Beth Portesi - Claude for EAs Workshop with Beth Portesi and Jeremy Burrows

Intro · 0:06

The Leader Assistant Podcast exists to encourage and challenge assistants to become confident, game-changing leader assistants.

Jeremy · 0:21

Hey friends, welcome to episode 393 of the Leader Assistant Podcast. It's your host, Jeremy Burrows. And just a quick intro to this conversation with Beth Portesi. Beth and I hosted a webinar called Clawed for EAs recently. And I really just wanted to share it with you all because I thought it was super helpful. And I know the AI landscape changes so quickly. So I wanted to get this out while it is still fresh. So you can check out the link to the actual video of the webinar. So you can see our slides and all of Beth's demo and presentation. If you go to the show notes at leaderassistant.com/393. Leaderassistant.com/393 to check out the actual video of this conversation. But I thought it was helpful enough that the audio version is still a valuable use of your time to listen to. So hope you enjoy this replay of the webinar that we recently hosted on Claude for EAs. I'm going to kick off, though, with a quick overview of AI. I'm going to share my screen to get us started, and then I'm going to hand it over to Beth for some awesome clod for EA's content. So let me actually share my screen. All right, so you should be able to see this. You can see the kind of menu on the left as well, probably. Go ahead and give me a thumbs up in the chat if you can see this okay. This is actually kind of fun because I actually designed the... I've had the slide deck built, but I decided to redesign it using Claude Design. So Claude Design is one portion of the Claude ecosystem. And I went in and uploaded a PDF of my slide deck to Claude Design this morning. And after a few iterations and chatting with Claude Design, this is what we came up with. So... I'm actually pretty impressed. This was, again, this was probably about 20, 25 minutes of Claude Design taking my content and redesigning it and then me just kind of giving it some feedback and some suggestions. So I figured I'd use, as we like to say at my company capacity, eat our own dog food or drink our own champagne. So I figured since this is a Claude for EA's workshop webinar, that I would use Claude to redesign my slides today. So if Again, this is Claude Design. So Beth will get into the different types of Claude, but this is actually an additional type called Claude Design that I use to do this. So my name is Jeremy Burrows. I am a longtime executive assistant. I'm currently senior executive assistant to the CEO of Capacity. I was hired number one when we launched the company in January of 2017. And now we are nine and a half years or so into the company. We have over 400 and something employees all over the world. We actually just publicly announced that we have crossed$100 million in annual recurring revenue, which is pretty awesome being part of this journey from day one to$100 million in revenue. So fantastic. Super excited to be part of the AI revolution, but also excited to be an assistant. I've been an assistant for about 20 years overall. I wrote my book six years ago. This last week was my sixth anniversary, sixth birthday, whatever you want to call it, for my book, The Leader Assistant. Then the podcast has been out for about seven years. Really excited to be here and talk a little bit about AI. I'm just going to have a quick overview of AI for you today. All right, so there's three types of AI. When you see AI in the news, it just kind of often gets bundled into just AI in general, right? Well, I'm going to kind of walk through a quick overview of the three types of AI. There's automation. AI or non-generative AI, and then generative AI. So what is automation? Automation is software that operates without intervention. In other words, automation can be a replacement for manual labor. And then non-generative AI, or I'll just shorten it to AI for the purpose of today's session, is software that learns and applies patterns. In other words, AI or non-generative AI is a replacement for cognitive labor. And then we got generative AI, which has kind of been all the hype with chat GPT and large language models and all that over the last few years. Generative AI is software that creates new content based on data. In other words, it could be a replacement for creative labor. So again, generative AI is Is creative labor non-generative or just ai is cognitive replaces cognitive labor and then automation replaces manual labor but there's this new newer term called ai agents and so ai agents are software that combines all three types automation non-generative ai and generative ai In other words, AI agents are a replacement for manual, cognitive, and creative labor. Or to shorten that, AI agents are a replacement for labor. So to recap, automation replaces manual labor, non-generative AI replaces cognitive labor, generative AI replaces creative labor, and AI agents are kind of a combo where they use all the different types of AI to replace labor. So how do we choose the right type? You've got a problem you're trying to solve. You're trying to implement AI in your day-to-day job. How do you choose the right type of AI? So here's kind of a quick chart on how I have used, have determined which AI to use to solve the problems that I'm trying to solve or to work on the use cases. So the first tool is automation. So automation is really good for connecting to key systems, doing the same thing over and over again, predefined logic, not so good for making new decisions, right? Non-generative AI is good for finding a needle in the haystack. It's good for improving pattern recognition, but it's not so good for making up new content, hence the name non-generative AI. And then you've got generative AI, which is actually good for creating new content, but it's not always great at finding the needle in the haystack. And you've seen these headlines in the news where generative AI or these large language models are hallucinating. They're making stuff up. So this can be not so great for generative AI. But then we've got the AI agents that are operating without intervention. They're learning and applying patterns. And they are actually determining which AI to apply. So you can actually work with an AI agent like Claude that we're going to talk about today. And the agent will actually help walk you through which type of AI. Is it automation that you need to set up? Do you need to Set up a non-generative AI tool or do you need to set up generative AI to solve your use case and solve your problem? What it's not so good for is walking your dog. AI agents aren't so great at walking your dog, at least not yet. So I like to think of when I have a problem to solve, Instead of thinking, oh, I'm going to go take this crazy long course from MIT that's a free course on AI and just kind of learn all the things about AI. No, I'm actually going to... So now that I have a high-level overview of what's the difference between automation, non-generative, and generative, and then what are AI agents, I'm going to sit there and I'm going to look and figure out, okay, what type of AI would actually be helpful for this use case? And I'm going to interact with an AI agent like Claude to help me discern that. And then now that I can kind of pinpoint, okay, which tool, which type of AI is going to be helpful, that's where I'm actually going to spend my time and energy learning and doing workshops and trying to train up on how to use that. So hopefully that's helpful. Again, just a quick overview. This has been really helpful for me as I've dove into the AI world and specifically how it applies to my role. I do much longer sessions where I go into very specific examples of each type of AI and what type of– Workflows i have set up for my job but that'll be for a different session today we're going to try to focus on cloud for eas but the bottom line is uh automate or be automated so we are very uh in the middle of this conversation because assistants a lot of times unfortunately these companies see assistants and they're like oh you know these are we can replace the assistants with these AI tools, right? And so I would rather be the assistant that goes to my executive and says, hey, look at how much time I've saved. Look how much money I've saved. Look how much energy I've saved for you using these AI tools. Look how much I've automated for you and the team. Versus being one of the assistants where your executive comes into you or asks you into their office and says, hey, sorry, but we've automated a bunch of your job and we're not sure what's going to happen and we're going to move on. I would rather be... Proactive, and be the one to automate things. And I love as a transition to hand it off to Beth, I love that story. She shares more about this in the podcast that we'll publish here in a couple weeks. But I love her story of she was an assistant, and now she's an AI enablement lead, and I'll let her kind of share more about that. But she wasn't going to just sit around and wait for AI to automate her out of a job. She went out and embraced AI, learned it, spent a ton of time figuring out how to use it and implement it. And now she's actually helping other people in her organization do the same. So that's my quick summary. Hopefully that's helpful. Again, add questions to the Q&A portion and we'll get to those later. But I'm going to stop talking and hand it off to Beth. So Beth, have at it.

