The 4 Levels of Vibe Marketing | How to Go from Basic AI Prompts to Full Agent Automation

0:00All right, there's four different levels of VI marketing. As you progress from one to four, you gain efficiency, leverage, and quality. As you move through these steps, you also expand your surface area. So you go from being able to do research or, you know, basic content creation to building, you know, full-blown websites or micro tools,

0:27internal automations, things like that. And truly, one marketer gains the power of 10 people. I've seen it happen in real time within our community. And today, I'm going to walk you through each level of Vibe Marketing. Within these levels, I'm going to give you a few recommended tactics, how to get the most out of each level, and some tips uh

0:48for how to progress uh up to level four. So, level one is when you're using LLMs basically in your day job. You know, you're you're using Chat GPT03. Hopefully, you're using Gemini 2.5 Pro. You're using Manis. These are great LLM uh interfaces and tools for doing research, understanding what your competitors competitors are doing, uh understanding gaps and opportunities in your marketplace, creating, you know, content and brand guides, even different artifacts like dashboards or reports or things like that. The problem is that when you're working in one LLM, there's a lot of context switching. So maybe I use GPT 03 for competitor research, but I really like to use Claude because it

1:40does a better job at nailing my brand voice for content creation. So I'm switching between these two LLMs. Uh, and I call that context switching. So I'm going from where the research is done. I have to, you know, copy some kind of information or something like that and then go and paste it into Claude to give it the context of what I'm trying to do. And I'm essentially working in silos using, you know, various tools for different for different jobs. Um, you know, there's there's better ways to to do it. And we're going to talk about that in the next levels. But while we're here, I'll

2:17give you a few tips for getting the most out of uh level one. Uh, one is, you know, what LLMs you're going to use for different tasks. Uh, for me, I like to use 03 and Gemini 2.5 Pro, as well as Manis for all my research tasks. Um, I'll use it for things like, you know, competitor analysis, gap analysis, which is very important. What's happening in the market that, you know, my competitors are missing that I can take advantage of? I'll use Claude to create different types of content ranging from, you know, social toformational content on my website, maybe lead magnets, emails, things like that. To get the

2:57most out of level one, I talk about this a lot. You have to do two things. One, you have to know what questions to ask. And two, you have to know what good looks like. So if you don't know what questions to ask, you're not going to be able to allow AI to guide you along the way. So uh you know, for instance, I'll lead you through an example prompt and we can unpack it a little bit uh with these two guidelines. So I wrote a prompt here. This is kind of how I write prompts. Um I'm saying you're a growth marketing expert that works with proumer

3:32AI products targeting SMBs. You specialize in elevated direct response that blends a creator's relatability with conversion focused content. I'd like you to analyze my website and give me your 10 best content ideas that will attract and convert visitors in our ICP, which is growth marketing leads at tech startups to a free trial of our product. Consider organic social, organic search, LLM discovery. rank each channel and associated opportunities within those

4:08based on perceived effort and impact. So, let's unpack that a little bit here. So, I'm framing up the LLM. I'm saying you're a growth marketing expert that works in, you know, this certain industry, vertical, or niche. So, I'm trying to frame it to have some expertise that's relevant to what I'm looking for. If you don't know what channels to attack, you know, I'm already mentioning what it should consider. Consider organic social search and LLM discovery. You could say, "Hey, you're a growth marketing expert that works in, you know, your industry or your niche. What types of channels uh should I be focused on without having to

4:46spend money on ad dollars, for example?" And then from there, maybe you're digging in and asking it to recommend the right channel for you based on visiting your website and understanding uh your competitors and what's happening in your industry. So from there, maybe you're, you know, digging into a specific channel and then asking for strategies and tactics based on the amount of effort and impact that those might have. Um, so that's how I would kind of go about, you know, prompting uh an LLM to find some opportunities for for growth and for vibe marketing. From there, obviously, you can keep working with the LLMs. You can, you know, try to

