Adamma Ihemeson on Building Autonomous AI Systems You Can Trust

Speaker A: Foreign.

Speaker B: I'm your host, Carrie Gard. If you're new here, uh, I'm the CEO of MKG Marketing, where we help B2B tech marketing leaders grow their search visibility to build real pipeline. We're officially back from summer, like I said, and every Thursday we will be right here having conversations like these, um, until Cyber Marketing Con, where I'll be headed to Austin, um, here on November 4th 14th, to go hang out in person. So if you are going to Austin, hit me up. I would love, love, love to connect with you all in person. It's gonna be magical. But today, today I'm talking about a different community. Marketing Women shout out to our gals, Gabriella Delvey and Natalie Contave. Um, they were previous guests here on, uh, the podcast many moons ago talking about community and they built one, an amazing community called Marketing Women. I was like, I have to fill the calendar for the with guests who's going to be on my show. And I was like, oh, I know where I'm going to get some amazing, amazing, amazing people. I'm going to go hang out in Marketing Women and see who will accept the call. And, oh, man, did they show up for me. Um, this is the first conversation of many more. My entire Q4 is practically most, um, people from that community. So thank you, thank you to the Marketing Women community. Thank you to those who accepted the call, and thank you to Gabrielle, uh, and Natalie, who created the community to begin with. Today's conversation is with Adama Ahemisen and I actually, as I mentioned, I met her through Marketing Women and, um, I'm just so excited for this conversation. She is a founder and CEO of the Automated CMO where she helps marketing stop reacting to whatever tool is trending and start becoming the architect of an agentic AI system that actually thinks, plans and acts for them. Adama, welcome to the show.

Speaker A: Thanks for having me. I'm really excited for this conversation today.

Speaker B: Oh, uh, me too. Me too. And you had your own summer, uh, hiatus.

Speaker A: Yeah. And m went on a ten day silent meditation retreat. Um, I got back, what, Sunday? Um, and yeah, it is exactly what you hear. Um, I'm literally not talking for 10 days. I am completely disconnected and I'm meditating like eight to nine hours a day. Um, a great reset, an, um, awesome practice. Um, it's definitely a lot, but it's like the fourth time that I've done it. Um, I went to Barcelona actually, to volunteer at a center for 20, uh, six days actually. So, um, yeah, no, it's. This has Been a really great reset and it's nice to kind of jump back into things.

Speaker B: Well, you'll have to share some resources with us for anybody who needs that disconnect. We are on screens what feels like pretty much from this time we wake up to the time we go to bed. And so who wouldn't love a minute to not do that for 10 days? That sounds pretty magical. Um, yeah, so you're going to have to drop some resources for us. But today we're here to talk about, um, as I said, agentic AI and how you're using it. I think it's the big elephant in the room, so to speak, of how um, any of us are trying to figure this out. I feel like when you think about LLMs, um, and AI, I think we all jump mostly to writing, which is what is probably the wor. The worst at.

Speaker A: Yeah.

Speaker B: So I am super pumped for this conversation because we're not going to talk about that. We're going to talk about how it can help you in so many other ways. Quick warm up before we get started. For anyone meeting you for the first time today, what's the short version of what you do and how you and who you do it for?

Speaker A: My elevated pitch is that I teach marketers how to use AI as an assistant, not a replacement. Um, so really understanding exactly how to actually systematize AI within your marketing workflows, um, and teaching especially non technical people exactly how to go from zero to a production ready system workflow or AI app.

Speaker B: Oh my gosh. Well, you're going to teach us all about it today. It's going to be great. Um, you've made the move from one off prompts to low code AI apps that actually execute strategy on their own. What's the real difference between using AI and building something agentic?

Speaker A: Yeah, so the biggest difference, honestly when we talk about something that's agentic, it's actually kind of like working from some sort of knowledge base or memory. Um, so you can really kind of like go into utilizing rag. So if you use a tool such as Mind Studio, you might have that RAG might be your knowledge base of like background information or knowledge about your business, about your work. Um, when you're using like LLM, um, sometimes it's kind of just like a back and forth conversation. Um, so you might use it to like ask questions or um, and then if you want to take it a step further, what happens is like if you want to actually make it agentic, it's going to like work in a loop or it's going to be autonomous. Um, so it doesn't necessarily need your manual input all the time. Um, especially if you set up those specific triggers within the process in order to execute uh, at either a certain time or ah, depending on a specific trigger. Um, so that's really the biggest difference when you come and talk about something that's actually agentic is that it's more autonomous and it has background information where it's able, able to make decisions on its own based off of like the responsibilities that you're giving it.

Speaker B: Not gonna lie. Sounds a little scary.

Speaker A: Yeah, yeah, it can definitely be very, very scary. That's like a whole conversation when it comes to like governance ethics, um, AI safety which me personally I feel like that hasn't been put at the forefront. Um, but I feel like it should be. Um, and actually I'm going to talk about um, one of my processes. I have a five step phase when it comes to my build. Um, and one of the biggest things and key things that I've implemented is making sure that I'm going through like an AI security check for everything that I'm building. Um, so I definitely implore people to make sure to one always still incorporate human in the loop. Um, but then also, also being within the lead because like you should be the one that is the decision maker and the core decision maker when it comes to the processes and the things that you're building with AI. Uh, but it can definitely be scary. There has been actually some things that have come out this week. Honestly that has been very concerning to me as well. Um, but I do think that AI

Speaker B: agents colluding behind the scenes of a Chrome browsers. That one.

Speaker A: Yeah, that and an AI researcher at ah, Anthropic recently resigned and then like I saw that too but they're saying that like both OpenAI and Anthropic are being incredibly irresponsible with this like, with this AI rat race. Um, and that the real threat of uh, to humanity in like the next decade which is. Oh that's, that's very concerning. Um, so that is again that's kind of like pulls me into just like the fact of like needing regulations. To be honest, I think that the lack of regulations has been in order to kind of give like this free range and in the sake of or the name of like innovation and like this, this race um, for who can get to like super intelligence. But I do think that that's honestly incredibly irresponsible. Um, and I think that more people are having more of A conversation about, like, what the risk actually looks like. Um, so I'm glad that the conversation is starting to happen because, like, what realistically happened is that AI has been around for a while, but what happened in 2021, 2022, is that it became generally accessible to the general public is what happened. Um, so, yes, like, props to like, innovation and it has changed significantly in the last five years. Um, I would know because, like, I've been using it since 2021. Um, but I think that it's definitely high time that we start having like, serious conversations about like, what those parameters and what those regulations actually look like.