Beth · 10:49

Thanks, Jeremy. Actually kind of building off what you mentioned about, you know, the assistance and AI and like what direction. I actually wasn't even someone who was like, oh, I should go learn this. I just kind of started using it and then found that it was helping me so much that I just fell in love with it. And that's part of why I founded Vibe Coding for EAs. But yes. I'm Beth, founder of Vibe Coding for EAs. I work at a tech company. I am now the AI enablement program lead. Prior to that, I was an executive assistant for 12 years. But like Jeremy mentioned, it was just like late one night, I was trying to figure out a solution to something and I was using Claude. And it was so good at giving me a solution that I just kept going. I was like, oh, what else can I do? What else can I do? And I got so excited about it that I kind of disappeared for three months and just went down this rabbit hole of building and creating and finding solutions and creating automations. By the time I came up for air, I was like, every executive assistant needs to know about these tools. So I feel like my passion and kind of direction with this came from excitement and enthusiasm rather than like fear of being replaced. So yeah, so hopefully we can share some of that enthusiasm today. Alrighty, so what you'll walk away with today is at minimum an understanding of what Claude is, confidence that you can do this on your own after the session, and then lastly, reflex and inspiration. That once you experience this magic that Claude or any AI tool can give you, that you can ask, how can AI help with anything that comes across your desk? So how we're going to get there is we're going to go through basics and a demo. Then you'll open Cowork or Cloud for yourself in practice. If you don't have access to Cowork or to Cloud today, that's okay. You can still follow along. And then we'll talk about use cases. All righty. One second here. I'm actually going to stop sharing and pull up Claude and do a quick lay of the land. So it looks like most, about 60% of the folks on the call have used Claude before. The others have not. So I'm going to do a quick walkthrough, a lay of the land. So when I open the desktop app, which you'll need for today's session, you won't just wanna use the website URL. You'll wanna download the app if you can. So here, if you have an individual account, it's gonna look a little bit different than if you have an enterprise account through your company. But in the upper left-hand corner, you'll see Home and Code. You can just stay on Home for now. And then in the chat window here, you'll see Chat and Cowork. This just changed this week, so it might be new for some of you. But if you're on the Enterprise app, all of those icons are up here in the upper left-hand corner. So those are the three tools within Cloud, which is Chat, Cowork, and Code. Here you have your chat window. And then down below that you'll see a couple of options here like which folder do you want to work in and when you're using something like clod co-work you're going to always choose a folder on your computer that it's working from and i'll talk more about that in a minute or if you're just using chat you don't need to choose a folder and it's a little more simplified In the right side here, this is where you're going to choose a model. Choosing a model with AI or with cloud is like choosing a tool for the job. You can think of Haiku as like the teeny little tiny screwdriver you get in like an Amazon or like an Ikea furniture set. It's like good for that one thing. It's not going to last forever and it's not good for everything. So this is like your cheapest, most efficient model. Sonnet is like your middle of the road model. This is like your Phillips head screwdriver. It's like handy, very useful, can do a lot of different things. And that's kind of like your everyday driver. And lastly, Opus is like the power drill with all of the attachments. This is the best model for complex tasks. If you're building agents or automations or writing code, it's great for that, good for analysis. But you don't necessarily need a power drill if you're just doing something simple. So I'd say for the most part, you can get away with using Sonnet, which is like a middle-of-the-road model. One other thing to know is that it's not just how good these models are. The better they are, the more expensive they are. And whether you're on an enterprise plan or an individual plan, you likely have a cap, a usage cap for how much you can use. And you can check your usage by clicking your name in the lower left and click settings. And then usage on the left is where you'll see what your current limits are and how close you are to hitting those limits. So just pointing that out is something to keep an eye on, especially if you're working on a big project or something to keep an eye on your usage and what models you're using. Okay, let's see here. Chat history. So on the left here, this is where you can find, like if you click Recents, you'll see all your chat history. And there's something else called Settings and Global Instructions. So if I click my name here in the lower left and click Settings, this is where I can give broad settings that I want to apply to every single chat or anything that I'm working with in Cloud. So here, for example, say you're somebody who always wants responses in bullet points. You could say, give me all responses in bullets. And this is where you can save all your instructions. Or if you want to give it the context that you're an executive assistant, you work at X, Y, or Z company, you can give that context here. One thing to keep in mind is you probably don't want to give Claude like 30 pages of instructions because that takes up your usage as well because it's going to reference these instructions every single time you send a chat. So this is where you'll want to paste some concise instructions for Claude that apply to everything. If you're working in co-work, you can also set special instructions for each individual project. So say, for example, you're planning an offsite. You'd come here to your projects, click new project, and you're going to put everything in here that you're using for planning your offsite. So let's say it's offsite, FY26, create project. Now on the right side, you'll also see instructions that you can add for this specific project. So you can say, you know, this is going to be in San Diego. We're staying at XYZ Hotel and give the theme or the objectives, whatever it is that you're planning for your offsite. So a recap of that is that you have general instructions that apply to everything that you do within Claude. And you also have project specific instructions so you can keep that separate from the different types of things you're working on. Okay, to go back to the home here, I'm just gonna click New Task, so it brings me back to this chat window where I can toggle between chat and co-work. So that's a quick lay of the land of Claude. I know that's kind of fast, there's gonna be a lot of concepts today, so I'm just gonna keep moving. But this is one thing that I face. Is it just me, or is AI terminology super confusing? Even to this day, I will be sitting in meetings, and somebody will say an acronym or term I don't know, and I'm Googling on the other side. What is MCP? I don't know what any of this stuff means. So I wanted to go through a couple of concepts that are foundational that I think might be helpful when trying to understand some of these terms. Yeah. The first one is a token. Every time you use Claude or other tools like ChatGPT, that's credits, this is tokens. I thought it would be great to do an interactive learning exercise on what a token is. So I'm going to drop a link in the chat. And this is a token simulator. So this is going to be an interactive experience to kind of walk through the definition of what these things are. What is a token? What is a context window? Let me drop that now. Okay, so you can click that. I try not to just talk at you the entire time. I want this to be interactive. So please click the link in the chat, the token simulator, and it'll walk you through a couple modules. I'll stop talking for a few minutes. I'll give it like four or five minutes and see how many completions we get through this. And then we'll keep going and talk a little bit more about co-work. Also during this time, if you have a question, feel free to drop it in the chat or in the Q&A and we can talk about it. And by the way, if you don't have Claude yet, you don't need it to do this token simulator exercise, so don't worry about that.

