Watch Me Automate an Agency’s Media Buying With Grokbot

0:00AI is capable of so much more than you're currently using it for. And in this video, I'm going to show you exactly why. I'm going to automate the media buying process of a marketing agency from start to finish and gatekeep nothing. I'm going to show you exactly how I do it, the exact tools I use, and show you how you can copy me. This doesn't just relate to media buying either. This exact process I'm going to follow from start to finish. Can relate to any aspect of any business. Just follow my steps and implement this for any department inside of your business to replace team members or enhance them

0:36to handle further workloads and make your business a whole lot more profit every single month. So sit down, relax, and let me show you how you can take your AI usage inside of your business to the next level. All right, so let's dive in. you know, first and foremost, let's actually plan out what this is going to take, right? So, when it comes to media buying, let's break down each of the

0:58stages in which actually go into this. And this is going to define what agents we're going to need to build this platform. Because think about it like this, when you're designing a system, when you're thinking about agents, every single agent has to have one specific goal. You know, every agent has to be specialized. We're looking for specialized agents, not generalists, you know. I want one agent who's amazing at one thing rather than one agent who's really good at 10, you know. So, we're

1:27looking for a really high standard here. So, let's break this down. So, in terms of the agents we're going to require, which is just going to be uh designed based on what it actually takes from start to finish to have a media buying strategy. Let's break this down. So, first and foremost, we're going to need an offer creator. Everything's going to come based on the offer. So, we're going to need an agent who's an expert at creating offers. After that, we're going

1:48to need need a copywriter. So, we're going to need an agent who can write the ad copy. The ad copy is going to feed into the rest of it. So, offer creator, copywriter. We're going to need a video scriptor, someone who can script out our video creatives. We're going to need uh an image creator

2:07so we can run some static images. We're going to need a suppose we're going to need like a landing page creator, right? So we can run our traffic to a landing page. So let's do that. Landing page creator. Number six, we're going to need So we've got our video scripts. We're probably going to need like an AI video creator, right? Something that can maybe leverage something like Higsfield to actually create those uh or turn those those scripts into real life, right? So at this point we have our offer, we have our copy in which we can run, we have our video scripts, we have our images in

2:44which we can run, we have our landing page, and we also have our videos in which we can run as well. From there, we're going to need a campaign creator, aren't we? Something that's actually going to put all of this into the campaign for us, we're going to need once that is done, we're going to need a like a

3:01campaign optimizer. So, campaign optimizer. So, something that can like keep an eye on the campaigns and optimize them automatically without a human having to step in. After that, we're probably going to need like a campaign reporter. Something that's really, really important is when it comes to media buying is reporting, right? So, we're going to need someone who can keep track of the metrics and track them in some kind of sheet. I think that's going to be essential as well. And then the 10th agent I would probably say is this is like an agent I always have in all the systems I build which is just like a QA agent which QA

3:38just stands for quality assurance. So when I'm saying QA basically what I mean is for example when the offer creator is finished it sends it finished offers to the QA agent and the QA agent reviews those offers and says yes I'm happy with those. I think they're a good idea or no I think you've missed a mark here or you've forgotten or you've forgotten some of the input information we're giving you about that business. Uh, same thing with the copywriter. Write his copy. It pass it to the QA agent. The video scriptor, same idea. It does its thing and then sends it to the QA agent

4:07who checks it. And this just really, really minimizes mistakes. So, this is our 10 agents. This is like the 10 things I think goes into media buyer. So, to recap, we have our offer, the offer we're going to run, we have our copy, which is written based on that offer. We have our video scripts which are then based on the offer as well. We have our images once again are based on a business and the offer. We have our landing page. We have our AI videos in which we can run which are based on a video scripts. We have our campaign creator um in which is actually going to

4:36you know put the copy, the images, the video ads, everything all together inside of a meta ads campaign. We have our campaign optimizer who's going to automatically be making adjustments on the campaigns as and when required if we're outside of KPI. We're going to have our campaign reporter who is going to be, you know, actually doing our reporting data and tracking metrics for us. And we're going to have our QA agent as well who's going to be just checking all of the work throughout the process to make sure we're not making silly mistakes. So that's going to be our 10 agents. And now let's kind of think a

