Intro to Claude Cowork: How to Get Started

Hey there, it's Mike Stelzner. I was just in downtown Franklin having coffee with Michael Hyatt, and we were actually talking about AI Business World and how excited he is to be attending. And he's not just going to show up and teach, he's going to be hanging out the entire time, and he's really excited to be hanging out with 1,000 other marketers who are committed to come and to learn. Because the truth of the matter is that AI is taking the marketing world by storm. And it's an incredible competitive advantage. And doesn't matter whether you are 70 years old like Michael, or in your 50s like me, or in your 20s and just getting started. I want to encourage you to come and experience the best training you will ever get in your professional career because I've recruited the absolute best of the best. It's taking place in the Anaheim Convention Center right next to Disneyland on April 29th and 30th. Grab your tickets now and commit to taking your AI marketing to the next level. Go to aibusinessworld.live and get your tickets today. I can't wait to meet you there. Welcome to the AI Explored podcast, helping you put AI to work.

And now, here's your host, Michael Stelzner.

Hello.

Hello.

Hello. Thank you so much for joining me for the AI Explored podcast. AI Podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzner, and this is the podcast for marketers, creators, and business owners who want to know how to put AI to work. AI agents are available to everyone now. Today we're going to explore how to get started with Claude Cowork. My special guest is an AI strategist who helps professionals establish their authority with AI and LinkedIn. He's host of the Trending Communicator podcast. He's co-author of The Most Amazing Marketing Book Ever, and his business is Lilypath. Dan Nessl, welcome to the show. How are you doing today?

Mike, it's great to be here. I will, I will not lie. I'm going to tell you, I've been waiting for the invitation for ages. I am so happy to be here and I appreciate it and appreciate all you do for us and for the industry.

Well, it's my pleasure. This show's not been around that long, but I know it feels like ages, so I'm super excited to have you here. And you probably want to be on my other show. Yeah. Yeah.

Just the chance to sit with you for on any of your shows. I'm here.

That's super cool. Well, let's start with your story. How'd you get into AI?

Sure. I don't know, you know, like I think I got into it, um, like many people who can't sit still and who have, I guess, institutional and professional ADHD, uh, like me. It was a new thing, it was technology, and it promised to make my life easier. So sign me up. That was like the beginning.

When was that exactly?

That would've been November, the end of November of 2022, just as soon as ChatGPT launched, you know. But the reality is I was part of a group of, of of people called, which has evolved into what we call the Rise Community with Mark Schaefer. And you know, of course, you know Mark. And we've been, we were talking about all this stuff. And as soon as this thing launched, the texts are going buzzing like crazy. And, oh gosh, you gotta see this thing by Shelley Palmer. And you gotta like, so we start looking at it and I just, I felt like it was like putting on an old leather shoe. It was so comfortable for me. So from that moment on, I have been like, I don't know, I want to say, I want to say neck deep, but it's more like totally submerged in the world of generative AI. And, you know, I was a corporate guy at the time, so finding ways to use it and implement it was really challenging. And I had to, you know, make a lot of, I suppose, uh, new excuses, uh, during my job and try to figure out ways to bring it into work. And it actually ended up working in my favor for the most part, and it allowed me to experiment even more under the auspices of my company. And then, you know, I went on my own eventually, and here I am now. So it's been a, it's been —of a journey.

Well, awesome, super awesome to get you on the show. I mean, we, we've connected at so many different conferences over the years, so it's really great to have you here. All right, when it comes to creating agents, which is effectively the bucket, the category under which we're going to be talking about today, what do you see as one of the biggest misconceptions that are out there? Because I know you're talking with a lot of people, they're talking to you, and there might just be some things that we have to dispel before we get into this today.

Let's talk about that a little bit. For sure. And I, I just want to I'll put the caveat out there that I'm not an agentic AI guy. I know, I just experiment and do what I need to do. But that said, I think the first issue that we've seen and the first misconception is one just of what the heck is an agent? Right. Nobody can agree on what an agent is, or maybe we can, we're starting to agree now. But you know, you hear things like, oh, it's something that you just kind of let loose and runs around and does whatever the heck it does and ends up getting your credit card numbers and you're done, right? That's one version. Another version is like for our Copilot users out there, Sorry, they see agent and they think basically custom GPT, you know, like, so there's that whole range in between. And, you know, to me, basically an agent is any AI, like semi-autonomous or autonomous AI, let's call it action or activity that you start and you let it go and it has a degree of— it could do multiple things. It's not just working on one thing. It can actually, you know, when the AI actually accesses something else that's not in the immediate, like, interface. For example, when it can access an app, when it can access your web, when it does one of those things, then it becomes agentic. There's an argument to be said that it's— that's like, oh, that's just a basic automation. But I think it goes, you know, far beyond from that moment where you're just, okay, I think I'm going to create an agent that's going to collect newsletters for me, all the way up to I'm gonna just leave my computer open and let, uh, Cowork and Dispatch just run my day so that when I run my— run overnight so that I wait— when I wake up in the morning, I've got all kinds of things done. That is extremely agentic. So, so there's, there's a range. Love it.

