Claude Projects Run Your Business (Exact Setup)

Author: AI Foundations · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Claude Projects Run Your Business (Exact Setup) · Channel: AI Foundations · 29:55 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Claude

Summary

The video presents a four-step system for using Claude Projects to support business work: choose a project with a clear goal, provide relevant business context, work with Claude on a task until the result is right, and then schedule the workflow as a routine. The creator demonstrates setting up a newsletter project, adding business and writing materials, and asking Claude to read the context and save useful memory rules. He then works with Claude to draft and revise a newsletter, responding to its questions and correcting gaps. Finally, he turns that collaborative process into a weekly scheduled routine, emphasizing that automation should grow as trust and confidence build.

Key points

Give each Claude Project a specific mission or North Star goal rather than asking one project to run the entire business.
Projects can support ongoing work or work with a clear end date.
Add project-specific context, including a business overview, current priorities, brand guidance, writing rules, and examples of past work.
Ask Claude to read key context before starting tasks and use project memory to retain useful instructions across threads.
Let Claude identify missing information and answer its questions before relying on it to complete work.
Work with Claude on a real task and refine the result before scheduling the workflow.
Routines can run on a schedule and guide the user through a process; they can also be tested and edited.
Increase Claude’s access and level of automation gradually as confidence in the workflow grows.

Chapters

Start	Chapter	Summary
00:00	Why Claude Projects can help with business work	The creator explains how a coordinating agent can break a task into jobs and introduces the four-step setup.
01:00	Step 1: Choose a project and set its goal	He distinguishes ongoing projects from projects with an end date, then creates a newsletter project with a clear goal.
03:04	How the project interface and agents work	The creator describes the coordinator, task threads, and the different roles of the coordinator and thread models.
05:06	Step 2: Add business and project context	He uploads a business overview, priorities, brand and writing materials, and past newsletter performance to the project library.
11:15	Step 3: Acclimate Claude and work through a task	The creator asks Claude to review the context, save memory rules, work in parallel, and identify missing files or information.
18:28	Draft and refine a newsletter with Claude	He has Claude draft a newsletter, reviews its first pass, and steers revisions to improve the preview line and content.
23:38	Step 4: Turn the workflow into a routine	After completing the task collaboratively, he asks Claude to create a scheduled routine that guides him through the newsletter process.
27:18	Review routines and scale automation carefully	The creator shows how to inspect, test, and edit a routine, and explains that different tasks call for different levels of human involvement.

Quotes

Projects usually work best when they have a mission behind them or a North Star goal. — 01:00

It doesn’t know who you are, what you sell, who you serve, what your problems are, what tools you use. — 05:06

And step number three is to work with Claude. Don’t automate just quite yet, but work with it. — 11:15

step number four of the system is to put it on a schedule, and to build routines around the way that you work. — 23:38

The more confidence I build in these routines, the more access I can give Claude. — 27:18

Concepts and entities

Claude Projects, Claude, Claude Desktop, coordinator agent, thread agent, memória do projeto, routines, automated C-suite, AI Foundations, Drake

Full transcript (original language)

00:00 Most business owners watching this right now do not have a work ethic problem. As a matter of fact, you are the hardest working person in your company because you are the one doing everything. The content, the research, the inbox, the same tasks every single week, and at the end of the day, your to-do list is still longer than when you started the day. So, working harder is definitely not going to fix that because there’s only so much that one person can do. But, Claude Projects changes that. Now, with most AI tools, you are the manager. AI is doing the work, but you still have to ask for one thing, wait, [music] check

00:30 it, and then ask for the next. But, a Claude Project [music] has a coordinator. You hand it one task, it works out what needs to happen, splits it into jobs, runs them all, and brings you back the finished result only alerting you when [music] it absolutely needs you. You never have to tell it how to run things. Today, I’m going to show you the exact four-step [music] system to set this up so that Claude Projects can begin running your business, which leads me to step number one, which is [music] picking the job you want to create the project around. Starting with

01:00 step number one, which is selecting the job you want the project to handle. Now, the biggest mistake you can make when using Claude Projects is coming into Claude Projects and selecting new and then saying, “My business.” and expecting Claude to run your entire business out of a single project. I do not recommend that because what is the ultimate end goal? Projects usually work best when they have a mission behind them or a North Star goal. And so, you can have two types of projects in Claude. You can have an ongoing project,