5:05little bit about the architecture. So now that we've got our agents defined, the way I would h build this from here is all of these agents are going to be built inside of Grockbot. Of course, Grockbot is the best agent harness in the world right now. So we'll build all of these inside of Grockbot. Grockbot is going to communicate through uh to our UI. I would say it's going to be like a to and from relationship. Our UI is going to talk to Grockbot and

5:29Grockbot's going to talk back to the UI. The conversations are going to be happening through web hooks and Grockbot's going to be talking back to the UI through callbacks. So to break this down, a web hook, when I say a web hook, it's basically just, you know, if you've done anything on Zappy or Make or any of these automation platforms in the past, you probably heard of these. A web hook is just a way for my UI to send information to Grockbot. Grockbot's going to have like a URL. I can send information to through a JSON format. I'm going to send that information to Grockbot. Grockbot's

6:02going to then send information back to my UI to a call back URL, which is just like a web hook inside of our UI uh with the JSON response. Okay, so we send a body request and we get a response back. And then uh to be honest like that's going to give us full functionality here through the UI and then our UI could probably be linked to something like our on boarding form on boarding form and which could actually fully automate this process, right? Because if I was running an agency, I would want like new client fills out the onboarding form and this just happens. I don't have to

6:40go to the UI to fill it in. I want this fully automated. So when that onboarding form is filled, their campaign is live within like two hours, you know. So the only way to do that would be have our onboarding form linked to the UI or even our onboarding form linked all the way to Grockbot uh through these web hooks

6:55in which we could just tap into. Right? So that's how I'm thinking about the architecture. We have 10 different agents. The agents are all built inside of Grockbot. Grockbot talks to our UI through web hooks and callbacks. And our UI communicates back to Grockbot through web hooks. Um, the UI is linked to the onboarding form. So, this can be 100% automated. And I think that would cover our entire media buying system if I'm honest with you. Let's get a little bit more granular then. So, let's say the each of these agents. What information do I need to give them? I think we're going to need references for what a

7:29typical offer looks like. I don't really want the AI guessing what kind of offers to run. The way I kind of see this agent functioning is the offer creator should probably create like three offers and and potentially serve them. I'm trying to make this a fully automated system doesn't need much human input at all. Maybe some final approvals in terms of like can I put this campaign live? I can approve that. Maybe some like maybe some approvals around like can I send these these reports into Slack or can I make these campaign optimizations? But I think that's really the only human input we should have, right? Uh so offer creator, I'm just

8:08going to say just create one offer in which you think is best and then build all of the the information like build everything, you know, pass all your information on to the rest of the 10 agents and have everything finished rather than serving me offer options. I think you make the decision on what you think the good offer is and then me and my team can just sell it to the client, say this is the offer we need to run to get the best results, right? uh copywriter. Once again, I'm going to have to send uh references in here of the typical copy in which we run and

8:35what works. I think that'll be really useful. Videos script and once again I'll give it some references of the video scripts we typically use and what makes them good and and you know script them in the same way. Uh image creator once again I can probably give it some references of like static images or at least some instructions of the static images we typically run. Landing page creator I can probably create a template landing page in which it can copy and

8:56paste from. AI video creator. I'll probably I think what we should do is this should obviously be very very tightly linked with the video creator. These should communicate quite closely. The video scriptor should probably when it when it's scripting creatives be like creating scenes as part of those scripts, right? And those scenes should obviously then be passed to the AI video creator. So when it's creating those those videos, it looks at the script, it looks at the scene and it turns that into, you know, into like a into the actual video, right? It doesn't have to guess how how the script should look. I think that's going to be the best thing,

9:30right? So, the video creator is actually going to be like the the video scriptor plus scene creator. That's going to be agent number three. Campaign creator. Once again, I can give this some instructions uh of the typical campaigns we set up. Maybe we can say like all enhancements turned off, multi-advert multi-advertiser ads turned off, no instant forms, use this pixel, different things like this. We can do