Okay, so let's talk about Claude Cowork. Uh, we're going to get into the how side of it in detail, but what are the benefits? What are the— what's the upside when someone embraces Claude Co-Work? And I'm going to confess, I have really not done very much with it at all. So I'm asking just as much for myself as I am for everyone who's listening. What, what does it open up? What are the advantages or benefits?

Well, first of all, I don't think you're certainly not alone. I mean, it, it's still technically a research product from, from Anthropic, I believe.

It's less than 6 months old, right? I mean, it's less than a few months old.

and, um, people look at it and go, okay, what, what, what is this? Why did my, uh, desktop app for Claude all of a sudden change? And it's hard to kind of initially understand, well, why do I need it? You know? But I think that the advantages and the kind of the, you know, the, the benefits, uh, so, so to speak of jumping in aren't apparent until you start to jump in. You know, it's one of those kinds of things. But what I learned real fast, uh, about that, and when I started playing around with it, I thought, okay, this is, this is an interesting way to centralize some of the activities I have going on in Claude. And then I realized, wait a second, this is an interesting way to centralize some of the things I have going on on my PC. And then it sort of started to dawn on me that it was a way to, uh, enhance and augment the building and the work that I've already done, whether it's in Claude, on my PC somewhere else in, you know, in my ecosystem, all from one convenient location, because they— the agent that is Cowork is actually a, a system of many agents, and it just does the job. I mean, that's— it's incredibly useful in that way. But I think you start to see the benefits immediately because, you know, it, it, it becomes a one-stop shop, like I said, for everything you've been doing in cloud. And, you know, especially if you're a cloud user, the benefits are pretty clear immediately. But even if you're not, it— the fact that you start to work within files on your PC, the fact that you start to, you know, that you can see that it's connected to not only, you know, to your Google Drive and to things like that, which is nothing new in the world of AI really, but you start to be able to connect dots and understand the workflows that you have better than you did before because cowork helps you understand them. And, um, it goes all the way, I think, from there to, uh, to this kind of ultimate benefit, which is— and I heard, I'm not going to take credit for this because it was a brilliant quote from, from Nate B. Jones. I don't know if you ever see any of Nate B. Jones's videos. Uh, guy's a genius, very technical fellow, but he said that it gets work off your desk. Like, that's what it does. It gets work off your desk and it, you know, that's I think a killer benefit.

For folks that are regular listeners to the podcast, the week before this week we did a podcast episode about Open Claw and this week we are doing it specifically on Claude Cowork. Now, while Dan was talking, I looked, Claude Cowork came out in January of this year, so it's extremely new. Open Claw came out like literally I think November or December of the prior year. So these are very, very, very new applications and Claude Cowork is by the company Anthropic. Right. And Claude, as anybody who listens to the show knows, is one of my favorite apps that I, that I use all the time as far as AI stuff goes. So I just want to kind of set the groundwork that this is very, very new. And I work really hard to try to find people that understand this stuff, that can speak to it with a certain level of competency because they've done it. And when Dan and I connected and we decided to do this topic, I was very excited about this. So, Dan, what do we need to understand about Claude Co-Work before we get into the actual how side of it, because you've already made the case that this goes beyond your typical AI tool. Yeah, it goes beyond ChatGPT, it goes beyond Google Gemini, it goes beyond Google Claude because it has the ability to do things that frankly those tools cannot do on their own. But just what, what exactly do we need to understand? Just help people kind of wrap their brains around. Kind of some basic stuff they need to understand?

Yeah, I mean, wrapping your brain around is, is, is probably an apt way to, to, to start that, that discussion, because I think initially you need to have the right mindset before you jump into something like this. And that is, you're not going to go in there and chat with it like your buddy, or maybe you could eventually, but you're not going to treat it like a ChatGPT or like a regular interface. You want to be a manager, you want to be a supervisor and a manager. The first thing is you're going in there imagining that now you have a team. And maybe I should have said that in the benefits section, but essentially co-work becomes— it's the fulfillment of that AI team that we have been talking about for ages. So when you have an AI team, the team needs a manager and a supervisor. So you got to go in that with that managerial and supervisory intent. It also means you need vision. You have to be clear about your intention. The better you are in describing and understanding everything that you want to do in a given session, the better that Cowork will work for you. It's the same with any AI. Of course, the context is key, but with Cowork in particular, it's gonna go in a direction and it's gonna keep going. You can stop it, but you wanna make sure that you're not wasting your time, uh, and you're, you're focused on your goals and on your tasks and on things you wanna take care of. Um, the other things that you, that you need to take care of, more kind of housekeeping logistical type, type things. Like, it works on your computer, like it, it works in your PC. You have to be comfortable with that. You have to have the security on your PC or the security settings in mind when you do that. And it's worth noting that when you are using Cowork, if you're using the, the Claude Pro or Max plan, you know, you need to be actively toggling off whether or not to make it available to training. So you have to actively opt out, let's say, of the training. And if you're on an enterprise version or a team version, it doesn't share the training, but it's not logged yet for compliance purposes. So anything you do on Cowork is pretty much, you know, on you. Um, and you should check with your security levels at your company or wherever you are that— see if it's allowed. Um, the next thing that you have to understand, of course, that you're allowing it to run free on your computer to a point, but it keeps asking you for permissions. So again, if you let it into some place that it shouldn't be and it pulls out data that it shouldn't have, it's your fault. Um, just remember that. And then the fact that, you know, Mike alluded to it, to it earlier, it can operate different apps within your PC or, you know, through the web. And we're not talking about MCP connections, although those exist. We're talking about it, it can actually go and, you know, take your API and build something for you someplace. You know, like there's a lot that it can do. You have to be mindful of all of the security and, uh, all of the compliance risks that might come with that. So you are giving up a lot, but at the same time would say different from OpenClaw, which you, you've just spoken about, Mike, Anthropic provides and, you know, with Claude, a much safer environment in which to do this. So your Claude desktop. And by the way, this only works through the desktop app, doesn't work through the web. So you use your Claude, you have to download Claude desktop if you don't have it already for your PC or for your Mac, and that's where it works. It's kind of a sandbox. It's not going to go anywhere. It's not going to cause any problems unless you let it loose on the world and tell it to.

