Think AI Podcast - Fix the Process First: Two Founders on Automating for Real | Ep. 19 with Noemi Kis

Episode Date: September 1, 2026

🎙️ Fix the Process First: Two Founders on Automating for RealMost founders build a business that ends up owning them. Noemi Kis did the opposite. She built a seven-figure industrial water treatment company with a team of two, systemized it until it ran without her, and now spends about one hour a week on it while teaching 200,000+ non-technical founders how to run lean with AI.In this episode, Dave and Noemi compare notes on what AI automation actually looks like in production, where humans still have to stay in the loop, and why refining a process before you automate it is the difference between leverage and expensive chaos.In this episode:00:00 A seven-figure business run by two people02:02 The moment she realized effort had stopped equaling results05:14 Why the business model matters more than the hustle08:05 Running a company on one hour a week with EOS09:40 The "lazy CEO" mindset that protects your energy15:25 Can they do it, and do they want it? Hiring filter17:22 Don't automate chaos: fix the process first20:35 Voice DNA, redlining, and how to avoid AI slop22:34 Creating content for non-followers, not followers26:57 Her live trend-scouting agent stack with human review30:35 The $2,000 lesson in AI cost control37:01 How Dave runs multi-agent memory across Claude, Obsidian and Notion40:49 What happens to a job description when AI takes the taskIf you are sitting on a business that owns your week, this is the one to listen to. Subscribe, drop a comment with the first task you would hand off, and send this to one founder who needs it.🔗 Links & ResourcesGuest website: https://www.fastestgrowingtrends.com/Noemi on LinkedIn: https://www.linkedin.com/in/noemikis/Noemi on Instagram: https://www.instagram.com/kis_noemi/Dave on LinkedIn: https://www.linkedin.com/in/davegoyal/#AIAutomation #FounderMode #LeanBusiness #AIAgents #Entrepreneurship

I actually remember the moment that I said, I'm never going to scale this because I was so busy. You're doing a lot of order fulfillment, shipping labels, you go to the warehouse. It is a lot to do for one person. Welcome to the ThinkEI podcast. Each week, we talk about the most exciting AI research, tools, case studies and more. I'm your host, Dave Goir, and I've been working behind the scene in data and AI for over 30 years. whether you are an AI expert, skeptic, or something in between, this podcast is for you. Welcome back to the ThinkAI podcast. I'm Dave Goyle.

And my guest today is Naomi KISS. Naomi built a seven-figure industrial water treatment business with a team of about two people. Wow. And then did what most founders never do. She turned the playbook into a media brand. Today she teaches non-technical founders how to run, being on AI and automation to an audience of over 200,000 people across LinkedIn. That's the number.

And also on Instagram. Her post did more than 9 million impressions in the last 90 days alone. She is a TEDx speaker. My take before we start, most founders overcomplicate this. Naomi did the opposite. She says your business is supposed to fund your life, not own it. I truly believe in that. And she runs on the idea that you refine a process before you automate it.

Again, I agree on that. And you keep whatever makes customer happy. I've spent 30 years in data and AI being a disabled kid and entrepreneur. I am running an AI crew of my own. I do sell the same discipline to enterprise CIOs and CTO. So I want to compare notes on where AI automation is actually real and what it really takes to build a business that does not need you and where human still is in the loop. Naomi, welcome to the show. Thank you, Dave, for having me.

Your story is inspirational and I'm excited to share with your audience. Thank you. So let's get started. One commonality, we both are immigrant and you left Hungary, grew up across three countries before you turn 18 and then walked away from corporate sales. Walk me through the specific moment when you decided that you want to build your own company instead of working for somebody, especially on a corporate tech. Absolutely. So the moment that I realized that my effort is not correlated to the results. So I only realized that when I had a lot of ideas, and I wanted to do a lot of innovation. And then I didn't get promoted.

I didn't make more money. And so I'm like, why am I doing all this? I have so many of my own ideas. And I remember the moment that I finally started my very first business, which was actually my water filtration company, I said, wow, this is what it's supposed to feel like to follow your passion. because I felt so good for the first. I'm not having to be on somebody's schedule and really follow my own vision.