10:00campaign optimizer. I think we should give it like a KPI tracking sheet, right? So we can give this like our typical KPIs like here's what our KPIs look like. These are the KPIs we follow. If we're outside of KPI in this way, you should make this adjustment. We can give it some direction instead of letting it kind of have free reign. Campaign reporter once again like we can maybe I don't think we really have to give this any information. It's just pulling all the data and it can display on our UI or create reports through Google Sheets or something along those lines. And the QA agent I don't really want to give that

10:30any direction at all. I kind of want that to call its own shots um and do what it thinks is necessary. And I think that about covers us for this this back in that architecture and design. So the next step from here now that we've got our our actual I call this like our

10:44architecture. So let me zoom in right here. Architecture. Excuse my handwriting. Once we've got our architecture, I'm going to pass this into a design. We're going to turn this into an actual formal design in which we're going to build. Think of this like a UI design, right? this UI design when we have that finished we're then just going to feed that into something like clawed codes to actually build the finished product right so this is like the road map in which we're going to follow so think about it like this if we create this architecture and I feed this diagram into cloud code or codeex or

11:21whatever you want to use we also have a UI design and I literally feed that code into cloud code or codeex or once again whatever you want to use how much easier is is that going to be to get a finished product without claude codes having to guess what it thinks it should build. It knows because we've done the architecture and the design. It doesn't have to guess. It just has to make it a reality. So that's exactly the road map we're going to follow. So let's dive on to the computer and let's build out this UI design, shall we? All right, beautiful. So I have fed the claude code

11:54agent right here. I've fed it the design which we just created on the iPads. I've set the model to Opus 5.5 and medium. This is the best model by far as at the time of recording this, right? And I'm going to use whisper flow and I'm literally just going to talk to it and I'm going to tell it exactly what we're going for. So watch what I say here. I want you to create a front-end design as an artifact here on clouds. I've given you a screenshot of the architecture design in which we've just created. I want you to create the UI. You're not going to be creating all of these

12:21Grockbot agents and the onboarding form and things like that. You're literally just going to create the UI and which is going to communicate with Grockbot. You're just creating a design. It's not going to be any backends. You can use all pilot and fake and demo data to create this. Think very very deep about your UX design and the UX experience and how the UI looks as well. It's very very important that this front end looks extremely good. If we have uh an amazing back end but a front end that looks terrible, everyone is just going to assume the app is bad. So, it's essential to me that we have a very

12:57smooth and nice looking front end UI. I want the UI to be very simple, not too much on a navigation bar. I know there's a lot of agents that Grockbot is going to have that we're going to be communicating with through the web hook, but much of this stuff can just happen on the back end and just be then served to us on the front end. It doesn't have to be like everything visible. So, for example, I think we should maybe have a client's page. This is going to be for a marketing agency which can have multiple clients. It could have dozens of clients. There needs to be a dashboard

13:28um probably where we can filter based on clients. I really like the design where it's typical like sub accounts like a go high level design where in the top left you can choose all clients or you can change it to a specific client. Um there should be a dashboard in which we can see essential reporting data that the campaign reporter would have fed back to us. just some of the primary metrics, h the primary KPIs and things like that. Obviously, if we have all clients selected, it's going to show us

13:55the data across all of our clients. Maybe we can then have a table on that page which shows us all of the clients in a table list as well as cards on that dashboard, too, for easy viewing. The table should be filterable. You should be able to sort it and different things like that. Like a nice UX experience. Obviously, if I change to a specific client, that table will go away because I won't see all the clients in that case, but I should just still see the cards. I reckon there should be a create function on there where for a specific client, I can create things manually in

14:25case I want a new copy or a new image or new video scripts or new AI videos, you know, different things like that. I should be able to communicate with the Grockbot agents as I desire inside of that create page. But this should also be fully automated. So, if I add a new client on the UI in the clients page, click new client and create the new client and upload all the information it needs like logos and color schemes and different things like that. Um, it should automatically then call all of the agents in order. The QA agent should be called obviously like offer created, QA agent is called, copy is written, QA