Okay, I want to ask a couple clarifying questions. Um, if somebody already has a Claude account, they're paying their $20 a month or whatever it is, is that good enough, or do they need a higher level tier to use a tool like this?

What's your reaction? Well, initially it was only rolled out to what they call the, uh, the Max plan people, but I think now it's available to anybody with Cloud Pro. Yeah. So you need to be paying. Uh, it's just you run in, you run into uses, usage limits. Um, when, you know, you're at a, if you're not paying the most, et cetera, you know, the way you would expect.

Yeah. I have like whatever $20 a month plan and it's available to me. Yeah. I know that is available. Okay. So the next question has to do with, um, a little bit of security stuff. Mm-hmm. The discussion that a lot of people are having specifically about utilizing tools like OpenClaw and maybe even Claude Code is, hey, maybe find a separate computer because it's going to have access to a lot of things and it could do things to your computer that could be not good. Is that not the case with Claude Cowork? Is Claude Cowork a little more safe to use on your everyday work computer? Talk to me about that.

Yeah, so, you know, I, I would defer to our friend Chris Penn for some of these discussions. What are you doing though? You know? Yeah, but I believe that, you know, and I know that Claude, if you run Claude code or, or co-work, it is fundamentally safer than an OpenClaw or one of these other open source products, right? In some, and in some cases you do get what you pay for. You're paying for security, you know, you're opting out where you need to opt out. You have the limits that you can place. And of course, if you're on a team or enterprise version, much, much more secure. Um, so it's a matter of, of how far do you let it in to your computer. I wouldn't recommend— like, I don't think you need a separate PC, a new PC, to run something like, you know, Cowork. I would if I was doing OpenClaw. I would take out one of the PCs I haven't picked up in a while and dust it off and say, you know, maybe it'll run OpenClaw. Yeah. Um, and— but yeah, I don't think you— the, the security concerns— while you should always be concerned about security and act as if you're gonna be hacked any second, right? You should be more at ease with Cowork than with the other ones.

Here's what we heard so far. Uh, you gotta have a paid account with Claude. Any one of 'em should do. You might hit limits, you might have to upgrade depending on what you want to do with it. Mm-hmm. You have to manage this thing. Mm-hmm. Uh, kind of like you manage a person. Yeah. And it does work as a downloadable standalone app on your computer. Um, and just make sure you have the settings in Claude turned off to train the model because obviously it's going to get access to things. Presumably that you don't want the Claude ecosystem to get trained on. So then as far as actually what can it do? Yeah. And look, I'm assuming you've been a regular Claude user before you started Claude Cowork. Yeah. So there's a lot of people that already use Claude like me, and we already know that you can do a lot within Claude with projects. It's kind of a very powerful thing and you can connect things to projects using connectors and all that kind of fun stuff. So what is like the extra push that comes from CoWork that is not available, you know, on its own. Yeah, exactly.