5:28create those assets there based on the voice or whatever you're trying to achieve or based on what you need to go and execute against one of these marketing channels. Um, so recommended tactics to squeeze the most out of level one, uh, would be setting up MCPs within Claude. So, um, you know, the ones that I use, I I really only use four MCPS when I work with LLMs, whether it's, you know, working with Claude or something like that or working in cursor, which is in level three. Um, I recommend perplexity, uh, data for SEO, ampify, and firecrawl. These can completely handle 99% of your research and strategy

6:15building tasks. Uh if you're activating those in Claude, for example, uh how might the prompt change? You might say, you know, you're a growth marketing expert that works with proumer AI products targeting SMBs. Use the perplexity MCP to go and identify my top 10 uh competitors for my website, which is whatever uh.com. So, it'll use Perplexity, which is highly uh efficient and skilled at research to go out there, analyze the market, look at competitors, and most likely find things that you aren't able to to think about uh on your own. So once you find those competitors, maybe you'll use the data for SEO MCP to say, "Hey, uh, use the data for SEO MCP

7:00to identify, uh, you know, the top 10 keyword opportunities for me to go and rank for in our niche that our competitors are currently not ranking for. Um, so that's a way that I would do that. Uh, you could use the Appify and Fire Crawl MCPS to go and scrape information from competitors. Maybe you want to go and get an idea of what they're writing about, what kind of brand voice they're using, what they're doing in terms of copy on their website, what they're doing on social, and get all that information as context within your LLM. This prevents, you know, hallucination and, you know, Claude, for

7:38example, in this case from, you know, making things up and it creates a more focused and targeted data set for you to work off of to create these assets and artifacts that you can actually put to work. Now going through that and going from this initial prompt and research and asking questions to actually arriving at something that I can go put into action is, you know, kind of disjointed. And that's where level two uh comes in. In level two, you build workflows. You know, if you're online, if you're on X, if you're anywhere these days on social, you've seen workflows. You've seen the screenshots of inade and gum loop and

8:20things like that. Now, this allows you to escape that siloed approach uh and automate these manual uh se or semianual tasks that you're doing. Um you know, reducing the amount of context switching you have to go through and especially if you have something that you know, let's say you're doing this research on a weekly or a monthly basis to come up with new content ideas, you can automate that. those can be sent in to where you're working with your team or something along those lines. So, it's

8:53about chaining these actions together. So, if you start at level one, you really have to kind of audit your process. Uh, one of my best tips for doing that is if you have a research flow to to to start with, let's say, understanding what your competitors are doing to then create a piece of content,

9:11record yourself doing it uh with Loom. Uh, and then you can upload a video or something like that into um into the LLM and say, "Hey, how could we automate this with Gum Loop? How could we automate this with inate? Help me think through the logic of these steps that I'm doing and recommend uh the different nodes that I could activate in this workflow to automate it. Um, help me go from zero to one." Now, it's unlikely that it's going to create a workflow uh out of the box like one shot that you can go and like paste into inad or something like that, but it's a starting

9:49point. You have to learn how to use these tools. You have to go and learn how to configure the nodes uh etc. So, um you know, it's kind of like a series of if uh then uh statements. So, you know, if a new competitor uh comes up in my market, research what keywords they're ranking for, uh understand the gaps, etc. Alert me when that's happening. Um, and then, you know, if that's one of my target keywords, create a blog post and publish it on my website, something like that. This is where you really start to unlock like a

10:23bunch of new efficiency and leverage. And this is where most of the VI marketing that you see online is kind of happening right now. However, I think that a lot of people are trying to sort of retrofit uh very complicated things that could be done in code into workflow building tools, which we'll talk about in a bit. So, an example uh workflow that I would recommend looking into would be uh go to Gum Loop. They have a SEM Rush keyword node. Um, have it go and find 10 keywords for you. Figure out how to organize those in a Google sheet. Use an AI node to loop over each one of those

11:06rows. Create a targeted content brief based on your brand voice. Send it to Slack for approval. And then once it's approved or something like that, uh, create a draft uh that you can review. uh if you like it you can go and publish it on your website. So to do that that manually to find the keywords to organize them to research them to create you know optimize briefs for each one and then actually write or have an LLM create uh that content would would take you a long time and uh this can really free you up from doing a lot of manual work that otherwise you may have been