15:00agent is called, video scripts and the scene creator runs, the QA agent is called. The QA agent is called at every single step to validate um you know everything that is being created by these agents to make sure it's correct. But all of these other agents should run in sequential order right there. Obviously besides the campaign optimizer and campaign reporter, they should probably be set on cron jobs. This is going to be something for the backends, but maybe you can leave placeholders for that because I'm eventually going to like pass your front ends UI codes onto another claude code agent who's going to create the back ends and then set up

15:31Grockbot for us. But I'm just trying to give you all of the information. My color schemes at this agency is black and an electric blue. So, make sure to follow that color scheme. H make sure none of the colors are kind of competing with one another. I want everything to look nice, nice fonts, aesthetic uh kind of page, minimalistic design. Ask me any questions in which you need to create the perfect front-end design. Thanks a

15:56lot. Okay, so just like that like this is like a very very large prompt. Imagine me trying to type this out. I use Whisper Flow, as I said, which is the tool, and I simply just talk. I'm going to run this prompt right now. We're going to give this 15 minutes or so. This is going to spit us out the perfect front-end design. If we have to go back and forth on it a little bit, we will,

16:16but we're going to test this design out. We're then going to download that code. Like, we're actually going to download the code in which it writes, and then we're going to pass that on to Cloud Code backends agent, which is going to build out the actual codes in the background. So, let's answer some of these questions and we'll uh we'll continue onwards. All right. And beautiful. Just like that, we have our code. So, let's have a quick look through here and just make sure we're on the same page here. So, dashboards, we have this. It's followed that sub account design we were talking about where we can choose a specific client

16:47and see all of their information or choose all clients and see all of the information. We can choose different time frames. We can add clients. We get all of our Grockbot activity to see what it's currently working on, what it's smashed out recently. This looks really good. Really, really good. And it's a nice aesthetic UI as well, which is what we were going for. Yeah, we've got our table down here. You can see all of our live clients. Sort by leads, CPL versus

17:11target. Yeah, this is great, man. Really, really solid. Let's have a look. Clients page. Oh, yeah. We can even see their assets in here. Yeah, this is awesome. Beautiful. Exactly what we're going for. So, this is what we like to see. Uh, create. So, yeah, here we can do our video scripts, create our offer, get some ad copy, even make our static images in

17:37here. Yeah, this is what we like to see. Automations page. What's in here? Yes. So, it looks like this is when we can schedule when our ad optimizer runs, when our client reports run. Yeah, really nice. Nice and simple. Cool. So, literally what we were going for, a nice simplistic design, minimalistic design, nice and aesthetic. Looks great, man. So, what I'm going to do now is I'm going to download this code. So, if I just ask Claude code to give me the code, I'll download that code. I think we can even probably download it from up here. Yeah. So, we choose downloads and I'm then going to

18:08feed that to a Claude codes thread and we're actually going to turn this into a real app. I'll build the back end of the app first and then once that's done, we can literally have Claude code write our prompts to give to Grockbot to then build out all of the agents which are going to power this thing. So let's move on to there. All righty. So as I explained earlier, we've given it the front-end codes and we've also given it our architecture diagram, too. I was explaining earlier on the iPad about how if we just give Claude codes more context. Here's what we want it to look

18:37like with the code so you don't have to guess. And here's the actual system planned on paper so you can see it and visualize it. The output is going to be 10 times better because it doesn't have to guess. It understands exactly what we need. Right? So I've given it the information. I'm going to do another whisper flow. We'll do another like large prompt two or three minutes of speaking just to give it all the information we can possibly give it. And then you know we'll let it just build. We'll give it half an hour. We'll give it an hour. will build out the system and then

19:05we'll build out the agents on Grockpot. To wrap up, we're going to be working on a new system. This is going to be a system which is going to automate media buying inside of Harry Media, the fake agency we're going to be setting up. You are going to be using fake data to begin with and pilot data just to get this

19:23system set up for the purposes of demo. However, I want you to be able to support real data too. So, tell me what en variables we need to set up to make this happen. I want to deploy the app separately for a front-end folder and a backend folder folder structure on the back end a Python back end. We should follow cyber security best practices like putting end variables on the back

