YT YouTube2Text

YouTube2Text → Transcripts → ChatGPT Codex: Automate 90% Of Your Business on AI Employees

ChatGPT Codex: Automate 90% Of Your Business on AI Employees — Transcript

by Rashid · 10,499 words · 1,402 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Everyone is talking about cloud code and
0:01how good it is with getting work done
0:03inside of your business, but very few
0:05people are actually talking about
0:06codeex. Codeex is chat's version of
0:09cloud code and it's just as powerful as
0:11cloud code if not more powerful. I've
0:14been using it for the past 3 months and
0:16I've been giving it all types of work
0:17inside of my business from writing
0:19emails, helping me create YouTube
0:20thumbnails, write up some automations,
0:22and help me create plugins and custom AI
0:25employees that can deliver to my
0:26clients. And no, I'm not talking about
0:28just Chat GPT. I'm talking about Codeex,
0:30which is a completely different app that
0:32is built on chat GPT. Now, you might be
0:34asking, what's the difference between me
0:36using the normal Chat GPT and using
0:38Codeex? The normal chat GPT is great for
0:40getting one-off tasks done inside of
0:42your business. You want to ask to do
0:44research, you wanted to explain a topic,
0:46you want to learn something, or you want
0:47to create a one-off deliverable like a
0:49PDF report or an email, you can use Chat
0:53GPT. But if you care more about scaling
0:55your work and building and deploying and
0:58shipping digital assets and systems that
1:01do the work for you, then Codeex is a
1:03better choice. That's because Codeex has
1:05more capabilities that allows it to
1:07almost act like it's an employee inside
1:09of your business. So, in this video, I'm
1:11going to show you how to set up Codex
1:12from scratch. First, I'll show you how
1:14to set up your project so it knows about
1:15your business. Second, I'll show you how
1:17to teach it your own processes and your
1:19workflows inside of your business so it
1:21doesn't guess. And finally, I'll show
1:22you how you can give it access to your
1:24own real tools so it can run processes
1:26inside of your business. Then I'll go
1:28over a real demonstration on how I hand
1:30codeex a real job inside of my business
1:32so you can see an example of how it
1:34works from start to finish. And
1:35throughout this video, I'll break down
1:36the fundamentals on how codeex and AI
1:39agent actually works and how it
1:41interacts with your files, your folders,
1:43and the things that you can customize
1:44with it so it works best according to
1:46your business. Now, by the way, this
1:48video is a long one. So if you want to
1:49learn the fundamentals of how to get
1:51codeex to run business processes inside
1:53of your business just like an AI
1:55employee then I recommend you to save
1:57this video if you don't have the time to
1:58watch it to watch it later. So the first
2:00step is we want to download the chatpt
2:03for desktop which will download an
2:04application. So you can click download
2:06for Windows or if you use Mac you can
2:07download it as well and then you can
2:09download it and then install it. So when
2:11you install chatpt this will be what you
2:13come across with. You'll see chat and
2:16work. So what we're going to do is we
2:18want to use chatpt codec. So on the top
2:21left where where we see chatpt over here
2:25we're going to click on the drop-own
2:26menu and then we're going to select
2:27codeex instead of chatpt. So if you read
2:30the description over here it says create
2:32learn and explore and codec says build
2:35debug and ship. So it's called codeex
2:38because it was initially created for
2:40developers who code. But as a business
2:42owner having a let's say per se
2:45developer or AI developer codeex in your
2:47business is very high leverage because
2:49now it can build tools for you it can
2:52build systems it can debug tools debug
2:55apps debug systems and ship things
2:57inside of your business. So instead of
2:59you creating and doing things onetoone
3:01Codex which is an AI agent or can act
3:03like an AI employee that builds stuff
3:05inside of your business and operates
3:06your business. So, we're going to click
3:08here, codeex, and instead it's going to
3:10open up this page um right over here.
3:13Now, quick thing. If you're watching
3:14this video and you're wondering and
3:15you're lost on where can you implement
3:17AI inside of your business, I created a
3:19free skill that'll show you exactly how
3:20to do that. It's called the AI employee
3:22map. All you have to do is ask your AI
3:24agent of choice to map out your
3:25business. It will interview you. It will
3:26ask you a question about your business
3:27and then it will nicely map out your
3:29business in swim lanes. And on the swim
3:31lanes, it will show you where the
3:32bottleneck is. It will show you where
3:33you can implement AI and whether you
3:34need to implement the digital asset as
3:35well. You can get it for free by joining
3:37my email list by clicking the link in
3:38description. It's completely free and
3:40you'll also get access to other bonuses
3:41as well. Links in the description. All
3:42right, so before I dive into using CEX,
3:45I want to explain the fundamentals on
3:47how an AI agent works and how Codeex
3:50works behind the scenes. Cuz if we
3:52understand how Codeex works behind the
3:54scenes, then we can understand how it
3:56can act like an employee inside of your
3:58business. And the bonus thing is if you
4:00understand how CEX works, you'll also
4:02understand how cloud code works by
4:04extension cloud co-workbot and all of
4:07these other AI agents and how they work
4:09on the back end. So if you've been using
4:11AI, you've probably been using it this
4:13way. You ask chatbt or whatever the AI
4:15model is of your choice. This could be
4:17chatbt claude. You give it a prompt and
4:20then um that model gives you an answer
4:23and then from that answer you can use
4:24that answer to help you get work done.
4:26So this is why prompt engineering was
4:28all the craze because everyone was
4:30looking at okay how can I get the best
4:32prompt to get the best output out of the
4:34model. Now things has improved. Models
4:36have improved way better. You don't need
4:38to prompt as well. You can just speak
4:39clearly what you want and a model will
4:41probably get give you a good answer that
4:43you can ship out. Another thing that has
4:45changed is a lot of these models have
4:47included other uh tools and
4:49capabilities. Let's say there's image
4:51generation which you can ask. There's
4:54artifacts which can create PDF uh tools
4:57etc. and also capabilities like being
5:00able to control your browser. So this is
5:02all has improved how AI uh performed and
5:05a lot of other stuff like connectors or
5:08what I like to call plugins that allows
5:10you to connect your model of choice to
5:13your tools that you use inside of your
5:15business whether that's notion etc and
5:17all that stuff. Now, this is great, but
5:19the problem with this is that we don't
5:21have a system and we're using the AI
5:23model as a as a calculator. Okay, when
5:27we use AI this way, yes, AI can give us
5:29answers, but we're still working from a
5:32onetoone perspective. So, what's one:1
5:34and why is it a big problem as a
5:37business owner if you're trying to use
5:38AI? So, one one means that I put in one
5:40prompt in or one input in and then I get
5:43one output out. Now the problem with
5:45this is that if you use AI this way,
5:47anytime you want an output, you need to
5:49constantly prompt it. And that means
5:51it's not scalable. The only thing
5:53scalable here is you. That means you
5:55need to prompt more to get more output.