Speaker B: Uh, yeah. So how do you help people in terms of making that leap from using it just as like a prompt, uh, chat buddy, essentially, to being something more autonomous and navigating their fears around like, for how I feel of, like, I would love it to do stuff for me, but I'm not sure that I trust. Trust it.

Speaker A: Yeah. So the way that I usually recommend people to like, go about from going from like one off prompts to actually building systems is I would actually do an audit, um, of how you use your LLMs right now. Um, and then again I use it as an assistant. So I would actually ask it to kind of go back and forth as to like, what does my process look like right now? And having a conversation about like what your either manual process is or how you use the LLM in order to get from point A to point B and ask it to actually create a playbook for you. Um, so it is again, you have like your own accounts. Um, and it's something I've like definitely put into practice recently is that I asked it to like go over the entire chat session. Um, because sometimes it like it goes like, you know, this back and forth conversation, um, and it's helping with, helping with the execution side. And I ask it to either like recreate it into a skill or go back and then create it into a playbook where I can actually start systematizing it. Um, and in order to actually build a system, you do have to like, be knowledgeable or be aware of exactly like what type of tools you're specifically like using. Like, everyone's systems don't look the same because everyone has like, they are not agnostic to specific tools. Right. Some people use HubSpot, they might have Mixpanel within their tech stack. Um, they like, there are just like so many other tools that like, that are available for people that people are using within their marketing. Um, so what I do in order to like utilize that and create that playbook. It helps with actually being able to like actually building those systems, um, step by step, basically based off of exactly what your tech stack is. Um, and then reiterating it and making sure that the outcome is to the desired outcome that you're specifically looking at. Um, and then another thing also that I think that is really, really great. Like just honestly common practice in my opinion is that you want to always have oversight when it comes to AI. Like I don't believe in just like blindly trusting it. Um, so definitely like if you use Slack, there's definitely a way that for you to like get notifications when it's done with like whether it's like creating a draft or whether the results of that specific system has been executed and you're able to see exactly like what the actual outcome looks like. Um, or if it runs into an error and then like again you can create some sort of like Slack agent that is again having conversation like based off of a specific system that you're using that it can tell you about those, the specific error notify you if there's a specific error that is happening within the execution of that system. Um, so those are just like a couple of things that I usually recommend in terms of like actually starting to like systematize like what AI within your work. Um, if you don't want to do a playbook and that might be overwhelming, like say again, um, you have like too, too many contexts in like a chat. Um, then I would try breaking it down actually. Um, this is another thing that I've done um, and that has actually helped me with building, um, is that I have broken down in phases, um, and then actually designing what the architecture actually looks like. Um, so in this phase we're focusing on these specific tasks and this specific responsibilities and then we're moving on to that next phase. Um, so that is can kind of like gave a little bit of a breather. Um, when it comes to being able to like systematize it and like instead of being very confused and trying to like go like blindly having the elephant. Yeah, it's great to kind of just like break it down into phases and reiterating on that those phases as you, as you go. Um, so again I use it, again I really use it as an assistant. Like it is literally like my, it is my co builder. Um, so like I'm going back and forth when it comes to troubleshooting. Um, but it already understands what its general responsibility actually is.

Speaker B: Yeah, I, um, I have like certain Skills that I haven't run. Um, but I. I'm always that human in the loop right where I have. I, like, look over it. My biggest problem with AI right now, I don't know if you're running into this too. I feel like it's shifted a bit recently, or maybe just the way I'm using it has shifted. And so I'm getting. I feel like I don't always understand it. It's so verbose and kind of overkill that I actually have to, like, tell it to simplify. I don't want to say dumb it down, but like, like, why are you trying so hard? Like, can we just have a conversation? Have you noticed that shift? Because I feel like that's where I get overwhelmed. I'm like, I don't even know how to tell you. Like, if you're asking it to build a playbook, then to sit there and have to read it and understand it and know what steps it's going to do. I almost feel like people, myself included, get overwhelmed and just like, just run it. I don't even care. Just go do it. Which feels like the wrong thing to ask of it. But, like, I feel like I sometimes have that moment where I'm like, I. I can't sit here and read all this. I just. I, uh. How are you navigating that in terms of. You said to break it down, which I think is smart. Um, but how, like, just the man. Does it like to talk a lot?

Speaker A: Yeah, no, it does. It can be. It's a yapper, honestly. Like me, to be honest. Um, and this kind of goes back again. A lot of my core concepts, like, and my practices is really a step by step. And honestly, that has a lot to do with the context that you're giving it. So, like, giving it context in terms of, like, assume I'm like a complete beginner. Like, I don't know anything about this. And even if you want to get even as, like, speak to me in a sixth grade, like, reading level, like, speak to me in a very simple way. So giving it as much context as possible. Possible. Even when it comes to like, say you have like a project on Claude, you can give it like, um, you can give it actual instructions on like, common, like, general practices that it should follow when it comes to giving their, like, their specific output or their specific, like, how they're having a conversation with you. So context is honestly going to be like your best friend when it comes to like, trying to reduce, like, how overwhelming things are. Could be. Um, and Then also if it is giving you like very complex and just like too, too much like information at once. Um, again I go back to like, again talking to like slow down, take a step back. Like let's pause for a second. Um, and then like just like giving me either maybe just like the core things, the key things that I need to know that is going to be applicable to me depending on exactly what my role is when it comes to how I'm using AI. Am I using AI in terms of on a strategic front? So then it's going to give you kind of like strategies and recommendations on um, that angle or are we talking about it more on the execution aspect? So what are the specific tools? How can I actually build specific workflows and making sure that it's properly being able to produce exact outcomes that I am looking for. Um, so definitely making sure that you're giving it as much context as, as possible. Um, whether it's like within that specific chat and giving you instructions or it is again and I have this even with my skills, um, because like I train it in a way that I specifically work. And like the way that you train AI, like it really is the best way to train it is like you would know from how you actually function, like how you actually learn how you receive information. So the way that I've like trained some of my skills and some of like my systems is you have to give it to me step by step. Like I cannot like you need to break it down as much as possible because I need to focus on one thing at a time. I can't be able like it's not possible for me to execute all of this information all at once. That's very overwhelming. So let's like just like go back like and like just like break it down. Um, sometimes they may have like a standard on like what are, like what like step one is this specific thing, step two is this specific thing. But that still helps at the end because again you have one focus, one focal, like, like one focal point that you're actually looking at. Uh, and like you're trying to execute or you're trying to strategize on when it comes to using AI. And I, for me I think that that has been a huge changing game when it comes for, when it comes to me for using AI versus just like creating these like high level prompts and then like having it go kind of like free for all in terms of like what the actual output looks like.