19:44end database calls on the back end. Front end communicates with the database through uh roots. Uh we should use a Postgress database using Subbase. We can deploy the back end on render. We can deploy the front end on versel. I want the front end to be runnable through um VITE and I want the back end to be runnable through Unicorn. Everything should be able to be run locally so we can test this without having to push to production or staging environment. We're going to be putting the code onto GitHub as well. I've given you the architecture diagram which explains the system we're building here as well as the code for

20:22the front-end design that's already been created. Below this prompt, I'll paste in the prompt I gave to another claude code thread that created the front-end UI just so that you can see what I told it so that you are fully aligned on what it has built. We're going to be using Grockbot to create all of the agents and the kind of like implementers who are going to be doing all of the work on our back end. You're going to be communicating with Grockbot through web hooks and Grockbot should be able to call back to you when it finishes things with its outputs. Grockbot isn't set up

20:55yet. That's going to come next. I want you to create the backend harness and then we'll create Grockbot afterwards. You're going to give me all the prompts I can give to Grockbot to set it up. However, you know, we'll set up your side first before we do create Grockbot because I want you to to firstly define what what callback URLs uh Grockbot is going to send back to and what you want Grockbot's web hook routts to look like as well, like what body requests are you going to send it, what um call back, what is it going to give you and its call back result JSONs and

21:27different things like that. As you can see from the diagram, there's going to be 10 different agents. I explained to the front end how those agents work. I'll give you that prompt as I said. H other than that, let's just make this happen. As I said, tell me all the end variables you need me to get you to set

21:42up on the on the back end. And for now, let's build this a demo and pilot data that just works so that this can be fully tested without me having to configure all of the M variables and things like that. Get started on a build. Obviously, ask me any questions you need whatsoever so that we can have the perfect system and you don't have to

21:59guess things. And just like that, we have our prompt. So I also copied the prompt for the front end which I am going to paste in. Yep. So that's our prompt for the front end below and our prompt for the the back end at the top. So that's the prompt I'm going to send. Literally I'm just going to hit enter. I just want to make sure that we are on the right setting. Let me get whisperflow out the way which I can't move. I think we're on I just want to make sure we have bypass permission set. Well, whisper flow is in

22:27the way. No worries. So I'll hit enter. Uh, as I said, we're going to give this maybe 30 minutes or 60 minutes. This is going to get to work. This is going to take a little while here, but that's absolutely fine. I think we can change auto. Yeah, I wanted it to be on bypass permissions just so it can it's not going to ask me for permission to do things. I fully trust it to do what it needs to do. Um, and yeah, we'll see if it asks us any questions here we need to

22:47answer so that it has all the context. That's like a really good thing to add into your prompts to be honest just because if you don't say that it's, you know, it's just opening yourself up for some AI hallucination and things like that, things that's going to guess that you might not want, right? So add that to your prompts. It'll ask you questions. You can answer them. This might already have all the context considering we've given it a front-end UI and an architecture diagram plus the

23:09prompt we gave to create that front end. But let's wait and see what it builds for us. As I said, let's give it 30 or 60 minutes. We'll skip till we get his first response and we'll have a look at what it's built. Okay guys, we have our claw code finished. So um now I want you to give me the prompt in which I can give to Grockbot. I'm going to set up one agent on Grockbot first which is going to be like our our CEO agent who's going to create the rest of the agents. I want that prompt in which I can give to that

23:38agent who's then going to set up all of those agents for us. It needs to be detailed please. So that uh and it knows exactly how all of these agents are going to function, what we're going to send them, uh what their web hook routts should look like, what their call back um you know what data they should call back to us and where they should call it

23:56back to. Um for now we can let's just have them calling it back to the on render link for now. I haven't deployed this on render yet or Yeah, we'll have them call it back to the on render link for now. Maybe we could set up an endrock. I think that might actually be a better idea so we can test this before putting it on produ onto production. But maybe you can set up an endrock that links with our our backend localhost h and which they can call the data back to. I think that's a better idea. So put that into your prompt as well. And then once we've

24:34fully tested it and it's good, we can update that endrock link over to to our on render link. H give me that contract now please. All right beautiful. So there's our prompt. So when I am saying mrock and backends and web hooks and things like this like what I'm trying to figure out is I want to test as much of this as possible before I start to pay for backend services and compute use and different things like this. Like if I run everything locally, I want to have it working there first before I then go and put it somewhere and pay for the compute power. It's just going to save

