands are basically the same thing. So we actually got rid of them. What we did add, though, was the ability for you to say this is only invoked by the user or only invoked by Claude or both. Because a lot of the best, I'll show you an example. This is our productivity plugin. This memory management skill is basically direction for Claude on how to manage its memory. And so in this case, this is not something that the user would ever wanna know. Almost think of it more like documentation for Claude. So we say, okay. This is only gonna be invoked by Claude. And you can see that there's not really a task here. It's more just a bunch of kind of documentation. So we got rid of slash commands. I would say the I would say sub agents, I basically think of, like well, I would say sub agents and hooks are the other two features of plug ins. They're both much, much more advanced concepts that I think most users would never use, but then also just kind of cover conceptually different things. So sub agents are almost like subcontractors. Like, say, a salesperson, you wanted to have a sub agent that was, like, deck proofreading or something. There's a lot of benefit to having separate context windows, and the sub agents allow you to do that. And then hooks kind of allow you to do things in response to certain things. So at the beginning of every session, do this. At the end of every session, do this. But would say, like, probably 98% of the plug ins that we see are just the basics skills and connectors. So I think walking out of this, really, the only two things you need to know are skills, connectors. You take a bundle of both of those, you make it a plug in, and that's basically it. Does that make sense, Amelia? I feel like Yeah. yeah. That's how how at least I think about. it. They're like the composable parts that help Cowork do what it needs to do. One analogy that's been landing with customers is thinking about this in terms of like you have an electrician or you have a plumber. And both of them are going to have separate tool kits when they show up for a job. And those tool kits in Cowork land are gonna include, like, skills and connectors that are specific to their roles. And then maybe they use subcontractors, i.e., sub agents or, like, need a hook to do something if something else happens. But those are almost besides the point. What's important is they're showing up to their job. One's an electrician, one's a plumber, and they each have different toolkits. Yeah. So Exactly. And and the idea is that we kind of have one of these for each of the different functions that are kind common within organizations. So this data plug in is actually a bit of a banger. let's see. it. I just I I use the data plug in all the time. This is not Anthropic Real Organization, obviously. But, yeah, the data plug in is just, like, so, so, so cool. But, yeah, I'll I'll maybe I'll send it back to you, Amelia, and you can check back in on some of the demo stuff we were doing. Yeah. Definitely. And I wanna kick kick off, another element in a project too. So let me share my screen. Let's see. Okay. So we're in we're in projects here and we have the test project that I created as an example. But we also have this customer work project. And I've hooked this up to a customer folder that is set up and nicely, like, isolated. I actually just keep it on my desktop because this is, my home base for work. And just so you have some intel on how I got started in co work, I actually did something a lot like what's in this chat where I said, it's my first day in co work. Like, ask me about my role, what tools I live in, and where I'm always behind, and help me get started. And so this is a great way to just jump in and start doing something in co work. But the more advanced version of this is also having Cowork that works for your job. So what Claude Code and also Cowork are really, really exceptional at are like reading and navigating file systems and writing code. And I wanna explain, like, a little bit more about the file system piece here. So in my customer work, I have one folder for each customer. And within each customer folder, there's, like, a folder of all my call transcripts and then a specific, like, instructions folder that's tied to that customer. And so what this lets me do is, like, tasks like this where I need to understand themes and patterns from all of my call transcripts across my customer book. And I want it to read and pull out specific product feedback that it has. And so when you're in a project, you can actually just, like, stack a bunch of tasks all in a queue and do, like, multiple things at once, which is really nice because you can kinda just have this rotation of, like, checking in on the tasks that are underway. But I wanna point you to, like, its thought process. So it thinks a little, understands the task at hand. And then I like that it calls out. I see 12 customer folders. That's the, like, architecture I I explained about how I've set up folders to kinda work for work for my job. And within customer book, we can see that it's, reading all of my call transcripts and the notes and starting to kinda get to work on a doc for me to summarize. And, Amelia, yeah, maybe maybe. could I ask you to just click that folder icon next to customer book? Because