Speaker B: Yeah, I think the, the context is definitely key and how Much information you give it. I think it's a balancing act, um, which I'll ask about in a second. Um, and the, the rules, like the. It sounds like you're giving it specific upfront rules of how you want to like rules of engagement almost in terms of do you do that on the global level or do you do that on a project by project?

Speaker A: This global. I do it on a global. Especially when I'm building it. Like I literally have a phase and a process, um, and is entirely on a global level. Like I practice a step by step process, um, even with like my. Especially with my builds. Honestly. Um, just because like I and each project can be different. Um, like when I go into like starting to like build a new project, um, or a new app, uh, what I usually do is like I already have kind of like a vision or kind of like an idea, uh, as like how I want it to be and how I want like the end goal to be. But then what that does my process actually helps with like breaking that down and architecting it step by step. Again, like, it's really like kind of like building blocks, right? Like you do need to like have to. You can't really build something without like a, like a strong foundation. And those actual like that guidance, those like step by steps. Those are my, like that's my foundation. Those are my building blocks that is going to help me continue to go up, um, and move throughout the process, my build process. So that's just like something that I, again for. For me especially and I think that it would help for other people um, to like really just like go back and just like kind of like break down things like on a step by step basis. Because again, AI can be very overwhelming with like how much information it can actually give you. Um, and again the thing that's also great, especially the fact that I have it on like uh, it's on a global front for me, is that I can reiterate it as I'm going through each phase, which I reiterate it all the time. Like I may have a skill, it's working a specific way. I've been using it for a while and then I'm reiterating on like things that I'm noticing that it's not picking up on and I needed to pick up on. So like, maybe there are common questions that I'm like troubleshooting all the time and I'm like, okay, you have to realize that like let's add this like within the context and the knowledge base so that you understand that, like, this is something that I always need help with. Um, and I need you to always be prepared to give me that information as well. So on a global front and just, like, breaking it down, I think that that helps so much more. Um, and it's a lot less overwhelming. I would even say that's even less overwhelming, especially when it comes to, like, no coding tools, honestly. Right. Like, and I think this is, again, this is kind of, like a little bit based off, like, my own grievance with, like, the messaging when it comes to AI is like, build an app in 10 minutes, build an app in 30 minutes. I can't stand it. I'm sorry.

Speaker B: That's a terrible idea.

Speaker A: I think it's ridiculous. No, it's actually ridiculous. So, like, having, like, within, like, actually having those phases and being able to, like, one catch when there are errors that happen within that specific, specific step or specific phase, rather than, like, going over and then finding that error way down the line and having no idea or no direction as to, like, how it happened and how to actually fix it. So that I do think that it's just, like, it's so much more digestible and it's just, like, so much more at ease with your. Again, using it at your own pace. And like, I. I mean, I feel like, again, I. My brain, it works a million miles a second. Like, uh, it is like, it. It is. Obviously, sometimes it's exhausting for myself as well, but the way that I've honestly trained it, it's like, at this point, point especially, is really. It's been something that has been at my own pace, and that's based on, like, how I specifically learn, how I specifically function. Um, so I think it's very important. Like, I think people really do need to understand, like, how do they learn? How do they, like, receive and retain information? This is honestly kind of like when I notice, like, oh, I think I'm incorporating learning sciences within the work that I'm doing now. So I think that makes sense, a very core thing that people should also start incorporating and, like, understanding. Because, again, people learn in different ways. They receive information in different ways. They might be more visual learners. So, like, maybe you want it to give you, like, to search, like, a YouTube tutorial or demo, and then it helps you in that way, like, by giving, like, showing you in the right direction. Or maybe you're more so of a reader. Again, I have been more so of a reader. So, like, I ask it to, like, write out the instructions for me very clearly. Step by step. So it really depends on like, what type of learner are you? How do you actually retain and execute information that is going to actually really help when it comes to how you're utilizing AI effectively?

Speaker B: Uh, that's such a. Such a great idea. Um, my head's going. Because I'm trying to think, like, I need to go do this. There's so many instructions that I give it that I'm a not giving it that I really should, that I forget that I can. Like, to your point, in terms of how I learn, I'm more of an auditory, visual learner. So sitting there and reading those blocks of text totally exhausts me. But if we can chunk it down and I can get bullet points and I can skim a little easier, like, even that would be a huge help. And then if it throws in some YouTube videos or some things like that, I could totally see that being, um. Yeah, I need to go tell it to help me out with that. And as well as the reading style, like, I think it's. I appreciate it. That thinks I'm a very smart person and I can read at such a high level and I can, but it's not something I can do all day. So if you can bring it down to like sixth grade level, I think that'd be great. So I'm gonna go make that happen. I also think the steps you do on a regular basis, right? If you're. You mentioned this, right? If you are repeating the same building blocks over and over again, like, how do you give it those instructions as well to say every time you start something, this is the order of operations in which it needs to roll? I, um, think is. Is so incredibly helpful and cool too, and not something. Again. I keep repeating myself. I'm like, why am I repeating myself? I can go give it global instructions. This is silly. Such, um, a great reminder. You, um, man, you've had quite the career, Adama. You reset your career and shipped eight apps in six weeks. I thought it was a miracle that I built a website in six weeks, whether alone, you know, eight of them. Um, you know, you reset your career because you felt burned out. What burned you out enough to make that reset necessary?

