Lunch & Learn - Session 28

Well, hello everyone. It is Lunch and Learn. I'm Tom Powell at the Founder's Office, and joining me is our General Counsel, Jackie Kuiper. Hi, Jackie. Where are you today?

I'm in Minnesota.

You're in Minnesota. Are you guys getting some snow or no? Is it back to beautiful spring weather?

It's nice. It's still like brown and gray here, but no snow, thankfully. And it should green up in like the next few weeks.

Okay. And I'm in Utah today. I just finished up skiing with a group called Blister, which was super fun. For 4 days, we got to demo new equipment. So I just fresh off the mountain. And Candice, I think you're in your library.

I am.

I love it. Well, Candice, thanks for joining us today. You were doing a lot of the research on our guest today, which is— I'm super excited about. I know you are as well. And Jackie, you're the one who was the first one in our company to actually build a bot. So not even our tech people built a bot before Jackie did. So that was the fun part about this. You know, most founders right now aren't struggling with access to AI. They're struggling with how to actually make it work inside their business. And our guest today, Nate Herkelman, has spent the last few years solving that problem. He's built and scaled an AI automation agency past 6 figures a month before the age of 24. He's had already one exit, which was fantastic. He leads Up at AI and has built one of the largest AI AI automation communities in the world. But what stands out is how he approaches it. Nate doesn't just treat AI like a tool, he treats it like a layer inside the company. Culture, workflows, behavior. He's showing teams how to move from curiosity to daily use and how to turn that into real operational leverage and revenue scaling. So today, this isn't about what AI could be, it's about what it actually looks like today and what it looks like inside of a business and how do you build a business around it. Nate, also goes by Nate Hurk as his trade name. So, Nate Hurk, we're so glad to have you here. Thanks for joining us.

Yeah, thanks for having me. I'm super excited for this chat today.

So, I know Candice was popping around on different questions as we were getting ready for the show. But how did you get started? When did you recognize? Do you have that moment when you said, "This is something"? Absolutely.

I remember it very, very clearly. You know, 'cause I think a lot of us probably remember when ChatGPT came out in November of, I think it was '23. And at that point I was in college and I remember I was sitting in the business building with some of my friends and we saw this thing come out and we all just got on there and we were just playing around with like writing poems and writing stories and just kind of the like fun use cases, you know, we didn't really see much value in it, but I slowly started to figure out how I could use it to help me with my homework. And at that point, what was happening was it would take me longer to do my homework with AI than it would've if I just would've done it manually. But I was so obsessed with the idea of figuring out how this tool could help me because I knew if I kept learning the skills, it would, it would ultimately save me a lot of time. So that was when I first kind of started to realize it's, it's not great yet, but there's definitely value here. But then what ended up happening was post-grad, I, um, started working for Goldman Sachs and I was in a business intelligence role. So I was basically just spending all my time automating things, automating reports, automating emails, pretty much just internal automations. And I really loved it. I loved the idea of having an end goal to work towards and not exactly knowing the way you get there, but knowing that you can piece together different things and sort of build that logic. But what started to happen in my automations that was really frustrating to me was there was pretty much always a point in that automation where a decision had to be made and it took a little bit of either, you know, subject matter expertise or just some reasoning. And even if it was a simple decision, even if it was, you know, read the sentence and tell me if it's positive or negative, we needed some sort of decision. And that's when I was talking to my, my team and I was like, I just wish that we were allowed to use something like ChatGPT because I, if I could send off the automation to ChatGPT for just this one step, have it return a decision and then let the automation keep going. That would turn this automation, you know, we're already making this process from 1 hour to 5 minutes, but it would go from 1 hour to 0 minutes because there would be no human in the loop at that point. So that's when I started to realize the value there. And at work at the time, there really wasn't any AI that we could, we could work with. I mean, a big part of my job too was I was the AI meeting note-taker. I sat there, I took notes, and I sent out action items and meeting summaries. and I realized that there was a tool for that. And so then I really, really started to dig into it outside of work. So that was kind of the, the light bulb moment. And then from there, um, I would just, I would run home and I would build agents and build automations.

And you started your YouTube channel?

Yes.

Right.

Yeah. That's when I started.

Was that, was that kind of the first step? And how did, so now just to put it into perspective, how many people do you have registered or view or have viewed videos on your YouTube channel? It's in the hundreds of thousands.

My YouTube channel, I think, has 30 million views on it.

That is awesome. 30 million views. That's an incredible number. And Nate, is that the best place for people to plug in when they first start with you?

I would say so, yeah. There's a lot of content out there on my YouTube and also just YouTube in general. So much high-quality free education out there.

That's incredible. Well, Candice and Jackie, I know you wanna jump in there. Let Candice jump, 'cause she's been biting, I mean, chomping at the bit here for a little bit. So jump in.

No, I just, you know, I love your videos. I've told you that. And I was excited this morning because the video I watched before we jumped on, your new video, and we talked about this a little bit, but I want our audience to hear because I think one of the things we hear in AI is people aren't always getting the results they want when they're using like a single platform. And one of the things I noticed, and it was the most obvious in your video that you posted, I think today— was it today's video or yesterday? Um, it's a really cool video because you—

the Avatar one, AI version of—

am I talking to really you?

I should check.