01:31 something that never really ends, like my YouTube success project here. I’m always going to be creating YouTube videos, that is ongoing. Or, you can have a project with a clear end date. So, if I was launching a new course, then I could make this a project, and this could be a strategy room where Claude is completing things for this thing that has an end date. Either one works, but first, in order to create a project, make sure that you number one have Cloud Desktop installed because it

02:01 works much better off your desktop. It’s easiest upgrade you can make. Step number two, you want to go to your left-hand sidebar and select projects. If you don’t see it, you can hit this more button and maybe it’ll be hiding right there. But for now, I’m going to select projects. Then I’m going to close my sidebar. In the upper right-hand corner, I’m going to select new project. This is where you’re going to name your project and give it a goal. Don’t worry about contacts right now. There’s a better spot to add that in the future. For now, you just want to think about what is the name of this project going to be and what is the overarching goal

02:33 that you set for this system. So maybe I want to make a project that is ongoing. I could call it something like weekly newsletters. It gives me a little icon here. I can just select that if I want it to be something different. Maybe I’ll just make it this little mail icon and maybe I want this project to be blue or actually maybe I want it to be gray. I can do that. Now, I want to give a goal, a North star for what this project is being made for. So the goal I put here was create newsletters and track performance of past newsletters for AI

03:04 Foundations. And I’ll have a clear goal that Cloud can go, “Okay, I know that I’m going to be creating newsletters and I’m also going to be tracking performance.” So this is the coordinator agent’s main purpose and main job. I’m going to select create project and then we are going to be brought to this screen here. Now that we have our project set up, let me quickly show you around the project interface before getting to step number two. Just so you know your way around and you know how this system is working with this new Cloud Projects feature. So the interface does look a little bit different. If

03:35 this is your first time seeing this project interface, I’m not surprised because it is a new Cloud feature. And also, if it’s not on your account yet and it doesn’t look like this when you create a project, then I would wait a few more weeks and see if it’s rolled out to your account. They’re rolling it out in batches. So your screen might look different if you create a project, and if so, that’s probably why. But, over here on the left, you have your coordination agent. You no longer have to decide how many chat threads you need for specific jobs in your business. You

04:05 tell the coordinator over here what you want to complete, and then it creates chat threads for you over on the right. And threads are individual agents. And so, if I go to my settings, as you can see, we have different models within the same project. We have a coordinator model for managing and creating threads, and then we have a thread model, the one actually completing your tasks. So, you’re going to notice that the effort level is actually higher on the thread model compared to the coordinator effort. Because the coordinator effort,

04:35 all it has to do is like go, “Okay, we should create a thread for this. We should create a thread for this.” Well, the thread model is the actual agent completing the tasks that you set forward. And so, you can change those here if you want to get crazy. I created a full Claude Projects guide. If you’d rather watch that first, I’ll leave that in the upper right-hand corner, somewhere up there. So, make sure you go check that out, and then come back to this video. That video explains all of this stuff in depth and shows you the basics. But, now that you have this set up and you understand how Claude Projects is working, let’s get into step

05:06 number two. Step number two is adding context. Now, I want you to picture this. Picture you hire somebody very intelligent to your company. As a matter of fact, this person that you’ve just hired doesn’t even know the name of the company that they’re working at. Well, that’s exactly what Claude is doing right now if you don’t have any context about your business. It doesn’t know who you are, what you sell, who you serve, what your problems are, what tools you use. It doesn’t know anything about the person behind the keyboard. And so, step number two is just getting your system up-to-date with all of the context it

05:36 needs to perform the job as a human would. So, in order to add context after you’ve created a project, what you need to do is you need to go go this little button right here that says library. If you do not see library, it’ll probably be under the plus button. Maybe it’s just hidden for you right now. But, make sure you click library because this is where all of your files live for the project. You can upload up to 50,000 files in here, and you can see them in this nice list view. Also, anything that Claude creates for you, it’s going to be stored right in this library as well.

