[0:00]If you're a business owner still manually doing tasks that a machine could be handling in quite literally seconds, you're not just wasting time. You are actively losing to competitors who had figured this out months ago. So this video, it's going to be about Claude Code, but it's not what you think. Now, most people, they hear Claude Code and they immediately think that it is some developer tool that they can't use. They're wrong, and that misconception, it is costing them dozens of lost hours per week and tens of thousands in operational inefficiency. If you're new to the channel, my name is Nick. I've been working with companies for over two years, helping them implement AI, either driving significant amounts of leverage or increasing their bottom line. We've seen firsthand how impactful implementing AI actually is for organizations and where this is all leading us. Now, I'm going to be covering the exact framework that companies like Telus and Zapier use to automate 500,000 plus staff hours. I'll also cover the three fundamental differences between Claude Code and every other chatbot that just forgets everything every single instance. And I'll provide you with examples with real numbers showing 25 to 75% productivity gains. And of course, the step-by-step breakdown of how to actually set this up and start using it across your entire team. With practical walkthroughs for content, proposals, outreach, operations, and much more. So by the end of this video, you will know exactly how Claude Code works, how to set it up, and how to get your entire team using it for the daily tasks that are constantly eating your time right now. Now, this isn't just going to be a quick tips video. This is technical where it actually matters, but if you are serious about scaling without proportionally increasing head count, stay with me. Let's get into it. So let's first just start by destroying the biggest misconception when you actually hear Claude Code, you know, you are imagining having to do some sort of programming or needing some sort of technical prowess. Now, that's not wrong, but it is catastrophically incomplete because Claude Code, it is simply an AI agent that operates inside of your local file system with the ability to read documents, write files, execute any commands, and connect to external tools and systems via MCPs. And most importantly, maintain persistent memory across every single project that you are working on. Now, all this is being done through simple natural language. So you just simply type in plain English what you want done and the system will write the code for you. Now, why does that actually matter for your business? Because if we were to map the current AI tools on an access. So high autonomy on one side, meaning the tool actually does real work for you and ease of setup on the other side. Well, Claude Code, it is sitting at the top right, and that is where we want to be as business owners. We simply just want to tell something to do the work and it does it the way that we want it done. Now, if you compare this to Chat GPT, Chat GPT, it is pretty easy to set up. It's actually incredibly easy, which is why they have so many users. But it does not do the work for you. It just answers questions, it might write some text, but that's going to be it. You know, they have some other features, but you are very limited to its capabilities. So you are still the one who is having to copy and paste, still the one executing everything. And if you compare this to other no-code tools like N8N, Zapier, now there is improved agent capabilities, but the problem is you need to build every single path. Node one, then you have node two and then build out and map out node three. While with Claude Code, you simply just describe what you want in plain English, as I mentioned, and it figures out the plan and actually does the work for you. And here's what makes this completely different is that traditional AI tools, they suffer from three fundamental flaws that actually make them useful for real operational work. So the first problem, this is what I call context window amnesia. This is where you're spending maybe 30 minutes explaining your business to Chat GPT. So let's say you do get good answers and it is helpful, but tomorrow, it forgets everything. Now, you can leverage their inherent memory in what they have in the system, but it is very limited in its window. So you are essentially just starting from zero every single time. Now I've watched business owners literally waste hundreds of hours re-explaining the same processes, the same client requirements, the same brand voice, just over and over and over again because Chat GPT, it does not have a great memory system. That's not what it's really built for, at least as of right now. Problem two is the siloed sessions. Now, this is where your sales conversation does not connect to your support conversation. So you have two different systems or two different departments not communicating with one another. So, for example, your support conversation does not connect to your operations. Everything, it is quite literally isolated. Now, nothing compounds. It's also just like having a few different consultants who are never talking to each other and you have to explain your entire business to each one every single day. Problem three is no operational execution. So we already touched on this very briefly, but Chat GPT, it does and can give you advice. As you shall know, it can draft documents, but it cannot actually do most of your work for you. So it cannot read your client