I think it really what finishes the the really what finishes the mental model for me is this is really just a folder like any other folder on your desktop. I think maybe. you're not sharing your whole screen, so may maybe Yeah. Yeah. I'll I'll share the the just the, like, file window where we can see a lot of stuff within customer management. But the key thing is this customer book where each of these folders has an account.md dot m d. .md is just short for markdown. And it's like the the structure that Claude is best at reading, basically. It's like code like text, Yeah. but don't really have to worry about it beyond that. And then these, like, transcript folders. So it's a nice little, like, organizational choice that is very, very compatible with how Cowork thinks. Did I miss anything there, Matt? No. No. It looks great. Alright. So we can share again and see how it's progressing. It's still, like, in the writing the file stage. We can see that it's pulling in, like, docs skill to be able to write a word doc for me. So we can kick off, like, another task and then revisit this one. Another thing that's, like, on my mind actually about my customer book is, like, what should I be worried about? So let's just ask it that and see what it comes comes back with in terms of at risk accounts. And, yeah, we'll see what happens. The other maybe we'll oh, sorry. yeah. I was gonna say. maybe we take. No. a question then. can take a question, yeah, while these load. Cassandra Wu asks, hello, Cassandra. Thank you for joining us. Can I have it linked to a folder to Google Drive? Not yet, but we are we are hard at work making that possible. That's yeah. It's one of the top kinda questions that we get asked. But, yeah, not yet. How can I connect Cowork to my Outlook on Office 365. I think we actually already have that. So, Amelia, do you have the Office 365 connector set up? I don't have, it. I can show it quickly. Yeah. I don't have it here. So so grab the screen. Yeah. Let me, let me show quickly. So, yeah, I think we are actually this week, maybe maybe the question is around pro and max plans. We're launching to pro and max, I think, this week, actually. So if you're on a pro or max plan, like, get excited because it's coming this week. But, yeah, our MS 365 connector allows you to search Teams chats, SharePoint, Outlook calendar, Outlook email. We're looking at adding more kind of right tools. A lot of this is more reading. But, yeah, I believe that Microsoft three sixty Office 365 is already largely supported. So that's very exciting. But, yeah, if you're on a program max plan, you should be excited because it's coming this week. Okay. Let's see. So we've got some progress, and we can check out what's been cooked. Here's where we're at with the, like, customer book analysis. So it's built me this this doc that I can open in Word or soon we'll be able to open in a Google Doc. And it basically explains, like, all of the themes from at least 20 transcripts from the past week. And it's telling me, like, exactly what accounts they're tied to. Looks like we have some, like, NetSuite integrations in terms of sync timing that we're going to need to fix. And without even I didn't even give it, like, much specifics here on, like, coming up with account, like, renewal risks. But it built this, like, beautiful status for me to know what customers are, like, most at risk based on these product feedbacks. I've given a doc is great, but I also think one of Quad's strengths is, like, building these, like, more interactive dashboards. So I just had it say I just kinda instructed it. Like, can you make it more visual? And we'll see see what it comes up with. The other one that I had kicked off, and I can always see these from the, project page where it also shows any outputs that are tied to the project in one place, kind of like a Google Drive within Cowork. The other thing I wanted was for it to identify what customer accounts are at risk. So we can see that it's pulling from Slack, my calendar, and Gmail to identify these, like, at risk customers. Now we know HubSpot is connected here. So I wanna make sure that when it's doing an at risk analysis that it also enriches the data with the data that I have in HubSpot. So I'm gonna tell it, like, specifically, can you make sure to enrich this output with the, like, contract size and other details about the account available in HubSpot. And this will, like, specifically invoke that connector. And fast forwarding a little bit, I'll then make this whole work stream a skill that's called red flag alerts. And it will encode the instructions to include HubSpot. So I shouldn't have to encode or tell it to do that again. Yeah. This is this is so awesome, Amelia. Yeah. I think, honestly, the making custom HTML artifacts or dashboards is probably one of my favorite things. Do you mind if I take over screen share? And I'll also answer answer Nathan Strang's question around, Claude building docs in corporate formats, Yeah. Yeah. because I think that's one of my favorite things about, one of my favorite things about Cowork. I would also say that, like, I live, breathe, and sleep skills. Like, I think skills of the future, skills are everything in my opinion. So you'll see that almost every single one of my conversations with Cowork