Yeah. So the project— a Claude project, for those of you out there who, who may not be familiar with what a Claude project is, it's, it's sort of Claude's answer initially to what a custom GPT is or something like this. But you create a fixed environment within Claude, like a little Claude within a Claude that has its own system prompt and its own resources, etc., right? So, and you can work within that to achieve very specific and certain goals. They're really robust, and there's a lot you can do with them. They kept adding more and more to them, right? They kept adding connectors and different things. Then where, where Co-Work comes in is it takes that project and then adds an entire layer of, okay, let's bring this project out into the world in a way and really enhance everything that you can do with it with access to the web, access to apps, access to other things, and bringing Claude code into the picture and bringing building and creating into the picture and finding new ways to maybe take your project in new directions. You know, I think so many people have built great, great Claude projects and, um, you know, through extreme effort and ingenuity have create— have transformed those projects into apps, sometimes skills, which is something we'll talk about. Um, which makes a lot of sense. But now you have a whole new frontier, or, or expansive unknown frontier, where you can take what you've built in the project and really kind of blast it out. So like, let's say for example, you have a Claude project that focuses on— we'll just say content creation, very simple, right? You have a content creator for yourself. Uh, you might have built a skill on that before if you were really advanced, or you might have, you know, kind done a lot with it with connectors. With Claude Cowork, you can connect Cowork or connect the project into Cowork, so the project, the entirety of the project, becomes context for what you're going to do in Cowork. Oh, then you can use Cowork as if you're using the project. You know, you can tell it, I want you to follow the rules of the project, or like, do you know, look at, check out the, uh, the instructions I have for the project. That's what I really want you to do. But You know, what am I missing? What else can we do? How can I automate parts of this? And then because of the way that it uses resources in context, you can go a lot longer in a co-work conversation than you could in a project, you know? And I am not super, you know, familiar with exactly how the technology behind that works, but when it approaches the context window, it automatically will compact the conversation and continue to keep going, and it will look for ways to, uh, to use less context. So the conversations go a lot longer, um, you can get a lot more done.

You know, as I'm thinking this through, I think I have an analogy. It's kind of like when you work with a project or any kind of AI bot, for lack of better words, you have to micromanage every single little task. It does what you ask it to do, it takes a few minutes, and then you have to look at it and you have to go back and forth and back and forth. It seems as if What Cowork does is it can tackle larger projects because it can actually go out and do all the pieces kind of on its own instead of doing them step by step. So it requires a little bit less active management, a little bit less micromanagement. It's the equivalent of hiring an employee who knows absolutely nothing and is constantly standing in front of your desk asking questions versus one who's a little bit more experienced and is only going to come back when it you need something. Is that kind of what I'm hearing you say?

Yeah, I think so. And I think, you know, a good way to think about it is you— so let's go back to our content creation project, okay? You've got, you've got the Mike Stelzner bot, or whatever you want to call it. Whenever you want to write something in that, you upload some document, or you kind of think about, okay, here's what I want to write about, and you bring in some resources, etc., right? And you go through it, miraculously you get some great content out of it, I hope, right? Well, with Cowork, let's say you connect Cowork to the project You don't have to go searching for documents anymore. You just start to say, look, here's what I want to do, uh, Claude Cowork, uh, for my content engine. I want something every day. I want a new article on this topic. I want you to go out and search for it. Then I want you to bring it, bring it here. And using the guidelines I have in my project, I want you to create 3 summaries and 2 LinkedIn posts and so on, right? So you can do this, have it all done for you by the time you wake up in the morning, or, you know, while you're getting a cup of coffee or whatever it is. And you've just sort of radically expanded your capability and maintained the quality gates that you've probably put into your product or to your project.

I love it. Let me ask you this. Is it taking over my browser so I cannot work at the same time, or is it all done behind the scenes?

Well, it's interesting. If you want it to take over your browser, you have to allow it to, of course. Um, but when it does, you can see it. You can— it opens up a tab group. If you're in Chrome, it opens up a tab group and it's working over there. And you may get, you know, you may get these kind of Cloud is now debugging your browser type thing, and initially that freaked me out, but you can continue doing whatever you're doing while it does its thing in its browser windows, and you can go look at what it's doing. You can open up, you can go click over into its tab group, and you can see it moving the mouse. It's slow, but it does the job. Yeah, so you can continue working on everything else while it's going in the background.

You've heard me talking about AI Business World for a little while, and I know some of you that are listening probably don't have the budget to travel to Anaheim, California. And I want to let you know that AI Business World is a subset of Social Media Marketing World. And we happen to have something called a virtual ticket. And with the virtual ticket to Social Media Marketing World, you get all of the content from our AI Business World content live streamed. So we're talking more than 20 AI sessions, multiple main keynotes, including myself, but you also get the recordings of all the other content that's taking place. At Social Media Marketing World. So if you really, really deeply want to learn, but you just cannot afford to travel, grab your virtual ticket to Social Media Marketing World by going to aibusinessworld.live. Go to the register page. You will see the virtual ticket option there, and you have a chance to experience some of the best training that you'll ever experience. It'll be live and it'll be recorded. Again, aibusinessworld.live. All right. You have a couple of examples that you and I, uh, agreed on. Uh, one is about WordPress and the one's about SEO. Let's give those examples and then we'll dig into some of the granular details. Yeah.

And I will say that this is, this is kind of, you asked earlier about how you wrap your head around it and you wrap your head around it by experimenting. And my first kind of run with it to kind of really see how it could go, especially with browser takeover, it was okay. All right. Cloud Cowork. Let's see what you can do. Um, I have a WordPress website, uh, and I'm not a WordPress guy. I, you know, I've learned a lot. I can do a lot with WordPress, but I was having some particular issues with, uh, plugging in snippets of code, which incidentally, 6 months ago, snippets of code would never have been in my vocabulary, right? So the fact that I'm saying that is, is a miracle.

Code snippets, I think, is what they call them.