database, update your CRM, analyze your support tickets, generate reports, and update your project management system. They have recently released some different features where you can connect to your CRM and stuff like that, but again, it's pretty limited. So with all of this, you cannot tie it to one connected workflow. So what I often say is that Chat GPT, it's really just like a smart parrot. Where it can tell you what to do, but you still have to do the work. Claude Code, it actually solves all three, and the best thing about it is that Claude Code can actually program and code these things for you, hence its name, Claude Code. But when you connect it to external tools and you give it specific agent skills, it can do real business work. So you do not need to build any infrastructure for these agents. Claude Code, it will handle the entire orchestration for you. Simply describe in plain English what you want and Claude, it will build and do everything for you. So what you're getting from this is developer-level power without having to be a developer. You know, I think it was Andre Carpathi or maybe one of those other CEOs. He was saying, quite literally, just a couple weeks ago that software engineers, like, they're going to be completely non-existent in about 18 months from now. And that's completely true because from what I'm seeing with these systems, you can drive significant amounts of leverage without having to be a developer. Because where AI is nowadays, it can do incredible amounts of programming for you. And that is a complete game changer as a business owner. Now, let me show you guys exactly how this works visually, so that when we actually get into the walkthrough, you will understand what is going on. So on the left hand side, we have you. Let me zoom in a little bit. We have you, you, of course, being the user. So you can interact with Claude Code in two different ways. You can either give it a prompt like a question, or you can give it a task. Or you can have it packaged into a slash command. So we'll get into that in a second. But I like to think of slash commands like a button. So when you click this button, it is going to just run a workflow that you have previously defined. So instead of explaining everything over and over again, you can just package everything once and Claude is going to run it predictably every single time for you. Now, moving on, we have the Claude.md file. This is going to be your business brain. So this is where Claude's memory actually lives. So the Claude.md file, all this is is just a permanent instruction file that gets loaded every single time that Claude Code actually starts a brand new session. So it's going to contain all of your business rules. It's going to even contain your SOPs if you have them, your brand voice, how you want to write in your language, how you want it to create your emails and format it to talk like you. How your data structures, where tools and files live, what text stack you guys use. So let's say hypothetically Chat GPT is a consultant that you hire every single day who forgets everything overnight. Claude Code, it is just going to be that employee who shows up with that entire employee handbook, all memorized and will never forget a single detail. And the most powerful thing is that you commit this file to your shared repository, and now every single person on your team, they will get that context and that same exact system automatically. So, for example, you can upload everything into GitHub, which is just a database where you can store things securely amongst your team. Moving on, we have agent skills. This is just going to be your AI playbooks. So skills, they are essentially just packaged instructions that tell Claude exactly how to do specific tasks inside of your business. So you can think of them as standard operating procedures all for the AI system. So Claude, it doesn't come out of the box knowing how to write proposals, exactly in your language or in your way, or qualify leads using your criteria. But when you give it a skill, it can do that work predictably, even consistently in the way that you have done it previously and how you want it to done, more importantly. And with that, skills, they do stack. So your lead qualification skill, it can trigger your proposal generation skill, which triggers maybe your follow-up skill. So you're essentially just building a library of reusable components that all are going to be working together for you and your team. MCP, it just stands for Model Context Protocol, and this is just what allows Claude to connect to your entire text stack. So again, like Gmail or Outlook, Notion, Slack, Salesforce, whatever, I mean, there's over 8,000 MCP servers available right now, and there's only going to be more and more. So with MCPs connected, Claude can actually read your CRM and send emails, it can update your databases, post to Slack channels or Teams, and automate multi-step workflows all across every tool that you are using. And you just simply have to add an MCP with one command and Claude, it will handle the rest for you. Last but not least, we have the sub agents. So these sub agents, these are all just parallel executions. So this is where it gets extremely powerful because Claude can spawn separate AI instances to handle focused sub-stacks, each with its own context window. So you can have one agent just researching any competitors, while another one is going to be drafting proposals, while this other one updates your CRM and it's all running simultaneously. And because each agent has its own context