It was a great feeling. That's an amazing story. And, you know, I'm still curious and we're going to jump into that a little bit later on, industrial water treatment, pretty unique and novel idea. And where did you? How did you lend? But before we do that, did you always have aspirations to become an entrepreneur, or it just landed due to the corporate world and now things were moving?

I did not always have aspirations. I was born in Hungary where very kind of obedient people, culture, where you follow the rules, you don't say anything out of the ordinary, you know. And so when I came to America, I was going to follow the path, the stable path, go to college, I went to college, I got a degree, I went into corporate. and I was going to become a CPA. But it just started weighing on me. I just, it didn't fit.

Like something was off. It didn't fit. And I did always have ideas. Like, what if I started this business? But I never took action on any of it. And then finally, I decided to take action on that idea and it worked. That's amazing. And, you know, the power of taking the first step,

this is that story where you just have to, to jump into the water to navigate through it and figure that out as you go along. I did similar things. I built like nine businesses and five of that were miserable failure, but I never give up. You know, being a disabled person, you always deal with uncertainty, limitations, issues. And despite of all that, you know, what you do using your willpower and strength and grit, that's the example of any business, I don't take it lightly, whatever you will build, It takes a lot of time, effort, sweat, hard work, honesty, integrity to get what you

get. So I commend you on that. Thank you, Dave. Great. So that leads me to what I was trying to ask before. You build a seven-figure industrial water treatment, supply business from a very small team. What we through the exact systems you put in place and let two people run most founders would staff with actually 10 or even both.

Yes. So at the beginning, I was doing everything. I was trying to get the business of the ground. I was doing a lot of cold calls, cold emails, relentlessly going to trade shows. Nobody would sign up. It was very painful start. And I honestly didn't know if it would work. But eventually, I started getting customers.

And my cold outreach started to pay off. And people started trusting me and buying from us. And then I grew the team. and then I hired my first full-time person. The one thing that's, I think is really important we're choosing a business model that I think I got lucky in the water filtration is it is a recurring business.

So whenever you have recurring revenue, you don't have to constantly do as much outreach and constant when you're not project-based revenue, right? Because when you sell a product that's consumable, like water filters, people need to replace it. Second, is it's an industry that has a high barrier to entry. So it's not like drop shipping a product from China

and selling it to Amazon where you have 100 competitors. But it's like where these are critical equipment where people, big companies have to trust you. But once they buy from you once, you become their trusted supplier. They don't want to switch because it's too high risk. So the customer retention is not only very high, it's also a recurring business.

and that makes it a much easier business to run. Less marketing, less sales. I actually remember the moment that I said, I'm never going to scale this because I was so busy. You're doing a lot of order fulfillment, shipping labels, you go to the warehouse. It is a lot to do for one person. And I remember the moment where I was like,

I need to hire somebody. Then I hired somebody and I noticed that I was still busy. I was still doing admin. And she still came to me for decisions. So I was still just as busy. And then she didn't have a very clear systems as well. And so that's when I learned about something called EOS. Have you heard of EOS?

Yeah, we are trying to implement that. So I'm going to pick some notes from you. It's incredible. So I joined the entrepreneurial organization in San Diego chapter. You have to have a business right now. It's $2 million to join. And you have to be the founder. So it's, it's, you know, business owners. And they've scaled to a certain point. So everyone almost was using EOS in there. So I said, what is, what is EOS? I have to learn how you guys

are scaling with this. And what it is is you put your business into these systems. And every single week, you have something called a level 10 meeting. And that's currently how I run my business. I mean, this is now 10, 12 years later, right, that I started the business and I run it probably one hour a week. And the reason I can do that is because I have a really good manager who I've empowered to not have to come to me for decisions. The way I think the best way to do that is if they make a mistake, you're very empathetic. Because people come to you when they are faith to make a mistake and they don't make decisions. So then they have to come to you for everything because you're micromanaging them as a leader,