5:57And that's not what we want as a
5:59business owner. Because if you do it
6:00this way, well, then you can't scale
6:02yourself. You can't prompt more because
6:03that means you need to spend more hours
6:05of your time. And that means your
6:06business is still tied to your hours
6:08that you work. Instead, what we want is
6:10we want to be able to do all the work
6:13for us without needing us to constantly
6:15prompt it. And we only prompt it when we
6:18want the AI to build a system to do
6:20that. So AI needs to sort of be embedded
6:24inside of our business and not us just
6:27talking to AI like it's a separate
6:29consultant or separate tool or
6:31calculator that helps us get more work
6:33done. AI should be incorporated inside
6:35of our business. And this is what codeex
6:38allows us to do and what other agents
6:40allow us to do. So what we want instead
6:42of one to one, we want what I call one
6:45to n one to n which is one to many. That
6:48means I build one thing and then AI or a
6:52system can do this task for me over
6:54time. That way I don't need to have it
6:56tied to the hours that I work. Now the
6:58way Codex works is that Codeex can build
7:01digital assets inside of your business.
7:03So we can build, deploy, and debug stuff
7:06in your business. What does that mean?
7:08That means you can have systems that run
7:10without you being there. So let's say we
7:13go over this example over here. Let's
7:14say you want to turn every YouTube video
7:16into a weekly newsletter. You could ask
7:19that to Codeex. And what Codeex should
7:21do, ideally, it should build a
7:23newsletter system that lives inside of
7:25your business or your work that every
7:27time you post a video, it will transform
7:29it into a newsletter. And this time you
7:31won't need to constantly prompt it. So
7:33this system will continuously work for
7:35you over the long term n amount of times
7:38or many times. I'm building one thing
7:40once and it's running for me over time.
7:43So this is a concept to what I like to
7:45call one to many. And this is what I
7:47like to coin the term digital asset.
7:49Okay? It's an asset because it works for
7:51you over time and allows you to scale
7:53your business. Now, I will be posting
7:56more content on my YouTube channel
7:58explaining how digital assets allow you
8:00to scale your business. So, if you want
8:01to learn more about that, feel free to
8:03subscribe and check the link in
8:04description where I'll have some other
8:06resources breaking it down. If your goal
8:08is to use AI to build a leveraged
8:09business, a business that provides you
8:11wealth and freedom and tons of profit.
8:13Well, then I have a newsletter where I
8:14talk exactly about that. Every single
8:16day, I'll share with you an insight on
8:17how I'm using AI inside of my business
8:19and how my other clients and other
8:21business owners are using AI inside of
8:22theirs. specifically how AI does 80% of
8:25the work for us inside of our business
8:26and how we use AI to regain back more
8:28time and serve more clients without
8:29working more hours. So, if you want
8:31that, you can click the link in
8:32description, put in your name and email,
8:33and you'll automatically get signed up
8:35to get a daily email. You'll also get
8:36access to AI employee builder free
8:38portal where you can read all of my
8:40previous emails and insights on how I
8:41use AI as well. Not only that, but
8:43inside you'll also get access to a free
8:45plugin that you can install. It'll give
8:47you some free skills that'll help you
8:48map out your business, find out where
8:50your first bottleneck is, and show you
8:51what you should build out first in terms
8:52of AI employees. You can also get access
8:55to free replays of exclusive workshops
8:57and master classes I've hosted before
8:58and other resources on my help you
9:00including the digital assets playbook,
9:01which shows you 50 different digital
9:03assets you can set up inside of your
9:05business to have it run more leveraged.
9:06And in there, you'll also find exclusive
9:08offers I'm running where you can get
9:09more support to build AI and employees
9:11in your business. Now, joining is
9:12completely free. Again, if you want to
9:14join, click link in description and you
9:16can put in your name and email can join
9:17for free. But basically, what this is
9:19going to do is that additional asset
9:21creates outcomes for us without us
9:23needing to operate that specific part.
9:25So this might create us different
9:28newsletters each week. So week 1, week
9:292, week three, week four, it is creating
9:31deliverables for us without us needing
9:33to do it. This is also what turns AI
9:36into an AI employee. So in this case,
9:38Codeex is acting like our AI employee
9:40over here or more specifically our
9:42newsletter. AI employee. So all the
9:45outputs that Codeex creates, now we can
9:47review it and then approve to send it
9:49and then could ship it out to the real
9:51world for us. So instead of you running
9:53your business, you're just reviewing the
9:55output that Codeex does. Okay. So before
9:57I dive into building anything on Codeex,
9:59I want to go over the basic fundamentals
10:01on how Codeex works and how coding
10:04agents like Codex or Cloud Code actually
10:06work together to be an employee inside
10:08of your business. So, the first thing I
10:10want you to think about is I want you to
10:12think about if you were going to hire a
10:14person inside of your business, what
10:16would be the things required for that
10:18person to pretty much act like an
10:20autonomous employee to get work done
10:23predictably and correctly every single
10:25time? Well, then you it probably need to
10:28have multiple things. It needs to
10:29understand about your business context.
10:31It would need to understand from its
10:32previous mistakes so it can improve.
10:34It'll probably need to have access to
10:36training or standard operating
10:37procedures. It probably needs some sort
10:39of quality assurance so it understands
10:41if it's doing the work correctly and
10:43probably needs access to tools like your
10:45computer softwares and so on. So all of
10:48those things Codeex has already built
10:50inside of it as an agent and is highly
10:52customizable. So you do not need to
10:55think about building an agent. All we
10:57need to do is learn how Codex works and
11:00then we can build custom employees
11:01inside of our business. So if we were to
11:04go under the hood to see how Codeex
11:05works, it would work like so. Now, Codex
11:08is just a coding agent similar to cloud
11:10code. You're given access to a code
11:13space or in better terms or easier for
11:15business owners is a project or a
11:17specific workspace with your files and
11:19folders. Now, when you talk with codeex,
11:21it's able to interact with these files
11:24and your folders and get work done just
11:26like an employee does that. Now,
11:28remember all the elements I shared
11:29before? These are the elements that
11:31Codeex works with to act like an
11:32employee. So, let's dive in and break
11:34down each one step by step. So let's say
11:37this is you and you give a prompt to the
11:40model or the chat which is the agent.