window, the main conversation, it stays super clean. So I know this might sound a little bit complex and confusing, just stick along with me. So with all of this, you're not going to be losing context, you're instead multiplying your output. So what would take you eight hours of work, it will take you 30 minutes because sub agents, they do all of the lifting for you. All right, that's enough theory. Let me show you exactly how to get started, and I want you to know that you do not need to be technical. Once again, I'm just going to keep reiterating that and instilling it inside of your brain. So first thing is to just download an IDE, whether that's going to be VS Code, Visual Studio Code, Anti-gravity, Cursor, or even Win Surf. They're all going to be completely free. All you have to do is just search up Anti-gravity and you can find download it right here.
[11:05]Or you can be searching up Cursor and you'll just go ahead and download it onto your computer. As I mentioned, they are completely free, but they do offer their own paid plans. Although if you're just using Claude Code, you will just be utilizing your Anthropic subscription or the API. I recommend using the subscription instead of the API because things can get pretty out of hand in terms of the cost. So once you have it installed, you will be prompted a window like this. So I'm just using Cursor and we can either open a project, clone a repo, or connect via SSH. We're simply just going to open up a project. We're going to navigate inside of our AI folder, create a folder wherever you would like. We're just going to call this Biz Demo YouTube and we're going to open this up. Now, this may look a little bit different for you, but if you are using Cursor or even all the IDEs, they all stem from VS Code. So Cursor, Anti-gravity, Win Surf, they are built off of VS Code. So VS Code is the kind of foundation. So they're all going to be inherently similar. You can change the layout where you can have the agents, we can remove this, we can remove the chat on the right hand side, and we can include the panel on the left, which is just going to show us this is our computer. These are all of our folders. So if we were to add new files inside of our File Explorer, you would be able to see everything over here. Anyways, on the left hand side, you'll have to find whichever system you are using, whether that's Cursor or Anti-gravity, you'll have to search for an extensions section. So inside of Cursor, we just have to click on this drop down. You'll see extensions. I know with Anti-gravity, it'll just be instead somewhere down here. It shouldn't take you too long to find it. And we just want to search up Claude Code. It's going to be available on every single one of these. You may have to do a little bit of scrolling, but make sure you are finding the one that has the highest ratings, because that's going to be the real one. From here, feel free to read any of the description, but all you want to do is just click on the install. Now, to actually start using Claude Code, we are just going to click on these three dots, click on Claude Code open. If you are using this for the first time, which I would imagine you would be, you would have to input your own subscription, you know, log in to your Claude or Anthropic account, or you can apply it just an API key. Now, I again would recommend just signing in with your Claude subscription. So I would recommend to either have the $20 a month, $100 a month, or the $200 a month. Start off with the 20, see if you needs to upgrade later. You actually do that, you simply just search up Claude on your browser. You'll be able to create an account, you'll go to the settings and be able to upgrade, do whatever you would like. And if you want to utilize the API, then you just search Claude developer API and you'll be able to log in to the Claude developer platform and get your API key, you know, put in $10, $20, whatever you want to do. Going back into our IDE, we're just going to remove this chat on the right. This is Cursor's native chat. So we can just utilize any of these models that we would like. We don't have to utilize Claude Code or anything like that, but it's not what we're focusing on. We're going to be focusing on using specifically Claude Code. So we're going to close out Cursor's native AI features. Now, moving on, we want to actually create a Claude.md file. So again, this is going to be our memory. To do so, we just have to type out /init. So this is going to initiate our instance and auto generate our Claude.md file based on our existing files. But I do recommend customizing it. So what is going to be going in here is your company overview, so what you guys do, who you serve. So this can also be your key processes at a high level, even your brand voice, how you communicate with customers, how you want to talk, your data structure, so like where different types of information actually lives, and of course, your tool connections. So like what external tools that you use and how they connect. So just start simple. Do not try to document everything right off the bat. Just start with one process that you want to automate first and you can expand from there. So it's going to ask us if we want to allow this bash command. So do we want it to give it access to our computer and execute some, some code for us? So we're going to continue allowing. Now, you can, if you would like, you are able to bypass all of this, where it's not going to have to ask you for permission every single time. In which case, once we allow all of this, we can just