25:06us cash. It's also going to save us time. So, I'm going to deploy it locally here. Uh, Enro is just like a it's going to set up a URL in which another computer can hit that's going to give that information to my local computer. Right. So, let's give this a minute to write this. And uh in the meantime, I'm going to load up Grogbot. We'll get that CEO agent created and uh and we'll start building. All right. Beautiful. So, I've opened up Crockbot here. I literally just clicked in the top left here and chose Harry Media CEO. This has created created an agent. That's all I've done

25:38so far. When Claude code gives me this prompt, I'm going to paste that into here and that's going to get to work. And that's how simple it is here to set up Grockbot. So, I've got Crockbot here on the top plan. You probably wouldn't even need all of that usage to be honest. I think the usage is pretty fair on Grockbot in my opinion. Uh, but basically what Grockbot is going to do is for any of you who have never used this, the way you can think about it is Grockbot is like openclaw if you guys have ever heard of that. But it's controlled on its own

26:05computer. So SpaceX has like a cloud computer in which this uses. It has its own machine. It doesn't run on my machine, which just means that this can even run when my computer is switched off and do things. It's extremely smart. The browser use is amazing. The computer control is amazing. Um, it also has a ton of different models in which you can use here out of the box. And it's super user friendly. Like I don't have to tell it what to do. I don't have to tell it what models to use or how much effort to use. It makes all of those decisions itself. Um,

26:37based on what it thinks it's going to require for this certain task I've given it. Okay. So, you're going to see how this works real soon. It basically just controls itself. So, I've created the CEO agent here first. I'll give this the prompt. It's going to go away, create all of the other 10 agents, and get them all set up fully, and then just tell me when it's done, and tell me what to give

26:56back to Claude Code. So, super smart. Looks like we've just got a brief through. Uh, so first and foremost, we have to set up Enro. So, let's get this done first. So, I'm just going to run this in the terminal that's given me here. [snorts] Let's give this a second to load. I don't know what happens. Let me try one

27:12more time. Do we have it? I think that's H. I'm going to have to Oh, I see. I see. I see. Your domain enrog free app. So, that's probably the issue. Uh, so let's change this over. Enro HTTP 8000 HTTPS something. Enro free app. So to connect it, I'm just reading through it prompt real quick

27:46in your step number one claiming the free enrock domain. What should I change to your domain placeholder value you've left in with? Okay, we'll get that changed. H and then it wants us to generate a secret key in which this can use. So, we'll get that done. So, log into dashboard.enrock open domains in the left menu. Okay. Let me

28:06grab mine. Okay. Okay. So, we have Enro set up. I've just grabbed our contract and which has gone straight to Grockbot. I'm going to open Grockbot back up over here. And I'm just going to paste this to our CEO agent. And then we'll just leave this to get to work. This will probably give us a confirmation in like 30 seconds or so

28:23and tell us it knows what it's doing. There we go. Got it. That's a big build. Starting with the router and job store on my computer. working out how to give it a public URL and I'll set 10 agents and run the self test. So let me actually whisper flow to this real quick. I have another agent which we're using which is using like a cursor web hook endpoint which I can hit and then you then define the bearer token which I have to send you. That's working really

28:44well. Let's uh do this in the same way. Okay. Just giving it some direction on how the web hook endpoint should look. That's going to set up these web hooks. Yep. Absolutely. Okay. So, I'm going to give this about 30 minutes or so this is going to take uh to be 100% finished. Once it's done, we'll regroup and see how our agents are looking and we then we can probably run the final test on

29:10this system. All right, we got our word back from Grockbot. Uh let's I've been through this a little bit already. It has created all of the agents, but I think we need a couple adjustments. So for example, offer copy script and QA is all using OpenAI which is fine. Landing page is trying to create these custom. I think we should probably do this through lovable. I think it's going to get us a better result and be a lot less complex. Image is using its own image tool. I'm probably going to get this to use GPT 2.5 sunburst which is chat GPTs. I think it's it's a whole lot better. I think