06:08 So, I’m just going to go ahead and add my context. Context is going to be entirely project-dependent, obviously. I can’t tell you what the best thing to add for your project will be, but what I can say is there are some generic context files I recommend every business owner, no matter what project you are in, and I’ll show those to you now. I’m going to select add in the upper right-hand corner, upload from device. Then, I’m just going to select files. Now, in my automated C-suite, I have a business context folder. And this context folder holds a lot of things for

06:38 my business. Number one, it holds my entire brand guide. So, if I ever want to do anything with branding, I’ve got it there. But, it also has these two documents right here. My business overview and my priority document. So, I’m going to open both of those before I explain them and upload them to the library. As you can see, two files uploaded, and I’ve only used five of 50,000 files because these are a couple pages long. So, I’m going to open up my business overview and show you what that is real quick so that you can add one, too. Basically, all this document is is

07:09 an overview for my business. And this is what I want Claude to read before it helps me with any task. So, it knows and has the context about who I am, what are my problems, who do we serve, what do we sell, and everything so that the guidance that it’s giving me is actually directed towards my business, and it’s not just using general training data that everyone has access to. And so, it has what do we do, who do we serve, what do we sell. It has my team, how the business runs today. It has all the repeating work that happens in my business in one file. It has my audited

07:40 tasks, the things that I’ve began automating, and how much that’s going to be saving me per year based on my hourly rate. And it also has a few more things in order to keep me organized, like a review interval, a last updated at, a context status, whether this thing is active or not. This is all front matter for the AI agent to read and understand before it actually looks at the business overview. And so that’s file number one. I recommend every business creates their own business overview, and also a priorities document. So I list my 90-day priorities in here, and these are

08:11 updated every 90 days to see if I actually hit them. This file is in place to keep my agent on track, so that it’s not recommending anything outside of my main objectives as a business owner and the current direction of the business, so that even my newsletters are geared towards those goals. So these are the two files I recommend everyone uploads, and I hope that this gave you some ideas about some certain context that you can add to your project. Now, obviously, I want to add some newsletter specific context, so that I’m not just uploading a bunch of documents, a bunch of general

08:42 stuff about my business, but I want this to be a project where I am drafting and tracking past performance of newsletters. So I’m going to add a few more context files here. I’m going to search around my marketing folder, since that is where all of my marketing context lives. I’ll select context, and then I’m just going to start adding some things. So I have my voice markdown file. How do I sound? I want to present myself in the emails? My point of view on life, my life stories, things that have happened to me so that I can interweave that in my marketing, even if

09:13 they are embarrassing. I think that’s great to have a document full of your life stories, especially for things like marketing. So I’ll add in those context files real quick. I’m going to add some more. I’m going to add this copy email skill that I’ve created that teaches Claude how to write emails in my voice, specifically for my audience. This is a little bit different than my voice MD. My voice MD has to deal with tonality. This has to deal with email structure a little bit more. And then I’m going to upload another thing that accompanies that skill. So, I’m going to hit add

09:44 files, go to my references, and going to copy in my email formats. And so now I have added plenty of context for my AI agents to start operating like me and completing work in the way that I would. Since it knows my business now, it’s now acting as an employee because it can read my overview, my formats. I can even map in past emails to the context that have performed well, so it can use examples. I can even map in past examples of emails that have performed

10:14 well on my newsletter, so it sees what success looks like. There are so many things I could do here. Context is mainly where you want to spend most of your time because if you give Claude everything it needs to do the work, it’s really good at doing the work now. I actually thought that was a pretty good idea getting all of my recent emails. So, what I did is I connected my email software to Claude, and I said, “Hey, can you go get all my campaigns and put them in a markdown file for me with all their stats?” And then it did that for me and found everything. The

10:44 deliverability, the summary of all of the most recent campaigns that I’ve ran, and then it put all of my email body copy for each email. And so I’m going to put this as my context as well. Going to head back to my library, going to add that file, going to go to the email folder here, and just throw that in. So, now has my last five emails that I’ve sent off and how they performed. Whenever I give context to Claude, I like to think about it like it is truly a human employee. What are all of the tools and resources they would need in

11:15 order to complete work on my behalf in the way that I know gets results already. That’s kind of my mindset going into giving Claude context. But now that we have that handled, it’s time to move on to step number three. And step number three is to work with Claude. Don’t automate just quite yet, but work with it. It’s very hard to hand Claude a task and tell it to run it without you in the background if you’ve never done the task with Claude to see how it behaves with your context. And if you’ve never seen if it could even get you the results you want, then it’s going to be very hard to