press edit automatically and it would be able to do everything for us without asking for our permission. Now lastly, it's going to ask us if we are allowing it to write the Claude.md file once it was able to read all of our files and everything. We're just going to press yes. Okay, awesome. So it said create a minimal Claude.md file for this empty project. Since there's no code yet, the file is a stub, it should be updated as the project develops with build commands, architecture notes, and key conventions. Now if we just navigate back to our File Explorer, we can see we now have our file. And once again, this is just going to be associated with our File Explorer. So if we come into our Finder, we can go into AI, and if we find where our YouTube business demo was, you'll see here's the Claude.md file. Simple as that. Back in our IDE, this is just going to be mapped on the right hand side, but you can move this around wherever you would like. So if you want to put it at the bottom, you can put it at the bottom. If you want to put it on this left window, you could do so as well. And the point of all of this is because if you are working simultaneously on different projects or different processes, working with different agents or sub agents, rather, then this is where it's going to be important where you can see everything all at once. Obviously, as I mentioned, because we didn't have any files or anything like that, SOPs, documents, guides, inside of our folder. It wasn't able to really do anything, but if there were, it would look a little bit something like this. In which case, I just duplicated another folder that I had, which is basically helping with creating long form scripts. So we have our Claude.md file. All I did is above, I typed init, and it's just going to instantiate our entire environment. So we have the Claude.md file, we have the system overview, operating principles, directive quick reference. So directives are essentially just all of my instructions. So important files that it should understand like our ICP, who we serve, our language that we like to employ. Now once again, this is going to act as our entire brain. So it's going to be the memory. Any time you're adding something into here, so if I want to provide it with a new file, it would then instill this into our entire Claude.md file and continue just expanding upon that. Now next up, we're going to be building our very first skill. So this is where Claude learns to actually do work your way and what I use 99% of the time when I'm actually using Claude Code. Maybe not 99, but a lot of the time, it is able to do a lot of my work for me through the skills feature. So to actually showcase like how this is actually best used, I'm going to be showing you guys how we actually do this and use it for our own company. So I'm going to say create a skills folder at the Claude Skills Exploration Proposal Generator and create a skill.md file inside of it. Here's the content for the skill. So everything I want to create a proposal from a meeting, you should have them look like this. Now, from here, I already have a document of our SOP for how we actually create proposals. So we can see an example of what a proposal should actually look like. We have the output structure, what it should be containing, situation appraisal, what is the exploration, so on and so forth. So this is everything that we want to include in our proposals that we send to our clients. So I'm just going to copy all of this and I'm going to go back to Claude Code and just paste in the content for the skill. And we could just be a little bit more specific. So every time I want to create a proposal from a meeting, you should have them look like this. All right, let's make sure we are using Opus. It's actually on Sonnet, let's go to Opus 4.6, and we'll run this. So now it's going to ask us if we want to allow some of the commands. It'll probably ask a few more times. You can see I just saw something got populated. So we have the Claude, skills, exploration, proposal generator. All right. Now here, it's just going to ask us to allow it to write the skill.md file. We're going to allow that, of course, we can actually open this up and we can see now it is finished up. So we do need this to convert to YAML format. So YAML, this is just going to allow us to actually call this every time we want it to be used. Now, we want to add onto this as well, and we want it to be able to every single time that I have a meeting, or I want it to be able to pull from Fireflies and take that meeting transcript and put it into this proposal. So what I can say is I'm going to be using something called Whisper Flow. Now, Whisper Flow, it allows me to dictate whatever I'm saying into this text box right here. So I'm just going to type out. I need you to also create a skill where it is going to, one, when I ask you to create this proposal for a client, you should be able to pull from Fireflies, take that meeting transcript, put it into this skill to create the proposal generator. So we may have to delete some of what I said at the beginning. Okay, now, that looks good.