But what, code snippets, you know? And thank you. So yes, still not properly in my vocab, but it's all good, you know, code snippets, anything. And I didn't, I, I didn't even know how to do it. And I said, okay, let's do it. We can do— so I opened up, uh, Cowork, and I said, all right, here's my problem, and I want you to go to my WordPress website, and et cetera. It says, and then it goes to the website, can you log in for me? Sure. I logged in, and then, uh, it proceeded to go through the dashboards in the backend of my WordPress to locate where I needed to make the changes. And it said to me, oh, what you need is a plugin that allows you to insert and inject snippets into the page code. Would you like me to install that for you? Here are your choices. And it gave me 3 choices and said, this is the— and it told me this choice is probably the recommended one because it has 5 bajillion daily users. These are a little less. So I say, fine, do what you want, right? Make it happen. And it wasn't like instantaneous. It walked me through the process. But by the end of the, of the session, I had the snippets were where they needed to be. It didn't break my website. And in this case, it was snippets for, for some advertising code, you know, for some code, for some tracking for code. So like it put it in there perfectly well. Now, success. I thought this was tremendous. If it could do this, what else could it do? And then my brain started going, you know, so I said, okay, still in the same session, you must have noticed, Claude, that I have Yoast SEO on there, but I also have Google Tag Manager and all these other things. I have all these things. Help me set up all the things. Uh, and, um, you know, to my, uh, great pleasure, it helped me come up with all of the ways to, you know, when you build websites, sometimes you don't fill all the stuff in. You're just like, okay, I'm just going to do this and let Nope, it gave me everything for everything and gave me, at least at that level, what I needed to have an SEO-optimized website. Now, of course, then you need to have the human who's really an expert to come in and say, well, you know, you should do it differently. And that's— I would totally recommend that. But if you're just getting off the ground, oh my gosh.

Well, I mean, you could technically tell it to learn how to do that and be the expert and have another co-work session come in and actually review it, could you not? You absolutely could.

And maybe that's a good segue into our discussion about skills and about, uh, about plugins and so on with, um, with, uh, with Cowork.

Yeah, before we go there, before we go there, first of all, we've talked about projects and, um, one of the questions I've got is has to do with like costs, right? Because if anybody uses Claude, they know that at least as of this recording, Opus is the greatest model, but it also kind of burns through your credits pretty quickly. And then Sonnet is the next most powerful one. Um, with Cowork, do you kind of get to choose, or how does that work? Like, let's talk about that a little bit.

Sure. Just like anything else, any other Claude interface, you'll be familiar with the interface if you're familiar with Claude. You have the box and you can choose your model and, you know, make connections and all this kind of stuff. Just some of the options are different. But yeah, you can choose your model and decide where you're going to burn your tokens if you're, if you're concerned about that. Um, and if you're not on a Max plan, you should be concerned about that, probably, especially now that at the time of this recording, you know, there's been an influx of users. So yeah, Anthropic is tamping it down a little. So yeah, you can be careful, and there's certain things you definitely don't need Opus 4.6 for, and you know, it's on.

So, um, you were telling me in preparation that you had some projects that you had created that were not optimized for Opus. That's right. Talk a little bit about how you used Cowork to help you with that.

Yeah, well, once I discovered that, um, that you could connect projects into Cowork, and you could do that in one of two ways. You can go from the project. There's a— if you're in Cloud Projects, you could say open in Cowork. Um, but now Cowork has a little add button at the bottom. You can, you can add a project in, and, um, I believe you can add multiple projects into one session, but don't quote me on that yet. I've been using Cloud Projects since there's been Cloud Projects, whatever that is, to— yeah, 6, 12 months, 18 months. I don't know what the number is. You, you too, right?

I feel like it's been forever. Yeah. Yeah.

So you go back and you say, okay, well, I designed this project when, when we were still on, you know, Sonnet 4.1, and, or I, you know, whatever. And now we're on Opus, and the prompts or the system prompts that you have that you've built under Sonnet don't work the same way in Opus, or it's not necessary to work the same way in Opus, and vice versa. So I had a few projects that I was working on a lot, and I said, oh well, maybe cowork— you know, normally I would, I would take the project instructions copied into an MD file and say, okay, hey, you know, let's go through this, let's, let's tidy this up, but open the project in Cowork. I said, Cowork, this project was designed for Sonnet 4.5, um, it will do better with Opus 4.6. I want you to optimize it for me. Great, will do. And then it just went, uh, and it gave me a complete new set of instructions. It revised the, the system instructions or the system prompt, which was able to cut it down by almost half. Wow. It told me to get rid of a bunch of the resources that were are like now considered to be redundant because Opus 4.6 has a lot of inferential capabilities that you don't need to spoon-feed it everything. And, um, yeah, it ended up like I went— I had one project, I never used the max amount of space allowed, but I had one project that was up to like 30 or 40, 40% of usage of like, um, allotted space. After optimizing, went down to 15%. The thing works a lot faster and it works better. I mean, I didn't necessarily trust him. Why don't you need that guideline that I spent so long creating? You know, "No, we don't need it." I love it. Claude is principle-based.