right? And so that was and training, right, really good training and systems. So we have a, you know, when you go to the doctor and they want to know how healthy you are, they take your blood pressure, your temperature and vital signs. You have those, I have those vital signs for my business, those numbers that when I I look at it week to week, I know exactly like, oh, something's off here. And so she presents to me all of those vital signs, the KPIs, key performance indicators. And these could be sales, how many leads you've got on how many quotes you got out because the number of quotes we have

is a indicator of the future sales we're going to close. And then of course, also things that have already happened, like how much closed we sell, sleeves we closed. And that's, yeah, that was it. I think you have to know as a founder, you have limited energy. And it's almost like a lazy mindset. Like, I don't want to do any low value tasks myself because you're visionary. You are the one talking to, you know, partners. You're driving your business forward. You have to have the CEO head on, the high value CEO head. So don't. don't take on low value task, make sure you give it to other people or you automate them. That makes total sense. And, you know, I, so earlier I had no experience, so I kind of fail through and done through. Then I had a bunch of coaching from either one-on-one or in the group. One of the good things I've learned about CEOs is that you need to identify what are

your trainers, energy trainers and energizers, right? And the, that energizes you. Now, it could be training people. It could be meeting with customers. It could be building processes. And you need to identify and focus a lot more over there. Everything else needs to be delegated, isn't it? And once you do that, but then you have to have some, you know, KPI's matrix,

whatnot, right? So that, you know, what's the health of your business, just like you've given the analogy of a doctor. I've also read a book. I'm a big reader. So, you know, this. book, I forgot the name already, but it's about the four principles of execution. There's a short form called 4DX and there's a coaching on it too.

And what it essentially means is that you need to have a leading indicator rather than a lagging indicator. That's right. I love that. Yeah, a revenue is just a legging indicator. So revenue is going up and down, but it's too late because if it's down, what are you going to do about it? But if there is a math where you have number of leads, that gets into qualified leads that get into conversions and that gets into revenue.

If the leads are not coming in on a week-to-week basis, then you have a problem. I'm sure you may have, can you give me some examples like that, which will be beneficial to the listeners? Yeah, so I don't really care about orders below a certain size. But in these meetings, I want to know how many proposals we got out, which is we quote customers on products they want to buy. And then those quotes turn into orders, right? So a certain percentage always converts. So that's one of them. But I want to know high value orders and customers and make sure they close. And I want to hear from the order fulfillment person, what are some

blockers to this customer? Who is this customer? Is this a regular customer, a new customer? Because I know that my highest ROI time is spent on, if we close this big order, it's going to be 80% of our revenue, right? So instead of going through every order, I'm going to trust the team with those. But high value orders and also any cracks. So I want to know issues. Like bring me problems.

Like I missed this customer email because it went to spam. Well, if this happens again, we could lose sales, so called the email company, you know, call the system we're using, the CRM we're using order fulfillment system and make sure that doesn't happen. Or we've been getting a lot of these orders for this product where our closing rate has gone down. And we realized that the manufacturer, we had a higher price than other people.

And so people started buying them from a Chinese version, right? And so that's the only way we realized it is somebody brought me the issues they're having and when we can solve it. So I want to hear issues. Please tell me all the issues because we solve it and we move forward. So this is good if you don't mind expanding on it. So what, so I see, let me start talking about some of the challenges we see in our business. So every business is either about a product or a service. That's one thing I've learned.

Well, learn it's the concept. And then the thing I learned is you actually build relationships. So you build the trust with your employees, your partners, your customers, and then you establish yourself as the authority rather than the business. And when you do those two things, I mean, larger organization, yes, they will have the brand as the authority. But in most cases, for some of us who are, you know, $10, $20 million or less, you have to build yourself as authority, right?

So now when you're putting so much time on building yourself as authority, you're hiring resources. Sometimes it works. Sometimes it doesn't. In our business and consulting world, our people are the product. So everything needs to be pictured perfect. That's what you're trying to tell or pitch your customers and they will not trust you if you're not delivering on that value. And then there is a lot of training, management, performance, evaluation has to. happen and things still doesn't work. How do you handle that? Because system is one thing,

but then people, this people process technology, people always come first. So if people are not able to perform, either they do not have training or worse, they are not willing to perform, you know? So how do you handle those situations? You know, there's these questions we can ask for hires. Can they do it and do they want it? Sometimes they really want it. Like I love this, like, they're so, like, oh, I really want to, but they're just not capable of the job you're trying to train them on. Like, they're just never going to be, right? So it's this capability, but also passion. Like, do they actually like it? There's a third one I forget. But those are the criteria. I usually look for team members. And I don't mind some letting somebody go.