11:42Okay, codeex is going to work in the
11:44background. First thing it's going to do
11:46is going to read this document called
11:47agents.mme. Now what agents.mme it
11:50pretty much is a highle document that
11:52every single codeex chat or agent that
11:55you open reads through that file so it
11:58understands how to do things inside of
12:00your business. It teaches every single
12:01agent how your business runs and reads
12:03it every single time. So this gives AI
12:05the valuable context required to do
12:07work. The next thing is that it can read
12:10and write to your files which includes
12:12your skills and tools and anything else
12:14that you have stored. So this could be
12:16digital assets, this could be landing
12:18pages, anything else that you have
12:19stored in your work. AI can then adjust
12:21it and work on it. Another example could
12:24be like Excel sheets, PDF, any type of
12:27media or code, image, video asset. AI
12:31can then adjust it and interact with it
12:33and build upon it. Okay. Now, if we want
12:36AI to be careful with how we edit things
12:39to make sure that it reviews its output,
12:40we can also have hooks which kind of act
12:43like guard rails. Now, as you work with
12:45AI, AI is going to make a lot of
12:46mistakes. So here's where it has access
12:49to memory where it gets notes from chats
12:51and it can read through the memory to
12:53understand how it can do better work
12:54more tailored uh to how you do things.
12:57So if you think about it, these are all
12:58of the elements required for an employee
13:00to work. It needs to understand the
13:02context of your business. It needs to
13:04have access to your files. It needs to
13:06know how to not make mistakes or how to
13:08review its work with hooks. And then it
13:10needs to learn over time with memory.
13:12And then at the other side is that we
13:14also need to be able to ship work. So
13:16here's where we can have uh codeex
13:19connect to the external world or which
13:21is other apps that you use. So let's say
13:23you use notion, you use slack, you use
13:25GitHub or any CRM or email, Google
13:28drive, we can connect that to codeex or
13:30your coding agent of choice which is
13:31cloud code. And then if you have any
13:33custom plugins, you can also install
13:35that as well like skills and connections
13:37can be installed to further customize
13:39how your employee works. So this in turn
13:42adds full customization on how you can
13:44build an AI employee and Codex provides
13:46that and these other coding agents let's
13:47say cloud code as well. So in a nutshell
13:50this is how it works right over here and
13:53what this allows you to do is you can
13:55build and run systems. You can have
13:56codec build code write and run code. So
13:59this could be like automation. So you
14:00could build tools you'd otherwise pay a
14:02developer for. You can also create a
14:04schedule trigger or create triggers to
14:06have an agent run at specific times. For
14:08example, instead of you needing to
14:11prompt all the time and and the agent
14:13relying on what you say, you can have it
14:15interact with triggers that goes inside
14:17of your business. So, let's say um every
14:20day at 7:00 a.m. you want uh Codeex to
14:23review your YouTube videos and then
14:25create a newsletter. It could do that
14:26for you. And then because you have
14:28access to skills over here, it can run a
14:30specific skill to do to do that specific
14:32work. And then it can go ahead and
14:33upload it for you if you have it access
14:35to your social media as well. So that's
14:37an idea. It can also spawn sub agents.
14:40So it can split a big job across
14:42helpers. That means you can get more
14:44work done in less time. And also
14:46plugins, I mean skills and connections
14:47can be installed very quickly. So it's
14:49very easily, very flexible. Um, this is
14:51why I recommend that. So here's how it
14:53look like to spawn sub agents. The exact
14:56agent would then be exact copy of this,
14:59but then it would just spawn another
15:01agent and you can spawn as many agents
15:02as you want in parallel as well. Okay.
15:04So let's get started by building
15:06something on codec. So the first thing
15:07we need to do is we need to create a
15:10project that allows us to store all of
15:13our systems, all of our files, folders
15:15and context so that codeex can work
15:17inside of our business. So I already
15:20have a project over here. I called it
15:22rashet business. So if this is your
15:24first time opening up uh codeex, you
15:27need to create a project. So the way you
15:29can do this is you can click on the plus
15:31button over here, right over here. And
15:34what this is going to do is going to
15:35open up a new project. So ideally I want
15:38to put a project name. I'm going to call
15:40this uh codeex demonstration. And then
15:44we want to create a folder on our
15:46computer. So you can click on add. And
15:49you want to find a place on your
15:50computer where it's convenient for you.
15:52I can select desktop over here. I can
15:54create a new folder over here. New
15:58folder. And then what I'm going to do is
16:00I'm going to name this folder the exact
16:02same thing. I'm going to rename it
16:04codeex demonstration and I'm going to
16:06click on select folder and great. So now
16:08I have a folder. So if you have multiple
16:10folders you can add them over here as
16:12well. So I'm going to create project and
16:14I'm going select trust folder. Now on
16:16the left side I'll be able to see this
16:18project over here. So the nice thing
16:19about uh codecs is you can have multiple
16:21projects. So if you work on different
16:23businesses or you don't want to separate
16:25things you can separate them in terms of
16:27projects. So over here I can then open
16:30here open up a new chat like so. And now
16:33I'm just started with codeex
16:34demonstration. So the first thing I want
16:36to go over is what is actually in a
16:38project. So we just created a folder,
16:40right? And within this folder, it's
16:42completely empty. So AI is not going to
16:44be really helpful here because we didn't
16:46give it access to files or folders.
16:48There's no skills. There's nothing. So
16:50the first thing we need to do is we need
16:51to run this command called init. So what
16:54this will do, this will create an
16:55agents.md file with instruction for
16:57codeex on how we do our specific
16:59project. So, we're going to write
17:01slashinit and then we're going to pretty
17:03much tell uh codeex what we're doing
17:05inside of our project. Hey Codex, this
17:08project is for me to demonstrate exactly
17:10the features on how codeex works. We're
17:12going to be building out some uh example
17:14digital assets, some skills and so on.
17:16So, viewers on my YouTube channel can be
17:18familiar with how codecs work. So, I'm
17:20going to put that and I'm going click on
17:22enter. And what this going to do, we can
17:24then see what codeex is going to do in
17:26terms of building out our uh folder
17:29structure over here. Now, this is what I
17:31would do if this is the first time
17:33you're creating a project, you haven't
17:35used my agent before. If you have used
17:37cloud code, then what you can do in the
17:39same time is you can click on the bottom
17:42left on your user and can click on
17:44settings and then you can click on
17:46import. And then if you are using cloud
17:49code you can then go ahead and over here
17:52and import uh what you've used before.