Is it you, Nate, or is it your AI? Yes, I can confirm this is Is it you? So, but one of the cool things that you do, and you talk about it, and you spent— I can imagine you spent hours figuring this out. You troubleshooted various platforms, and then you walk us through it so that we don't have to do the troubleshooting. But I'm sitting here going, how did Nate figure this out? How did Nate figure out that, like, this single platform that says it can do all these things didn't really do all the things. So then he went to this platform and it didn't quite do all the things either. So he went to this platform to feed into this platform to feed into this platform to get the results he wanted. You do that in a lot of your videos and you spell it out for us. But how are you troubleshooting this? And how would you advise people who are implementing AI into their businesses and/or building their entire business around AI? How would you advise them on how they should be troubleshooting and figuring out these solutions? Just a really simple, easy softball question.

Absolutely. I mean, it's a really good question because I think when you are in that mindset of, I know what I want, you know, we're all very good at knowing what we want, but the path to get there isn't always super clear and it's very overwhelming at times. So what I've always done from the beginning And I think because my background of building in Alteryx was the tool I used at Goldman, building in Alteryx, you know, making those automations, I got really good at being able to really break down a process as granularly as you can get. And when you start to do that, you realize, like, let's say we have an overall process that we want to automate. That process itself has maybe like 6 baby steps. And if you look at just one individual baby step, it's a lot clearer as far as how you would do it. So for example, in this scenario that Candice was referencing, I was using Claude Code to help me with HeyGen to create an avatar of myself, ElevenLabs to clone my voice, and Remotion to edit all the videos together and add motion graphics. And so it's kind of 4 tools working together, and I just thought about it piece by piece. So I realized the first thing that I needed to make a convincing video is my voice. I think the voice is like the most important thing. So I just started doing research on voice tools and I already knew ElevenLabs, but I tried out a few others and I ended up just realizing that my voice clone on ElevenLabs was the best. And so that was kind of step one. I knew I had the voice. Step two was how do I get a clone of myself? And HeyGen just dropped a new avatar model. And so I tried that out and it's, I'd say, by far the best on the market right now. So then I had step two figured out. So I kind of just took it piece by piece. And then the cool part is that was the creative side. Like I got to be creative and test out tools and try things. But from a, on the technical side, you're asking me how I was able to connect these tools together. I said, hey, Claude Code, go read the documentation for these 3 tools and figure it out for me. And it came back with 3 approaches. And you know what I did? Cause I didn't know which approach would be best. I said, try all 3. And so it's basically doing the, the brainstorming, it's doing the experimentation for me. While I sit there and just make sure that it's not going crazy off the rails. And that is how I usually approach these massive ideas. I give the AI, obviously, the ability to do research, and then I just have it go do research and I have it teach me things. And that's kind of my framework for figuring out what to do and then how to do it.

Thank you, Andrew.

Amazing. Dad, you're muted.

Yeah, no, I want to jump in there for just a second because you said, Nate, when on the video, you said something super important that I've said for much of my life, which is the one thing that we are all equal in is we have 24 hours in a day. And you were clear about how you were using the various tools to save you time and execute faster on the things that make you human and make you Nate. And you were even talking about, hey, yes, I can use AI automation over here. I can have an agent that's acting as me on the screen. I'm still Nate, and there's reasons that you want to be dealing with me. One of the brightest things I think you said, though, was I made this video, and if it was a, I think you said, 2-minute video and it cost you $50, but you could replicate that video in multiple places. You couldn't ever before now make a video for that, especially if you counted your own time in it. And so that's the brilliant part for the founders that are listening here. We're seeing a lot of founders that are able to move through their complexities, but also their constraints so much faster because it can help them think through that. Can we jump over to that? Because again, we've got a bunch of founders here. By the way, as founders, please look at Nate Herkelman out there. If you guys have questions, this is the time you have him live, and you're gonna be able to ask some questions while you're on here. So Nate, constraints, businesses, how do you see this actually in this revenue model? It's not just cutesy making videos, it's how do you actually turn this into a business?