11:45 trust it. So, that’s why step number three is to work with it. And so, you’re going to be handing off a real task to Claude here, but let me show you one thing to do before you start anything. If you want to pull up your threads whenever you’re in a project, you can just hit this little overview button and you’re going to be able to see where Claude’s workspace is over here on the right side of your screen. Now, after you add all of your context, one thing I really like doing is getting Claude acclimated to the context. So, I’m going to send off a prompt. I’m going to dictate, as you can see in the bottom center of my screen, I’ve got a dictation tool where I can speak into my

12:17 microphone and it types for me, which is kind of nice. But, I’m going to do that and I’m just going to tell Claude to get acclimated to all of my context and to always read the business overview before starting tasks. Hey there, Claude, I just uploaded a bunch of context to this folder here. This is going to be a project for my weekly newsletters. I want you to read all of my context, get acclimated to the workspace. Maybe you can read it in parallel and then have some summaries be brought back to you by different agents. And what I want you to do is just always read the business overview, no matter what, before you

12:47 start work. So, make sure you add that to your ongoing memory and add some other memory rules that you think would be important when helping me achieve the overall goal of this project, which is to write newsletters and review past metrics eventually when we get the writing side of things down. So, I just said that huge prompt. What you’re going to notice, uh, some things I want you to understand, as I said, maybe you can read it in parallel and then have some summaries be brought back to you with different agents. I don’t want Claude to go through individually and read the file one at a time, like, “Okay, file

13:18 one, read. File two, read. File three, read.” I just want it to spin up multiple agents, have those agents bring back summaries, and then that’s how it gets its knowledge. And I said, “What I want you to do is always read the business overview.” The unique thing about these Claude projects is that they have an ongoing memory. So, if you tell it to do things, it will remember that across all of your threads, all of your chats. Because if you go to the settings, what you have in the memory tab is an auto memory. So, not only do you have project instructions that you can add where you can tell it how to

13:48 think, but you have memory that gets added in notes that Claude writes itself as it works in this project. And this is just one huge memory file that Claude will read before every new chat. And so, that’s what I’m kind of telling it to do is like, “Hey, get its memory file right.” I’m going to send off this prompt. And what you’re going to see is Claude is going to start working in both panels. Remember, over here on the left-hand side is your coordinator agent, and on the right-hand side is the agent that is actually doing the work, the threaded agents. And so, it has a one task right now, read context and set

14:21 memory. As you can see, there are three things it needs to complete right here. Number one, it needs to read all of the updated context files, synthesize what matters for writing the metrics, and then save the memory rules. What you can do is you can even click in here and see what Claude is doing in action. As you can see, it just put three agents to work in parallel for me. So, if I click in here, I can see what it’s doing. Right now, it’s using sub agents and going through and reading all of my context so it gets done faster. But, this is my thread agent. This is the

14:51 thing doing the work. This is the thing that prompted the agent to do the work. So, it’s a unique little way about how Claude goes about working in this new projects feature. But, when you have multiple tasks that you need to complete, it’s amazing watching it break it down into five different agents or five different threads working all at the same time. I could also type in another task here and get it going. So, I don’t want to do that right now because it’s getting its memory and kind of acclimating itself to the environment. But, if I have multiple things I want to work on, I I send off multiple different prompts and multiple

15:22 different queries until they’re all resolved or waiting on me. As you can see, that one just got resolved. It read the context and set the memory. It said, “Memory rules I saved. Read the business overview before any work.” This is the top rule. “No made-up facts, a map of what each context file holds, and which reference files are missing, and where the metrics live, baseline numbers, and the lessons so far.” And then it said, “Three gaps that you may want to fill.” I uploaded a file that is referencing files that aren’t here, like copyengine invoice.py, newsletter drafter. I’m

15:54 doing all of this stuff that isn’t here, so I need to add that context file. I can get rid of voice.md since it doesn’t matter in here. And then it says I need to answer the question about why the list dropped so much. So I’m going to fill in all three of those gaps, and then I can start handing off my work to Claude. First, I’m going to go in here. I’m going to delete my voice file since he said it’s unnecessary, or she, whatever you want to refer to Claude as. And then I’m going to add the rest of the files that this skill needs. It’s this entire copy engine thing that I’ve built that helps me train AI to write