is a skill that I'm invoking. It's almost like, if you do something more than once, you should make it a skill. And so one of my favorite skills that our sales team uses a ton is this create an asset skill. It's basically you know, if you're a salesperson in which you know, I'm a product manager, but I actually have been doing a fair amount of sales. You know? Amelia and I go into a lot of these customer conversations. And in every customer conversation, you want that really high polished deck, one pager, landing page, just something that's gonna, like, really wow the customer. And so what this skill does is it basically creates assets. It asks some questions. It's trying to get a big sense of the pain points, like, lot of the stuff that a salesperson would do every day. And so you can see I actually ran this for one of my customers, and so it created this amazing landing page for Jones Capital, one of my customers. It shows, like, how we help, kinda key use cases, why Silvern. And so I can literally just send this to the customer, and this is gonna wow them a lot more than any deck. It's something they can share with their executives, stuff like that. But the best part of skills are how you can mix and match them. So I actually also have this skill called brand guidelines. And if I click it, it basically says, like, hey. Can you apply Anthropic's brand identity to this? So, you know, it shows a bunch of colors. It shows accent colors. It shows what typography you should be using, how to style text. So exactly what you were asking, Nathan. So then what I'm gonna do is kind of mix and match. So I'm gonna say, hey. Slash brand guidelines, can you use this skill to update the landing page? And so Claude is gonna, you know, take that skill. It's gonna read it. You see it just read it down here, and it's gonna take that thing that it just produced and almost remix it in the style of this brand guidelines. So, yeah, like I said, like, eat, sleep, and breathe, you know, skills. I think skills of the future. I have hundreds of skills that I'm using. Everybody at Anthropic has tons of skills that they're using. It's really just an amazing way to kind of encode a lot of the busy work, the stuff that we're all doing every day into kind of repeatable bundles. And then, yeah, I think, you know, taking taking skills even one step further, you can even schedule skills. So, Amelia, I don't know if you had a I'll let this cook, but do you wanna show some of the scheduled stuff that you were planning to show? Because I think I think scheduled is just amazing. Yeah. You can stop screen sharing. I'll grab I'll grab the screen. My scheduled task actually just went off, so this is perfect timing. So I I created, like, a a daily haiku DM that gets sent to my personal Slack. And we can see that this ran this morning when I scheduled it. And now I set it to, like, run daily at 01:30PM. And so if I go, I can screen share my, like, Slack as well and show exactly where this lives. I actually have no idea what it's going to have said today. This time, it was kinda cheap. Hopefully, something good. Yeah. Okay. So we've got our our Slack. This is not Anthropic Slack. It is a Silver and Capital, example org. But, if I go to DMs, I can see I've got some little Haikyus ready for me. It looks like it went off PST and eastern time. So it actually sent two, which is interesting. But, yeah. This is all to show how you can schedule a task, have it sent outside of Cowork to, like, hit whatever tools you need updated or messages sent to. And I'll show you how to compose one yourself. I wish we had a poll kicked off to, like, have people vote on what Haiku they like. But this is about DC. This is about general ledgers in finance. I think the first one from this morning is my personal favorite. But, anyways, let's go back into, Cowork and see how you actually create this. Alright. So when we're in, like, a new task let's just start this from start this fresh. I want to create a scheduled task that analyzes my transcripts. Actually, we'll do this in the project. Sorry. So in customer work, we're gonna set up this, like, at risk accounts piece, as something that's that's scheduled. So we are looking at, like, what are the what are the at risk accounts? What's going wrong with them? And I asked it to, like, draft messages to Slack and also to help me prep emails to those accounts. So I'm going to have things once these are connected, I'll have things sitting in my inbox or my, like, drafts within Gmail that are drafted for me to actually just, like, approve and send off to the accounts. And I'll send these into to Slack as well. But it it's really changed how I work. Like, I don't know about you, Matt's Screenshare, but I just sit in Cowork all day and I'm, like, interfacing with all of these other tools, like, from Cowork as my kinda, like, central console operating system. Yeah. 100% agree. I think those MCP apps that you just showed, if if you notice kind of that box that had the send button, our partner we are great partners with Slack, and they built a new MCP app that basically allows custom Slack actions to render directly inside of Claude. So it enables a lot more of this interactive UI versus just pure text, I think. Yeah. Stuff like that has been so magical. And so many different