Yeah, I mean, there's so many different ways that we're seeing it help businesses and also internally the way we've been using it. I think the first thing that you need, you know, I think we've all heard this term of AI readiness or having your data be AI ready. And that can mean a lot of different things. But to me, what that means is at this point, I basically live in cloud code because I can talk to my cloud code and it knows pretty much everything that's going on in my business. And because of that context, it makes it actually able to be useful because a lot of people have had experiences where you end up talking to AI and it takes you longer because, you know, you have to give it information or you're repeating yourself. And so when you get to the point where you have your data, your processes, your team, um, it's a game changer. So I guess real quick, what that looks like for me is my cloud code is plugged into my entire Google environment. So it can read all my emails, it can read all through our shared drive, and it's also plugged into ClickUp. So those are basically the two places where I have data that's like, it needs to know. Um, it can read my DMs, it can read our project management board. And also I have it set up now so that all of my meeting recordings automatically go in there as well. So basically if someone was able to talk to my Claude code, they could probably get an answer. They could get a better answer from that and a quicker answer because if they shot me a DM and I couldn't get to them in 2 hours, um, and I also might forget something, whereas my Claude code would respond instantly and it knows everything. So that's like, that's a huge unlock right there. So like we're helping our team right now get set up with kind of their own executive assistants because we have, you know, we have a framework that we like to use for our quarterly sprints called OTAs and we have monthly tasks and the managers have to, you know, like we have to follow up with people, we have to make sure that people aren't getting blocked and there's a lot of things to keep track of. So my system is able to basically follow up with people and nudge them in the right direction and then if they have anything that's blocking them, it lets me know immediately. And I think that's just one kind of like general productivity thing that everyone can use no matter what, you know, once you're plugged into the right data. But then you really have to start to think about your specific workflows. And there's a big difference between automations that are triggered by you, the human, or automations that are running all the time. So for example, I have tons of automations that I use but I only use them in certain times when I actually want to sit down and talk to something. So for example, like this, this one that records or creates videos for me, I'm only going to do that if I have scripts to record or, you know, something like that. And then I'll, I'll send it off. And that still turns the process of potentially 5 hours into 30 minutes. But then there are the ones that you want to set up that happen, you know, every Monday or at the end of every day or on very specific events. Like the Fireflies one I have, as soon as I finish a call, it automatically sends that over and gets, you know, ingested into my call transcript database. So it's kind of a way of thinking about your quick wins, because I think if you're coming from a world where, you know, you've got this stuff set up but you've never really built an automation, what is the one thing that you could do right now that gives you the quickest win? And for a lot of people, that's maybe just more like a a cloud project or if you've built like a custom GPT in the past. So I don't want to go down a rabbit hole of like a super specific answer, but that's kind of the phase we're in right now is we've got everyone set up with this kind of operating system of context and data, and now we're helping them dive into, you know, break down your day-to-day for me, break down your week-to-week, and let's identify the opportunities that we think are, you know, kind of the highest value, but also the quickest wins that we can get for you right now. And then once you get a little more comfortable, we'll start to build out those, those more the automations that kind of sit there, set it and forget it, and provide value all the time.

So, Nate, you talked about this. I'm sorry, Jackie jumping in here too at one point because I want to talk about how she builds her automation in the legal side because legal tech is really a big part along with our med tech. And, you know, I come out of the fintech originally space, but You have free videos. There's so many free videos out there, I don't even know where to start. And we were talking about this before we got to start, but you also, more importantly, have a program that people can join in and follow you and actually learn real time. Is that relatively accurate? I mean, when you've got a new release, you're talking about it right there and people are watching it the next day and involved in it, or how does it work to get into one of your training programs?

Yeah, I mean, honestly, it's, it's basically that the, the program that we offer, um, AI Automation Society Plus, it's— I think the value there is the community and the support and the structure. Like, there's nothing inside of my community that's like groundbreaking that you wouldn't get on YouTube as well, because like everything that I learn and all the experiments that I run, I want to share on YouTube. I think it's just more fun that way. But inside of the community, that's where you get Yeah, support. You get to talk to other people doing the same things. And it's basically like, watch these, you know, 50 videos in order, rather than you kind of have all these thrown at you all the time. And every day there's 50 more. And that's where you kind of get the overwhelm.

Yeah. Jackie and I sat next to each other while I'm trying to build a bot and she's building a bot right next to me. And she knows that I'm like, ah, and I come out of Silicon Valley, right? I was building— I worked on a 512k computer way back when called an Amstrad and tore apart an Apple Lisa. So, and so, but I'm thinking this is just absolutely thrilling. It's like the Star Trek days and Jarvis from Iron Man all tied into one. So, question for you, you said speaking to it. I think there was big news this week. Did you, you did a partnership with a company who has some great voice automation or voice, tell us about that.

Yes, I have partnered up with a company, a tech startup, and it's kind of like a Whisperflow if you guys have used that before or been aware of it, but it's just going to be a lot more agentic. So definitely more details to come, but I'm super excited about that. I think that the vision for this company is it's going to be able to move a lot quicker, and I think it's thinking a lot more It's thinking more on the Jarvis landscape than something like the other competitors out there. So yeah, I'm super excited. If anyone listening right now does not yet have a tool that transcribes so that they can just talk into their computers rather than typing everything, then I would say that would instantly kind of like give you a massive productivity boost. I know for me, it's like I don't type en ever anymore, even if I'm just saying thanks.

So when I talk to Siri, sometimes she cuts me off because I get so frustrated with how poor that transcription is. As a scenario, this stuff is light years ahead, right? As we've been talking about it, like you're really able to do programming, talking to your computer. You don't have to move your mouse. You don't have to touch the keyboard.

Yeah, yeah. That's kind of the goal of where we're going right now. It's more so like you click into any text box and that's where the text will appear. But where I think it's going to go is you're talking and it understands enough to be able to, you know, it's already looking at your screen, but it will move your mouse and it will type and it will do things like that. When I say right now that it's agentic, I meant more like it will understand commands. So if you start off by addressing it in agent mode, it will be able to translate things, or you can ask it questions and you can do more of sort of like the— it's more like a chatbot. When the— with the Agentic right now, but soon the Agentic is— we're going to add tools into it so that when you say, hey, you know, can you, can you send an email and can you grab that Google Drive folder and, you know, attach it, instead of just giving you like the text, it would actually click on your, your different tabs and drag things around and like it would actually kind of control your interface a little bit more.