29:46the images from there just come out nicer. Uh videos it's using Hexfield which is good. We've got Hexfield connector in here we can use which is fine. uh meta marketing API. So I've already got a meta access token which is in here which is also perfect. So let's do this. I also have a few reference documents in which I'm going to feed this. So I'm going to grab these on my other monitor right here. So I'm going to feed it a few reference uh things like I've got some reference

30:14copies I want it to create based on. I've got some direction around landing page creation and things like that and which we can use. So a couple notes for you. Thank you for getting those agents created. Uh firstly landing pages. Let's use a lovable connector which I've already connected. Uh and we will create our landing pages through lovable on the Harry Harry's

30:37workspace. For images I want to use GPT2.5 sunburst. use our OpenAI API key which is already connected in here to create those images. Don't use your own image tool. Uh I think that is fine. Besides that for I've got a few instructions I want to feed you. Firstly is a campaign creator instructions which uh basically just breaks down how I like my campaigns to create uh to be created, how I want them targeted, how I want them named, how I want the ads set up, enhancements off, different things like that. I've got some copy templates in which our copy

31:13agent should be creating copy based on. I've got some image instructions which just detail how the image uh creative should look. Um I've got some offer templates in which you can use as well based on just typical offers in which we run. And finally, I've got some video scripting templates in which our video scripting agent can use when creating those scripts. So I'm going to feed them all to you below in this prompt right

31:36now. Okay. So feeding it all of our resources which I've just created on the side literally just Google Docs just got some examples like the copy I want to use the instructions I want to use for creating images. This one is video scripting templates. So just feeding all of that information and then I'm going to send that in so that it can update our existing agent. So this is

32:02what I'm talking about when using Grock. I'm literally just communicating with this CE CEO agent and he's setting up all of the rest for me. So, this is why I'm saying Grockbot is so user friendly. Right. So, uh when this is finished, we should be good to go and we'll start running a test to our UI. All right. And finally, Grockbot has finished cooking up all of the agents. So, I'm going to go back to Clot. I'm going to ask Clot. Grockbot has just finished setting up all of the agents. Is there anything in which you need me to gather from Grockbot for you to start sending test

32:38requests or can we start sending test requests the agents? Now the test I want to run on our UI is just going to be an endto-end test of me creating a client through the client's page and seeing everything come to fruition from copywritten to offer created to landing page setup uh into a final campaign

32:57which is created. Okay, we're going to send that uh prompt over to Claude. Claude should be pretty quick since we're not asking him to build anything. We're just asking a question and we'll see what we need to give them. All right. And just like that, it is time to test. So, I've logged into the front edge UI right here. And to do this test, I'm going to h create a new client inside of the new client dashboard. So, let's go ahead and do this. We're going to create a a random client. Let's call him

33:25Harry HVAC. And we're just going to fill in the form. Industry, we'll go for other. Main location, we'll go for Houston, Texas. Contact name. Phone number. Okay. Logo wise, let me grab one. Thrown a random logo on there. These all look fine. So, I'm going to leave these. Tone of voice. We'll go for bold. Let's continue onwards. Main service. We'll do H track replacement

34:00lead magnet I'm going to leave blank. Current offer we'll do like a $1,500 buyback. Um plus a free smart thermostat with AC replacement. Who are we targeting? Homeowners in Houston, Texas area. Biggest problem they have. an AC that keeps breaking down. Customers served, we'll say 20,000. Testimonial from David Young. And best testimonial, yeah, we'll just say installed in a day. Came

34:40out on short notice. Beat the price of our competitors by far. Okay, continue. So, ad account ID. Let me throw an ad account ID in here. All right, so we got all our assets filled out. Monthly budget, we'll go for $100 a day, $3,000. Target CPL, we'll do 75. And average deal value, let's put $15,000. Primary goal, we'll do the calls. Uh, so this all looks good. So, let's go for launch on boarding. And