11:43doing in the past. So a few tactics um you know gum loop in learn how to use these AI and agent nodes you can essentially arm uh an AI with various tools that it can use to help complete your task to get relevant information and input and uh do it uh in a in a highly efficient way. um using these AI nodes also reduces the amount of steps and complications that you might have uh in your workflow otherwise. So uh really in my view one of the main benefits of this workflow automation stage is you start to understand the logic of how these tools work together, how data

12:30flows, u you know how to automate things. you can really figure out what might be good for an automation or a tool or something like that. And eventually uh you start to move to level three once you see some of the inefficiencies of workflow automation in level two. So um level three is when you really unlock vibe coding. Um so most people with vibe coding they start with replet bolt lovable vzero something like that. I'm a Replet fan for uh for beginners. Um it handles a lot of vibe coding uh things for you that otherwise you would have to go and learn uh and and set up if you're kind of unfamiliar

13:14with like development and getting a development environment going. Um you can deploy it handles a lot of security stuff for you. Um you can easily set up a database. Uh store your API keys, things of that nature. Um, now I saw this quote from Naval that really got my head spinning uh the other day and this is one of the reasons why I think level three is so so important. So he said that AI is a form of leverage. Leverage increases the returns to those who use it. Software engineers are gaining leverage relative to everyone else. And uh that's because vibe coding is the most mature agentic use of AI available

14:03right now. So if you think about um you know being able to prompt and get something to completion that can have an impact on your business that's really uh what vibe coding in my view uh is all about and where a lot of the opportunity uh lies in it. So after you, you know, start using Replid and Bolt and things like that, eventually you start to see the power of, you know, these these coding agents and to truly unlock those, eventually you will move into cursor and set up Claude code. So I'm in Cursor

14:39right now. I just have a notepad open. Cursor is really an interface for a lot more than just building and coding. Yes, there's a coding agent inside, but you can also work with your LLMs. you can call the MCPs that we talked about prior. Um, and you can have one place, one interface to do your research and to put that insight into action and actually deploy it into something that can directly impact your marketing efforts uh, essentially right away. So from from that point of view in terms of using AI to really unlock leverage in my opinion true Vibe marketing is much closer to vibe coding than workflow

15:22automation. Um so uh that that's a kind of an insight that I've been having after going deep into into cursor and cloud code over the last five or six weeks. Um, you know, when you're in this stage, two important things that are, you know, thematically similar to each level. Um, if you know what questions to ask, uh, the agent or the LLM, uh, you can essentially solve any any kind of problem. And then you have to know what good looks like, you have to know what you're trying to get to. You have to be able to provide kind of a goal to uh, to this agent to give it additional context

16:02uh, so it knows how to hit your goal. Now, you're going to have a lot of bugs and a lot of errors no matter what vibe coding platform you use. There's no true kind of oneshot ability out there to, you know, throw in a prompt and and get something to the finish line unless it's a very basic like landing page or something like that. As you start interacting with other services like, you know, databases and uh AI APIs and and things like that, you're going to run into errors. Uh that's why it's very important to understand the logic. Um it's an uphill battle at first, maybe

16:36for the first like week or two, and then uh you start to gain a lot of momentum uh as you unlock these different aspects of of vibe coding and you begin to understand like how it all kind of works together, which I think is uh is really a huge uh learning in terms of how you can get the most out of AI. Um when you start to use claude code uh you'll start to see the power of agents and where a lot of this is heading. Cloud code lives in your terminal. Uh you know it has access to um all the files that you're pointing it to uh in your file directory

17:15like the different code files that you're working on for a project. Has access to the files on your computer. uh you can give it prompts and it can come up with a plan and you can watch it work in real time to solve these different problems and things like that that you know uh that it might be facing along the way. So um this is this is vibing and this is working with AI and this is what agents are like. Workflows are not that. Um there's some tools that are coming out that allow you to prompt to workflows and things. But in my view like having uh cloud code in the