That sounds super cool and, uh, very exciting. Uh, Jackie, I think we— do we have some questions coming in?

Looks like we've got a lot of questions coming in. And before addressing some of them, Nate, one of the things I'm hearing you say in this, in the context of like, it's moving so fast, it can be so overwhelming. There's so many tools. How do we layer them together? You can go to YouTube and watch all these fantastic videos. How do you watch them in the right order? Right? It can be very overwhelming. What I'm hearing you say though, I think, is it's super powerful to go in with a specific goal. And if that goal is really big, break it into smaller goals, know where you want to end up or where you want to save time or add an efficiency or add like a productivity amp or a boost, right? And then go in and approach it that way. I'd love to hear your thoughts on that if I'm not kind of stating that as part of your approach. And also, like philosophically, you're, you're uniquely situated in that you are teaching AI and kind of living it and you're part of this. And there's so much rhetoric out there about AI replacing people and you're the person who's built a whole business on this, where do you still see like your most impactful human ad, right? And same question, like within your team, where are the big, like the necessary and really human components that aren't going away in this?

Yeah, it's a really good question. I'll start with the second piece. I think, I mean, that's the big question in everyone's mind, right? Is like people being replaced. And I don't do as much consulting and implementation for businesses as much anymore. But even when we did that, I would kind of even feel that, you know, we'd come in and not in person usually, but we'd have these meetings and we'd meet with other stakeholders and we'd meet with the people that own that process we were trying to help automate. And sometimes they'd be like, you know, like these are the guys that are trying to like, get us fired, basically, like, why would we help them? So, you know, change management is one thing, having things come down from the top and having the leaders of organizations explain, we're trying to make your quality of life better so that you can do 10 times more stuff. And you can remove the, you know, the actual stuff that drains you guys, the stuff of copying and pasting and the manual stuff you do, the repetition. We're trying to let you focus more on the high-level stuff. That actually, you know, makes you tick. So what's really important though about this is AI agents or AI automations will be really bad unless you actually give it the right subject matter expertise and you guide it in the right way. And I kind of think about it like when I was just explaining this to my team the other day, some of the people that have never really built an automation was that you have to kind of approach it with the mindset of teaching a kid to like ride a bike where when you start, it's going to be bad and the kid is going to fall over and you're going to say, okay, you were leaning too much on your left side. You're going to need to sit a little bit more in the middle. And then you try again and then, okay, cool. Like you, you, you, you've pushed them off a few times and now you're to the point where you put on training wheels and you don't have to hold, you know, the handles anymore and you kind of just slowly upgrade it. But every time, every time you add an upgrade, you're still watching it and you're still making sure that it's, it's doing the right thing. And then you eventually get to the point where you take off the training wheels and you, you can let the kid just kind of ride, but you're never going to just like send him off and then like go to bed. Like you're going to send him off and maybe just watch through the window and you're still going to kind of keep tabs on it. But the reason I say that is basically just to say that I right now could not go build a really good SEO AI agent because I don't know anything about SEO. And I don't think it's going to be this world where you have a founder with an AI CFO, an AI CMO, an AI everything, because you don't know enough about each of those verticals in order to be able to make those agents good. So I think it's more so having the people who are right now editing your videos or who are right now writing your copy, having basically take what they have in their brain and how can you get that into a system that's able to think like them and do like them. And that's not a quick process. That's not just a simple, extraction of their mind and you get an automation in the next day and they're gone. It'll take a while, but, you know, I guess the point is, I think a lot of the value there is people's, people's intuition. People have gut feelings about things, and it's really tough to replicate that with an AI system. You know, it'll end up— if you have that expectation, then I think that you'll end up getting frustrated when you try to build that out, even if you do the method of I want you to research everything that Alex Hormozi has said ever and then build me an AI Alex Hormozi. It's just not the same, you know.

That's a fantastic answer. Thank you. And great analogy. At the end of the day, you still always need the parent watching through the window, even after all those processes have been put in place and the kids successfully riding the bike. I love that. Thanks, Nate.

Absolutely. And then the first piece, I don't think I answered the first piece you said about— you remind me it was about— it was just about kind of breaking down processes.

Yep. So in a roundabout way, you did kind of answer it, but yeah. Am I right on that where it's really helpful to like start big picture vision, but then break it down and work backwards almost?