16:25 like me, which is kind of nice. I’m going to upload that. And then I’m just going to open up the thread where it asked me the question, right here. And then I’ll just say a quick message. “Hey there, Claude. I deleted my voice.md since my voice lives in the newsletter drafter and the copy engine skill, and also I’ve uploaded the copy engine skill, so you have that context now. And one more thing, the reason my list dropped to 45,000 to 9,000 was because I decided that I should test on a smaller list in order to re-engage my audience so that my newsletter software didn’t start flagging me as spam if I was

16:56 getting a lot of hard bounces to start.” Okay, so I’ve answered all those questions. I can send that off right away, and then we’re all good to go. And so while that’s working, I just want to show you how memory is being updated. And so memory, if I come in here, as you can see, it’s adding all these memory files so that anytime I start a thread or a conversation, it’s going to read all of these files. And whenever it has a question for you in projects, it will say waiting on you. So right now, it says blocked. It can’t do anything until I get this one thing right. So I forgot one more thing that I needed for my newsletters, and that was a skill I

17:27 created to draft them. I’m going to upload, and now that should be out of waiting on me and into the working and then into the resolved, and then I can start completing work. I know that was a big preliminary step, but now Claude is acclimated on my environment. There are no open questions that it might have. It has everything it needs in order to write newsletters like me. And don’t be intimidated when you come in here and you see all of this amazing context I’ve got. This has been years in the works of just working with AI and understanding what it needs. If this is something that

17:57 you want to set up for your business like from the ground up and you don’t have any starting point, then I recommend checking out the first link in the description because that’s where I show you how to create your automated C-suite and get all of these set up in this way. The threat is done. I’m going to click. Oh my, it asks for another thing. But I do like this. It asks me questions along the way. I’m just going to answer, and then we should be good to go. So this was really like our first task with Claude, and this is our first time working. Now that task is resolved, that’s great. I can react with a nice

18:28 little emoji. Maybe I’ll send off a little thumbs up. Thanks, Claude, for the help. And now I can get into actually creating a newsletter. And once I get this project to create a newsletter I like, I can say, “How did you do that? I want to schedule that out to be a weekly task that we do or a bi-weekly task do or a daily or an hourly task that we complete.” Because the last step I’m going to show you is setting things up on these things called routines, which make your threads work on a schedule when you’re not even there. And now we can get into the actual step three, which is sending off

19:00 a task. And so I’m going to prompt Claude once more. I’m going to say, “Hey there, Claude. I want you to create a very engaging newsletter based on my other ones that have I’ve really good open rates and success. I want this to be based around Claude skills or Claude projects because I think Claude projects, especially the new Claude projects feature, is really good. And so, everything that we’ve done in this thread today, I want to make a newsletter around. And I’m also filming a YouTube video about this at the same

19:30 time as we speak. And so, create a newsletter around everything that I’ve been doing in this project for my newsletter. All right, so this is very meta. I’m creating a newsletter about my weekly newsletter project and about everything that I’ve shown you today in this project. And that’s what I’m going to create a newsletter around. And so, I’m going to actually add a few more things. I’m going to say, “I want this to look like my other emails, and I want it to stress the importance of having really good context in your project and kind of our uh four-step system that we’ve been doing. Create a project with

20:00 a clear goal. Step number two is adding context to the project. Step number three is completing a task with the project until you get a result you like. And step number four is setting that up on a schedule so it runs without you. Okay, very quick. Now I’m going to send that off. And remember, I’m sending this to the coordinator agent. The coordinator agent over here on the left decides the tasks to create over on the right-hand side. So, right now it’s just doing one task, Claude projects newsletter. But there’s four subtasks within it. So, all it needed was one

20:32 agent, but it has four tasks. I’ll be back when that initial draft is done. All right, it looks like it got done creating the first draft of the newsletter. So, I’m going to select the waiting on me thread over here on the right. And I’m going to read through it. It’s completed all four tasks. Says your draft is ready. Subject: Claude caught a mistake in my own files. My file said 30,000 subscribers. The real number is 58,000. So, literally, the mistake we just covered. That’s the thing it’s using for the email. And so, I can look here at the email body. Today I set up a