apps are starting to implement these MCP actions within the chat directly. So, yeah, it's been amazing. Yeah. So now our team knows that, like, people are having serious mobile receipt upload issues, and I haven't had to kind of switch context or switch tools. And then within Gmail, I won't switch screens and show it, but there will be, like, drafts, sitting that I could approve, edit, and send. Now I'll just call out, like, I was super nervous about trying this for the first time because and even connecting Slack or connecting, emails. Because I was like, I don't want Slack I don't want Claude just, like, sending stuff without me being able to check it. And it's very important to note that, like, I'm still in the loop here. I sent them off really fast because I, like, trust Cowork a lot now. But you can always edit and check the work. It's not, like, automatically sending without your approval. Yep. This is amazing. Yeah. So we've got some work done. That's awesome. Let's see what else oh, we wanted to schedule this. Okay. So once again, back to the theme of the call, like, can you can you help me put this work stream on a recurring weekly schedule? I want red flags surfaced every Wednesday at let's see. We'll do 01:55. So maybe it hits right before this this webinar ends. But we're gonna get some prompts on, like, Claude is running the schedule skill. And it'll prompt us to kind of approve approve the schedule task. Matt, what what tasks do you have, like, scheduled for your work right, now? I I do so many. Maybe I can take over screen share while this works. Sure. Yeah. I mean, I think so I'm a product manager, so I'm gonna I'm gonna show things through a very product manager lens. Also, I just wanted to close the loop. So if you remember, we ran this brand guideline skill. And I don't know how much you know about, Anthropics brand guidelines, but this is pretty spot on. It's the fonts. It's the colors. It's the orange, the black, the tan. So this deck is now totally in Anthropic colors, and you can kinda tweak it to your heart's content. But to me, you know, if I were going into a meeting with Jones Capital, I'd feel totally confident sending this to them. But the way I I like to think about Cowork is, like, Cowork is everywhere where I'm not. So it's almost like, what if I could clone myself a 100 times and have it be in every room, in every Slack channel, in every place that I'm not? So a lot of the kinds of things that I like to do scheduled tasks for are kind of like monitoring. So this is just an example of the kind of thing that I'm doing in co work, but I have this, I have this skill called user research synthesis. So if I go to product management, basically, I have a skill called user research synthesis. And, basically, it says, oh, like, you know, read these transcripts, like, look at the things people are talking about, try to pull out insights. And, you know, there's so much amazing user research going on around Anthropic, but I just don't have time to read it all. And so, basically, what I have is Claude do it. I said, hey, Claude. This is how I synthesize user research. Can you look across all of our Slacks, find any user research that has been sent out, and then summarize it for me and send me a message. And then so, basically, every day now, I can come back and say, hey. What was all the user research that happened? Give me a summary. And, you know, this is like yeah. I almost think one of the best use cases for these scheduled tasks are, like, monitoring and, yeah, being where I'm not. Mhmm. Because there's happening in Slack and across so many different tools that I just literally can't read it all. But the good news is that Claude can read it all. And if I give it just a simple instruction and a skill, what I what I want, how I want it processed, what how I want it to be delivered, then it basically helps me be in a bunch of places that I'm not. And so this kind of monitoring, scheduling tasks, all that stuff is just, like, so, so, so useful to me. And I know a lot of our just to finish the swing on all the other functions, I know a lot of our legal teams do very similar stuff. So, you know, they're looking for NDAs that are being discussed and scheduling these. They're looking for things that may be legal risks and trying to flag them. So it's almost like, what are all the the mental model is almost like, what are all the places that you wish you were and the things that you wanna be monitoring but aren't able to? And if you know, make a skill for it and then schedule it. Yeah. I think that's that's how I think about it. And I think somebody in the somebody in the Q&A had asked what the difference between a skill and a scheduled task is. A skill is what you want done, and then the scheduled task is when you want it done. So they actually combine really well. So you can see that this scheduled task basically just is me running a skill at a certain time every day. So that's the idea. Perfect. Do you while while you've got the screen, do you want to discuss, like, how to share skills? I know we just launched something exciting on that front. Oh, yeah. Yeah. Absolutely. So within a large organization, you know, Anthropic has many thousands of employees. We kind of think of there being two separate ways to share things