If I'm seeing that it's just not going to be a good fit, especially if it's a customer-facing person, then it's a no-go, right? So because like you said, it's everything. Your customers look at your company as the person that's representing it. And so I think it's important to have the right team in the right seats. Yeah, I remember I went through a couple of just, you know, one-off sessions on EOS. In right person, right seat is the concept, which is what you're leaning towards, right? Exactly, exactly. And usually the person, it will be happier elsewhere where they can find a position that's best for them and they can excel in that.

Okay. That's amazing. And that leads me to another point. So, you know, you talked about building your business. You talked about your passion coming from corporate to business. Then you thought about how to automate it, find the things that makes you happy, which you want to do and then delegate every. to the right person in the right seat, but then there is still chaos. So you tell founders not to automate chaos. Walk me through a process where you automated too early. And I have some examples too. And then you had to unwind. You have to backtrack, which is fine.

You know, it's a learning curve. I have gone through several times. Yes. Sometimes the best way to automate chaos is to take it out of your own plate as a founder because founders are usually visionaries. It takes a certain person to start a business, right? Like you have an idea and you like think big and you take risks. And that kind of personality is not usually super systematic and super organized and like head

down in paperwork, you know, most of the time, not always, but you have to have an executor on your team that's going to make sure that whatever is not. not getting done, gets done without you being in the loop. Like if you have to write daily LinkedIn posts for your company as a founder and you're not doing it, you miss days, give it to somebody, train them how to do it, and then all you do is do a quick approval. And then you have to make sure that something gets done, take it off your plate, create a system for it, hand it to somebody else or hand it to an AI who I've trained all my people to use AI. They're not allowed to do anything without first planning it out with AI

because you're just 10x more productive. And so yes, I think that don't automate chaos. So if you don't have a system of how you do things, don't automate yet. Write it down, create it. Go to Claude, Fable or Opus, right? You open it. And you say, these are the things I'm doing.

right? And this is what my team is doing. Help me to create a system where I'm not in there. It happens every week. And I think that's what will make the most difference in helping somebody scale. Yeah. That's very true and I can relate to it. So I build my own agentic OS system. And I love writing. I've written books. I've written music. And so I initially started working on LinkedIn. I started posting a lot of particles, but then finding time is always a challenge. And I hired a few people. They started using AI too because AI started being in our life, but in the wrong way, you know, and like you're saying, there's a right way and there's a wrong way. And so research doing through AI is the right thing to do because you're going to spend hours and hours finding all those links,

but then you have to enforce AI to validate that. So I have a research agent and a validator agent. And if I don't find all the right sources with the fact, I'm not going to use that. And then I pick up the topics, write at least 5,200 words on it. And then I've built my own voice DNA. So the DNA forces it to think it like me and write it like me. Still, it hallucinates, no matter what. So then you go in and fine tune. But something that used to take me three to four hours now takes me 10 minutes or maybe 15

minutes into two buckets. And that's good enough. I mean, and then the quality is still consistent and you're consistent in posting it. I'm sure that's what you're explaining in your side of the world too if you're doing things. That being just an example. Yeah, absolutely. What you said is really good because AI slop is so easy to create, anybody can. So you have to have your own voice set up for your team to use when they're writing for you or doing anything for you. And yeah, I agree. And I have this something called redlining that I do where my AI agent does the work that I ask and goes back through it and criticizes it and redlines anything that's like it doesn't fit. I shouldn't have said that. That's not valid. Just like you said the validator

agent, maybe that's more for checking data sources. But yes, research. strategy, writing, I think are really, you know, creating images. Those are really great AI use cases. That's, that's beautiful. And so this leads me to the next one. You talked about building a media brand and, you know, building media brand as an in-down. So you grow on, and you know, if I'm wrong on numbers, correct me, I have dyslexic, so I always get into that. So you grew to over 200,000 people across LinkedIn. Is that kind of? correct? 200 across Instagram. So LinkedIn 885 or something and then Instagram 30K. That's amazing. I'm only able to get to like 12K right now.