Absolutely. Yeah. Um, people always ask about that and before you even jump onto, you know, an end-to-end canvas or before you even open Cloud Code, I think it's really valuable to write down what it is you do every day. And I remember actually when I, I did like a 2-week audit. I was reading Buy Back Your Time by Dan Martell, and he did like, you know, the buyback matrix. And I was writing down where am I spending my time. So I wrote down everything I did for the full week. The idea of like why he told us to do that was, you know, to buy back your time. But when I was writing down those things, I realized there's so many of these things where instead of buying back my time by hiring someone, I could actually just build a system to automate that. Because once you actually write down things and you look at it, you'll realize how much time you spend Like, if you spend 20 minutes, um, moving data from one platform to another, that should be like an instant red flag, you know what I mean? But sometimes, because you're just in the day-to-day of that, you don't even, even notice that you just did that, or you don't notice that it took 20 minutes because maybe it felt like it took you 30 seconds. Um, but sometimes, like, there is this, this learning curve type of feeling where, let's say, let's say I wanted to build an automation to, um, I guess just like, let's say, respond to my YouTube comments. Me building that automation might take me 2 or 3 hours sitting down, focused work. And in the time, that feels unproductive because I could just go spend 10 minutes responding to YouTube comments. But of course, the 10 minutes stacks every single day, whereas if you put in the investment of 3 hours right away, then you might not ever have to think about it again once you've put in the time. But that's why some of the people on my team, like I was talking to one of our community managers and trying to help him figure out how to automate, you know, every single person that leaves the community, he would reach out to manually, he would DM them manually. And he was getting frustrated thinking about how to set this up and how to maintain it. But I think it's kind of one of those things where your progress, if you're doing it manually, is very linear. You're trading your time for output, I guess. And what happens when you start building automations is your, your progress is going to feel less than linear because it's going to feel like you're going nowhere for multiple hours. And then you get that huge exponential curve and then it just never stops. And it's, it's only exponential from there because the systems just kind of feed into themselves.

I think that's— I love your explanations. I also love the ease that you have with making explanations. You're wise. And knowledgeable beyond your years, which is super fun. And that's why we love working with you. It's a cool thing. So, I'll give you a little background. For myself, I have had a long time of saying, "Okay, where can I replace myself with someone human?" And so, like, I want to get better at yoga. Best thing I can do is hire a yoga coach. Jackie is a yoga coach. She does her own yoga, but I need a yoga coach that comes in and she's a yogi, right? I don't know if that's the right term, Jackie, but I need that. I am a competitive skier and I coach people how to ski, yet I still, I learn more when I coach, but at the same time, I have my coaches within that space. And obviously within the business, I have just a wealth of people that I've had that have both come in and relieved a who, not how relationship, which Candice and I oftentimes look at. It's like, who already knows how to do this? You said something really important, which was just because you may be able to feed all of Alex's materials into that AI, it's still not Alex. It still doesn't know how to see around the corner of the parts unless you yourself were already in that space and knew that. Such as if you came, you were in finance, if you went over to construction building, you'd make so many errors potentially. That's the saving grace, right? I mean, you're going to have a faster learning curve, but I think that employing those bots are there. We've got a whole bunch of questions, but before we do, I have said, you know, in the space of being a founder, there's going to be very few, if any, business that AI integration isn't just part of the business. We won't even talk about it being AI any further. It'll just be, it's like, you know, I'm a Google pro. Nobody would say that as a consultant. Even in the early days, they would say I'm a Google pro, right? But right now they say I'm an AI pro and they come in as a consultant. Pretty soon that's not going to be that. And Candice, I think I'm actually stealing that from our conversation last night. I think you said that last night. Last night. Peel away. I totally agree, right? So you take 35 years of banking experience and lending experience and capital raising experience from our founder's office and all the things we've got, we still can't figure out all of the different bots that are going to be able to take over those things. Because if we're talking to you as the business owner, you've got a unique situation. And we may have seen it, but there's still parts to it because you're still a human. So in the consulting side, I think that's still going to be an opportunity. I don't know what you think about it, but instead of just dialing up ChatGPT and trying to figure out what Tom's going to tell you, are you going to get a different answer from the actual real Tom?

Yeah, I think, I think so. Because also, when you really boil down, like, and I'm not like a super technical machine learning guy, and I never want to understand like the parameters and how all this stuff works. But it's a black box, you know, and if you feed in 100 of the same prompts, you're going to get 100 different answers. Like, they might all be different variations of correctness, but you will get 100 different answers. So yeah, I mean, absolutely, I completely agree with what you're saying.

So there's a huge value for that. Well, Jackie, do we want to go through some of these questions?

I don't know if all of them are relevant, but some of them are more comments than questions and a friend request, Nate.

Awesome.

Alex says, hey Nate, big, big fan. Since new models come out on the week, how do you recommend selling automations confidently when a new model or service could come out next week and outperform it? And also, can we be friends?

Yes, I would love to be friends. Absolutely. It's a really good question. You know, I think that is honestly, it hurts you and it helps you because if you are kind of having this specialty with AI, then a business owner or, you know, people that are dabbling with it but don't kind of live and breathe it as much, they're even more overwhelmed than you are. I mean, even Andrej Karpathy and all of these guys that are top-tier experts of AI are still overwhelmed too. Like, one of them just put out a tweet recently. I remember I showed that to you, Tom, but he was like, if you— like, I'm overwhelmed every day. And, you know, He was one of the founding engineers of OpenAI. But that, you know, we would actually pull that lever quite a bit because it's like, if we're on a retainer with you guys, every single month we're going to be continuously running more optimization tests and testing every new model and making sure that your system is both balancing the quality and the speed and the cost. So that's kind of the way that we would pull that lever. But it also does hurt from a perspective of if you are building something that's a bit more of a productized solution, solution, you know, you have that fear of, is this going to be around in a few months? Or what if it gets so easy to build the solution that people just start building it in-house rather than working through us? So I've never, you know, built and sold and scaled a SaaS or more of a product. It's more so been custom consulting and implementation. And I think that in, in that world, because it moves so fast, that actually helps you out because you're the one that the businesses will look to when a new model drops. You know, we were getting that all the time in all of our Slack channels with our clients. A new tool drop or a new model would drop and we would get 20 messages. What do you guys think of this? Should we use this? Should we not? Is this hype? You know, all these different frameworks, MCP, and we were just kind of making sure that they were just staying informed. And because there's a lot of hype out there on social media and You know, I'm going to make a video about basically every big release and big feature, but that doesn't always mean I'm saying use this. Sometimes I might be making a video to say, hey, you know, like, this is cool, but I don't think I'm going to use this. I don't think it's very practical yet. If I wanted it to be practical, I would want to see X, Y, and Z. So, um, yeah, I think that it actually is, is good for you if you're kind of more of a consultant.