17:55So as you can see over here I can import
17:57from cloud code cloud co-worker cursor
17:59and it's pretty straightforward. So if
18:01you want to port to uh uh codeex it will
18:04pretty much port everything that you've
18:06used there. Okay. So now we can see that
18:08codeex created two files. It created a
18:11/getit with hooks info objects reference
18:16and so on. And then it also creates an
18:18agents.mmd folder. And the cool thing
18:20about agents.md is that codecs always
18:22read this before it does work. So if you
18:25want a standard way of work done or you
18:28want to hold um information about your
18:31workspace, codeex would use the
18:33agents.md folder over here. All right.
18:35So we don't stop here. What we need to
18:37do is we need to get Codex to build out
18:39the infrastructure for us so we can get
18:41work done. So I'm going to ask Codex to
18:42do that. Hey Codex, can you create the
18:45infrastructure required for our uh
18:47workspace to work including the skills
18:49folder, agents folder and everything?
18:51You can look online at the open AAI uh
18:53documentation on how to create this
18:55folder structure for us. Thank you. All
18:56right. So now Codex created a folder
18:59over here called agents and within
19:01agents it has all the skills. So here's
19:03where skills will be located and skills
19:06are basically standard operating
19:07procedures that allow Codex to do
19:09predictable work inside of your
19:11business. Then there's the code which
19:13has all agents and so on over here. Now
19:16it also created an assets folder and an
19:18examples folder. So this is how we can
19:20start using codeex in a basic level. All
19:22right. So now we have the basic setup on
19:24codeex. The next thing is we want to
19:26connect it to your tools. So on the left
19:29side if you click on plugins well
19:31plugins will allow you to look at all
19:32the different applications or software
19:35that you use. So, if you are currently
19:37using a specific software and you want
19:39AI to have access to it to do work for
19:41you, here's where you'd connect it. You
19:43can choose Gmail, Google Drive, GitHub,
19:46uh, Notion, Google Calendar, etc. So, I
19:49have a few connections I've implemented,
19:51which are over here. All of my installed
19:53connections are here. You can always go
19:56here and manage your plugins, and you
19:58can also add custom plugins as well. So
20:00if you want to create a plugin, you want
20:02to add a marketplace, so there's a
20:03plug-in online or you want to add an MCP
20:05server, you can do that as well over
20:06here. All right. So next, let's go into
20:08the next feature, which is getting uh
20:11CEX to do work for us, which is creating
20:13a skill. So a skill is really helpful
20:16because a skill is what teaches codeex
20:17to do something for us. All right, so
20:19going back to the visual, the most
20:20important thing that allows AI to get
20:22work done correctly is one word,
20:24context. So there are two places you can
20:26have context stored. You can can have
20:28context stored in your files and skills
20:30or you can have context stored in your
20:32agents.mmd. Let me explain to you the
20:34difference. So the difference to when to
20:36update your agents.mmd and when to
20:38create a skill depends on how you want
20:40to use it. If you want every single
20:42agent or every time you work with codeex
20:44to remember a specific rule or a
20:46specific approach to how you get work
20:48done, then you'd put it into agents. MD.
20:51But if you want it to be specific for
20:53how you do a specific task, then you'd
20:56store it into a skill which is loaded on
20:58demand. So agents.m MD, you want to
21:00think it's always on and then skills can
21:02be loaded on demand. Now the reason why
21:05you'd want to use skills and you
21:06wouldn't put everything into agents.md
21:09is because of context bloat. So in a
21:11nutshell, what skills allow us to do is
21:14to pretty much package the specific
21:16context at the right time. So the agent
21:18can be very uh efficient with how it
21:21does its work because if you spend way
21:23too much tokens, it's just going to get
21:25confused. It's going to compact and the
21:26quality degrades over time. So the way a
21:29skill works is that it's just an SOP but
21:32for an AI model. So let's say you want
21:34to ask guy, hey, write the April
21:36newsletter and you have a specific way
21:37of writing your newsletter. You don't
21:39want AI to do random work. You give that
21:42to Codeex. What Codeex is going to do is
21:44going to look at your skill folders. Do
21:46you have a skill on this? and then it
21:47will find the exact folder required to
21:49do that skills. Then it will run through
21:51the skill and then it will follow that
21:53process and then it'll do the work just
21:55like how you wanted to get it done. Now
21:57in order to create the skill we want to
21:59use this command called skill
22:01slash-creator.
22:03We can use this and invoke it at any
22:05time to create the skill. Uh if we want
22:07codeex create a skill then when we
22:09create a skill we can also invoke that
22:12skill directly by using this. So let me
22:14show you how that works. So, let's go
22:16over an example. I'm going to ask Codeex
22:18to write an email uh just a basic email
22:21uh without skill. I'm going to give it
22:23some sample work that I like and then
22:25ask it to create a skill from it so that
22:27further emails I create can follow that.
22:30I'd like you to write an email, a quick
22:32short email on the difference or the the
22:34power of using codeex compared to chat
22:36GPT. And the power is that Codeex can
22:39create digital assets for you. It can
22:41build, debug, and ship stuff for you,
22:44which means you can build systems that
22:45run run without you. If you have codec
22:47running your business, well, then you
22:50can pretty much break the tie from your
22:52income to your hours and have an AI
22:54employee that does the work for you. So,
22:56I'm going to share that with uh Codeex
22:58and let's see what it does and then we
23:00can compare it later on. All right, so
23:02this is what it wrote and this isn't an
23:06email that I would write. This is very
23:08short. It's not an email that would send
23:09to my email list. And I double checked
23:12to make sure that it wasn't reading my
23:14skills. So, it didn't read my skills.
23:16So, what I'm going to do is I'm going to
23:18open up a new chat over here. So, way I
23:20can add a new chat is like so. And
23:22that'll open up a new chat. So, what I'm
23:24going to do is I'm going to find some
23:26emails that I like and I'm going to
23:29share that with uh codeex to create a
23:32skill so that emails sound more like me.
23:34So if I go to my notion database here, I
23:37have a lot of emails I've sent
23:39previously. Let's say this is one. I'm
23:41going to copy this email. I'm going to
23:43copy paste that over here. And because I
23:45have notion connected to my codeex, it
23:48should be able to pull these up. All
23:49right. So I got three example emails.
23:51And I'm going to run this command called
23:54um you're going to use the dollar sign.