And it took a lot of effort to get there. But I did it in like 18 months. So I'm happy and proud about it. That's really good, Dave. So you did the 9 million impressions, I believe in like last 90 days. That's some number. Yeah. And then you said that turn this content into an inbound engine.

So help me understand that. I also want to build it for myself and I get boxed into certain things. So those things are very inspirational. And what do you did differently to stand out? Absolutely. So there's, you know, you can scroll on LinkedIn and you'll see people creating content every day for the last five years and they are not growing. and then you have people who post whatever and then they have like a thousand likes

and they have hundreds of thousands of followers. And when I saw that, I said, what is that? Like, what is the difference between these two people? Because I don't have the patience to wait, you know, five years. So I said, how can I cut the learning curve? So I hired the coach that had like 200,000 at that time. And I said, teach me everything that you know. So she literally taught me,

there is a skill of content creation that is a different skill set. Like you have to learn it. And it's different for every platform, right? So I cannot post something I posted on LinkedIn on Instagram and get really good traction. Because each platform has its own vibe and algorithm and personality. Right. And so if you're creating on LinkedIn, you have to learn, you know, how do you write a hook? what kind of image, what size image is going to, what do you write in your headline that's going to

have people want to follow you, right? So you have to kind of look like a creator and you have to create content for non-followers. So if you create content that's like, here's my team or I got promoted, that you're going to get a lot of engagement from your followers. You're not going to get new people. But if you share something high value, like, hey, these are the three things I automate. and you put it into a format that goes viral on LinkedIn, you know, with a really great image, you are now going to get people who start to follow you. And if you do that every single day on LinkedIn and every single content, you look at the analytics and you say, why did this not do as well as yesterday? Or I have now 30 days of content. Let me do analytics. Let me put it into

Claude and say, do a strategic analysis on this. And you have, you have a 30 days of content. Let me do it. have to kind of be like a crazy scientist and make your content better and better every day. And then eventually you don't have to do that. You can kind of put it on autopilot after a while because you're just going to know what does my audience like, you know, and it's going to be easy. So that's the point I'm at. It's kind of an autopilot now. Well, that's pretty good. And what you're saying is while there is an art, which is what I'm targeting, there's a science behind it. And you need to learn the science from the right. people who have done it and who knows the secrets behind it or the theory behind it so that you can

implement on it. And yeah, this is a great tip for me and for the others who wants to really create good content, which a lot of people can use and follow rather than just create an AI slop, which is crazy. Exactly. And content creation takes time. It takes, you know, hours per week to do it. So if you're not seeing results, don't keep doing it thinking it's eventually going to work. It's not. Only replicate what already worked. If you get five likes and there was one where you get 10 likes, you just doubled it. Like, do only that now. Something about that worked better. So don't go back to just repetition. People think, oh, if I just keep doing it, it's going to, no, it's just going to keep failing and you're going to waste a lot of time. So that's a great tip. And, you know,

not changing anything will not bring different results. You change nothing, nothing will change. Exactly. Consistency only works on social, if it's the right content. Then double down on it, yes. A great conversation here. And this leads me to my favorite topic. I've been into data and AI space for very long time, but been technical enough.

But then you went on the different directions, which is very valuable to the listeners here. you teach non-technical founders to run lean with AI. Can you name one workflow as an example that you actually run in your live production model live mode, where you still tell founders to keep a human in the loop, if that's an example that you are willing to share? Yes, on Instagram, I am known for talking about trends. You know, I love fast-growing trends. I think it's very valuable for founders. I actually use, have you heard of OpenClaw?

Mm-hmm. So there are two agents that are really popular right now, which is OpenClaw, and then there's a more user-friendly one, which is Hermes. Hermes I use a lot. I have used OpenClaught too, but yeah, Herbys a lot, though.