We've talked about— so, you know, we're in the risk management business and a lot of what we do, and when, uh, one of the Cloud agents came out, immediately our guys were like, oh, this will be great. And I said, you absolutely cannot run that on our machines. For finance service, you cannot run that till we run risk management on it, which I think is also going to be a large business within this business of how is it interacting, where is it going to impact the financial services. We saw that there was actually testimony at Congress, but then went to the Federal Reserve and said, this is what is happening and how fast someone could break into bank. Securities that we have today. So that's another part of it. And for our own implementations, how does an entrepreneur— you said you let it run your emails and you let it run your direct messages and you have these things. How do you keep it from being able to take over your whole system? And we live in, you know, Terminator days.

Yeah.

An old deal. Most everybody that's listening to this probably doesn't even remember what Terminator was. But, you know, that was the highlight movie for me. Okay, grandpa. I am a grandpa. That's okay.

Yeah, I mean, there's a ton to do with security and privacy and access, and that's why everyone was getting those Mac Minis back a while ago when OpenClaw came out. It was because the agent was basically fully autonomous on your local infrastructure. And so people were putting them on Mac Minis so it couldn't touch any of their real stuff. So I never did that. I did more of the VPS approach. So I was putting my OpenClaw on a server. But the best practice that I would say just about agents in general is don't treat it like a second version of you. Treat it like an intern. Treat it like an employee. Would you give an employee your passwords, or would you give it your credit card information or your bank information? Probably not. You would give it an account and you would give that account access to read it or to view things or to comment. Um, so that's kind of the approach that we've always taken. They have their own ClickUps, they have their own email accounts. They have their own logins for things like that. And we can also then have an audit trail of what did they do if they made changes. You know, I also have my, my settings configured so that they can't ever delete anything. They can suggest things get deleted, they can move things, but they physically cannot delete anything because that command has been turned off for them. So things like that as well. There was a— I forget who it was, but there was a— I saw a tweet. I think it was someone at Google. One of their open clause deleted like thousands of files that the company needed. And I just like— that's just a perfect example of how careful you do have to be. But then also, like, when I do have an agent that responds to comments on YouTube, I have an agent in my community. And whenever I'm doing anything where an agent is speaking on behalf of me, it's basically forced. It's written in the script to sign off with Nate's AI agent, or people have to know it's an agent just in case for some reason it says something that, you know, I can't actually live up to, or it's maybe sassy to someone and I don't want to give off that impression. I just kind of look at those very human way of thinking of like, if this was a person and I was having this person do things for me, what would I do with them? Like, how would I trust them? How would I be able to sleep at night? You know?

Well, it's funny because we have— this happened this last week. We have a virtual that's been out. We were early adopters of Delphi where you have a virtual Tom and you could go ask questions. And our, a number of our team members have what's called almost Tom. So they have the, you know, it's taken all of the data to, you know, terabytes of data. And so they can ask it questions to say, what does Tom think before they come to me with a problem to say, hey, this is what we think the solution is going to be. But I wrote 3 Substack articles on SAFEs, which is really popular. Y Combinator has made SAFEs really popular about raising money, and I'm not a fan of it. Candice and I won't invest in SAFEs. So if you bring us a deal, it's a SAFE, we just don't invest. I mean, that's one of our go-to scenarios. You need to take time to actually figure out what the business model is and how equity is going to be within that space as part of it. So I say that, I put out these 3 Substacks, and then somebody comes up on LinkedIn and says, yeah, but your virtual you says they like SAFEs because we didn't update the virtual me with the new Substack information. And so it went through there. It's like, yeah, this is a creative way. And it sort of went going through there. So it was interesting, but they knew that they were on a virtual me as opposed to what I had actually put out there. Which is, which was fun.

Yeah.

Okay. I'm going to go through a couple of these. So, Gary, I'm going to take yours. Just as you know, we're the Founder's Office. What we do is we're specifically founder-centric. And so that's what we're talking about here is how do you implement these into there? And a lot of our questions that people talk to us about are how they raise capital. So they're in a capital raise scenario. So this is Gary and he said he's in a seed round of $1 million. To raise to build an AI platform for inspectors. Don't know. I think we've had Gary ask these questions before. Construction inspectors. Any advice? I've done all the legwork, verify the product is needed. You can also build custom ChatGPTs. I've built 600 of them. I'm not sure I'm understanding Gary's question specifically for you. For us, what we'd say is if you're in the seed round, Gary, what you've got to get really down and you can check out on our website, but you've got to be able to tell the problem that you're solving first. And then how does what you're doing solve that problem for them? And then what does it mean for them in that scenario? And that's just the basics of storytelling that we do as our 6 C's. But from an AI standpoint, if someone's going to actually make, let's say, an agent to help inspectors do something, where would they start? What's the first thing that they've got to do?