23:56And then you're going to type skill
23:58creator. I'm going to click on tab. And
23:59that will open up skill creator. So now
24:01I'm going to ask here I'd like us to
24:03create a skill um so that we can write
24:06emails uh in the future. So I gave that
24:08context and if I open up the right side
24:11again here and open up the files what is
24:13going to happen is um codeex is going to
24:17create a skill inside of our skills
24:19folder and because it's stored in my
24:21skills folder this can be a digital
24:23asset or a system that codeex can run in
24:26the future. So, anytime I ask AI to
24:29write an email, it'll run it the
24:30specific way. Now, as you notice here
24:32how I'm opening up the chats, I'm
24:34opening up one chat at a time and each
24:37chat pretty much is its own separate
24:39agent. Just keep in mind because I'm
24:41going to break down how you can have
24:42multiple chats communicate with each
24:44other and how you can get more work done
24:46in less time. So, uh, Codex has read the
24:49emails and it says the pages show a
24:51useful split. The skill should hold a
24:53repeatable writing process while voice
24:55examples and changing offer facts stain
24:57references. Build a small email workflow
24:59around that distinction. Use the three
25:01emails as examples without copying their
25:02dated offers with standing rules. Great.
25:05So let's give Codex a little bit of time
25:08and then we could see what it comes out
25:09with. All right. So CEX created it and
25:12it created a project skill rashed emails
25:15and a voice profile drawn from your
25:17three notion emails. It guides future
25:19drafts from a concrete business orange
25:21situation through one clear ID and
25:22relevant next step. The dated offer CTA
25:25stay in example. So new emails need
25:26current campaign facts. The skilled past
25:28validation to use it. Use rashed emails
25:30to draft an email. Okay, great. So if I
25:33go to skills here, I can now see Rashid
25:36emails. I can see the references and the
25:38voice profile and skill.md. So if I open
25:40up the skill.mmd, this pretty much
25:42explains to the agent how to use this
25:45skill. And the best thing about skills
25:47is in again um AI or codeex is only
25:51going to read the name and description
25:52and it will only call the skills when
25:54it's relevant to the specific task. So
25:56if I ask AI to write an email, it will
25:58know to run this specific skill because
26:00it can read through the name and the
26:02description. So Rashid emails create an
26:04email draft blah blah blah read voice
26:06profile for the voice standard and
26:07source examples inputs writing process
26:10exactly what to do and then you example
26:12use right over here. And then if I open
26:14up the voice profile, it pretty much
26:17tells the agent on what to do to how to
26:19write it. Then it goes through the
26:21source examples over here. Actually,
26:23it's smart enough to copy paste the
26:25links to my notion. So, it just tells it
26:27to open up the notion in the future.
26:29Now, ideally, if you don't use notion,
26:31you'd probably want to have all these uh
26:33emails located here as references, which
26:35could say like sample emails. So, I'm
26:37going to ask Codex to do that. Can you
26:39uh remove these source example links to
26:41notion and store them into a um example
26:45folder? I think that's better because it
26:47will save time and tokens so future
26:49agents don't need to always open up
26:51notion. So all right so it updated the
26:53skill inside of Rashid emails under
26:55examples. It has all of the emails copy
26:59pasted over here. So now we have a skill
27:01and we have a functional digital asset
27:04or system which is an email writing
27:06system. So, what happens if I want to
27:09ask AI to write a an email for me? What
27:13I could do is I can open up a new chat
27:15and then there's two ways I could pretty
27:18much trigger this skill. I could force
27:20trigger it by click by putting a dollar
27:23sign in the chat over here and then I
27:25put Rashid emails tab. Now I can force
27:30use that skill or I can just ask codeex,
27:33hey, I'd like us to use this skill to do
27:35this work to get this done for us. So
27:38I'm going to force use it over here. Um,
27:41but what I'm going to do is I'm going to
27:42be smart here. Instead of asking this
27:44one chat to do it, I'm going to ask it
27:46to write three different emails for me.
27:49And I'm going to have this chat act like
27:51my orchestrator chat. So I'm going to
27:52tell Codex just that. This is a good way
27:54to work if you want to get multiple
27:56things done at the same time. Okay, so
27:59this is what we're going to do. Hey, I'd
28:00like us to write three different emails
28:03that I'm going to be sending for a
28:05campaign that I'm going to run. It's
28:06going to talk about how learning codecs
28:08and cloud code is one of the highest
28:10leverage things you could do because if
28:12you use it as an operating system, you
28:14can have it build AI employees, build
28:16digital assets, and those AI employees
28:18and digital assets can then work
28:20together to then scale your business
28:22over time without you needing to work
28:23more hours. So, I want us to come up
28:25with three email ideas. And then once we
28:28come up with these three email ideas, I
28:29want you to create three different chats
28:32locally over here that you then assign
28:35them to write these emails and then
28:36you're going to review the output with
28:38me together in this one chat. So what I
28:41did was I'm going to have this chat act
28:44like the orchestrator chat. So what how
28:46I can do that to improve my productivity
28:48on codeex instead of needing to open up
28:50multiple chats over time, I can have an
28:52orchestrator chat that works for me. I'm
28:54going to call this orchestrator. So now
28:56we can see how the agent is working.
28:59It's reading through the skill. It read
29:01through my email skill. It read through
29:03the voice profile. It read through all
29:05the sample emails. The three angles can
29:08form a sequence while I learn codeex or
29:10cloud matters. How to turn one recurring
29:12job into an employee and how shared this
29:13asset let that work compound. I'll give
29:14each email its own concrete business
29:16example and reply base next step. Okay.
29:18So it's using my uh sample emails to
29:21then create outputs for me. Okay. So
29:23Codex created the three chats. So if I
29:25go on the left side, I can also reorder
29:27my chats as well. So if I want to do
29:29this and this and this is a technique
29:32that I like to use with writing uh with
29:34having more visibility to what other uh
29:37chat or agents are doing. So these are
29:40the outputs that I created. I can read
29:42email one, two, and three. And if I want
29:45every single agent to get good amount of
29:47context to have good quality output and
29:50not to mix in what other agents are
29:52doing then I could do it this way. Now
29:54this is version one which is how I give
29:56it visibility. There's another way where
29:59I can ask claude to spawn sub agents in
30:01the background and can do work for us
30:03and we won't be able to see what it it
30:04does. But because we have all these
30:06chats over here, I can then see and I
30:09could then directly prompt each specific
30:12agent myself as well. And these chats or
30:16agents can then talk to each other to
30:17then coordinate to get work done. So
30:19what I can do here is I can read through
30:21the email. I can give some some review.