17:48terminal that has full understanding and context of all the different steps within uh a workflow or a project or something that you're trying to accomplish is way more powerful in terms of you know making updates uh adding features gaining extensibility of this tool or workflow or whatever it is that you're that you're building or working on. Um, so just an example of this insight toaction flow compared to, you know, both of the, uh, the prior levels that we talked about. Um, just to remind you, uh, we talked about sort of some prompting here to get content ideas and then you'd have to go back and maybe

18:27execute it in a different LLM or something like that. Uh, we talked about chaining that together in level two with workflow automation. uh still manual work to go in set these up um optimize it uh keep it working uh things like that uh but in this level it's more like hey use the data for SEO MCP in cursor to find programmatic keyword ideas which are you know bulk keywords that are relevant to your brand that people might be searching for. So like someone might be searching for Chhat GPT verse Claude, ChachiPT verse uh Gemini, you know, things like that. Uh semantically similar keywords, uh where there's a lot

19:08of them. Um so go and find those relevant to my brand uh with the data for SEO MCP. Then use a perplexity MCP to research the best programmatic page templates. So what are the best companies doing out there when it comes to designing these pages? You might look at like Zapier or something like that as a as a great example of what of what good looks like. Extract a brand guide from my website. After that, then I want you to create a PRD, a road map that will allow us to create 25 web pages on our site focused on these programmatic

19:43longtail search keywords. Staying as close to our brand voice and design as possible. Our goal is organic traffic, then converting those visitors to a free trial of our tool. Uh you're going to review the PRD. If it looks good, guess what? You can go pass it right into Cloud Code. Um you know, you can you can do that in one interface

20:05right here. Claude Code can get to work. It can build the pages. Uh it can optimize the pages. Um, you know, from there you can go back and have it look for any kind of like technical improvements in terms of speed or title tags or, you know, adding other information that say LLMs would would

20:25like to pick up or something like that. So, it's amazing to me that you can do the research, you can build the thing, you can deploy the thing, you can scale the thing all from here. Um, all from one place. uh the the prior levels, you simply can't do that. You're you're you're limited in terms of the

20:47strategies that you can go and execute. Fewer people are doing this also, which means there's more alpha in it. Um you can uh go deeper. Uh you can go faster. You can have more velocity than your competitors if you figure out how to do this. There's so many things that you can do in level three. So you can go and uh scale content production on your website like I just mentioned. You can uh automate like AB tests. You can feed data back into uh the loop here to improve on your strategies. You could create micro tools as lead magnets. You could create a quiz. You can create new

21:30offers. You can create internal tools for your team that go and plug into, you know, workflow builders if you want to, you know, send out notifications or have a human in the loop review process or something like that. It's very difficult to do that type of work uh in level one and level two. So, it really opens up a whole new uh area for you to uh for you to dig into. So insight to action, one interface, way more that you can do. Uh no franken flows. You're not trying to retrofit workflows as kind of SAS or uh software uh use cases. I I don't think

22:11that's optimal. So recommended tactics to get started here, sign up for something like Replet or Bolt, build a basic landing page or a simple tool. understand how to integrate an API like let's just say the open AI API into your tool or whatever it is. Um layer on authentication and a database. Understand how that works. Learn basics of security when it comes to vibe coding. It's very important. Start to play around with things likema uh to to take a Figma design and sort of match your vibecoded output as close as you

22:47can. Uh there's a steep learning curve. I I'll be honest with you, like at first I was getting frustrated. It took me some time to figure all this out, but once I did, I feel so much more empowered and uh to me like workflows feel kind of like uh not quite the stone age, but uh but you get the point. Um so yeah, it's it's it's worth the initial pain for the long-term gain. And I think you're setting yourself up for true leverage. again, uh, software engineers are gaining leverage relative to everyone else because they truly have agents that can go and do the work. So,