Jeremy · 19:34

One thing I've done before, I don't know if this is like a... It seems to help a few of the times when I've tried it, but if I have a long... Prompt I'll put it in a different AI like a Gemini or something and that's free that I'm using and then I'll just kind of have it make it more concise and then I'll put it into my paid my paid cloud account so that could be one way to help you get better at making things concise. It's basically the idea that you get charged by the token usage, and the token usage goes up if your prompt is longer and has a lot of filler words. In other words, I don't know if you're able to go to the token simulator site and try it, but basically you're just trying to optimize your prompts so that you don't hit the usage limits on your plan as quickly.

Beth · 20:26

Louisa asks a question. She says, I'd be interested to hear any suggestions or guidelines for inputting general instructions into Claude. Project instructions are easier for me to wrap my head around since they tend to be pretty specific, but it sounds like the general instructions could be really helpful if done correctly. Yes, Louisa, absolutely agree. There is a guy called Ruben Hasid and he has a sub stack and he talks all things Claude and like setting it up. I will drop a link in the chat right now to a blog post that he has on setting up coworker. And it probably takes about 25, 30 minutes to do it, but it is so worth it because he walks you through every step of setting up your global instructions and your about me so that everything is customized to you. Super easy to follow. I didn't do it for this session today because it's a little bit lengthy, but I'll drop that in the chat.

Jeremy · 21:22

And I will say that it's actually a good practice for us as communicators to be more concise, right? And so this is just really good education on how we communicate verbally and in written form. I found it to help me be more concise and to the point in my written communication for emails and slacks and whatnot as well.

Beth · 21:43

So Dee, if you have Claude Enterprise for work, Unless your company has specifically turned it off, it should be included unless they're just giving you access to chat. But yes, the instructions in the setup, his about me could help. You could paste that as your instructions for CloudChamp if you wanted.

Jeremy · 22:01

Lisa has a good question about why is AI managed on a token-based system? Yeah. I think the short, Beth I'll let you chime in, but I think my short answer would be that it requires processing, AI processing in their servers and everything. And so as it parses your words that you're putting in the prompts, it's processing it on their end. And then the more processing it takes to parse your prompt, the more expensive it is for them to do that.

Beth · 22:35

Yeah, I think it's essentially like it's mathematical the way that it processes text.

Jeremy · 22:41

Acknowledge Stephanie's question about curious if there's pros and cons or a better use case when thinking about clause models versus ChachiBT, Gemini, etc. I would just try a little bit of everything and see what works best for the use cases you're trying to work with. But also like maybe some of you are in the Microsoft ecosystem and you only have Copilot. Maybe you don't have all the cloud models connected to the Copilot. There's still a lot you can do with those. But I, in my personal experience, I've tried the multiple different types of vibe coding tools and, you know, The design tools and all that. And I've just found that Claude is, is just works better and more efficiently on all of the stuff that I've had to do with vibe coding. And so I would try it out and see which one works better for your use case.