30:23I can adjust it. Then I could tell uh
30:26Claude what to change and then whatever
30:28I change I can ask it to then update the
30:30skill to improve it. So over time the
30:33more you use AI the more skills you use
30:36you can then update and upgrade your
30:38skills over time to learn more about how
30:40you write things. Now personally would I
30:42send these emails? Probably not. Um
30:45because I have a way more developed
30:47skill that I use to write emails. I'll
30:49probably record another video breaking
30:50that down. uh but this is a good way to
30:53start to show you the capabilities of
30:55how things work. So let me show you
30:56visually how this looks like. Okay. So
30:58working with codeex you have three ways
31:00to get work done. The first way is you
31:02can open up a chat and you could just
31:04tell the chat to do the work itself. The
31:07next way is if you want to get more work
31:09done in less time, then you can have an
31:11orchestrator chat which is the main chat
31:13you talk with and that chat shares with
31:15you the results and then you can have it
31:17spawn other chats that you can then open
31:19those chats so you can see their thought
31:22process and how they work. So this is
31:23good if you want full transparency on
31:25how all of your agents are working. The
31:28third way is if you don't want to know
31:30what the sub aents are working, you can
31:32ask a chat to act like an orchestrator
31:34and you can have it do work by spawning
31:36sub agents for you on the background and
31:38then you just get the answer. You don't
31:40care about what the sub aents do. You do
31:41not care about the process. So those are
31:43the three ways. First way, you're
31:45probably going to do this if you want to
31:48get something done quickly. Second way,
31:49if you're building something more
31:51detailed, let's say you're running a
31:53campaign, you wanted to put things
31:54together. You want to get more details.
31:56Maybe you're putting together a YouTube
31:58video. Maybe use this. If you're doing
32:00research and you're doing some uh
32:02background analysis, probably don't need
32:04to know what what the sub agents does.
32:06You just need the numbers. You just need
32:08the date details. You can get Codex to
32:09spawn sub agents for you. Now, there's
32:11this one problem you should be aware of.
32:13I think a lot of people come across with
32:15is you do not want to just open chats
32:18all the time because your attention as a
32:21business owner and using the operating
32:22system is very important. If you are
32:25working with one chat, then you can put
32:27100% attention onto it and you
32:29understand all the context into it. If
32:31you're trying to manage three chats,
32:33then you're getting each chat is getting
32:3433% of you. And then the quality of the
32:37attention and the context you give to
32:39the agent to get work done in your
32:41business degrades. And then if you try
32:43to have 10 different chats or 10
32:44different agents, well then you're just
32:46going to be jumping from chat to chat
32:48and you're not going to be shipping
32:49anything. So you're just going to be uh
32:51talking all the time and you're not
32:52going to be shipping everything. So, in
32:54order to get work done, I recommend you
32:56to to try to stick to as little chats as
32:58possible. One, two, three, maybe five at
33:01max. And you definitely want to be using
33:05sub agents or order agents to help you
33:07manage other chats for you. So, you
33:09don't need to constantly open chats. And
33:11you can pretty much talk to an
33:13orchestrator or manager of specific
33:15chats, review the output, give it
33:16feedback, and then that agent or your
33:18main orchestrated agent can then talk to
33:20other agents for you. Okay. Okay, so now
33:22you understand the fundamentals and how
33:23you can get uh codec set up. I want to
33:26go over a demonstration of how I use
33:28codecs. Uh just so you guys can follow
33:30along to see how this would look like um
33:32using it on a real business like mine.
33:35So right now over here I open up my
33:37files on the right side and I do have a
33:39lot of stuff built out over here. Um as
33:42you can see I have a agents and cloud
33:44because I do use both uh agents which is
33:47for uh codecs and I also do use cloud
33:50code most of the time. So, if I open up
33:51my agents, you can see that I have a ton
33:54of skills that I've built up over time
33:56that allows AI to do uh specific work
33:59inside of my business. And I also have
34:01access to digital assets, uh different
34:04departments of my business that has
34:05different things tracked as well. Uh for
34:08example, marketing, operations,
34:10uh products, sales, and strategy, and
34:13some other stuff. Pretty much everything
34:15in my business is broken down inside of
34:18this folder structure where AI can
34:20pretty much access this. So if I open up
34:22my agents.m MD, it's very comprehensive
34:26and it pretty much breaks down
34:27everything about my business. So it
34:30explains a bit about my world model,
34:32what my business map is. It breaks down
34:35my current offer, my products, how do I
34:37get leads, how to get appointments, how
34:38do I get presentation, sales. Um, it
34:41pretty much routes to where everything
34:43is located as well. So like, uh, AI
34:46knows where to find stuff. So, it knows
34:48where to find my offers. So, pretty much
34:50any agent, uh, that wants to do work
34:53will know where to find relevant stuff
34:55inside of my folders. Also, like notion
34:57databases as well and department indexes
34:59and workspace maps. So, this is very
35:01powerful because anytime uh, an agent
35:04wants to know what to do business, it
35:05will know how to [clears throat] do it
35:06automatically. And then if I pull up any
35:07of my skills, let's say I go over my
35:10right skill. Let me pull this up over
35:12here. Let's say this is my right skill.
35:14This is a skill I use to write content
35:15that sound inside of my voice. If I open
35:17up the right skill, I have a skill.md
35:19folder. And this is pretty
35:21comprehensive. So I have the name at the
35:24top, write description, write to revise
35:25editorial emails, etc. So it helps write
35:28emails for me. And then it's way more
35:30robust where I have different workflows
35:31depending on what type of uh writing I'm
35:34doing. So if I'm doing one email, email
35:36series, newsletter articles. So within
35:38these workflows, it breaks on how to do
35:40things. Then it gives like a first uh
35:42use voice pre-flight. So what to read to
35:44know how to sound like me, um how I
35:47write, what's my style, and everything
35:49is written over here. And then if I pull
35:51up the workflows and if I put on write
35:54email, it goes over all of the framework
35:56on how I write an email step by step. So
35:59this is a very high-level skill that
36:01teaches Claude or uh Codex to do the
36:04work uh predictably. So let's go over an
36:06example. Let's me show you how I would
36:09write an email or get Codex to write an
36:11email so you can understand how things
36:12work. I'd like us to write an email. How
36:14in order to create high quality skills,
36:16you need to be able to pretty much give
36:19your taste, your judgment, and your
36:21instruction step by step into a skill so
36:24that AI can do the work. and you must
36:27give it the frameworks and I want this
36:29email to be quite comprehensive. Let's
36:31go. So I simply asked AI over here to
36:35write an email. So it should be able to
36:37call the relevant skills and do the
36:38right job in the right uh way that I do
36:40it. So we can see here that Claude is
36:43being very uh resourceful. It's reading
36:46through my big ideas. And in my big
36:49ideas is where I break down how I do
36:52work in my business. So if I type here
36:54big ideas. So go to big ideas. It will
36:57open up what my big idea is. So um
37:00everything that I talk about in my
37:02business. So everything I talk about in
37:03my business is laid out over here with
37:06like everything that AI can know how to
37:08write in my way. And then it's able to
37:10read different other skills that I have
37:12over time. So what's going on here is
37:14Claude is chaining skills back toback.