I love Hermes. So I've set up Hermes in such a way that my team can talk to it in Slack. So I have a daily Crown job, which is like something that happens every day without you. So it just runs on its own. You set it up. You can do that. Also in Claude,

you can do that. I think in Claude they're called something other dispatch. No, not dispatch. Anyway, it's called something else, but it's a repetitive task you do every day. I think they call it loop,

a loop or something like that. Something like that, yes. Yeah, yeah, loop, I think. And so set. So I want to know what are the hottest trends. So there's a daily thing that scrowse the internet, what went viral, what videos went viral, and what topics they were. News like Bloomberg, Wall Street Journal, all these checks.

And then just scouring like on internet, right, for trends. And then it has to kind of filter it, make sure it's for business owners, not useless trends. And then it checks the sources. and then it brings us every day five, things like that. And then my team from Slack can work with them to write up the scripts that I'm going to record on Instagram. And then they also repurpose that and write it up for LinkedIn if it works on LinkedIn. But that's kind of an automation that still has my team involved or myself involved.

That's a great example. And then you're following through with something which is already working because it's a trend. So obviously a lot of people are already looking for it. So if you're going to build something, a very high chances that your content will show up as opposed to a lot of others.

So beautiful example. Thank you. It's what I tell founders. If you want to find a business where you have built in demand, but it's not yet saturated, go after a fast growing trend because once it's growing, movie fast,

it's moving fast because there's demand there, right? People are driving it because they want something. There's like an obsession or a niche and not many people have started businesses in that area yet. So that's kind of like a cheat code. Mm-hmm. So, you know, using Claude or any open, well, either open or big companies, whether Chet, GPT, Open AI, Claude or not Grogpot and then Open Klahermes and Nemo Kla, which is from, from blanking on the name of the company, Nvidia, right? So these are some of the big names that has been there. One of the challenge, and I have my own ways,

but I'm curious to know, how do you manage the cost? Because if you really put the cloud on API, like Opus, great model, Fable, a little iffy for me, but Opus is amazing for research and other things. Then I also use perplexity and, you know, integrate that through APIs or MCPs. but cost is a big factor. How do you handle that?

Yeah. So I got in trouble at first because when you use a third-party agent like Hermes, you have to use a third-party authorization for an LLM, right? So you can just go into your max account, like you're already existing subscription with TredgeyPT or Claude or Gemina. You can't use it through that, right? You have to have extra credits through an API. So you're actually paying for every single thing.

unless you download your own local model, which I've been planning to do, but I haven't done it yet. And so I am planning to do that. However, the way that I know I spent, I think, $2,000 that first month with Hermes, and I said, okay, we can't, I can't keep spending this kind of money.

And so that's when I implemented OpenClaw, and the way I set up OpenClawe is that it's inside of my Claude Mac subscription. so it doesn't cost me anything extra because whatever we do, it's currently under the Claude subscription. The Hermes, I try to use Sonnet most of the time for even writing scripts. It's been good enough. So that's how we try to do it for using strategic models.

Like, I'm sure you know this, right? You don't want to burn through your credits doing menial tasks with fable or opus. So you want to use haiku or sonnet, which are the lower, kind of like your assistant, right? You don't want to use your CEO to do admin task, and you don't want to use your assistant or like your data entry person to do strategy. Kind of that I would look at it that same way. Funny enough, I have the same example when I talk to the customer. You don't want to deploy a PhD person doing your assistant work, you know.

So not that they cannot do it, but you don't want to pay them that much to do that kind of job. So similar example. Also, you know, Claude is really good harness in the sense that the agentic flow is really amazing. So I've done similar things. I bought a Mac Mini, which is kind of disconnected from the world. And I have this thing called LM Studio. And not only I have my own reg where I can just, you know, look into all the content that I have personally created or saved or referred to,

but I've also downloaded models like MIT's deep seek model or Owen. And then so I have a multi-model set up, but that's all running through Clod. And I have these cron jobs or launch jobs also on my Mac. So I have Clod running these models, including Chet, 4.0 mini. I love that model for a lot of tasks. It's amazing. And it only costs like few cents a day. And so the combinations, I still have OPAs doing the research and everything. So the combination is like costing me about two to three dollars a day. It's just not too bad. And I can scale up. I know if I scale up, it's going to go a little bit further. But I don't think it's going to go beyond