So did it make a new project for you, or did it actually just give you the system instructions and you had to do it yourself?

I'm just curious. Well, at that time when I did it, it didn't create a new project for me. I don't know if I could do that today. Maybe it could. Okay. But no, it didn't. It just told me, here, just cut and paste these over what you got and make it happen. Love it. And I did. It was easy.

Okay, so talk to me about connectors, because we've dropped the word connectors earlier. We mentioned that connectors are something you can use inside of, um, almost anything in Claude. Yeah. What do connectors do and how do they differ specifically when it comes to Cowork, if, if at all?

I wouldn't say they differ that much when it comes to Cowork, except Cowork uses them a lot more efficiently. What are they exactly?

Describe what they are.

So, okay, connectors are generally, you know, usually MCP-based to—

for the non-technical audience, what does that mean? It's a—

I knew you were gonna ask me that. It's— you don't have to be accurate. Yeah, something context protocol. Yeah, yeah.

But, but, but like describe it in non-technical terms. What is MCP?

So it, you know, there's, there's different ways to connect from app to app.

Model Context Protocol, by the way. I just looked that up.

Model Context Protocol. Thank you very much. Yeah. There's different ways to connect from app to app. Right? You can use an API, which is like— was never for the faint of heart before. Now it's— now it's not that hard, but you can do it that way. But an MCP connection is like a USB that goes between one app and another. Like, it's just a universal connector.

Okay, it's just— it's like a connector. Okay. Yeah. And what— why would you want to use them at all?

Because when you have different apps connected in by MCP or, you know, or by API— but here we're talking about the connectors, the MCPs, generally speaking it gives Claude seamless integration and, and use of these different tools. And then it can make decisions on the fly about whether or not which, which connector is gonna be necessary. Or, hey, I'm gonna try, you know what, you, oh, I see you have Airtable. Let's, let's see if we could build this out in Airtable, right? Let's build this out in Google Sheets. Let's, where do you want the output? So it gives it a lot more, many more options from which to, uh, to build whatever you're gonna build or solve what you're gonna solve. And enables you to really just expand your tool ecosystem in ways that you may not have thought of before.

Well, and for folks that are curious where you find these, if you go into Claude and you go into your settings, there's a thing called connectors and it looks like there's like 1,000 of them. I mean, loads of them. I don't even know what half these things are, but, um, what are some of the ones that I'm seeing? Some stuff in here like Ahrefs and ActiveCampaign and Airtable and Asana. By the way, that's one of the ones I've used cuz we use Asana. Um, but like, uh, GitHub, Gmail, Google Drive, Google Calendar. What are some of the ones that you recommend and, um, how are you using them? Just so people can understand that.

The beauty of MCPs is that, and the reason there's so many of them, is that it's situational. So you, as a marketer and a communicator, what are the things that you're doing all the time? You know, like I don't need an Asana connector, so I don't have one. Mm-hmm. You know, it's available to me, but why would I do it? But I use Airtable a lot and I use Google Drive all the time and I use my Gmail and all these, you know, these things. So I chose the connectors that are connected to the things that I use all the time for work. So by connecting to Google Drive, now I give Claude access, easy access to the folders and to the, to the files I have saw— I've saved there, but I also give the capability to write to it so it can do outputs that way. Um, by connecting to Airtable, there's a lot you can do with an Airtable or with an Asana or some of those things. So it can You know, Claude now helps me to build out bases in Airtable and reporting and all that good stuff.

You used to have to use third-party tools like Zapier and Make and all that stuff, right?

This kind of bypasses that, uh, to a point, right? So there's some native integration through MCP that you can do. There's some that you can't, and it varies app to app. What I've found is that I try whatever I can do to avoid automation, like to avoid dealing with Zapier and Make and, and, and like what a pain in the neck. And that, by the way, is maybe my favorite thing about Claude Cowork is it just allowed me to skip right over all that stuff, you know? But yes, sometimes you do need an additional, an additional connector. And when you're working with Claude, it will tell you, you know, you really should connect this with Make or with Zapier, right? Which, by the way, also have MCPs within, within Claude.

I wanna talk about Control Chrome because this is a critical one and we've kind of alluded to a little bit earlier. Explain. Why that's so valuable.

How much of your work are you doing in Chrome, right? How much of your work are you doing in your browser?

Like almost 100% of it for me, right?