37:16It read the right skill. It read my
37:18writing voice skill. It read my headline
37:20skill and then it read all these uh pre
37:24precision principle skill, writing
37:25format skill and everything. So it knows
37:27how to put in this email together for
37:30me. So it's getting all of the context
37:32at the relevant time. So we got the
37:34central point is clear. A skill needs to
37:36carry the decisions behind the work
37:37including examples of good and bad
37:39output exceptions and review criteria.
37:41I'm drafting that as a practical
37:42teaching email with concrete build
37:43sequence. Now, it's going to create a
37:45high quality email that I would send to
37:47my email list. And ideally, I've built
37:49these skills over time, and I've built
37:51my operating system which has all of my
37:54context where AI acts like an employee
37:56where I can write and do a task like
37:58this uh without me needing to constantly
38:00handhold and tell it what to do and back
38:02and forth prompting. Okay, so I wrote
38:04this email, which is pretty good. So,
38:06you ask, "Hey, I write an email. Gives a
38:08decent job, but change opening, cut the
38:09claim that goes too far. Move the
38:11example early. Rewrite the ending so it
38:12actually follows from the point you
38:13made. Then you ask to write the next
38:15seam. We make the same corrections
38:16again. It's tempting to think the model
38:17needs a better prompt. Sometimes it
38:18does, but look at the corrections you
38:19made. Each one reveals a decision you
38:21knew how to make an AI didn't. Why did
38:23that opening feel wrong? What made the
38:24claim too strong? So this is good. And
38:26then it breaks down everything like this
38:28is a highlevel email. All right. So this
38:31is a good email and it's pretty much
38:33ready to go. The only thing I would
38:35adjust it is ask it to format it. And
38:37then I'm going to share this to my
38:39notion database, which is the platform I
38:41use to pretty much store all of my
38:43emails and I use it to send it to my uh
38:45kit. I have an automation setup. So
38:47let's go ahead and do that. Hey, I'd
38:49like you to run the formatting skill on
38:51this. This is good. And then I want you
38:53to sync this to my email database on
38:55notion, please. Okay. So this is how uh
38:58codeex can interact with other tools or
39:01connectors. So, it's connecting to my
39:03notion and then it's able uh to sync
39:06everything there, which I'll show you in
39:07a moment. So, this is my notion
39:09dashboard. I use this to get everything
39:11done inside of my business. So, over
39:13here I have all of my previous emails
39:16that I've sent to my email list. It has
39:17the open rate, the title, the subject
39:19line, and everything. And it's able to
39:21find the relevant Notion database
39:24because I had it uh written in the
39:26agents.md. So, it knows everything about
39:29my notion and where to store things. I
39:30don't need to tell it what to do or to
39:33copy paste a specific notion database.
39:35Every agent knows how to do this
39:36automatically. So, as you can see here,
39:38it's opening the email database and it's
39:40going to do it for us. And for some
39:43reason, it didn't the MCP connection
39:46didn't work, but I guess this is better
39:47because now it's showing browser
39:49control. Um, so if the MCP server API
39:52doesn't work, it can pretty much control
39:54your brows, which is what it's doing
39:55right now. So you can see the cursor
39:56over here and it should go ahead and
40:00scroll down or add stuff for us
40:02controlling the browser. Okay. So now
40:04it's checking the notion connection. So
40:06it actually is using the browser over
40:09here using the chrome but now it's using
40:10the browser in chat GPT. So there's two
40:13different browser could use. It could
40:14use the browser inside. If I'm logged
40:16into notion over here it's interacting
40:18with it over here or it can connect from
40:20the browser. Okay. So chatbt uh puts in
40:23the email into notion. Now, ideally, we
40:25don't want it to use a browser. Uh, you
40:27wanted to use MCP connection, but for
40:28some reason was broken, but eventually
40:30got it done. Um, I guess it's good for
40:32demonstration purposes cuz you get to
40:34see uh chat PT controlling the browser
40:35and how it can get work done that way.
40:37But yeah, this is the email. So, next
40:40step is let's say I want to create an
40:42automation where every time I finish an
40:44email, I wanted to transform that email
40:47into a carousel. So, what I'm going to
40:49do is I'm going to open up a new chat
40:52over here and I'm going to ask uh
40:55codeex. So, I'm going to copy paste that
40:58notion link and I'm going to ask it to
41:00create a carousel. Can you turn this
41:03into a carousel that we can then post
41:06onto social media using the email to
41:07carousel skill? And then uh once I
41:10approve it, I want you to create a
41:12schedule task where every day it will
41:13check for new notion uh pages and emails
41:17that I published and then to turn that
41:19into a carousel as well and then post
41:21onto social media for us automatically.
41:23So uh this would be an example of having
41:26AI employee work on a trigger. So I just
41:29want to show you the output and how this
41:30would look like and then how this would
41:32look like on the trigger. But for uh
41:35just to show you how this look like, I
41:37do have some scheduled task already to
41:39have every day where I have a daily
41:41background sync uh daily AIB improvement
41:45review where I daily review all the
41:47skills that I've built out and make sure
41:49that I update my plug-in so my clients
41:51can get access to the skills I've
41:52developed. And then a weekly review
41:54background prep which pretty much
41:56reviews my whole week and uh gives me a
41:59report on how my business is doing. So
42:00those are uh things that I have right
42:03now where I have AI doing stuff. So I
42:04can open these chats and I can see what
42:07it's gotten done previously. So this is
42:10the prompt. I can pick the models using
42:12GPT 5.6. I can pick the reasoning and so
42:16far and so on. So I can review all of my
42:19scheduled task or like automations per
42:21se over here. All right. So this is what
42:24uh codeex created for the carousel. So
42:26it took that email and created a
42:28carousel. It looks pretty neat. though
42:29if you keep fixing the eyes work your
42:31signs are still in your head. So there's
42:33a slide one, slide two, slide three,
42:34slide four, all the way up to 10 and
42:36then there's a CTA at the end. So this
42:39is something that I could then send onto
42:41um all my social platforms and so on. So
42:44I can create a scheduled task and then
42:47CEX can do that and then the schedule
42:49task would then uh send that carousel uh
42:52to other platforms. Okay, so Codex
42:54created two different carousels. This is
42:56the one that is one to one that'll be
42:58posted onto YouTube and this is a four
43:00to five that can be posted onto
43:02Instagram and LinkedIn. So these don't
43:04look too bad. It pretty much transformed
43:05an email into a carousel and this can be
43:08scheduled out. So I can then say approve
43:10this finished carousel. Yes, approve.