$100. So $100 with clot and $100 with this. I think there's a pretty good sense. set up knowing that you are do you sorry to interrupt you also have your max account and in addition to that you're spending extra hundred yeah i i do have a max account so i have that hundred dollar not uh i don't know it's pro or max the hundred dollar clod account so so which is like five x of the value which is amazing amazing the max and it's so worth it yeah and then you know whatever i build on clod, I have either GROC or ChetGPT or Codex, verify that, right? So two different engines validate that too. I've heard Codex has gotten really good. I use only Cloud Code. Do you also do that? Like, I don't use Chad. I just do everything in Clode.

Yes, similar. Yeah. And then I've heard that a lot of people, like you said, I haven't done this, but they check with Codex as well. And I've heard great feedback on that. Have you? Yeah, lately they got better. It burns less token, but the prompting needs to be much more precise. Otherwise, it will hallucinate and become a yes man. So you have to be cautious about that. Right.

Interesting. This is, it's fun. Yeah, we started geeking out on this. See this? But no, this is a great conversation. And, you know, I love somebody practicing AI. I hear a lot of fluff about AI and people are just talking

without knowing how they use it on a daily basis. And, you know, there's a lot of refactoring rework of the work that I do too. So I build my own AI C-suite. And I even have given them goals, you know, the CMO or CSO, chief sales officers, chief marketing officer, CFO. A good example. So I am a CFA, you know, by education, did the MBA and CFA. So I understand accounting, finance control, really, really well. So I really need to know what's my run. way for next six months or so so that I have that budget saved into my bank accounts, then where I can spend more money and every time we can spend more money, we want to hire more people. That's how it works other than you spend time and money in the buying the tools such as AI. So that connecting to my quick books, not directly connecting, getting the data, I've downloaded the data in a database. It just looks into everything. So I've just like what you do

in EOS, I've given them some goals like you give 90 days. I think we call it rock, right? So you give them those things. And so it has goals. So it has advisors. So I've pooled a lot of advisory, like from Tony Robbins to so many others. So every time I'm building a new 90-day goal, I pull in advisors and apply those advisors, then it goes into 90-day goal.

And then through KPI as we are tracking it. I unfortunately did not do it on humans. So something like EOS, I think I'll take up on that and do that. So I have a question for you. Yes. I've had a hard time doing different agents because I find that they don't talk to each other and I have to remember like, oh, this is the, and then I have to say, look it out because I already told the other agent this and then like, oh, let me search because you're right.

You did tell that. So I have to remember like who was doing which task? So I just merge and I'm just using one that does everything. Do you find that works better to separate them? Yeah, so it's a challenge. And I have, so there are two ways. This is a challenge of managing context and memory. And there are two types of memory.

One is pre before, even before you start a task. You know, they need to know certain things. While they are doing the task or once they are done with the task, goal. Then there's another part of memory like learning and, you know, issues and things like that. So I have three levels of memory now. Every session, I and Hermes in OpenClau, automatically does that, but I live mostly in Claude World. So it actually captures each session lock and then it creates the handover. I've created a centralized executive stand-up meetings between all agents. So it's captioning in one commonplace.

I have defined rules. So each agent has its own agent memory, just like the ClaudeMD. So they know their own thing, but they are also subscribing from a shared memory. So there is this long-term memory which is going in my obsidian. So I created Obsidian memory, which is like a note-taker, but most people are now using it. So I have long-term memory in Obsidian. I have my short-term memory or the temporary memory, within those repos or the clod code folders.

And then I use notion to interact with humans and also with the memory. So I have like single task management system for AI and for humans, both. And then each of them have their own exposers. So that's how you can centralize. Still it will fall apart. I'm not going to lie. But 80, 90% is going to work. But because when you say memory, is that the MD files that it's saving, are those the memory?

or is that the skill? Because it's creating skills, right? And I have to say, use the writing. So we use this. Sure. So I have a few things going on. So I do have an agent, which is nothing but an MD file instruction for agent.