Beth · 23:34

Yeah, I brought this up in the podcast that I did with Jeremy. Bruce Lee has a quote and it's be like water, my friend. And the way that this relates to AI in my mind is that these AI companies every day are changing in their features and capabilities. One day one is doing this, the next day the other one's doing better. So it's like always fluctuating. And AI is one of those things where a lot of times you don't just like lock into one. But it is super helpful to be able to flow between them and just choose which one for the task that you're doing. But I will say that my focus on Claude is because I find that as a non-technical person, it makes doing technical things very accessible. So ChatGPT, great for communications, research, drafting emails, things like that. For me, Cloud is great when you kind of start going down the rabbit hole, like generated AI. Maybe you want to set up an agent, which we're going to do just here in a few minutes. So I find as a non-technical person, it's easier to do these things for the first time and learn with Cloud. And then you can go to other tools like Codex, which is OpenAI or ChatGPT's version of Cloud Code. Maybe once you've kind of gotten your bearings a little bit with Cloud. So let's move on here. Feel free to bookmark this token simulator and continue using it or practicing with it after. But a quick recap below. So a token is about four characters. So if you think of 750 words or so, that's about a thousand tokens. When we talk about context window, remember that a context window is like a whiteboard. There's only so much space. So when you're in a chat with Claude or even ChatGPT or these other AI tools, it can only handle so much on that whiteboard. So when you start writing on the whiteboard, maybe you have instructions for Claude. That's a piece. Then you ask Claude a question. That's taking up context. Claude sends you an answer. Until you go back and forth so much, you get to a point where your entire whiteboard is full. And in order to continue the conversation, something's got to be erased. So what it does is it compacts it all. It summarizes everything. And then you have more room. But what happens when you compact something, it means that not everything is going to go with it. When you summarize anything, even if it's just communications, you're not going to be including every detail. So maybe in your conversation, you told Claude, only give me bullet points. But then later your conversation gets longer and longer, Claude might forget the thing that you asked it to do once it compacts that conversation. So just keep in mind that if you have a really long chat, the answers might not be as accurate or it might miss details that you had previously told it. So what does that mean for you? It means that it's good to start new chats. I know you can feel like, oh, but I'm losing a bunch of context, but it can actually sometimes perform better when you just open a new chat. Okay, again, if you have Claude Enterprise to your company, this is where you choose the three products that you can work from depending on what they have enabled for you. And if you're on a personal plan, this is where you choose those three products, chat, co-work, or code. OK, so to answer the famous question, like what is the difference between these tools? So chat is like the clod that you probably already know. You ask a question, it answers. You brainstorm, it can help you. But every step needs your input. So it's kind of asynchronous. It's back and forth. So you can think of chat as like a thinking mode or doing one-off things. You can think of cowork as your doing mode. And that's what we're going to talk, dive into today. The key difference is that, um, Claude cowork can read and write files directly on your computer. Claude chat cannot do that. If you want to send Claude chat a file, you click the little paperclip icon, you attach whatever you want to send it. And it's like one at a time, right? It can only accept up to like 20 attachments. So there's a limit when you're using cowork, you can point it to a folder directly on your computer where there's lots of files and it can access all of that. And it can do things on your behalf. It can run multiple-step tasks. It can make edits, deletes. It can change things directly on your computer. Cloud code is more of the building mode. This is what engineers or other technical people use to build and write code. I will say Cloud Code has become more user-friendly over the years. So once you get comfortable with chat or cowork, you can also start using code and code can do everything that Cloud Chat and Cloud Cowork can do. But I'd say for if you're a non-technical person, just getting started, cowork is a great starting point. So again, you can think, I like to do visual. So you can think of Claude Cowork as like your 20 something super high energy, super genius intern who's ready to do things for you. So you can go to Claude Chat to ask how to do things and to kind of go back and forth. But if you want to delegate something like an entire task, you can open Cowork and have your intern do it. I think of this as a great metaphor for an agent, Sometimes when I hear people use the word agent with AI, my brain freezes up. I'm like, do I even know what an agent is? Why is it so hard for me to wrap my brain around this? And Jeremy's definitions in the beginning really helped. But I think of an agent as like having an intern, something that can do things on your behalf. So let's give an example of when you would use chat versus co-work. So say you want to organize your downloads folder. I don't know about you, but my downloads folder is always chaos. I have hundreds of files in there. I'm adding to it every day. And maybe you go to Cloud Chat and say, how should I organize my download folder? And it can map out all the folders and give you a step-by-step plan of like, here's how you should organize this. But then you do the drag and dropping yourself and all the changes because Cloud Chat doesn't have the ability to do that. So for Cloud Cowork, that's where it's different. You've got your intern or your agent. It can do everything for you. You can talk about the logic with it of how you want to change everything. Then you point CoWork directly at your downloads folder, and it can go in there, read every single file. It could even rename the file based on what it is. So say you have a receipt in there. There's no title. It's just a picture of a receipt. Claude has the ability to open that file, read the receipt, and then title that file. So it could say DoorDash receipt,$15, and rename that file. So it can do that. And then it can also move things. So say you have now renamed all your files and you want to put them into subfolders and categorize them. Claude Cowork can do that because it has the ability to act on your behalf. I'm going to keep focusing on these because I know it takes some repetition in these tools since you haven't had as much hands-on experience to understand when to use them. So we'll talk a little bit more about what makes co-work different. First is the folder access. You always have to point co-work to a folder when you're working. It doesn't have to be your downloads folder. You can create a specific folder just for Cloud so that it's there. Because when it creates files on your behalf, it's not going to just send them to you as an attachment. So with Cloud Chat, say you ask it to build you a slide deck. It's going to send you that slide deck, and then you have to download it. With Cloud Cowork, instead of sending it to you in the chat, it's just going to put it on your computer. So that's where your file will be once it creates it for you. So again, Cowork has folder access to your computer. Second is parallel workers. So when a task has independent parts, like it's a multi-step process, Claude Cowork can spin up agents that work at the same time. So say you have a task and it has 10 steps. Claude Cowork can launch multiple agents to go work on those things all at once. So now your intern just turned into five more interns and they're having more interns help. With Claude Chat, it's more asynchronous, like one at a time. It's going to do one step and then it's going to do the next step in sequential order. So it's more efficient to use something like CodeWork for a multi-step task. One thing to know when you're using Cloud CodeWork is that your desktop app has to stay open when you're using it. Your computer can't be asleep. If you close it, your session will end. If you have a scheduled task, you have to have your computer open at that scheduled time in order for it to trigger and to go off. One other thing I mentioned, I think earlier in the call and in the chat is that yes, you have to use the desktop app for co-work. It won't work in just the website. So if you go to clod.ai, you will not see co-work as an option. It has to actually be the desktop app. And the reason is, is because it accesses files and folders on your computer and it needs that access as an application. Okay. Danielle, I promise we will go into some examples. Usage and cost. When you're using Cloud Chat, you're just chatting. It's just a text exchange, right? Like you ask a question, it responds. And so it's text. Sometimes it generates a file for you. When you're using Cloud Cowork, you've got your intern, your agent, and you assign a task to them. It's going to cost more because they're actually doing the work for you. Say they're renaming all the files in your Downloads folder and they're changing them and putting them into subfolders. That's going to inevitably take up a lot more tokens and a lot more costs than just talking about how you're going to do it. So one thing to just be aware of is that if you have a limit, you'll hit that limit faster if you're using tools like Cloud Cowork. Okay, I know I've mentioned this a couple of times. I'm just going to hit it home with this last slide of local versus the cloud. Sometimes when people say local, I remember when I heard this for the first time, I wasn't sure what that meant. All it means when you're working locally is you're just working on your computer. So you're working in a folder on a file on your computer, not in the cloud, not like Dropbox or Google Drive or whatever. It's stored locally on your computer. Because your files are local, if you log in to Claude here, let me see. If we're looking at Claude here, And you click co-work, and you choose a project or a folder to work in. Say you're working in your downloads folder. You're going to point it to that folder. Because I'm working on the files on my computer, that means if I switch computers or I go use somebody else's computer, I will not see all of the work I've done here because it's only stored locally on my computer. It's not accessible from a new computer or a different computer because it's not stored in the cloud. So if you're using Cloud Cowork, just know that everything that you do is stored locally on your computer, including your conversation history with Cloud Cowork. It's also on your computer. Okay, so another quick example of cloud chat. So if I click on chat here, and say I want to send a attachment, I click here on add files or photos, I can choose which attachment I