43:13And then it will create a scheduled task
43:15where every uh day it will check for uh
43:18notion pages or some emails I've
43:21written. And so if I open up my notion
43:22page here, I have some emails that it
43:25can pull up to then uh turn into uh
43:28carousels. For example, this email can
43:30be transformed into a carousel as well
43:32and then help me grow my email list as
43:34an example. Okay, so it's asking me if I
43:37can use the local posted CLI uh which
43:40will help me post to my social media. So
43:42it's going to do that. I'm checking the
43:44documented. So, the nice thing is that I
43:46already have this connector, which is
43:48post CLI, which allows um an AI agent to
43:52post on all of my social platforms. So,
43:54I'll have another video breaking that
43:55down in the future, but this allows me
43:57to pretty much post on Instagram,
43:59threads, LinkedIn, um YouTube, and
44:01anywhere online by having an agent to
44:04schedule it out for me. So, this is very
44:06high leverage stuff. So, it's creating
44:07the automation spec. It's doing
44:10everything. And then in a few moments,
44:11it should create a new schedule task
44:13over here. All right. So, it created the
44:16schedule task. So, daily publish content
44:18carousels. I can open it over here. And
44:21this is the prompt it created. Um, and
44:24it pretty much tells what to do. Runs in
44:26new chat each run. Uh, so it creates a
44:29new chat every time. Project Rashid
44:31business model GP66 reasoning high. And
44:35then everything uh will show up over
44:37here. The agent will then know what to
44:39do. it will look through my notion uh
44:42dashboard or articles and then it will
44:44transform it into a carousel for us. So
44:48I can also check in on the left side on
44:49scheduled. So I can see that it's
44:51available right over here. Daily publish
44:53content carousels. I can click on here.
44:55I can edit this and or I can ask an
44:57agent to edit this over time or I can
44:59ask it to run right now. So this is
45:01another way where um as long as my
45:04computer is open um codeex can work and
45:06get work done inside of my computer and
45:08get stuff done. So going back to our big
45:11uh picture view over here. What schedule
45:14task does is allows AI to work uh
45:17autonomously or when things happen in my
45:20business. So I can get work done in
45:21another department or another agent gets
45:23something done and then another agent
45:25can spawn or a specific time review its
45:28work and do work. So, for example,
45:30getting an email and transforming that
45:32into a carousel every day at 7 a.m. as
45:34an example will trigger the model. The
45:36model will be smart. There's a prompt
45:38that goes into it. It'll be able to
45:39understand what to do. It'll open up the
45:41agents. It'll uh call the right skill
45:43tools and then it will be able to post
45:45or do work for us. So, now we can pretty
45:47much have an AI employee act at any time
45:50that we want. And we can build AI
45:53employees by prompting our workspace to
45:55build the skills and to build the
45:57context required for an AI agent or I
45:59like to call AI employee to get the work
46:01done for us. Right now your business
46:03runs in three jobs. Marketing which is
46:05getting attention, sales which is
46:07closing deals and getting clients and
46:09delivery which is delivering the actual
46:10value to your clients so they remain
46:12working with you. And if you are doing
46:14all three right now as a business owner,
46:16well then you're working on a hamster
46:18wheel. Your business only runs when
46:20you're working in it. And every new
46:21client you sign will add more to your
46:23plate, making it harder for you to grow
46:25your revenue. Now, imagine putting an AI
46:27employee in each of these systems,
46:29marketing, sales, and delivery. An AI
46:31employee is basically AI that would do
46:34the same job that you would do yourself.
46:36It knows how to do the work because it
46:38has access to digital assets which is
46:40intellectual property that you use
46:42inside of your business which includes
46:43your taste, your judgment, instructions
46:46and the exact way on how you do work
46:48inside of your business just like how
46:49you would do it yourself. The AI
46:51employee would basically act like an
46:53clone of you running your business or
46:55you hiring someone head of marketing,
46:57head of sales or head of customer
46:59success. The AI employee does most of
47:01the work and all you have to do is
47:02review its output and ship it out. When
47:04you build your business like this on
47:06digital assets and AI employees, it
47:08becomes a flywheel. Your delivery AI
47:10employee helps deliver good results for
47:12your clients that allows you to retain
47:14your clients and get good case studies.
47:15Your marketing AI employee takes those
47:17case studies and those wins and turns
47:19that into proof that you can then share
47:21out into your marketing to get even more
47:23attraction and more views towards your
47:25offer. Then your sales employee will
47:28close sales for you without you needing
47:29to constantly do sales calls and use
47:31digital assets to build that. whether
47:33that's a YouTube channel, an email list,
47:35sales assets, and other marketing assets
47:37as well. So, what happens is the three
47:40of these AI employees or the three
47:42systems all connect together and feed to
47:44each other like a flywheel. This is what
47:46I call the AI employee flywheel and it's
47:48the next iteration of how businesses are
47:51going to look in the future. And it's
47:52exactly how I run my oneperson business
47:54today. And the best thing about this is
47:56that every single time there's data that
47:58shares between each AI employee, it
48:00improves how that AI employee works. AI
48:02employees learn from what's worked from
48:04each other AI employees and that builds
48:06an infrastructure that builds up over
48:07time. So, not only do your clients get
48:10better value, your marketing improves
48:11over time and your sales are easily
48:13closed because you have good sales
48:15assets that teaches what they can get
48:17using the client proof and using really
48:20strong marketing to build up demand for
48:22your offers. So, you're always sold out.
48:24And that's why scaling a business like
48:25this is effortless and it's pretty much
48:27exponential because we have all of the
48:29infrastructure, the systems that feed on
48:32each other that allows it to improve
48:33over time to constantly build up your
48:35systems so you can build up a solid
48:37business that provides you tons of
48:39profit. The hard part is setting up the
48:41infrastructure to where AI employees can
48:43actually do solid work in your business.
48:45Starting with documenting additional
48:46assets, creating specific skills,
48:49creating the AI operating system, and
48:51more. And that's exactly what I help
48:52business owners do. I help you install
48:54an AI employee flywheel inside of your
48:56business. Now, there are a couple ways
48:58you can work with me if you want to
48:59implement this inside of your business.
49:01You can check the link in description
49:02below. Some of the options might be
49:04available right now and some might be on
49:06weight list. And if it's on the wait
49:07list, you'll hear back from me via my
49:09email

About this transcript

This page contains the full transcript of ChatGPT Codex: Automate 90% Of Your Business on AI Employees by Rashid, generated from the public captions YouTube serves with the video. The transcript has 10,499 words across 1,402 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.