Agent are having their own skills, which is another bunch of MD files. Then the temporary memory is also dot MD files where I have, right, and session log. But then in Obsidian, it is memory. for sure, the MD files, but then it also has that graph linkage. So I'm linking everything in Obsidian. So any long-term interactions, either customers, people are meeting like you,

there is a file there and you know, just like CRM. So it will go in there. And then we have certain systems in place, right? Like QuickBooks and we have Microsoft teams going on in our organization. Then we have HubSpark. So it's using those as memory and it's using. So we have a little bit complex setup of memory. Most of it is dot MD,

but then certain things are going in certain databases too. So that's how we're deploying it. Very cool. Yeah, I love it. And interacting humans with AI, it's pretty fun concept. And it has its own challenges, but it brings a lot of value.

So the people who hate doing repetitive job, we can easily move. This is the right person, right-seed concept, right? The things that people hate give it to AI. It's as simple as that. And I've had to change the job description of somebody in my

couple, well, a few people in the company because they were doing things that is now done by AI, right? And so I said, you have to be so adaptable because this one person was so adaptable. She was just doing one thing in the company that's now

being done by AI. And she now able to kind of manage the AI to do these tasks. But she just evolved. I said, you have to think differently. You're not a taskmaster. You're a manager, right? And so it takes a mindset to not want to not be replaced by AI if you are doing those

repetitive tasks. That's, that's really good. Come to think of it. And yeah, so what I also do, before. I build an agent, I have a clot co-work project set of C. We are geeking out on all this instead of podcasting. But this is the value, right? You know, this is what people need to listen. So I have a project, I have a pretty amazing prompt. I create a company DNA where we look into everything, identity, soul, workforce, communication, etc. And those are the things that AI has a central

memory on. But then every time I want to build an agent, I create a proper job description, like what roles, what inputs, what outputs, what KPI, what are the gotchas, where are the dependencies. And that's the prompt going into building something. And it always create a very consistent result. I also have a method to audit what I have so that I can either optimize the token usage. And also where I have the gaps, because again, eye slop happens everywhere, even in your build process. And so where it became more sloppy, how do we control it? So that's my monthly, not weekly, but monthly review.

And I take a couple of days to optimize everything. And then I find a lot of new things like that got missed out. And it was quite shocking and surprising. So it's maturity, but it's not there where I wanted it. But I'm very happy what I have so far. I love it. It's great. So, Naomi, this is great before we close.

Anything you want to add to our audience. At one point, just to give a little bit of proud moment, I went into Top 20 on Apple Podcast, and I thought I could never do it. Oh, that's huge. Thank you. Oh, my God. And then I got lazy, so it didn't go up again,

and I'm going to put an effort. But anything that you want to talk about, because you have a media brand, you promote certain products as well, what the audience should hear from you and where they should find you? Yeah, if you're a founder and you're interested in getting all the new latest trends, go to fastest growing trends.com. And you can sign up for free and you can get notifications of every new trend

where there's business opportunities with built in demand. So it's really cool. You can also follow me on Instagram, Kish underscore Naomi and LinkedIn. of course. We all going to do that right after this. Naomi, thank you. The thing I respect about the way you talk about this is that it is boring for a lot of other people's,

but not for you. Traction before scale and refined before you automate is what we heard. Keep the human where the customer feels it, not the AI. And that's, again, an amazing thing to create. And this is the version of AI. most AI founders should get sold but not.

You know, they are sold on something else. It's the version that actually works and I love that. For everyone listening, Naomi is founder of future proof founders. Find her at her website. I will let you say it. I'm going to butcher it on LinkedIn and Instagram, where she posts daily. And if you're a founder sitting on a business that owns your week,

this is the episode to listen to and get your life back. Subscribe to Think Yeahi Podcast and send this to one founders who can use Naomi's help. Thank you again, Namib. Thank you. You have been listening to Think Yeah, Podcast with Dave. Take one idea from this episode and turn it into action.