Speaker A: I mean, shifting and changing all the time, like, isn't it? Like, it is honestly unsustainable and overwhelming. And I. The thing that really got me burnt out is the fact that we're one. And I, again, this is maybe just my personal grievance is the fact of, like releasing like a new model like every couple months. We like Relax and, um, allow us, like, digest and test out this model, like, for a little bit longer instead of, like, releasing a model, like, every, like, it started honestly, like a new bottle. It was like, okay, every year, and then it was like, every six months. And then it was like, every, like two, three months. I was just like, that this is not sustainable. Like, this is like, I'm trying to catch up on, like, a new model, but I'm still stuck on the utilizing specific model. And also, like, I was also subscribed to, like, a lot of, like, AI newsletters, which I. I'm sorry. I think some of them are great. I think that some of them are really, like, very useful. Others of them, I feel like. I think that they're actually doing a lot more harm than good, to be honest, because I don't need to know every single thing that is happening within AI. I'm sorry. Like, this is like, that's. And that's why I always tell people to, like, it's really. Again, AI is very overwhelming. If you really want to start utilizing it, then what I would really focus on, like, is what your domain expertise is and then really identify exactly, like, what you actually want to delegate to AI versus, like, what you want to actually focus and invest in within your own skill set. So, like, for me, yeah, it was just like. It was the, uh, new models, like, constantly being, like, brought up. It was the, like, the news, like, the new things, like, coming out when it comes to AI, it was like all honestly. It was also like, other people's information again. Like, there are so many different ways that AI is used. And you see people, like, sharing, like, their prompts. You see people sharing, like, oh, they built this specific agents and things like that. And it was just like, okay, I. I'm receiving this information, absorbing this information, but then now I'm getting into, like, analysis paralysis. It's like, okay, am I using AI correctly? Or like, oh, this is something that I can, like, improve upon of, uh, like, what I'm actually, like, how I'm actually using it, or it was just like a, uh, combination of all of those things. And it honestly got to a point where I was just like, I don't even know if I trust my own knowledge and my own expertise, and I've been using it for five years. Like, that's crazy. So I, I really just. I was like, I. This is too much. Like, I need. I need to take a step back. Um, and taking a step back, especially when it comes to AI, that is scary in itself because it changes so quickly. Um, but then what helps is that when I did, like, and I took a sabbatical, like, I traveled Europe for, like, three months. Um, and it was so refreshing. Like, it was. Oh, my God. Like, honestly, I think that was, like, probably one of the happiest versions of me I've seen in a long time. Uh, yeah.

Speaker B: Where'd you go? I'm going to Derailey for a second. Where'd you go in Europe?

Speaker A: Yeah. So I went to Barcelona, then I went to Lisbon, then I went to Amsterdam, then I went to Vienna, then I went to Berlin, then I went to Stockholm, then I went to Helsinki. I got really sick in Helsinki. Um, and then I went to Copenhagen, back to Amsterdam, back to Lisbon, and then I ended off volunteering at a, uh, meditation retreat for. It was supposed to be 10 days, and it ended up being 26 days. So three months. Yeah, it was great. It was wonderful. Absolutely loved it. Want to do it again?

Speaker B: Honestly, like, what, a reset. Yes. After. After all that, I. I can't agree more in terms of. Of just how much AI news and information is out there. My husband is an engineer and AI user, and he is constantly talking about, oh, there's a new model today. Oh, there's a new model today. How are you? I mean, I don't. I use Claude, and I sometimes use Gemini. I sometimes bounce between the models. Like, I just tested Fable for the very first time and was like, okay, I want you to analyze this website from an SEO perspective and tell me what opportunities I'm missing. And I gave it a ton of context and history around the website and the product and everything. And it came back with a pretty detailed plan. I was like, all right, we gotta. I'm not pulling the carpet out from underneath this thing. And I made it write the instructions for Sonnet. I was like, sonnet's gonna do the work, so, you know, but I want you to do the deep research. And then we organized it. I was like, I can't do all of this at once. I will blow up the algorithms, and it will. My website will die on a vine and have to read. I will upset the algorithm gods. Like, we need to chunk this out. And so then it built a timeline, and then we delivered it to Sonnet, and now Sonnet's doing one piece of it at a time with me overlooking every step of the way because I'm so terrified to disrupt the SEO Gods. Um, but how are you? Is. What models are you? My husband's testing everything from Deep Seek to, um, Ones I've never heard of to fable. So are you doing the same? Are you sticking to one platform? How are you. When you say models, what are you?

Speaker A: Yeah, so I am sticking to one now actually. Um, I used to be the person that was bouncing back, back and forth, so I first used to use Chachi pt. Um, and then I started like playing around with Claude a little bit. Um, I would bounce to Gemini especially like when it comes to, to like whenever I like ran out of usage on cloud. I'm like, I'm gonna go to Gemini. I don't. I need Gemini right now. Um, but I exclusively use Claude honestly. Um, and I've like, at times I would. I honestly, I primarily use Sonnet. I think that's the Sonnet model, honestly, is, is really good. I enjoy the Sonnet model. Sometimes I'm using Opus, but sometimes it depends on like, okay, do I want to like run out of my toilet tokens?

Speaker B: Because yeah, it goes fast.