35:08let's see how Grockbox agents hold up. So, remember, everything here is being communicated to Grockbot through web hooks. I've done this, the offer creator. We've sent a web hook request to Grockbot. Grockbot's going to receive it. They're going to create an offer and it's going to send that offer back to us. We're then going to send a request to the copywriter through another web hook. The copy is going to be written sent back to us through call back a call back URL. H and we're going to receive that copy and store it. Okay, so as we

35:37can see QA passed on the offer creator. Real real nice. H and let's just give this a couple minutes uh until we get the call back on all of our information here. real quick. We can see whilst this is working right here, if we go into uh assets, we can see the assets over here are being updated as it's going, which is super sick. It's row uh three three different offers right here. Uh and it's throw three different uh ad copy variations as well. All looking real real good. I like it. Offers, refer a

36:06friend there, and exclusive rewards. That'll do. That'll do. Where we at right now? So, we've moved on to the video scriptor. It's now scripting some video ads. We're going for one for the referral rewards offer, one for the buyback offer, and one for the fair pricing offer. Real nice. So, uh, let's continue on. Let this finish up. And just like that, all of our agents are finished. So, offer was created, copy was written, videos were scripted, and the scenes were set. All of our image creatives were done as well. Our landing page was completed, our AI videos were

36:39done, and then the campaign was created. So, let's have a quick look at some of the assets. So, it's given us the AI videos uh here inside of the landing page. Looks like they ain't displaying correctly. So, there's a little bit of work to be done. Uh besides that, it's got image creatives as well, which also

36:55don't display correctly. That's fine. Tells us the landing page is finished and gives us a URL to go and have a look at it. So, let's have a look at the finished landing page. As we can see, I mean, reasonably solid. It's got a form on there. Beautiful. It's got the one testimonial in which we wrote. So maybe there's some room to add some extra testimonials here. But other than that, I mean a reasonably solid landing page, right? Like I mean that's

37:19going to do the job right there. Um we can see all the copy it wrote, the video scripts it wrote, all the offers in which it's testing. I mean really really solid, right? This is one prompt that's got us to here, you know. Um so let's have a quick look at the campaign as well. So I loaded up the ads manager. Um, so I can see Houston plus 40 miles, which is fine. We said we're we're

37:42servicing central Houston. Uh, let's see. So, it shows the data set in which I shared. That's awesome. It set the budget to 100 a day. Um, we said it was $3,000 a month. That's really good to see. Just in Texas, um, let's have a quick look at some of these ads, how the image creatives look. Yeah, I mean a really really solid image creative that would perform well. You know, it's set up to go to our landing page. Oh, looks like it even set up three different landing pages, one for

38:13each offer. That's really good to see. Good practice. All of the AI enhancements are turned off, you know, like this is solid work, you know, like I'm truly impressed by this. This is one run through. So, I created the client. This all happens automatically. And, you know, that's been the goal of this video. What I have just showed you today is exactly what Grockbot and Claude Code are capable of. We have just autom automated the media buying of a fake marketing agency in this video. If this was a marketing agency, anytime you sign a client, that is your client good to go within the case of an hour or two

38:49as it's ran all of these steps. This is what AI is capable of. So, I hope this video was informational. I hope this uh this kind of showed you what truly is possible um when you're using the right tools and shows you what AI is capable of today. Obviously, there's still some work to be done here. The landing page could be slightly better. The image creatives could be sharpened. Uh I don't see any video ads in the campaign. These are small things to fix, right? Uh the assets aren't displaying correctly on the UI. But this is how I went from nothing to a finished product and walked you through the

39:28entire build process in this video of automating media buying inside of a fake marketing agency. And you can do the exact same. Listen, if you're interested in getting some support in building things like this, my company, Build Chip and Scale, actually trains and places AI engineers. That is our business. We literally train and place AI engineers into businesses just like yours. So, it doesn't matter what business you run. If you're interested in having an AI specialist who's trained on building systems just like this one I have done today and even better than these ones, you can just

40:07click the top link in the description. You can book a call with me and my team and we'll walk you through what this process looks like and we'll find an AI engineer who's going to be perfect for your specific business and the systems in which you're interested in building and place them inside your business so they can build them. Thank you so much for taking the time to watch this video today and we'll catch you very very soon. Thank

40:30you guys.