want to send. And then it sends it that one time, right in cloud co work, I'm going to point it directly at that folder with all of the attachments and files in it, and it can access those directly. So if I have a PowerPoint I'm working on, it's just going to make the edits directly in that PowerPoint versus send me a whole new attachment back. Okay, we went over this a little bit in the beginning, but I'll cover it again. So say you have a specific project you're working on. Say, again, the offsite example. Here I've got my, you can click new project, title it. I'm going to click here. I've already created it. So I have this here. So what it did was it created a new folder on my computer called Offsite FY26. And anytime I have anything related to this project, I can type in the chat here and it's going to pull from that folder and any specific instructions I gave it about this offsite. I can also add more context. Say you already have like a couple contracts for this offsite or you have more files, you can add those as context here and it's gonna reference those every time you send a chat. So remember, because your context windows are limited, you don't wanna add like, Hundreds and hundreds of pages of PDF that it's going to reference every single time because that context window is going to fill up. So just be mindful of the context and the attachments you put here. Gabriela, great, great question. Gabriela asks, are the projects that you have in chat mode the same as you have in cowork? The answer is no. So If you're using Cloud Chat, you have a projects icon here. This is different. Remember, everything in Cloud Chat is stored in the cloud on the Anthropic server. So it's accessible from any computer you log in from. Everything is stored in the cloud with your chats and with your projects. When you switch over to co-work, the projects are different. They are not the same. They are not connected in any way. So if you have an offsite folder or a project you create with Cloud Chat, it will not be accessible or shared with co-work because one is stored in the cloud and one is stored locally. All right, global instructions. Somebody earlier in the chat said, can you give some guidance on global instructions that I can give Claude? What's a good example of an instruction I would want to set that applies to every single chat? Sorry, just to be clear, global instructions for co-work are different than the instructions for all of Claude. So if you come into your settings here, And you click on Cowork, this is where you'll see Global Instructions edit. And this is going to apply to all of your Cloud Cowork sessions. So this is where you can add your About Me, any specific things you want to, preferences you have and how you want it to respond. You can add information about your company, where you work. And this is where you can put your global instructions that will apply to every single co-work session that you have. So just to be clear, there's three places in Cloud that you can add instructions. Number one is your instructions for Cloud Chat. Remember everything that you do in the Cloud. You click your name, you click Settings, and here it's going to be on that first page on General. These are your instructions for Cloud. Okay, so this is your instructions here. The second is global instructions specifically for cowork, for your intern or for your agents that you're going to create. These are the global instructions you want for them. And the way that you do that is you click settings, cowork, you scroll down to global instructions. And then lastly is projects. For every project, you can have a specific set of instructions for that project. I just wanted to clear that up in case it's confusing. I know there's a lot of different places you can add instructions. One of the instructions that I like to include is, you know, never delete any files without my explicit confirmation. Never modify any files outside of the folder that we're working in. For the most part, Claude is pretty good about not doing this anyway, but I'm going to be explicit in my instructions not to do this because those are more high risk things, right? If you start deleting things, that's where it gets scary. So I like to specifically call this out. And typically, Cloud will do this anyway. Naturally, it'll say, hey, are you sure you want me to do this? And then you'll click Allow. Another couple of instructions I have here are, show me your plan before executing any multi-step task. If you're unsure about any instruction, ask rather than assume. And then lastly, flag all assumptions explicitly before acting on them. So these are just a few examples you can use for your global instructions. OK, when you get stuck, which inevitably will happen when you're using these tools for the first time, there's a learning curve. Maybe you need Claude to simplify it for you. So don't forget that when you get stuck, phone a friend and that friend is still Claude. I have this visual in my mind of Claude is like a super genius in a van parked outside wherever I am. And he gives me a play by play of exactly what I need to do because I don't have the technical wherewithal to do some of these super complex things. But I know that I have the resource. I've got the bug in my ear of somebody telling me step by step exactly what I need to do. So when you get stuck, you can phone a friend and that friend is Claude. Okay, so let's do a quick demo. I know it's easier to see these things being done than just to talk about them. So I'm going to show you and then have you practice for yourself. Alrighty, this is like a cooking show. I've already cooked the meal. I'm not going to like make you sit and watch every prompt I send and Claude send it back because that takes time sometimes. So I've already done this and I'm going to show you the exact prompts I used to get something done. So one thing as an executive assistant that I struggled with often, or even just being an employee at any company in this day and age is we have so many tools, so many places where communication is coming in, Slack, email, Gemini meeting notes, or my AI meeting notes. I have so many action items everywhere. So I wanted to find a way to bring in all of this information from different places into a to-do list or a dashboard. One way you can do this with Claude is you can give it access to these tools. And this is through what is called connectors. So if I come down here in the chat window, let me just open up a fresh window so we can see this here. Here's a fresh chat window. If I click this plus sign, I can click on connectors. And I can see all of the different types of things I can connect to here. If your company, if you have an enterprise cloud account with your company, they might have things already populated here that you can enable or request access for. So I've got my Gmail connected, Google Calendar, Google Drive, Instacart. That was a fun one I was testing. Slack and all kinds of other connectors you can add. So for this specific example, I have enabled the Slack connector because I want it to reference all of the data in my Slack and pull from it to help me manage my action items and make sure that nothing slips through the cracks. So this is what I asked it. By the way, I am very lazy with my prompting. I call it lazy girl prompting. It's not some long extravagant prompt. It has like bullets and, you know, I've not prompt engineered anything. I have just simply asked for what I wanted. And the only thing I do at the end of my prompt is I say, ask me questions first. That way, if there's any missing gaps in what I've asked for, Claude can help fill those in. So here's what I said. I said, check my DMs, my Slack DMs from myself to create a dashboard slash artifact that pulls my action items from the last 48 hours. Ask me questions first. So what I wanted to do is create something that's called an artifact, which is just a live artifact that lives within Cloud. It's like a dashboard or whatever you want to create. And it can sync to Slack or to your Gmail or anything and pull information from those spots into your artifact. So Claude asked me a few questions. Let's see if it asked me a few questions here. So it's not showing the questions it asked, but I did say ask me questions. And then it came up, it asked me a series of questions. I answered them and then it built my dashboard. And again, it's just pulling from my Slack right now because that's all I asked it was to pull from my Slack DMs. So if I click here, this is the artifact that it created. I'm going to minimize or I'm going to maximize this so you can see. So it went into my Slack DMs and this is what it pulled. It showed me the time that it came in. It even tagged a few different things here like executive operations. It showed me due dates if those were included. And so now I can also connect it to anything else if I want. I can connect it to Gmail to pull in action items as well. So it's pulled in everything here. When I'm done, it's interactive. So I can cross these things off and it's going to show me everything that's been completed. Zlatomir asks, can I connect a mailbox different than the one Claude account I set up with? Yes, absolutely. So let's take a look really quick at the different connectors that you can connect to Claude to create something like this. Again, you can get there by clicking the plus sign and clicking connectors. Or on the left side, you can click customize. And connectors on the left. And this is where I see all of the connections I have enabled. I've seen a couple messages just in passing about security. So let's talk about connectors and security. Because if it's accessing your email, for example, maybe you don't feel comfortable with it having the ability to do something on your behalf. Like you don't want it to send an email accidentally or something. You can control what it's able to do here in the connectors plane. So say I click on Gmail. To answer your questions, Vladimir, this is where you can decide what account you want to connect. So if you were to disconnect it and click Reconnect, you would choose which email account you want to connect to. I'm already connected here, and let's see what Claude has access to. This is where it shows you your tool permissions. So I can say read only. Read only means it's only going to be reading things and pulling data. It's not going to be taking any action. I'm totally fine with that. So I will always allow it to do that and pull data from Gmail. Write and delete tools. This is where maybe you want to be a little bit more cognizant. This is totally personal preference, by the way, or maybe your company's preference. This is where you can decide what you want it to be able to do. So let's say... Create a new email draft. For me, I have this as needs approval. If it creates a draft, a little box will pop up and say, do I have your permission to create this email draft? And I have to manually say yes. Or you can set it to always allow or block it. Okay, so this is the connectors where you can add them and manage the permissions. Okay, back to my artifact that I created. So that was great. I have this artifact now. I've crossed things off. I can uncross them. But this was only for the last 48 hours. So now this is