You know what I mean? Exactly. And, um, you know, sure, you can be doing work through apps, and where do those apps open? They generally open in your browser. You know, when you want to do any kind of coding or design, if you're not using the terminal, if you're not a little adventurous, you're using your browser. Therefore, when you, when you let Claude control your you give it control of Chrome, you're basically saying, you can write this stuff for me, you can fill in this form for me, you can click these buttons for me, you can— this tool that I never heard of before, if you're saying that it's gonna really be the thing that solves my deepest and greatest desire, then by all means, open up Chrome and give it a go. You just generally need to sign in and give it authorization to do it. Um, and you need to be careful, but it is— it's been revolutionary in some ways for me. And, and as I said before, like, it can do it while you're doing other things, you know, and it opens up its own tab group and it goes, you know, hey, I'm checking this out now and I'm checking that out now. It gives you— it gives you the play-by-play if you want to check it out. But it's not only a productivity booster, which sort of goes without saying, it's actually a, I'd say, horizon expander, you know, like it just it will keep going in directions and, you know, take screenshots and it finds things that you didn't find before. And, you know, everything from, from proofing your own website, which is probably the most basic thing, to let's inject some code or let's, you know, open up this app and I want, you know, go ahead and build something for me. If it's, if it can be done through Chrome, then Claude can do it through Chrome if you give it control.

Love it. Okay. You mentioned skills and I think you might've mentioned the word plugins. So I don't know what the difference is. Are they one and the same? Let's talk about that a little bit. Luckily, we can—

we— I think we can talk about them together. Okay. So, okay. So a skill is— let's just think of skill as a task that you do a lot that you can now, uh, you can create a little mini package program. It's almost like a little custom GPT that you insert into Claude. So for example, I have a skill that I built, uh, called the Ideal Customer Profile skill. So I knew what my ideal customer profiles were. So I created the personas and I just said, Claude, build a skill from this. So Claude built a skill so that now if I create something or if I have an idea, I say, can you run this by my ICPs? And it invokes the ICP skill and then it does what the ICP skill tells it to do, right? And you know, a skill is, is, is a little package of several assets. That's all it is. Now you can have a bunch of different skills and Claude comes native with a lot of them. Like when it writes a Word document, it's using a Word, a, a Docx skill when it's You know, when it goes to PowerPoint, using a PowerPoint skill. Plugins are collections of skills. Okay. I think that's the best way to think about, like a bucket of skills. So plugins are a new thing for Cowork, but essentially you install a, let's say you install the legal plugin, which is I think what I talked to you about earlier, Mike, is the legal plugin is a collection of, I, I don't know off the top of my head, but let's say 6 or 7 legal skills. So one of those skills is like contract. Review, and one of those skills is, um, is risk assessment. So when you have a legal issue, you know, you can invoke the legal plugin and it will find what it needs to do. But in fact, you generally don't even need to do that because if you have the plugin installed, Cowork knows, hey, I'm going to use my legal stuff and I'm going to do this. Um, I should say, by the way, that between the last time we talked and Now, um, there's been a, a court case that was pretty— that got a lot of air about, um, and I can't recall the exact— it's U.S. versus Hadrick or something— but anything you do in Claude is not considered attorney-client privilege. So be careful with using legal plugins and then, you know, acting on them, um, before you consult your attorney.

Okay, so it's important, but there's different kinds of plugins and Yeah, I want to talk about this a little bit. So when you're inside of Claude or you're inside of the app, it's called Customize. It looks like a little toolbox, right?

Yeah, they changed it. It was Connector. Now, like, it was Connection to go to— you go to Customize and you can add these things.

And underneath there you have Connectors and you have Skills. So first of all, how in the world do we make a skill? And then how do we find these other skills as well? Because it's not super intuitive to me, or like right out of the gate, how in the world to do this.

Well, there's a skills marketplace just like there is for anything else. And you can, you can, uh, when you go and you add a skill, you go to like manage skills and—

Oh, here we go. Yeah, I'm in it right now. And I actually see like when you click on skills, there's a plus at the top. There's something that says browse plugins and I see productivity, design, marketing, data, engineering, finance. There's like just a limited amount. These are presumably things that these plugins are presumably made by Anthropic and partners, it says. So it tells you, right?

It tells you whether it's Anthropic created or not, you know. So those, those are the plugins. Okay. You know, and, and skills are just like, you know, one-off little packages that you can— there's a marketplace for them. You can, you can find a lot of them. You can create them yourself. It's easy to build a skill. Like when I had the ICP situation, I just knew that was something I kept asking Claude to do. Can you check this again? Here's my 3 customer profiles. I'm gonna load them into this chat. Can you, uh, check this against these 3 customer profiles? Perfect use case for a skill. If you keep doing that, just package it as a skill and just— it's in there and you just say, hey, now check these against my, my ICPs. Um, brand guidelines are a great example for that. You know, you have brand guidelines for, uh, social media, like Media Examiner. Um, you can package that in a skill and you can— anytime it creates a doc or something, you say, now run this against my brand guidelines. So, well, and my understanding is that—

I mean, I know I, I've used Claude to create a skill from a project. I said make it as a skill. That's right, you can do that. And it used to be that I had to download a file and re-upload it, but I think that's not the case anymore. Is that correct? You can—

that's right. Yeah, I mean, it's— they make it easy. It used to save it as a, like, as a zip file, and then you had to drag it over, or it saves it as a skill.md and you drag it over. Um, now it's easy. It's much easier. You can just kind of drag the skill into the conversation and it will, it will, you know, install it for you.

Yeah. Have you used any of the off-the-shelf skills, or do you prefer to make your own skills because you have a little bit more, you know, customization options? What's your thoughts on that?