Speaker A: It goes fast, exactly. Like, it, like the higher the model, the more tokens it's using. Um, so I think it really kind of like depends on like again, like your own individual use. Um, but I, I've pleasantly loved the direction that Claude has been going in, honestly, over the last couple years. Um, I mean, like, they are the ones like they. So ChatGPT had custom GPTs, which they have recently retired actually. Um, and then Claude had artifacts. Um, and I just thought cloud artifacts, they've had more of like a visual component when it comes to like building like a product. Um, and I think that was like the. Their first kind of like step and dip into like the vibe Coding industry was artifacts in my opinion, um, versus like custom GPTs. I feel like they were kind of just like personalized chat bots in a sense, in my personal opinion. Um, I think that like, again, I use projects very, very heavily on clawed. I love projects. It's excellent, honestly, where like, again it's so great just because like, again my. All of my sessions are housed under that specific project. So I absolutely love that. I um, love Skills. Honestly, I've been. Skills is spectacular. I think it's excellent. Um, and it's like the, the first thing that I kind of like started to like kind of get acclimated with when I got back. Um, and I think that is like, I think that Claude has done a really great job of really being designed for not just enterprises, but also like small and medium businesses or even solopreneurs, like people who are like very business minded or trying to utilize it within their business. I think that cloud has really done a great job positioning itself in that front. Um, so yeah, I've exclusively been using Claude also. I can't stand the conversations that ah, I just have like, I have like very, very negative feelings honestly when it comes to ChatGPT and like the conversations that are being had when it comes to like the future of AI. Um, what like that opening. I, it's. I'm not having a very like, it doesn't make me feel good and I just like, again I probably just like have like a more morality complex where I'm just like I, I could barely. I can probably count the amount of times I've gone on chat GPT this year, honestly. Um, so yeah, Claude has like, I, I enjoy their models. I really like their models. They, they've been able to, I've been able to execute a lot like to be actually be able to build so much um, with Claude. Um, so I've honestly really stuck with it. Um, and honestly once I started using cloud code that's when I was like, okay, yeah, Claude, yeah, I'm not leaving Claude anymore.

Speaker B: Well, let's get into, yeah, let's get into that for a second because you built eight apps in six weeks. So were you, were you using Claude at the time? Walk me through how you, how you did that.

Speaker A: Yeah, so I hadn't been using cloud code yet. So like I had heard about cloud code I think like back in 2025, um, and. But it was like exclusively on like you had to use it via Terminal and I'm not familiar at all. Like I was not familiar with Terminal.

Speaker B: I was not there. I was not ready for that yet either. Honestly.

Speaker A: I even like, I tried to like watch like a tutorial as well and I was just like I'm not following through with this. Like this doesn't make any sense for, for me at all. Um, so then what I was doing, I actually started with creating a skill first. Again this is kind of going back to like how I started like transitioning into more of like a step by step process. Um, and I already knew about uh, PRDs, so product requirement document, ah, documentation, like I already knew about that so I started with that. Um, I had an idea. I like created a PRD assistant as a skill that was like responsible for actually creating like um, answering like what the app is about, who the target audience is, like what are, what's the mvp, what is like the over scope, um, um, when it comes to the actual build. And um, this is kind of like Going into like the pretty much like the entirety of like what the architecture of the actual app that I'm actually building. Um, and then the thing that I added, um, that again based off like again how I learn and how I work is that I was like I need to be able to build this on a step by step process, like within a phase. So then I ended up building a code copilot, um, code build copilot skill. And what that does is that after it use, after I build the prd, it uses that PRD and it's giving me like instructions and like custom prompts I'm putting into cloud code step by step. And then it's not even just like okay, like you copy and paste the like the prompt into cloud code. Um, and then it actually executes. Executes it. Um, what it also does, it comes with a verification checklist. So that verification checklist, that is the checklist that you're going over to make sure that everything is working within that specific step before you actually move forward to the next step. Um, so that is. That was like kind of like just like the. It was those two like that were like my core that I just like kind of like continued to utilize and repeat. Like I literally rinsed and repeated.

Speaker B: What did you build? Yes, uh, what were you building?

Speaker A: So I, one of the ones that I built was skill path. Um, and then that one is um, it purse. Like it helps people look for curated, um like curated and personalized resources for skills that they want to learn. So say they want to use, um, want to learn Python and then it's to ask them a series of questions on like where you're on in your Python journey, like what's the timeline that you want to learn it um, how you specifically learn. So like are you a reader? Do you want like video tutorials? Like do you need a community? Things like that. And then it like what it does, it generates a curated plan and like resources um, for you to actually learn that specific skill. Another one that I built was actually during a build challenge, um, that was hosted via Maven. Um, and that one is pigeon. That one is like a student engagement um, app for like people who are self learners. Um, so they, you buy a self paced course but you don't necessarily have a, like a form of accountability, uh, or you don't have a community to kind of like bounce ideas and to make sure that you're keeping on track. So that one was my first app that I built for two target users. There's one angle for students that are like looking to either maybe you're learning, trying to uh, join a learning group for people who are learning either the same course or the same skill. Um, and then it's like a learning sprint. Um, and then they. You're able to kind of like you're actually able to send like personalized messages like of encouragement or nudges. And it also helps you keep track m of like where you're at within like your actual, like your actual course. And then there's another angle that is more so for the course creators where they can actually have be able to interact and keep track of the students that have bought their courses. Um, and that build challenge, it was a five day build, build challenge. So I built that in five days. And that was I, that was the one that was on definitely one of my most like, like complex ones that I started building. Unfortunately I actually the deadline was UTC time zone and I thought it was like. But it's okay because I was like, okay. I built this really cool thing so like that I have no problem with it. Um, another one that I built was a visa compliance tracker. Um, that one is for people who are traveling and they just like want to make sure that like they're in the country for the amount of time that they're supposed to be at.

Speaker B: Oh my gosh, the calendar. Tracking your days and making.

Speaker A: Yeah, exactly. Right. I was just like looking, I was going off of like just my calendar. I was like doing a guessing game. Honestly when I was traveling I was like, oh man, let me make sure that I, I know I got.

Speaker B: Yeah, Visa math. It's real.