where we're going to create an agent, our first agent. I want this to happen every single day at 5 p.m. So that's exactly what I type to Claude. I just say, schedule this to run every day at 5 p.m. And every day at 5 p.m., so long as my computer is open and the Claude app is open, it's going to go to Slack and grab the action items from the last 48 hours and add them to this same exact artifact, which is a dashboard. So it says done. It'll run every day at 5 p.m. And refresh the artifact with the latest 48 hours of DMs. I said mine says self DMs because this was just a demo. I didn't want it to actually pull from my company Slack or anything like that. So this was just my personal Slack and I DMed myself. But it can check DMs in Slack. It can check channels that you're in, project channels. It can check all of it. So you can just instruct it on what you want it to check. Okay, so it says one thing, click run now in the scheduled section once to pre-approve the Slack tool permissions. Otherwise, the first automated run may pause waiting for your approval. So the very first time you run your scheduled task, your intern just says, hey, do I have permission to run this for you? And you say yes. So once this is created, the scheduled task, where does it live? Because I need to be able to control it. I don't want to have to come find this chat when it's buried below in a bunch of other chats, right? On the left side, you'll see the scheduled section. This is where you have all of your scheduled tasks or your agents. Here is the one we just created, daily action items at 5 p.m. And this is what it looks like. You've got the title and you have a toggle here. Say you don't want this to run for some reason, you can always toggle it off and pause it and it won't be active. Below that, it has instructions. Remember, we didn't see any of this. I just told Claude, I want to do this thing, I want to create this dashboard, and I want it to run every day at 5 p.m. What Claude did in the background was wrote itself instructions, and those instructions are listed here. And I can come in here and edit these instructions anytime by clicking the pencil icon. So say, for example, instead of 48 hours, I want it to be the last 24 hours. I can come in here and just directly change this 48 to 24 hours, and that's what it will check for. So you can come in here, edit these instructions to be whatever you want. Here you'll choose which folder you want it to work from. There's one option here, which is ask before acting, meaning every single time this runs, it's going to ask for your permission to do this thing. Or you can give it permission to act without asking you for approval. And I'd say maybe in the beginning, you aren't so comfortable with this. So maybe you want it to ask for your permission every time so you can see what it's doing and approve that. That's totally fine for me with where I'm at with these tools. I have so many of them created and so many running that like I don't want my intern to come like knocking on my door every time it's doing something for me. So I have most of mine to act without asking. And again, it'll ask you which permissions you have here. You'll also choose which model for this. I would probably choose a model like Sonnet. Remember, Sonnet is like your middle of the road, everyday driver, good to do most things model. And then below that is where the frequency is. So you can set if you want this to run daily, weekly, hourly, and the time here. Remember that usage is not unlimited. So if you have this running every hour, it's going to take up tokens every single time. So maybe you don't want to run it as frequently because it's going to cost you more. Because every time you run a scheduled task, it costs in tokens. Okay. So I'm going to click save. I have this here. Again, I'm like a, I'm kind of lazy when it comes to them. This is why I love AI because it makes being lazy so much easier. And I have so much help with all these things. I mean, lazy relatively as an EA definitely couldn't be like lazy, lazy, but if there's a more efficient way to do something, I am the first person to be like, I would like the easy way, please. So for me, I don't even want to really have to come in here and edit my instructions manually. I can also just open up a new channel and tell Claude like, hey, I want to change my action items dashboard artifact. And then I'm just going to chat with it because it's like an easier way and it will go and update those instructions itself. So you don't have to do them manually. You can come here and just chat with Claude to update any of your scheduled agent tasks that you have. All right, let's check the Q&A real quick before I move on to the demo. Lizzie says, a lot of us don't keep files on our computer because we use Google Drive. Would love to understand how cowork works with Google Drive. Yes, Lizzie, great question. That comes back here to the connectors. So if I click on customize and I click connectors, I can connect directly to Google Drive. So then Claude has complete access to Google Drive. If I say, hey, go in my Google Drive and find all of the Gemini notes, meeting notes from this week and tell me what my action items are. It can go do that. It can find and edit files as well directly in Google Drive from Cowork. It can also do this from chat, by the way, because it's accessing a Google drive. It's not a thing stored locally on your computer. You can also use chat to do this, but if you want to automate any of that, like say you have a Google sheet, you want updated every day at the same time, you would use cowork for that because it's taking action on your behalf and chat cannot do that. Same thing if you are a Dropbox company. Dropbox usually lives on your computer, so you have access from your Finder or from your search. If you have Dropbox, you can, when you choose a folder that you want to work in with co-work, you can click here, choose a different folder, and go find, for example, Dropbox here on the left. You can click Dropbox. And then point it to a Dropbox folder. Even though Dropbox is in the cloud, it's available from your computer, so you can still point it to a Dropbox folder to work in. Susan Williamson asks, once you check things off... So that's back in our artifact here of my dashboard. Once you check things off, what if you need to go back for reference or proof that you did something? Will it still be there? The great thing is, Susan, is that this is just something I kind of threw together with that one simple prompt. This can be as customizable as you want. So for example, if you come back to... This chat I was using, I could say, add a section for me to add notes to task when I mark it done. Or something like that. And then it's going to make that edit. It can be whatever you can dream up, Claude can probably build it. And that's why I love it. And it actually is quite creative freedom-wise. You can do whatever you can dream up, Claude can probably build it. Rachel asks, should I ask questions about an existing scheduled task in chat if the task lives in cowork? Will it have enough info to answer intelligently? Good question, Rachel. If you want to update an existing scheduled agent, you will do that from cowork, not from chat. Remember that chat and cowork, they don't really interact because chat is cloud-based and cowork is local-based. So they operate independently. Monica asks, what about delegate mail access to someone else in Microsoft 365? Is this not controlled by IT? Every company is different in their security settings, and especially with cloud enterprise accounts and what they allow or don't allow. I've never tested this because I haven't supported an exec using Microsoft 365 where I have delegate access. So this would be worth testing. If I had to guess though, I doubt it. I doubt there's a way from Claude to connect as delegate to someone else's email account. I think the more likely solution would be you would have to connect Claude to their email. So you'd have to authorize here in the customize section. In connectors, instead of logging in with your Gmail or your 365, it would have to be with theirs. But you can fact check me on that or ask Claude because I'm not 1000% sure since I've never done it. Okay, Susan. So here is where you can add the note. Now it's given us a note box if you want to add like reference for something that's been completed. Alana asks a great question as well. She says, does having more connections turned on equal more token usage? So let me open a chat window. I'm going to click my plus sign, click connectors. When a connector is toggled on, this is all of the tools that Claude has access to during this individual chat. So say, for example, I'm going to do something with Slack and I don't need anything else connected. I'm going to just toggle those off for this because the answer is yes. It takes up tokens and context every single time it has to go check those connectors. So If I say check my Slack action items and I have my Slack connector on, that's all I need enabled. It's only going to go check Slack. It doesn't need to go Gmail, Drive, Instacart. So yes, you'll want to toggle those off because you don't want to eat into your limited context window if you don't need those tools. Okay. All right. I know that was a lot. Typically, I will say, I do not usually put this many concepts in, but I was like, you know what? We're going for it today. We've got over 800 people. Let's make sure this is valuable to all stages and all experience levels. So we had a lot. We talked about tokens. We talked about context, windows, pricing. These are all really big concepts. And then on top of that, did a whole demo of chat and co-work. So-Let's do a test for ourselves. And if you have Claude Cowork, you'll want to open that now. Again, you have to be on the desktop app to do this. And maybe you want to just try one connector. Maybe you just want to connect your personal Gmail or something just to test and get started. Choose one connector or one tool you want to connect Cowork to. And we're going to authenticate that. I'll go through this physically in a second. I'm just talking through the steps right now. So you open Cowork. You're going to select your folder. So you're going to first create a test folder. Maybe you want to call it Daily Brief. Once you open Cowork, we'll select your folder. Then we'll insert this prompt. I'm going to drop it in the chat right now. Which is check whatever connector, check Gmail, Slack, whatever connector you're connecting, check this to create a dashboard or artifact that pulls my action items from the last 48 hours. Ask me questions first. If you don't want to do this test, that's totally fine too. Maybe you want to point it to your downloads folder and organize your downloads folder or create a test downloads folder and throw five files in there and have it organize that test folder. You can do that too, whatever you're comfortable with. Vicky says, this is super awesome. Just created my dashboard based on your action item example. Awesome, Vicky. I'm very curious what connections you used, if any. Yes, Erin, you can do it one of two ways. Okay, so you can either come here and click create new project and call it daily brief. Create project.