So if you are trying to talk to an inspector to figure out how to build a system that provides value for them? Like, how do you approach that kind of the question?

Yeah, yeah, absolutely.

I mean, I think it— so this is a good question because I don't really know what an inspector needs to do. So me, what I would do is I would sit down and I would basically say, I need you to walk me through end to end your entire process for, you know, start of a project till the end. Like, what does that look like? And what happens is they'll start to sort of go down that path. And actually, Tom, like one of the things that you told me to do is exactly what I started doing is you have a piece of paper or Google Doc or whatever it is, and you're writing down everything they say. And you're constantly checking in to make sure that you're not missing anything. And that's something we teach with our wireframing is because they might tell you one thing and you interpret it one way. And if you haven't confirmed that you are interpreting it the same way, then you're going to try to build something that's way off the rails and then misalignment and, you know, lots of problems. So basically going through the order operations in order, asking any questions if anything is not clear at all. But what's going to happen is that that could be a huge process. So you have to just kind of drill down and try to figure out which of these are the steps in this process that don't require any AI. I always try to go no AI first. So if there's any data moving between systems, that's just a simple script or a simple automation, go for those first. And the reason I say go for those first is because if this is your first engagement with this person, your goal, at least in my mind, your primary goal is to build trust. And you do that by delivering value. And AI is, like I said, with the teaching a kid to ride a bike, it might not do it right the first couple of times, and it might take longer for you to get to that point where you really feel the ROI. But with an automation, those are typically the automations that I build that have— that are completely deterministic, meaning input goes in, you know exactly what's going to happen and you know exactly what you're going to get. Those are the ones that you build once and you can essentially set and forget because you've handled all the edge cases and it works the same way every time. There's no variability. And if you automate all the pieces in that process that are no AI first, you're probably alone going to save them at least 20 to 50% of that process. And then you realize, okay, now the next pieces after this person's already kind of, you know, they trust me a little bit. Let's scope these in a little bit more. Let me, you know, pull some of the knowledge from your brain, get some of your subject matter expertise to help us AI-assist the rest of this automation and kind of take it, take it by breaking it down into those chunks and really going after the deterministic pieces first.

So it's really about before you build a process, before you automate a process, you need to make sure that that process is actually getting the output that you want. Because if you build a process for something and it's a poor process, you're just going to speed up a bad outcome. I mean, that's essentially it. So, what you really need to do is decide what or find out what that is. And I love what you said because you and I've had this conversation a number of times. Just because you believe you heard what the other person said, you need to check for understanding.

Yeah.

Especially in negotiations or different things that are going on, right? Because they may have a totally different understanding of what was said. And it throws everything off track. That's awesome. Okay, Candice, fire round. What else do you want to— you've got Nate here personally. What do you want to ask him?

Well, Nate, you and I talked a little bit about— I just want to close this out with, I think what you're doing with your internal team, and I guess I had an assumption that obviously your internal team had to be like cutting edge on AI, and I'm certain they are. But you make such a great point about like, they're all in their specific lanes and doing what they excel in. And how can you do your internal AI enablement so that they're using AI to help them in their job? And then you guys have AI o'clock, which I thought was really cool. I want to just kind of close this out on how you are, because I think a lot of our questions are the same of like, There's a lot of AI, there's a lot going on. There's— it feels like there's a lot every week. How are you implementing these tools with your own team to get them up to speed? And then how are you guys kind of keeping that progress moving every week?