Speaker A: Like we gotta be careful about that. And then, uh, another one that I built, I actually productize basically what it is that I do, um, for the automated cmo. So what that does is that it actually helps you. It creates a step by step module teaching you exactly what type of AI marketing workflow, um, operating system or agent. Based off of your specific tech stack. You select what your tech stack is that you want, want to use in order to build, um, the specific AI asset. And then it goes, takes you through that step by step, like process to be able to go from zero to actually building that AI agent or that workflow, so on and so forth. Um, oh my gosh, what else?

Speaker B: And you go download that one.

Speaker A: Yeah, so that one. Yeah, that one. I've actually been like, like constantly refining that one actually too. Um, yeah, those are like the. I mean I'm forgetting about like a couple oh, Vibe labs. That one was an accident. Um, And I said it's accident because I met someone that was like looking to like, oh, they want to learn how to like, Vibe code and like use cloud code. Um, and I was like, yeah, sure. So I created a notion doc of like resources that I was familiar with to help them learn. And then other people started asking me for that notion doc and then I was like, why don't I just turn this into an app? So that, that basically is like a learning hub for people who want to learn how to Vibe code. Um, and it creates like, ah, like a learning hub of like video tutorials, like skills that I've specifically used as well, things like that. So there um, resources that are available, like when it, when it comes to Vibe coding. And then it also gives you like, you do an assessment and it points you in the right direction as to like, what type of like Vibe coding tools you're supposed to use. So yeah, those, like, uh, those are the like, some of the ones I've built. Again, I. And I have like a list of like, more that I want to build as well. Um, because I was also focused on like building my portfolio as well. So I was like, I. I'm obsessed with this now. So like, why not like continue to build this? To be honest, um, once you get

Speaker B: the bug, you kind of. Yeah. My husband is just constantly, constantly building something new. He can't help himself because it's just once you get one idea, you just rolls. It's amazing everything you've. So you mentioned this. You productize the process for other people essentially before someone builds their first system, what do they actually need? Skill wise, mindset wise. How do they. How should they show up to the table to get going?

Speaker A: Yeah, um, well, one, I would again def. I always go back to like doing an audit. Like, and I say audit as in like, what are things within like your marketing processes and like what you are responsible for that you are okay with delegating or that you honestly hate doing. Um, and then identify what those are versus like, okay, what are the. What is your specific skill set or expertise that you enjoy or that you want to continue building upon? Um, and I think that the thing that helps with that is that it also, it helps with like you kind of preserving like your critical thinking skills with the things that you want to continue executing with. Um, versus again like kind of giving AI the responsibility of being able to actually execute for you. Um, and then I always. The. And this has been my process now is that I always go back to like creating basically an architecture Map. Um, so talking through exactly, like, what my existing process is right now, and then identifying like, what is my existing tech stack or what I need to start implementing in order to build that workflow from zero to one. Um, so building that architecture is going to be very, very important. I actually have a skill that I've been using. Um, it was originally for. It's now an addition to my phases when I'm building, um, with AI, but it's helped with both building apps and workflows. And it's called a Vibe Architect os. And what it does is it basically goes through the process of like, what are you building? Who are you building it for? What are, like, what are the tools that you actually want to use? Like, that you want to use it to, to use to build, um, what is kind of just like the core goal of the workflow or the app that you're actually trying to build and then understanding exactly, like, what are, like, what are some, like, workflows that are going to be agentic, what are like, processes are still going to be manual, um, where do you want to have oversight, um, and have again, be the human in the loop. So kind of like having those conversations and making sure that the architecture of how its functionality is actually going to work, um, is going to be very, very key before you even get into the building aspect. Um, so I always go back to do an audit and then actually architect and map out exactly what your process is and identify, have a conversation with your LLM of choice and identify exactly, like, how you want it to be autonomous and how you want that workflow to actually work for you. Um, so I think that that's kind of like the best, like, the best way to kind of like, go about if, like, you're, instead of like, going in blind, just, like, start with that. And that's kind of going back to why I always tell people to like, lean into their domain expertise, honestly. Um, so that's definitely, again, auditing and architecting is going to really be your best friend because those are going to be your core, core foundations before you even go into building in the first place.

Speaker B: I got to know though, because you've done this so many times, you've worked with so many CMOs. What are some of the first things that you do end up building for folks to help them get started? What are you seeing any trends in terms of the first go? Like, I could see my own, myself being sort of stumped on, Like, I don't even know what I would build or where I begin and obviously the playbook and building the architecture and going through the steps you just mentioned is really important. We should. I'm not telling anybody to skip those steps. I'm just curious as to. Are there some clear go to's that end up happening where people go and build? What do they start building first? Essentially?

Speaker A: Yeah, I mean I feel I've seen a lot of it has been very much focused on content generation, honestly. Um, which I mean I, I really don't blame people when it comes to that. But I do think that also like lead generation and lead qualification has been a very big use case, especially with like making sure to actually measure and like be able to measure exactly like who is going to be a lead versus who isn't going to be a lead, um, and how to be able to have like a more personalized security message when it comes to those leads. So I think lead generation has been a huge, huge demand and huge use case that, that I've seen a lot when it comes to like utilizing AI. Um, there's even one that I, I built internally and I integrated my calendar as well within like the entire process. Um, but I think lead generation has definitely been like the biggest, biggest, huge, like enormous use case that I've seen. Um, I think market research has also been a huge one as well, honestly.

Speaker B: Absolutely.

Speaker A: Um, again, I mean AI, artificial intelligence, so intelligence is like a very core aspect when it comes to it. Um, so I think that that has been like really helpful for um, whether they already have like existing tools that they're using, um, or they want to pull from different sources. Um, again I would make sure to definitely verify the information that it's providing for you. Um, just because again AI is prone to hallucinating. Um, but I definitely see market research as being a huge use case, especially when it comes to like competitor intelligence. I think it's a very, very huge use case, honestly. Um, so like understanding exactly like how your competitors are actually helping your audience, their audience and then what are identifying exactly the gaps, um, where they're not actually helping out and that is honestly opportunity for you to take advantage of. So I think that competitive intelligence has been a very enormous use case when it comes to like building workflows and utilizing within AI.