Speaker 04 · 1:00:49

Oh, it's not, not found. Why is this not found? Okay.

Beth · 1:00:54

So better, better yet open up your finder or whether you're using Mac or PC, create a new folder, title it daily brief. And then when you open up cowork, you're going to select that folder you just created. So I created one of my downloads called daily brief. I'm going to select that folder. So it's working specifically there. Deidre, do we need to be on cowork for the test or can it be done with just chat? You can actually do this specific example with chat as well, where you can add a connection and then create an artifact. Yes, you can do that. However, you won't be able to create an autonomous like agent where it's going to go do this every day for you at the same time. But if you only have access to chat, yes, you can just create the dashboard and use the connection with chat. Alex, is it called dashboard or artifact when prompting? What it's doing is creating what's called an artifact. I said dashboard because I wanted to be specific and that I wanted it to create me a dashboard. But you can just say artifact. Artifact is just the broad term for any of these deliverables, like these live deliverables that Claude is creating for you.

Jeremy · 1:02:18

Real quick note, we are recording this, and I'm actually going to share. Beth, is it okay if I include a PDF of your slides? Absolutely, yes. So I'm going to share a PDF. I know several of you asked for that. A PDF of my slides, a PDF of Beth's slides, and the video recording as well. I'm actually going to post all of this in our Circle community. We have an AI for EAs space within the community. And I'm going to post the replay and the slide decks and all that links to the token simulator. Anything else we need to add later, I'm going to post that all in the AI for EA space. And I'm actually going to go ahead and publish the post. Now and I'll add the video later that way you guys can go to it and you know I can link to it in the chat so that you know where that post is but we'd love to I know we've still got a little bit of time on this session but we'd love to continue the conversation hear about how you have been using Claude hear any questions you have you know put it in the comments of the post that I'm about to post and share in the chat that way we can keep the conversation going even after this call so I just wanted to give that quick

Beth · 1:03:36

Thanks, Jeremy. I'm also dropping the chat. We still have time, so no rush to drop off. I would love to keep chatting and answering questions and then have folks come off mute and talk a little bit more about use cases. But if you have to drop early, I am going to drop the link in the chat to the post-session survey. This is great for Jeremy and I to know what's helpful for you guys, what isn't. So if you could fill that out. We had one in the beginning, and this one will help us see how things went. Um, I think it was Monica who mentioned, uh, they were lost because their enterprise account looks different. So on the screen right now, I have pulled up what it might look like in your enterprise account and where to find Claude cowork. And that's different from having an individual account. So Monica, hopefully that helps. Okay. I'll go back to our demo screen so you can follow along for the demo. All right.

Speaker 04 · 1:04:24

Let me check the Q and a

Beth · 1:04:27

Alyssa says, can Claude also connect to a company drive or only an individual Google Drive? Absolutely, Alyssa, it can connect to a Google Drive. When it accesses your company's Google Drive, it's going to access it with the permissions that only you have from your account. So you, Alyssa, say, for example, you don't have access to the legal folder, then you won't be able to organize the legal folder because you don't have access. So depending on your individual account permissions, that's the same permissions that Claude will see through your account. Aaron says, are there programs you can't add as a connector? Yes. If your company has an enterprise account, they can say, yes, you're allowed to connect to Gmail. No, you're not allowed to connect to Salesforce. So they have some control in that way, depending on your company settings. And there's a lot of connectors out there. There's thousands of connectors available in Cloud. You can click here, manage connectors, and you can search. Let me find where you actually search for all connectors. Not connected. You can search for all different kinds of connectors. There isn't absolutely a connector for every single thing. So I am curious. I would love to hear from some folks on the call in either what you learned today, specific use cases that you use for cowork or for quad, or if you've tested one of these demos, like a dashboard or an artifact and how it went. Any brave souls willing to come off mute? Jeremy, do we have the ability for folks to come off mute?

Jeremy · 1:06:04

We can add them as a co-host, yes. We can highlight them or whatever. We just have to manually pick the person and put them on the stage. It looks like there's a lot of people asking about specific use cases, cross-checking calendars. So, you know, I've done some specific... Things like set up, uh, I use, I'm in the Google environment, so I've set up Google apps scripts, which basically will look at the calendar and, um, Pull information and do stuff, take actions with information from my executive's calendar and information from a Google spreadsheet and, you know, either manipulate the spreadsheet data based on what's on the calendar or vice versa. I even had a thing where I added a whole bunch of events to my executive's calendar for a personal trip. And then his wife called and was like, yeah, actually, we didn't want all those on the calendar because we're not going to them all. And have it attached, and I was like, oh shoot, I've got to delete like 50 events off the calendar. Well, I was able to set up a quick Google Apps script using cloud code. I just said, hey, I've got 50 events for the certain event that I need to delete in a batch instead of having to manually do it. And it quickly built a script. I put it in Google Apps script, hit run, and it deleted all those events that I wanted deleted. So little things like that. Anything from little things like that to... More complicated prototypes for our software product on features that I want to add to our software. I've had to build code prototypes so that I can show our product team, hey, I really want our tool to do this exactly. To building apps. You know, I've built multiple apps now that are interactive, that have sign-in, that save the data, that, you know, to build a travel itinerary app. I've built the Assistant Habits app for assistants to kind of track professional development and build a street, kind of like Duolingo, where you get a street going and you don't want to, you want to keep that momentum going. I did all that with Claude Code. I even had my website, If you've got to build a website or a landing page for an event, like Beth has shown examples of this in prior sessions, where you basically just build an itinerary page or an event project Event itinerary page website, basically using quad, manipulating data in a spreadsheet, you know, updating the site based on the spreadsheet, you know, anything that you can think of that would be helpful to visualize some data that you have. Or to quickly manipulate data. I even did something as simple as I got a PDF of an itinerary of an event that my executive was going to and it had a bunch of attendees listed out on the PDF. Format was just plain text on the PDF. Well, I wanted to put that into a spreadsheet. So I put it in a cloud. I said, hey, can you make this into a spreadsheet so I can copy it into Google Sheets? And it did it within a couple of minutes. And all of a sudden, I have a spreadsheet of 150 attendees and their LinkedIn, their email, and everything that was on that PDF, but in a much more friendly format when you're looking at a list of attendees. So lots of options. I always like to ask... What's the manual, painful, repetitive task that I had to do today or this week that I could potentially offload or have AI do for me? And that's how I start exploring what to implement.