Yeah, 100%. Real quick, just wanted to say thank you, Alex and Gary, for the questions. And Alex, shoot me a message on LinkedIn and we'll be friends. But yes, it's a great question. The, the whole idea here was that most of our team was using Cloud Code really well, but some of them weren't because their core job was kind of centered around like community management or, um, content and just things that we really didn't want to automate. But that, that didn't mean that they shouldn't be learning how to leverage it still in other areas of their work for organization or productivity or, um, to just make them even go faster, even though we want that human touch in everything that we do that's external. So given this is kind of something that we started at the beginning of April, so we haven't nailed it yet, but the framework that we're working with is we first of all got everyone's baseline because everyone's starting from a different place. Some people might even feel embarrassed if they're behind what other people are, are where they're at. And so that's why we had these kind of one-on-one sessions and we had this survey to capture where everyone was and what they think they want to be able to do with AI by the end of this quarter. And then the next step was to help get everyone kind of on that same baseline. So everyone at the, at the minimum, we wanted to work with individually until we got them to the point where they had Cloud Code set up, they understood, you know, what they're looking at, they understood the interface, and they were connected into ClickUp and Google Drive. And because now everyone's kind of at that place where they do have context inside of their, their project. But now what we're going to do is we're going to start to work on the first session we're doing is like, we're not even touching AI yet. We're doing that audit process of having them write down what they do in their day and having them estimate how much time these different things take just to sort of put a spotlight on the pain that they have, spotlight on how much they are doing that's manual. Because when I've done enablement sessions or workshops with businesses, it was, you know, I wasn't in very much control of the way we wanted to do it. The leaders of the organization already had some sort of vision. And it was basically putting me in this workshop. I would share my screen and do some stuff, send out a PDF, and it didn't stick, you know? And what we wanted to avoid was people feeling like this was another sort of compliance training they just had to click through, you know, check, check the boxes. So we wanted to make sure that people were able to feel the ROI. And I thought that the best way to do that would be to give them a very, very, um, tailored project, because if we just told everyone to go build an email responding agent, they might be able to build it, they might be able to think it's cool, but if it's not something they're going to use every day, then they're not going to feel the ROI. So really going after— and I know it's, it's tougher as the organization's bigger because you'd need more resources to work with everyone individually. But I do think that's the best way to get them to actually feel it. And so now we're going to do like biweekly workshops. We're going to have a video come out on Monday. They have the whole week to watch it. And then at the end of the week, there's some sort of deliverable. And we still want everyone to turn in the deliverable. The video is not required, but the video will help you get there. Because some of my guys, you know, the deliverable they'd be able to do in 10 seconds because they already have it all set up. That's fine. That's great. But the people that would need that full video, they're able to go through that and not in a way that it's going to take them, you know, 50% of their week. But I basically said, hey, we're doing this new thing and we want to help you guys get caught up fast. So I just want you to expect that maybe about 10% of your week is going to go in towards— go towards learning. And it's not going to be, you know, just— this is not just a quarter project. This is a culture shift that we're trying to go through. And that's why I decided to spin up that, that It's AI O'Clock channel, just to have some fun with it, because everyone has such a different, you know, lane of responsibilities. But what's really cool about it is I said every Friday, I want you all to just drop a quick message in the channel. And say something that you learned or something that you did with AI this week. And it ended up being people were dropping in stuff every day instead of every week. And we're seeing people on the content creation team, you know, do things like my avatar. And we're seeing people on the community side automate like different reports and different like things like that. And you're able to see where everyone's using it. And not that you might, you know, someone on this side might never build an automation from someone on this side, but just to be able to see that is really cool. And I think it's opening. People's eyes. And it's now definitely the most talked-in channel that we have. And it's a lot of fun, like, because now people that I used to think didn't really care about AI are sending tweets in there of like a new release or asking questions. And it's, you know, like I said, we're only 2 weeks into this whole project, I guess you can call it, but I've already felt a massive culture shift. And people have been hitting me up and saying, this is actually so fun. I didn't think I'd have so much fun building these things. And and playing with these new tools. So that's kind of the whole, I guess, structure that we're going with. But we're, we're very much so, uh, building the plane as we're flying it, but we're trying to make it as tailored as possible, way more— trying to make it feel way more like a one-on-one session rather than just like a, a set of videos that we chop out.

I love that you've positioned this as like a value add for your team, and I think it takes away some of the scariness of AI and makes it fun. And like you said, it is a culture shift. It's, it's giving everybody the space to level themselves up and not have it be like this punitive, threatening, like, you better do this or else you're not going to be keeping up with our team. It's more of like, join us, we're doing this and it's awesome, and we're all growing and learning together. So I love it. Um, I know we've got— well, we're— we've gotten a lot out and we have not even touched most of our questions. And so we're so sorry that we didn't get— we had so many, so many questions today and we appreciate everybody's engagement. Nate, we are so grateful to have you on this. We just love watching your videos. We love anytime we get to connect with you.

We'd love to have you back if you'd come. I think the comment section, the Q&A was going off and we didn't get to everything. So if you would, we'd love to have you back sometime. Matheus, thank you. For the audience, you can find Nate on LinkedIn, you can check out his YouTube channel, tons and tons and tons of amazing content there. Use his tips where he says, you know, kind of watch them in order, have a goal, and then AI Automation Society is there as well. Thank you so much, Nate. This conversation is not only super relevant, but I don't know, sometimes around AI I get a little bit of anxiety. And talking to someone like you, like, really grounds it and makes me feel inspired by it as it's meant to be. So thank you again.

Awesome. Yeah, well, I'd love to come back. Had a lot of fun today. Appreciate all the kind words. It was fun to be here, and I appreciate everything that you guys have done to help me out as well. So—

Absolutely. It's been— and I've got a little background noise, so I'm sorry, guys. I'm sitting in the airport at Utah, but we've got probably another 12, 14 questions. We will get those to you, Nate. Where they're appropriate and get them answered out to you. Otherwise, we will be participating in there. We have next week's Lunch and Learn. To Jackie and Candace, of course, thank you so much for leading this and having some great questions. Jackie has been kind of my initial partner in crime in where we started implementing our own AI stuff, and she actually took one of my processes and built— that's what she built the bot off of, which is super cool to be able to see. And by the way, to Mateus, I know Candace said, or Jackie, you said thank you to him. It's because we We sent a whole bunch of our people off to a workshop for this week where they're going to be working for 3 days on building their— what their constraints are specifically and working on bots for that. And so if you want to do something like that, Nate's got the program for you. You need to sign up, take a look at what Nate's program's doing between the YouTubes and his Up@AI. It's phenomenal. It's user-friendly, and I can't recommend it enough. Nate, you're brilliant. Love working with you and thanks for coming on. Appreciate everybody. Thanks. Have a great week, everybody.

So much. Thanks, everyone.