23:26let's figure out how to use those for different niche uh, use cases, different verticals, different different work uh, that we have to do outside of software engineering. Again, true Vibe Marketing, it's way closer to Vibe coding and workflow automation. the closest thing to Vibe Marketing uh superpowers out there right now in my opinion is Claude Code. So what's coming next? What's coming after uh after uh you know sort of level three. So level three still requires a lot of friction. You still have to know what questions to ask. You still have to know what good looks like. You have to set up a development environment. You

24:06have to do the research and prompt for it and know what you're looking to build. You have to be able to understand the logic and help AI to identify where to focus its its context uh on solving different issues, bugs, problems, uh things like that. Um I believe that, you know, this is currently underway. We're we're seeing real agents uh being developed. Um, and basically what I mean by that is that level four is where real agents are going to be able to plan, act, and optimize autonomously. So, in a marketing use case, it might be like these agents are out there. They're uh crawling your

24:51competitors, you know, every single day, sifting through hundreds of different opportunities across a multitude of channels, uh identifying which ones could have the biggest impact on your business, and then going and creating those assets for you in your brand voice, sharing them with you, or if you're okay with it, just going ahead and, you know, publishing those or putting those live into the market. Um the big opportunity there is having a data feedback loop. So once these things happen, how does the AI get smarter over

25:24time based on how these are performing? Um this is all technically possible. Marketing's been behind development isn't even here yet, but as AI improves, a new form of leverage is going to be available to us where we're going to be able to set a goal. The AI does the heavy lifting. And uh I've been predicting this uh since February that this is going to happen. Uh I I think I said within the next 12 to 18 months from there. Um you know I'm building in this space. I think this is one of the biggest opportunities out there period. Um so it's a it's going to be closer to vibe coding. Uh

26:03but the future is going to look um like uh less onus on the end user, less dependence on the end user. That means that there's going to be mass democratization uh in terms of the strategies that everyone is going to be able to go and do and execute and scale. There's a lot of implications to that that need to be thought through. Um you know taste, yes, that's still important. Um talking to customers, understanding the problems and where to focus all of this power is kind of really what uh what you have to be doing. um you know distribution and and data and all that is is very

26:42important too but uh this is going to definitely upend some industries and uh there's going to be less uh advantage to some of the more traditional like marketing tactics and things are really going to really going to change uh as everyone is able to do them more and more. So uh another just idea to kind of paint the picture of where all this is headed. So, you know, you'll have

27:06specialized agents for different roles. This will help with context management. So, you'll have, you know, one agent focused solely on going and digging deep into competitors and researching them and understanding, you know, what kind of opportunities that you can go and do for your brand. You'll have another one that's focused on, you know, creating one certain type of asset. It's going to be really good at that. You're going to have another one that, you know, is really good at kind of understanding data and uh looking at trends and things of that nature. And then you're going to have a management agent, a manager agent. Uh that is the layer above those,

27:39which is the one that you're going to communicate with. It's not going to be it'll be kind of like working with a team. You know, you're going to have a direct report. You're going to work with them. They're going to go and manage all the all the other stuff. Um and you're going to be able to kind of stay out of the weeds in a sense. So, the tech isn't quite there yet, generally speaking. Uh but it's coming soon. the uh the tools, the piping, the logic for these multi- aent uh use cases are coming. AI is continuing to get smarter and smarter every single day. Uh this is where it's

28:14all going. And uh I'll just say that u that it's coming soon without giving away uh too much more information. So I hope this gives you an idea kind of what Vive marketing is all about. um the progression of what a Vibe marketer looks like from just using AI to starting to chain those together to really becoming a full-blown like Vibe coder and orchestrator of the closest thing to you know real agents that we have today. Um and then you know painting the picture in terms of like where all this is headed next. So it doesn't matter where you're at uh you

28:53know in terms of the levels right now. Uh keep working at it. Uh, I think level three is worth the pain for the long-term gain. Again, people like Naval are saying, "This is where the biggest leverage is. So, why don't you figure out how to um integrate that into your work and uh and take advantage of of kind of the latest uh technology?" So, that's all I got. Let me know what you think. What level are you at? What challenges are you facing? Where do you think all this is headed? I'd love to

29:19hear your thoughts. Until next time.