within your company. First is maybe the, like, top down, which is, like, some admin centrally make some stuff and then shares it out with the company. And so if you actually go to this section, these are actually plug ins that are 100% customized just for Silver and Capital. So if you go to Anthropic and Partners, this is the finance skill for the world, kind of our off the shelf template version, but this is actually our finance skill just for Silver and Capital. So you can put all sorts of confidential information here. You can tweak how things are done, you know, for your company, stuff like that. And so if I actually go into this organization settings and go to plug ins, you can actually see that I'm live syncing these plug ins from GitHub. And this, yeah, this is just one of my favorite things, and I think a lot of our enterprise customers have absolutely loved this. Mhmm. The idea that you can build plug ins in, like, some central GitHub repository, which is how a lot of our customers build large plug ins, and then just sync them whenever you update them. So this means that kind of like we talked about plug ins being like apps. This is basically your app store. You can have custom apps for your company. You can sync them whenever they're ready. You can assign them to different groups. So, you know, maybe I want oh, I actually already did this. I I made sure the product management, plugin was installed by default for product managers. Perfect. So this is very much a, like, top down kind of centrally configured thing. But, actually, we just launched yesterday is we actually think a lot of the best kind of innovation at a company happens more bottoms up. So say I build some amazing skill. You know? Oh, I I have this amazing skill I'm really excited about. Do you actually have a skill that you've created, Amelia? Maybe you could show this one, and then maybe you could share it with me. Yeah. Let's let's look at that that call prep one, and we can see if there's a way to share it. K. So window and Claude. Okay. So if I go to skills live in the customized menu, And I can go here. I can look at my, like, call prep skill. And then I think the way that we we share this is happy for the skills section. oh, good. So we'll share this, like, morning or this personal let's see. This morning briefing one. And Oh, I I can actually take a stab at it, Amelia. okay. Actually, let's let's take a different question. I need a second to cook on this one. Okay. So I really wanna hit on, like, this this topic that's come up. It has one of the most votes. And it's when do I start in chat versus co work? Now they seem like similar surfaces from the outset. Like both you can ask questions and get responses. But the real difference and like the real power of Cowork is actually a little bit hidden under the surface. So Cowork is built on like the things and the structures that have made Claude Code really powerful. At Anthropic, we call this the harness for the underlying models. And what that means is that when you kick off a task in Cowork, it's using the same type of elements that make Claude Code an agentic solution for developers. But now you can benefit from the same power. So what that actually means is Cowork will make sure it's doing things in parallel all at once in a way that's really fast. It manages context well and can handle a lot more information. One direct way you might notice this is if you have a really long chat and you're trying at the end of a long track to to get, like, a brand guideline to work, but it keeps messing up in small ways, like not adhering to a font or something or, like, not doing the data analysis exactly how you want. Cowork is, like, better in those situations because it's not just queuing up an entire thing of, like, context. It's managing it for you and distributing, like, to the sub agents Matt mentioned earlier, to get things done, like, in a more powerful and efficient way. Yep. Where to start? Like, if you just want to vibe and chat and, like, get quick questions answered with chat, but you don't expect to make, like, a work product, like a deck or agendas or docs that need a lot of connectors, like start in chat and stay in chat. But if you could foresee your discussion turning into context for actual work products, just kick off the conversation in Cowork as a task. And it'll be a lot easier to create stuff after the fact. Amazing. Maybe I can actually, we have five minutes left. I I wanna show one more thing, and then maybe we can just dedicate the rest of the time to q and a. Yeah. So maybe while I pull this up, I wanna answer one question. How do I limit the information available on my computer to only those things I wanted to have access to? I think if you remember the thing Amelia showed in the beginning, whenever you give Claude access, you have to give Claude access to the folder specifically. So by default, Claude has no access to the folders on your computer. In in fact, Cowork actually runs in a virtual machine. So if you don't give it access, those files don't even exist to Cowork. So there are absolutely no way that Claude could see those files. So I think this, like, sandbox virtual machine that we've implemented in the background makes this a lot more kind of possible. So projects, I actually think are