Speaker B: Is there anything that we're going to run over a little bit, folks? Sorry, we got, we got started some technical difficulties and I have more questions. So as long as Adama's got time.

Speaker A: Yeah, I've got time.

Speaker B: Okay, fantastic. Only a few more, but I just made this conversation. Okay. So that's what people should be building or could be building as a first. Is there any that people started to build and maybe shouldn't have? Like, is there anything that you're seeing people building that maybe they shouldn't be using AI for and shouldn't.

Speaker A: Oh, well, that's a really good question. I, I think that honestly, I think that's more so. Uh, like the thing that I think about is like sensitive information. Honestly, anything that pertains to like really sensitive information. I'm not, not really the biggest fan of integrating AI when it comes to like finances, to be honest. So that gives me a little bit, A little bit. That's a little bit scary, honestly. Um, and I would honestly kind of like measure what you want to be autonomous versus what you want to actually have like more oversight over. So I think that kind of like depends on the person. Um, and I think do think that I, I think that I've seen like customer Persona uh, generators, honestly that have been like, good, but I have also seen some of them that have been like, not really up to par. And I think that that's mainly because like, you do kind of have to have like conversations with people. Um, there might be like kind of like some missing factors when you're trying to like scrape what conversations are being had on like social media.

Speaker B: Gotta come from first party data.

Speaker A: Yeah, I think first party data is always going to be superior, um, when it comes to trying to like identify exactly like who your core audience is and who you're actually helping. Um, so I've seen good, like, I've seen excellent, like, like use cases when it comes to it, but I think that first party data is honestly very superior when it comes to it.

Speaker B: Yeah, Mar A lack Lily would agree with you who has built one of those, but she does it from first party data and ah, she built one for me and it is absolutely magical. Like I run everything by my, my panel to say, okay, is this prompt, is this post gonna land? What's missing, what's not? I've done um, presentations of like, okay, this is our pitch deck. What am I missing? What's gonna land? And man, does it find like just. No duh. How did I miss that? Things and just really tightened presentations. Copy everything up. Um, I will say for me, I've kind of stopped using it for writing. I feel like it just doesn't capture my voice. I feel like it takes it in a direction I never would have or doesn't quite understand. And so I've done a better job of using AI to take stuff off my plate so I can do more of the things like writing that is like, coming from a place of my experience and me.

Speaker A: Yeah.

Speaker B: Um, and I've seen it really up my game almost because I run it through and ask it, like, okay, I run it by my panel. I ask it in terms of data, like, what data is missing. I have it fill in some of those gaps that make it a little meatier than I ever could without it. Um, but the writing itself, I find I've gone back to doing it myself.

Speaker A: I think that people want other. I think it's the desire actually for people to go back to writing authentically, honestly. And I mean, Claude recently had like, announced that if you're right, anything that you're pretty much generating with like, with AI, like with Claude, it has a watermark now. So, like, it's going to actually indicate that it was written by Claude. Um, so. And I think that that's even more of a reason to just like, go back to like, writing for yourself. Um, and I also, I, Yeah, I'm not a very big fan of like, using it to write for me, to be honest. I'm actually not like. And I think again, I've seen people do really great jobs, but I think I've also seen kind of like very key, like key repetitive things or markers that kind of like indicate that it was written by AI.

Speaker B: Um, and so clear now you can just.

Speaker A: Yeah, uh, it's very, very clear. You kind of really can sniff it out, honestly. So it's just like, it's better. I. I haven't seen it do necessarily a really great job of actually being able to adopt, like, how I, like how I write, how I sound like, the intonations and things like that. So I'm not a fan of it when it comes to writing. Um, and I mean, I don't blame people that like, use it for writing. Like, it. It's able to spit out something that very quickly. But I think there's just like a really big advantage of being able to just like, take the time and just like again, going back to like, preserving your critical thinking. Uh, like, I think that being able to write for yourself, I think that that should be something that you should honestly continue to lean into. To be honest. I think people appreciate it.

Speaker B: Yeah, I mean, I've tried like, having it look at my, you know, I'll give it transcripts. I'm like, okay, I just blurted a lot of stuff out to you, like, rewrite this as a post and I find that it can't discern my writing voice from my talking voice. And it comes out, it would come out like, um, incoherent almost. I was like, I don't, you don't even sound like you're saying anything.

Speaker A: Yeah, yeah. It's like, here's the thing. And I'm like, oh, I hate hearing that. I don't say that. But you're always putting, here's the thing. I don't know where you're getting that

Speaker B: from, to be honest, is another one, I think.

Speaker A: To be honest, oh my gosh. Like, there's just like, again, there are markers that like, that come out when it comes to like, for AI writing for you. And again, I just like, I think I ran into the same thing of like, it can't really tell apart from like my speaking voice versus my writing. So it, again, there are people who have done a fantastic job and like, have like really cracked the code when it comes to like AI writing for them. But I, I, I haven't been able to really do that successfully. And it just like, doesn't make me comfortable either, to be honest. I don't mind writing like that. I really don't mind writing, like taking what, 30 minutes to write something. And it also doesn't have to be perfect. I think that that's also the other thing that is like, that maybe again, maybe just like the era of AI or maybe just like the era that we're in of maybe it's just like perfectionism or like making sure that you're not making errors or that it doesn't sound like, no, go ahead and make the mistake. Like, make it sound silly. Go ahead and like, do the, like make it sound like worm verb, like blurb or like word vomit? Like, just like, again, it's a kind of trial and error, right? Like you write and like with more practice, you get better at being able to capture what your thought, like your thoughts on paper. So I think again. And that's kind of like going back to like, again. I think people appreciate that authenticity and that like, just like that real, you know, aspect of just like writing what it is that comes to thought that comes to mind and then just kind of reiterating learning from there.

Speaker B: I love it. Yes. Uh, I, I stopped censoring myself to a degree. I was, I like sent an email the other day and I was. And I, because I'm doing these interviews around AI Search and somebody missed the appointment and so I was, I emailed them, was like, um, are you still able to nerd out over AI search. Like, I, you know, like, I just, I would have normally in the past, like, oh, maybe that's non professional and I shouldn't say nerd out. Like I should take that out. And now I'm like, no, people need to see authenticity and they need to know that like, I wrote this and that's just what I'm gonna say. Put it out. Yeah, so.