[20:03]We'll run this off and we'll see what it does from here. Now, while this is running, what we can also be utilizing is we can go inside of Claude and just use the browser version, or you can use Claude Co-work to interact with these skills as well. And you can create the skills inside of there, but I would recommend to do this on your own computer. Here's how my file looks. I believe it needs to be formatted better for it to be utilized as a skill, right? So can you give me the prompt for that if I'm correct? Now, it's going to ask allowed to write skill.md. We will press yes once again, and we can see, um, where would it be? Okay, it's still generating. Okay, so it actually just updated the skill. So it's now it's going to be expanded upon where it's going to convert those meeting transcripts into this proposal. Now, like I mentioned, we have to convert this into actually being able to be used as a skill. Now, this is only going to be for if we want to utilize this. So if we want to go into Claude and we want to utilize these skills, you can go into the connectors or capabilities rather. And if you scroll down, you can see the skills. So this is where this is going to be more so important because it's going to have to follow that particular structure. So, if you're just building these on your own computer, it's not going to be a huge deal, but all you will have to do is just type out read my skill and reformat to follow proper Claude Code skill format. And I'm just giving it some specifics of what I needed to be doing, but usually if you are just saying make sure to reformat to follow the proper Claude skill format. It's going to look up Claude Code, look up the documentation, and find out what it needs to be doing specifically and handle everything for you. So this part isn't even completely necessary. In which case, it's going to be more so applicable to be doing something like this, where you guys can see we have the pitch deck generator and the exploration proposal generator. So we also have some example skills. So if you want to set something up with like skill creators, where it's going to help you create new skills, where you can just copy this. You can try it in the chat and it's going to help you create a new skill and handle everything for you, so you don't have to. Everything that you're going to be doing, it's not going to be all done in just one way. There's a lot of different ways to get things done. So what I'm going to personally say is, I need you to grab my most recent Reprise AI automation inquiry transcript from Fireflies, and then I need you to create it into a proposal using the skill that we just created. We'll run this off. I'm going to be using Sonnet 4.6. It's a little bit faster and it is going to be doing just as good of a job. So we'll see what this can create for us. So it's going to interact with our MCP that we already have Fireflies connected to. So it just fetched the transcript. You can see the IDs Reprise AI Automation Inquiry. And it's going to start creating everything for us. We can even delete some of these folders. Quick note, if you press command W or control W, if you're on Windows, then it'll close everything out. So now it's going to start creating the proposal in the exact format that I have just described. So it's giving the primary contact, what the investment's going to be, what is the exploration going to be pertaining, what we're going to be doing for them. All the deliverables, everything. And we can even have this put it into a docx or Google Doc if we would like. Now you can be doing this for so many different functions inside of your business. So if you want to be doing this for sales, like going through lead qualification or proposal generation, or for marketing doing content repurposing or SEO analysis, you can absolutely do so. If you want it to be able to scrape leads from the internet. So you, let's say you have a process of lead generation in which the first phase is going to be going onto Apollo and scraping leads, finding all of those. And you can use Appify where you can just automatically scrape everything. So Appify is just an API marketplace where you're able to make calls to different third parties. So here you would be providing Claude Code your API to Appify where it's going to automatically scrape everything from Apollo. And then you just provide it directions on what else you want it to do. So maybe you want it to control your browser and then automatically create the emails based on the findings and maybe you have it as another step to do research into them and search their website, so on and so forth. So you can be doing 99% of your job just through making sure that there's an API available for at least some components of it, but if not, it's able to even control your browser. But with that, it's going to be simple as saying, I just want you to control my browser for this portion and it'll be able to connect to Google Chrome and control your browser, in which that is going to be called a headless browser. But anyways, there is quite literally no limitation to what skills that you would like to be building out. And again, a skill is just a markdown file inside of this Claude Skills folder containing structured instructions. So you give it a name, you give it a description, so Claude knows when to activate it, the tools that are going to be needed and the step-by-step instructions for how to actually do the work. And once it's created, Claude can pull it up automatically whenever the context matches. So maybe the first iteration or first time you build something out, it's not going to be perfect. In which case, just continue iterating. Tell Claude what you want to be done, how you want it to be done, what needs to be changed, what adjustments you would like to make, so and so forth.