this is kind of the main way that I use Cowork. I'm a product manager, so I'm working on kind of project deliverables often. So, like, say, I'm working on an onboarding project. And like I said, eat, sleep, and breathe skills. First thing I wanna do is make a feature spec. Second thing I wanna do is synthesize customer research. Third thing I wanna do is build a competitive analysis guide. So, really, when I started this oh, and then fourth thing I wanna do is make a road map. There's always more. There's. so much, more. really, all I did was as soon as I started this, I fired off four tasks on, you know, feature spec, research, competitive analysis, onboarding road map, and then you can see all in one place all of those different artifacts that Claude has created for me. So it helped me write the PRD. You could see it looks pretty nice. It helped me write the on, analysis guide, competitive analysis, It helped me build onboarding guide, or, sorry, the the road map. So this is an Excel file. Somebody asked a question why we don't allow you to do this in Google. It is coming very shortly. It's actually a Google limitation in that a lot of their APIs don't allow us to do the things that we want, but Google has been a great partner. So very, very shortly, we will be launching things that will enable you to do more of this in Google. But yeah. So for me, this is kinda really how it looks at scale where I make a project. I have a bunch of different skills that I fire off all at once. So, yeah, that's, yeah, that's one of my favorite workflows. Yeah. We're just, like, over here stacking up tasks within Cowork and having them run and then checking in on them. I even have more of a pattern of starting things at the beginning of the day, having all of my calls, and then checking in on work at the end of the day, giving it feedback and letting it run and seeing it in the morning. So there's almost like someone on our team put it really well that like, Cowork is letting you take longer strides with Claude. So longer running work, more things that are agentic, and more things that you're like offloading to Claude and then checking back in. Yep. I know we have some, like, wrap up questions because we're I could do this all day, but we're we have things to do. Yeah. What what question should we wrap up with? Oh, should we show some slides? Or no. Yeah. Let's let's take maybe one or two more questions, and then we can might be, like, polls too. So one of the things that's like, how do you we use Cowork today. We could hit on that and see. It looks like I think we opened it. Oh, Oh, this is great. here. Okay. Cool. So we've got I ask it questions like chat. I've given a folder once or twice. That's awesome. The folders are really, like, the unlock because you can start having it, like, save instructions for you to that folder. Like, I like things in bullets or for this customer, be really, like, conservative or always, like, say please and thank you, whatever you want. Like, you can once it has access to a folder, you can start just within the task saying, like, by the way, add this to my instructions or make sure you remember to do x or y. Have skill or two saved. I hope we can say, like, there are four or five saved next time we do one of these, Matt's Screenshare. And then, yeah, everyone try to schedule something today. It's, like, so magical to see Claude, like, working on your behalf in the way that you want. Yeah. If I can if I can really summarize, and maybe I'll take over and just show our kind of wrap up slide. Yeah. If I can really summarize, I think a few things to keep in mind. Folders are huge unlock. It's a shared surface that you and Claude can interact with together. So step one is pick your folder. Step through three is the trinity. Skills are what Claude does. Connectors are what you give it access to. Plug ins are the bundles. So step one, working folder. Step two, skills, connectors, plug ins. Step three, schedule stuff. So Claude can be everywhere that you're not. And then, yeah, maybe step four is, like, parallel sessions so we can do 10 things all at once and then start producing a bunch of amazing stuff. That that, yeah. that might be, like, level level six, but, like, have Claude help you. Like, when you are stuck or don't know how to get started, just ask, like, hey, Claude. I'm new to co work. How do I get started? Or, hey, Claude. I don't know where where you saved my plug in. Like, where'd it go? You can ask anything to it, and it will help you, like, close the loop on what you're trying to get done. Yeah. I'm gonna wrap us up here. So please, please, please, if you enjoyed this or even if you didn't, you should join the Claude community. There's a lot more people to connect with. You can join the subreddit. You can join the Discord. You can come to a community event. Give us feedback on this webinar, you know, things Amelia and I could do better, things you wanna see covered at the next one, less advanced topics, more advanced topics, something in the middle. And, yeah, for all the feature requests, like, please keep them coming. And, yeah, thank you again. And, yeah, I hope you get a chance to use Cowork today and and schedule that first task like Amelia said. Yeah. Exactly. Cool. Alright. Well, thank you all so much. Bye bye.