Speaker A: Right.

Speaker B: Yeah, I, I totally agree. I think we're all sort of searching for that authenticity now. That is just washed out so much of it. Oh, uh, I could talk to you all day, every day. Last question for you. Um, for now we'll circle. It'll be great. Yeah, last question for now. What's next for you and the automated CMO and where should people follow along?

Speaker A: Oh, wow. Uh, so I mean, first I have been refining the automated cmo, the actual product said I'm at the point where I'm like, I've built this so much. I'm just gonna just like put it out there and like have people use it and then get, get feedback because I think user feedback is very, very helpful. Um, I have like a list of. A list of other like, products that I also want to like, create. I'm actually thinking about this idea like, for like the rest the. Of. Of like Q4 of kind of like doing like a build sprint for the rest of the year, honestly. So I'm kind of like thinking about that a little bit. Um, just because like, again, I've kind of got. Been busy on like systems building a lot and I haven't been able to like, build as much as I would like to. Um, so that's kind of like a couple of things that I, I do have like coming up for the rest of the year, honestly, is like doing more like build projects, hopefully doing more talks too because like, I do like, I do enjoy talking about it. I feel like I. And uh, I've been having this conversation for like this entire year is that like, right now I feel like we're in this, like, this cultural divide when it comes to AI. Like one. It is a PR nightmare. I like that. I think that it 1000% is, to be honest. Um, and I do think that there's like one angle where people. People are like, AI is like the future. The AI loyalists. Like, this is the future. It's here. If you don't use it, you're going to be left behind. This is incredible. This is so innovative. This is the future. Right? And there's another side of People who are like anti AI, like this is environmental impacts, the increase of like, like of data centers and the fact of like the power, like power outages and specific cities how like these residents are being the ones that are have to, to take the front when it comes to the cost of these data centers and like their power usage, um, and things like the risk of being replaced. And then I do think that there's this gray area, there's like this gray area of people who are like I can see the advantage of it, I can see the power of it. It's very innovative. It's can I could see exactly kind of like the, like how the use cases when it comes to AI. But I do have concerns about like regulations when it comes to like the environmental impacts and things like that. So I, me personally, I believe that I fall within that gray area and I think, and this was honestly one of the dilemmas I was having last year as well is that like learning about like the neck, like the negatives of AI and then it's like okay, I use it but then at the same time it's like there's also. I cannot ignore the negative aspects and impacts of it. So like I have been really thinking about like okay, why not have conversations about both? So I think that that's like. Because again I'm hearing conversations on one end or the other. I don't really like hear people talking about both fronts of it. Um, or even having like putting conversations about like AI, governance and safety and ethics at the forefront which honestly in my opinion I think that it should be at the forefront like that I think that's ridiculous that like governance and ah, ethics and safety is not like it should be like the foundation when it comes to like AI use. Like to be honest and the fact that it's not is ridiculous. So like I feel like I want to have like more conversations like when it pertains to that, um, about kind of like the nuances when it comes to AI, like how to learn with it. Like is it like how it's like impacting when it comes to, to like even the literacy crisis and things like that. I want to have a more open conversations about that gray area. So that's been something that I've kind of been like mulling over a lot is just like let me not censor myself anymore because like I, yeah, it doesn't feel good to just be like oh AI, AI. I'm like I'm sorry. Like I cannot ignore the like the impact and where we're going and like the direction we're headed. So we should talk about both. Like I, ah, I think it's that we don't talk about. So that's been something I've been really.

Speaker B: I love that you brought it up. I love that you leaned into it. I had another conversation a few weeks back, um, before I left, uh, for my summer vacation with a gal who is part of a company who's, um, built, who goes into organizations and does just that, sets up all of the governance, all of the policies, the foundation, and then teaches the organization how to work together to build using AI. And she was very much talking about the importance of everything you are. So just want you to know you're not alone. And uh, yes, it's important and yes, we need more voices, including yours, talking about what AI can do for us in terms of making our lives better. But also we have to recognize the impact that it is having right now, um, so that we could all contribute, uh, to making sure it doesn't, um, do some scary things that anthropic, uh, security, uh, guy was worried about.

Speaker A: Exactly.

Speaker B: Um, oh my gosh, this was so amazing. Thank you so much for joining me, Adama. I'm so grateful. Um, where can people find you?

Speaker A: Yeah, you can find me on LinkedIn. I'm pretty active on LinkedIn. I will probably be more active on LinkedIn, honestly, in the very, very near future, honestly. Um, you can also find me on Instagram. Um, funny enough, my initials are actually AI. So I'm like, I feel like this was almost a inevitable, to be honest. Uh, but you can find me AI, um, with AI on Instagram. Um, and then also email is also another way to, to find me as well. Um, but yeah, Primarily Instagram and LinkedIn, you can find me.

Speaker B: All right, well, we're going to follow you because we are all on the edge of our seats for the, for the CMO product, I can tell you that. Amazing. Thank you so much. I'm so grateful.

Speaker A: Thanks for having me. This is great. I really had a. I love. I can talk all day about here on this one. So

Speaker B: to our listeners, if you got value from this, go follow dahmer hemsen on LinkedIn and Instagram and check out what she's been building at the automated CMO links in the show notes. And if you're watching live here on LinkedIn and YouTube, hit follow as you catch us every Thursday between now and Cyber Marketing Con. And that's a, ah, perfect place to land it. Adama just spent the last half hour on building AI systems people can actually trust. The next question is one I think about for every client we work with. When your buyer goes looking in search and an AI search, do they find you? That's search visibility optimization, and it's what we do at MKG Marketing. Start with an audit@mkgmarketinginc.com I'm, um, Carrie Gard. This has been back on track, and we'll see you next Thursday. Thank you again. Thank you again. Adamas. Amazing.

Speaker A: Thank you. Uh, thank you so much. Sat.