I've used both, you know. I, I haven't made that many myself because in, you know, in the realm of sometimes, uh, when I'm creating content especially, or when I'm writing, or when I'm kind of evaluating an executive profile or whatever it is I'm doing, a lot of that is eyeballing and my own knowledge that isn't really necessarily packaged into a skill. But when it comes to things like editing or like creating or just like if I'm putting things in brand guidelines, yeah, I have skills for those. Um, and checking against my ICPs or personas, you know, I start the, the new company that I started called LilyPath is, uh, is all about LinkedIn optimization at this point. There's— it's going to be more than that, but at this point, right, if you want to, we have a very unique way of, of, of evaluating and optimizing profiles That started months and months ago as a skill in my, um, Claude, you know, a project that became a skill. So like there's this kind of pathway that you can go if you start to play with this. The other thing about skills is that they're portable, so you can send them to other people and they can then install it on their Claude. And by the way, be careful about that, but you can do that.

Yeah, I wanted to ask about that because it is true. Like I've got multiple staff members. We don't have a team account, but they've got individual accounts. Bounce, and I've sent them a skill for my conference that was really well trained up with a big document of testimonials that it can pull from. You mentioned there's marketplaces for skills. Like, do you know where to find those marketplaces? And—

oh, it's just through Claude, through the, um— oh, okay. Because when you add skills, you can just kind of scroll through and you'll— got it. You know, you'll see—

I would imagine there are third-party marketplaces for skills, but like, I would be really reluctant to using them, wouldn't you? Because who knows what their instruction sets are, right?

Whenever it says something like, uh, hey, check out the Skills Marketplace on GitHub, I just generally don't do it, you know, like, yeah, you know, not that there's anything wrong with GitHub, I just, I'm like, okay, but I would imagine you could upload the thing into Claude and say, please study this and look for any nefarious stuff that's inside of it, and Claude's pretty good at that.

You could imagine you could certainly do that.

You can use, in fact, there's a skill called the code review, I think, in the engineering plugin. Uh, there are such things where you can check on whether or not something is safe or not. You know, whether you trust it is another issue, but you can.

How often are you using Cowork versus Claude itself now that you have access to Cowork?

Yeah, it's completely flipped. It used to be, um, when Cowork first started, it was probably 90/10. I was using my projects to Cowork. Now it is the opposite. I mean, I, I barely use the regular, the regular run-of-the-mill Claude anymore, and I feel Maybe.

Did you have to upgrade? Yeah, because you were using so much tokens or whatever.

No, I mean, since, since Claude has been the backbone of my business, uh, and my consulting and my training services for a while, I, I went full in right from the get-go. So I've been—

you had like a Teams account or whatever, or the Max?

What is the cost? Yeah, I, I, it's that— I went to $200 a month plan because it was absolutely worth it for me. I wouldn't recommend that everybody do that, right? But especially if you're, if you're going to start coding and building, it starts to make more sense to go to the higher plans. And now with, uh, with LilyPath, we're on a team plan, so we have, we have a little bit more leeway in what we do with token usage and so on. But yeah, it's— it, it took a little bit of an investment, I'm not gonna lie, you know, to, to do that.

But I would imagine it was a very wise investment because the alternative, uh, investment would have been an enormous amount of your time or an enormous amount of someone else's time, and both those things are very valuable. Dan Nessl, we have just scratched the surface of the amazing things that you're doing. If people would love to connect with you online, what platform is your platform of choice? And if they're interested in checking out your business, where do you want to send them? Thanks, Mike.

Of course, LinkedIn. You can find me on LinkedIn. I'm pretty much an open book. And, you know, just search for Dan Nessl or Daniel Nessl, either one. You'll see me. You can Google me. My business is LilyPath. That's lilypath.com, and, um, I urge you to check it out. And then of course you can find me on the Trending Communicator, uh, podcast, uh, every 2 weeks. Awesome. With an episode coming to you. Awesome, Dan.

Thank you so much again for sharing your insights with us. It was amazing, Mike.

It was a pleasure and an honor and a privilege to be on the show. I appreciate it.

Hey, if you missed anything, we took all the notes for you over at socialmediaexaminer.com/episodes. A103. Be sure to follow this show on whatever app you're listening to us on. And if you've been a listener for a while, we would love a review, and we'd also love you to share with your friends. Do check out my other show, The Social Media Marketing Podcast. This brings us to the end of the AI Explored Podcast. I'm your host, Michael Stelzner. I'll be back with you next week. I hope you make the best out of your day, and may AI help you become more successful.

The AI Explored podcast is a production of Social Media Examiner.

Do you want to go deeper in your understanding of AI? You've been listening to this podcast for a while, but did you know that we have a membership with lots and lots of marketers, entrepreneurs, and creators who learn every single month? Every single month we do live meetups. We have professionals coming on who teach training. And this is exclusive content only available in our AI Business Society. If you're ready to begin committing to ongoing development, join the AI Business Society right now by visiting socialmediaexaminer.com/ai.