21:04 brand new Claude project for one job, writing this newsletter. Before I asked it to write a single word, I gave it my whole business, my business overview, 90-day priorities, opinions, blah blah blah. Then I gave it one rule. So, I basically exactly what I’ve shown you in this video today. That part was cool, but that’s not the part that got me. It came back with questions I didn’t expect. One of my files said my email list was about 30,000 people, and my campaign stats showed I’d sent an email to 45,000 people back in August. So, it asked me, “Which one’s right?” “Neither.” Hmm, that’s actually a juicy email. I kind of like that. And then it’s talking about the system that I’m

21:34 showing you here today. Right now, we’re on step three. I really like this. So, I mean, it did a really good job of reading my voice from my context files, which is why I like throwing in my voice and my past actual written files because it sounds more human. Like, you usually don’t see AI writing lines like this. Read the business overview before you do any with any all caps work every time or things like this dot dot dot. But, with the three use there, it’s very human, very natural, and it feels like it was

22:05 written from an actual human. So, this is pretty good, honestly, for a first pass. I might change the preview a little bit. So, I could say that. I could say, “Let’s not make the preview really number specific before we even share what’s problem was. We can share that problem as we go throughout the thread, but what I want you to do is I want you to come up with a better preview line and then fix your mistakes so this doesn’t happen in the future with like listing these statistics when it doesn’t even matter right now cuz people don’t even know that I’m talking about a newsletter in the preview. I want it to be more juicy to get all business owners to click on the email.”

22:36 There we go. So, just kind of steering the thread a little bit, right? Editing the file that it created. And so, it changed it to it stopped agreeing with me and started catching me. So, it would read back-to-back like Claude caught a mistake in my own files. It stopped agreeing with me and started catching me. Not the best. It’s okay. Here are some other ones it recommended. I actually kind of like the I thought my business was set up right. It wasn’t. So, I’m I’m to say, “Let’s do this one.” And then I’ll send that off. And so, step number three of this system, by the way, if you couldn’t already tell, is

23:07 working with Claude until it completes something in the exact way that you wanted it done, and then you can start creating schedules around that. It looks like there is another task waiting for me. So, if I go back to my threads, as you can see, it’s in the same chat. It says, “The draft now uses this.” The only thing that it still needs is the video link. I’ll say, “Remind me in 48 hours, and I’ll get you the video link, because that’s when the video goes up is Tuesday, October 12th.” No, Tuesday, October 13th is when the video goes up. So, remind me then, and I can send that off. Another cool thing that

23:38 many of you probably didn’t know is that Claude can remind you about things. So, it’s going to set up that reminder 48 hours from now. It schedules that out. Set the reminder, so I gave it a little thumbs up, letting it know that I heard it there. After you go through with Claude, and you complete tasks in this project that you would normally do alone, or that you would normally do AI assisted outside of a project, step number four of the system is to put it on a schedule, and to build routines around the way that you work. So, as you can see, this wasn’t a completely

24:09 automated process where I just set it and forget it, but it was a process where I worked with Claude. And if writing the newsletter usually takes me 60 minutes or 90 minutes, but writing it with Claude takes me 15 minutes, I would consider that a massive, massive win, especially if I’m doing this once or twice per week. And so, what I want to do is set up a scheduled routine that kind of walks me through that exact process that I just went through on screen. That way, it reminds me, like, “Hey Drake, you got to do your newsletter. Let’s get on it.” way that you start automating tasks. You don’t

24:39 just throw it on a schedule and forget about it forever, not at least to begin with. You position yourself to work with AI in the way that you work best, in the way that gets it done most efficiently, so that you can at least do the task consistently, and at least to the point where you’re not forgetting about things, and you have an AI enhanced or assisted workflow. So, let’s talk about routines now. I can just X out of this thread that I was working in. We can go back to the overview. And this is still going to be waiting on me because it kind of has a timer going to where it