[25:25]You couldn't tell already, Claude Code, it's not just a coding tool. The question isn't like is this going to be a coding task, you know, trying to transform what manual processes you were doing right now? It's more so coming from a stance of asking like, can this be done digitally? And if the answer is yes, then Claude Code can probably do it for you. Now, we've been seeing the biggest lift and leverage being generated in content generation and marketing. So let's say you have a blog post that performed well, and you want to just repurpose that into LinkedIn posts, Twitter threads, and an email newsletter and YouTube scripts. So you could just build one content repurposing skill, you point Claude at the original piece, Claude will spawn sub agents, one for each platform, and then each one's going to apply platform specific formatting and tone. And 10 minutes later, you're going to have a week's worth of content in your hands. So as I mentioned, you can also be doing this for email outreach and follow-ups, operations and admin. I mean, you can do it for so many different things, like where with email and outreach, you can have it read your inbox, draft responses matching your your voice, categorize emails by the priority, handle the follow-up sequences as well. Now, moving on to one of the most important sections is getting your whole team to be using all of this. And this is something that a lot of people actually miss is you can build the most sophisticated AI system in the world, but if your team doesn't adopt it, you have quite literally just built nothing. You've created an expensive science project that just sits unused. So I've actually seen this kill more AI implementations than the technical failures. So the system works or I may be there, but six months later, nobody's using it. So, why? That's because businesses treat this as a technical problem where it's actually just a people problem where your team is scared that AI will replace them. They're skeptical because they've tried tools before that just made things worse. They're overwhelmed because they're already buried in work, and if you don't address these fears directly, your transformation, it will fail, point blank period. So what I recommend for you and your business is starting small, starting with something that's actually going to drive a significant amount of leverage, you know, trying to find like the biggest bottleneck that you can solve. Of course, you want it to be a quick win. You don't want to spend too much time on it. You want it to provide a result relatively quickly. Now, from there, if it is going to provide some use to your team to actually share it with them, so maybe if you have five people all having to do the same workflow every single week. Well, in that case, all you will do is you can go ahead and go to your Finder. You can find the folder that you are working, so this is called long form copy and you can just share it with them. So you can upload it to Drive, One Drive, or Google Drive. You can send it across email, Teams, whatever you want to do. They will just have to install the IDE and they can just open up this particular folder. That is as easy as it can get. If you do want this to be shared synchronously, in which case, every time you make a change or make an update, well, there you will have to be using a repository. So you can utilize something like GitHub to share this. And if you want something to go live. So maybe you have triggers, or you have web hooks installed where an email comes in, you want it to automatically run this system. Well, you can deploy all this code into applications where you can run this live. So you can push this to N8N, or you can deploy this code onto a VPS and it'll run this automatically for you. But I'm going to cover that in a different video. So if you guys want to see that sooner rather than later, just let me know if you guys have interest in that, and I'll be happy to create a video. So once your team is actually using this, your business, you will notice that you're operating fundamentally differently. And your role, it will slowly transform. So you go from being in the weeds, firefighting, responding to every issue buried in Slack channels to being the architect. You know, you're reviewing performance dashboards, you're identifying the improvement opportunities, you're even designing new workflows, and you're thinking about strategic outcomes instead of having to react to any operational chaos. And this is really the shift from operator to owner. So you're no longer the bottleneck. You're instead becoming the designer of systems that all run without you. And you'll notice that your hiring changes too, because eventually you'll add head count again, but the roles are completely different. You're not hiring someone to manually process invoices, you're not hiring someone to respond to customer emails, you're instead hiring someone to train the AI on better responses and handle the complex edge cases that require any human judgment. You'll also find that your margins will improve dramatically. So when you eliminate 30 to 40 hours per week of manual work, either your team accomplishes dramatically more work, or your revenue increases without cost increases. Or you can accomplish the same with far fewer resources. And your cost decreases without revenue decreases. So most businesses that do a combination of both and your system, it becomes your moat. Once you have built this infrastructure, a competitor, they can't just copy your website or your pricing. They'd have to rebuild your entire operational backbone, which took you six to eight weeks to build and optimize. And by that time, they, ideally for you, they cannot catch up. And you're another year ahead of them, and this is not hype.