Beth · 1:09:53

Thanks, Jamie. Yeah, kind of on that same calendar note, I'll show a quick example of... I have actually started to use Cloud for calendaring directly from Cloud. So I add the Gmail connector. Here's one example. I just said to it, add a calendar event for every day from 5 to 6 p.m. Titled DNB for Do Not Book. Make it blueberry color. So it does it. Now I said, add a lunch block every day at 12 p.m. So if I come over to my calendar... Let me drag that over to this screen, pull this up, and then I'll talk about skills in just a second. I don't know if a couple of folks have asked. Okay, so here it is. So here's my calendar. It added a do not book every day and then lunch every day. This is an EST and I asked it for a specific time, but you get the point. It's able to add things to your calendar and change things. No problem, as long as you give it the permissions. Okay, let's talk about skills really quick. Someone asked about this. So a skill is something that Claude is going to do repetitively, but isn't a multi-step process. So for example, maybe you have your company's... Well, actually, it could be a multi-step process. But maybe you have your company's brand guidelines, for example. You can come here and drop in like a deck. Say there's an entire deck or... A PDF of your company's brand guidelines, you can say create a company brand skill. And then you attach it here. And what it's going to do is it's going to write all these instructions. It's called a markdown file. All that means is it's just a bunch of text. It's like plain text. It's going to create all this plain text instructions to say... Okay, anytime they ask for something to be company branded, that means this color blue and that color white and align it this way, use this font. And your skills are going to pop up here in the customize section. And you can see them all listed here. So for example, I have one for the vibe coding for EA's brand. And here's what it looks like. Just has the quick reference of the colors, the typography, the design principles, the voice. And so that's here. And it will reference that skill anytime you build something moving forward, a slide deck, a spreadsheet, a report, or a PDF. You can say, apply the vibe coding for branding skill. So if I go back, let's say create a single page PDF for my next workshop with Jeremy Burrows. And then I'm going to forward slash vibe coding brand. Or I could even just type it out and we'll see what it comes up with. So this is good. Skills are good for repetitive tasks. Some other examples of skills you might have are like an NDA review. Say you're working in legal and you have certain playbooks that you follow. That's where you would have like an NDA skill to say these are the principles we follow when reviewing an NDA contract because it's repetitive. It can be done more than once. And you can also reference skills with your agents. So say you wanted to create a social media post with your brand every day at 3 p.m. And like reference what's happening in the industry. You can say, look up what's happening in the industry, create a social media post every day at 3 p.m., apply my branding, and it can reference those skills. Okay, Claude asked me a couple questions and we'll see what it comes up with. Let's see. Lisa says, was what you just did a worthwhile of tokens versus just making it a reoccurring event in Google Calendar? That's, Lisa, kind of like where the trade-offs are. I would say that in the beginning, if you've never used these tools, I would say doing things with Cloud and getting that hands-on experience is worth it because it's not just that task. It's like the muscle memory of learning to use these tools and what they can be valuable for. But then your question is like the graduation to the next step of... Is the cost worth it? Is the return on investment of whatever the tokens worth it for this specific thing? So I would say getting started, like just, just get the experience with it and the muscle memory, and then kind of have some of those more like grander conversations with yourself about like, what is it worth to automate and what is it not? Hannah, could you use Cloud to reset an exec's calendar? We just did a big calendar audit. Plug for a great use of Cloud. Anne would like to delete all current meetings and resend with new meetings and cadences. Yes, absolutely. As long as you give it the ability to delete those events, it certainly could. As Jeremy mentioned, there is also something called Google Apps Script. Google Apps Script is great for anything that you want to do in Google automated. So you can create what's called a Google Script. It's like a block of code. Claude can generate it for you that says it's going to write a bunch of code and that code, all it's going to say is delete all of the events on your exec's meeting calendar. Claude can tell you where to paste that code and then Google will execute it. The reason why you might want to use something like Google Apps Script versus Claude is Google Apps Script doesn't cost tokens. So in order to execute that task of deleting all those events, maybe it's a... A big lift for, for Claude, but for Google app script, it can do it in one swoop and it doesn't take any tokens. The only tokens you'll use for that task is for Claude to generate the code that you're then going to go paste in Google app script. So if you're not sure what Google app script is, no stress. That's actually how I got introduced to Claude in the first place was it walked me through how to do something in Google app script. And you can ask Claude what that is and how you would go about automating that task. As I'm working through this one example, it had asked me a couple of questions I answered. Sometimes folks, when they're working with co-work, one of the reactions I see from the beginning is, man, this really took a long time. I could have done it faster than co-work. And I think that's number one, a natural reaction to be surprised at how long something can take. But you can remember to think about co-work like your intern. You're not just going to stand there. You're going to ask them to do something. You're not going to stand there and look over their shoulder and be like, are you done yet? Are you done yet? So just know that you can delegate something, go walk away, do something else, or spin up 10 other agents to go do things or interns to do things for you and come back and check and see how things are going. But it's not instantaneous. It'll maybe take a few minutes.

Jeremy · 1:16:30

All right. Just a quick reminder that we are... Going to continue the discussion through our discussion thread in the AI for EA space and circles. So continue sharing feedback, asking questions via this post here that I'm posting again in the chat. I think it's pinned at the top as well. I will add the replay there once it's done processing and uploaded as well. So you can watch the replay. Our PDF decks are there. There's a little bio for Beth. You can connect with her on LinkedIn. Check out her vibe coding for EA's website. And then, yeah, she's in Circle as well. So she'll interact and you can interact with her there as well on that discussion thread. I think we're getting close on time. I really appreciate all you for taking time to learn about this, what can seem overwhelming tool because there are so many things that you can do with it. But Beth, I really appreciate you tackling the cloud ecosystem for us. And thank you so much for your time, everyone. I think we got to about 350 or so live on the call earlier. So that was pretty exciting. And I hope it was helpful. Again, stay connected. Go to that post that's linked here in the chat. Stay connected. If you're watching the replay, comment below. Tell us what you think. Ask questions. We want to keep the conversation going and be a resource for all of you.

Beth · 1:18:12

Thanks everyone so much for joining. It is my very favorite thing to talk about this stuff. I love nerding out on about it. So if you have use cases that you just want to share, you have like that light bulb moment, which is, I love seeing people get the light bulb moment where they create something for the first time and post about it in circle. I'd love to hear about it. Thanks everyone.

Jeremy · 1:18:29

Thanks again, Beth. Thanks everyone. And we'll see you in circle, um, in the discussion thread. So talk soon. Hey, friends, thanks again for listening. I hope you enjoyed that conversation. Again, you can go to the show notes at leaderassistant.com/393 to watch the actual video replay, the full version, the full length version of that webinar. But also you can get the links to the token simulator, the other resources mentioned. The slide decks from myself and Beth. And yeah, you can engage in the conversation, post in the circle post in the comments and ask questions, share how you're using AI in your role and all that fun stuff. So leaderassistant.com/393 to again, watch the video. And I'll link directly to that circle post where we replay the video of the webinar and have all the resources and the conversation continuing. Thanks so much and keep leading well.