25:10 needs to send me a reminder. So, it’s always going to be waiting on me. I believe unless I just say like clear or something like that or if I resolve this task. I can also resolve it in the upper right-hand corner. I’ll resolve it. It should notify me still. And what I’m going to do is I’m going to sit tell Claude over here in the operation chat, the coordinator chat. I’m going to say, “Hey there Claude, so this entire process we just did where I kind of gave you an idea and then you you walked me through it. You used all my context in order to create a newsletter. I want to put that on a scheduled routine so that

25:40 I can be reminded every week, maybe every Saturday to complete that task. Can you set up that reminder for me? And not just the reminder, but the scheduled task that runs that literally walks me through the process.” And so, what I like about what’s happening here is I’m not just out of nowhere telling Claude to complete something for me. What I’ve done is I’ve completed the task first. Claude is already in its working context, got an idea of the thing that I want, and now I’m saying that entire thing that we just did together, put it on a scheduled routine. So, I can send

26:13 that off. I can open up my overview in the upper right-hand corner, which takes me to my threads. And what that’s going to do is create a routine for us in this routines panel. And so, I can go to the working thread, click in there, and watch it work. It’s going to review how today’s newsletter came together so the Saturday routine can walk me through the exact same steps. So, first it’s doing a little bit of research like, “Hey, how was this last thing drafted?” Then it said find a Drake’s time zone, which is pretty important for a scheduled routine. And then it’s going to create the Saturday one for me. It looks like this task just went idle. If I click in

26:45 here, as you can see, it says your newsletter routine is set up. It runs every Saturday at 8:47 a.m. Eastern Time. Kind of an interesting time that it set up there for me. It says, “Each Saturday I’ll message you here. Before that, I’ll read your business overview, your writing rules, and the hook ledger. And I’ll check your channel for any videos you posted that week. Then I’ll ask what the newsletter should be about. You can give me a video or just an idea. From there, it works like today. I pitch three angles, drafted in your voice, and we go back and forth until you like it. I’ll only write drafts, and you’ll still be the one who sends them.” Perfect. And later on, this is what I love about Claude projects, is later on, if I’m

27:18 like, “You know what? I want to take this to the next level. I want this thing to actually access my email marketing account, and I want it to start scheduling these for me.” The more confidence I build in these routines, the more access I can give Claude. At least that’s the way I see it. Now I can go back to this threads panel, and in the upper middle section of my screen, you’re going to see that little clock here. It says routines. You want to select that, because what Claude has just done is created a routine for you and set it to active. And so, what this does is it

27:50 runs on a schedule. So, if I click in here, you can see the entire process and template prompt that Claude is going to send to itself on your behalf. It says triggers. The trigger is just a time-based trigger. Every Saturday at 8:47 a.m., the next run, gives me that, will be October 17th in the morning. And the instructions say, “It’s Saturday. Time to walk Drake through drafting this week’s AI Foundation newsletter in this thread. Treat this as a fresh start each week. Don’t assume last week’s draft is still in progress.” And then, we have

28:22 these entire instructions that it gave itself in order to do it exactly the way it did it with us here today. This is the scheduled routine. We can even run it now if we want in order to test this thing out. We just ran it, but we could run it again to see how it works. We could test it. We could rename it. We could change the trigger to run at a different time on a different day. And these are Claude routines built in. And this is step number four of the system. Eventually, you can have multiple routines all running on a schedule. Some of them completing more work for you

28:54 than others. For example, something that replies to support emails that might come in my support inbox would be a lot more automated. It would draft up the emails. It would have my FAQ. It would have my previous emails I replied to. It take all of that in its context. And then it would pretty much just be I review what has been replied to and I hit send from Claude. That is a lot more hands-off. Something like a newsletter is a little bit more hands-on cuz I want to have control over the process. But every task differs. This is how you set up a successful Claude project and how you begin working

29:25 with it. I hope you enjoyed. If you enjoyed, please drop a like and subscribe. I would highly appreciate it. Make sure you check out the first link in the description because that teaches you how to set up your automated C-suite, which is all of your business context, how to automate best in Claude and how to pretty much uh do everything so you can grow your business. It’s a course and a training that teaches you how to grow your business and automate it using Claude code. And so, highly recommend you check out that first link in the description. But until then, watch this video because I know this video right here is something that you

29:55 are going to enjoy if you watched this one. See you in the next one.