Ep31 Taylor Dondich—20 Years in Tech Now Building AI Fellows That Cost Less Than One Hire

Taylor Dondich  0:00  
You want to make sure that you're giving just enough to make sure that you're refining the processes that the AI fellow is doing exactly what it needs to do, just like you would do with a normal hire, except that they're 24 or seven, they're never tired, they're running multiple conversations with multiple people, you know, and they have immediate access to do tasks that you know somebody would have taken hours to do.

Scott Groves  0:24  
Welcome to Henderson HQ. This is the podcast where you get all the stories behind the businesses that make our community tick. Don't forget to subscribe to our weekly newsletter. Hey, Henderson HQ, it's Scott Groves with the Henderson HQ newsletter website, all the media stuff. Thanks for watching the podcast. We're here with my new friend today, Taylor Don Dige, who's going to talk about all things AI, building your business with AI, all the stuff you can and cannot do. And it's interesting because working in the business space, I know that AI has become the new area of snake oil salesmen. Like years ago, it was people doing Facebook ads or running media companies, or what have you, and if I told you the amount of money that I've spent trying to generate leads in my business through paying some of these agencies that didn't work, it would probably blow your mind. So I'm excited to jump in with somebody who's been native to this space in this language for a long time. Whether you're a business owner or you're just looking for kind of more information on how AI can affect your local ecosystem, stick around, Taylor. Let's go, man. Tell me about your background, and why I should trust what you're saying about AI and computers versus the 22 year old life coach that wants to sell me, that wants to sell me an AI trade program at Instagram,

Taylor Dondich  1:33  
who's like, AI is the future, but all your, put all your dollars into it. You know, I'm just older, you know, I've. I think that's like my number one qualification, right. 20 plus years of experience in technology, you know, using machine learning and AI before Chat GPT even hit the field, knowing what works, what doesn't, from, you know, a technology perspective. So, you know, I've worked in technology 20 plus years as a frontline engineer, software engineer, slowly building myself up into tech leadership roles, VP of engineering, CTO. I've worked with some fantastic companies who have applied machine learning and AI in very successful ways. I've seen other businesses apply it in ways that were catastrophic, and we could totally dive into that, you know, but the scene is always changing now at a rapid neck breaks and neck breaking pace, and if somebody is trying to tell you they have the golden ticket on AI and how to apply it correctly, they're going to be wrong the very next day. That's how quickly the pacing is changing, and people are using it in ways that can actually be detrimental to their business. If you don't know how to apply it correctly, you might be having data breach under your nose. You know, you've got a lot of these personal AI assistants that people, you know, whether you're a solo entrepreneur or a tech-crazed CEO, saying we're going to apply this right away without doing any proper due diligence, like you would use with any technology, right, whether it was AI or anything else, and then all of a sudden it hasn't gone through all the security audits and stuff like that. We've seen that with law firms, where you know you've had data breaches, there you've seen people with like costs, AI costs spinning out of control, and they didn't really recognize and understand like the impact. You've also got the cultural impact, right, if you've got a team that's not ready to adopt AI, you've got people shoving it down their throats, and people don't know how to use it effectively, and you're getting kind of poor quality output, like it's a crazy landscape right now that we're in. The important thing to, like, really recognize, though, is that this is a tool, and these tools have been around for decades. Machine learning has been around for decades. We've had Amazon suggestion engine suggesting products, and we've had tools to find cheaper flights and things like that. All of this stuff that you would be asking an AI assistant to do, we've had this technology forever. But with the introduction of large language models, that's where the game has kind of changed, because now you're talking to a machine in, like, your own language, and because of that, you feel like there's this realm of intelligence that you're working with, but to be honest, these are just another tool that we've kind of had in our bat belts for a while now.

Scott Groves  4:17  
Yeah, the way it's been described to me, because I am not a tech guy, is like AI, the large language models, is basically kind of like almost like spell correct, you know, you're typing a word, it suggests what the rest of the word is going to be. This is just doing that at scale, so it's like it's kind of guessing what I want, what I want, or what I'm going to type next, and then it's giving me, instead of a one-word answer, it's giving me, you know, a one page or a one terabyte answer, that is, that is the

Taylor Dondich  4:40  
most simplistic way of putting it, like you hope that it's like the best romantic partner that can finish your sentences, right? But it's still kind of like a guessing system, right now. Of course, it's a much more intelligent guessing system, because it's backed off of all the data that that large language model has been trained on, right. And use different large language models, depending on like what kind of scope you're in. You have your general purpose large language models like Open AI, Chat GPT, Anthropic Cloud, and things like that, and you have very specialized models that might be in the scientific realm, medical realm, maybe doing voice analysis, things like that, but it's all based off of the data that they've been trained on, and so when you're looking at, like, Open AI and Anthropic Cloud. They've been trained on tons of human language content, and that's how it has a better idea of kind of guessing your intent, guessing your what's the next word to present to make you feel warm and fuzzy, and that's the other thing, is like, in the end, these machines are going to try and please you, right? Right, it's very hard to tell an LM to fight back and push back, right, because you could say what's two plus two, and it says four, and you could say no, it's five, and with enough commitment you will get the LM to tell you, oh yeah, you're right, it's five. Oh, that's scary, right? Yeah, it is scary. You know, and so when you hear AI hallucinating and stuff like that, it's because of that, right? Like, it's only trained to kind of figure out or predict what you want to hear. It's like a self-pleasing feedback loop, and that's the dangerous thing that I hear about self-starting entrepreneurs, you know, these 20 year old somethings who haven't gone through the paces of decades of grinding through and figuring out what works and what doesn't, they put a lot of trust into the machine, right? So they say, "Hey, tell me what's the next business I should build, and LM just tries to guess, right, and unless you know how to steer that LLM to actually doing the due diligence that you would have normally done yourself, because you've got the experience, you've understood what you've had to do in the past, like the market analysis, the competitor research, you know what's the total addressable market, and things like that. The LLM is just going to try and present something right, just

Scott Groves  7:00  
what you want to hear, yeah.

Taylor Dondich  7:00  
And so, if you haven't taken your lumps in business or in technology and stuff like that, where you can spot the red herring, the red flag that the LLM is presenting, and going, that's questionable, that's not right, or I'm going to need you to do a little bit more research here. Yeah, you're putting too much trust into it, and I think we're seeing that a lot with, you know, the 20 something year olds who think they've got the next business idea, because Chat GPT told them so, you know, yeah,

Scott Groves  7:26  
because you have to have that experience, or that I mean, this is the difference between like data and knowledge, right? You have to have the knowledge to be able to say, okay, poke holes in my business idea, what is the total addressable market, what's this, what's the scalability path here? Like, tell me all the reasons my idea is stupid, and if you're just falling in love with kind of your own bs, and the machines just feeding you back how brilliant you are, I could see a lot of people going down a long path, a big rabbit hole of like, oh no, no, but ChatGPT told me that this was going to be an amazing business. It's like, no, it kind of told you what you wanted to hear.

Taylor Dondich  7:57  
Yeah, absolutely, and I mean that that falls in the realm of business and technology, but also in social matters too, right? Whether it's politics or self-image or anything like that, you know, you can have a teenager talking to Chat GPT, and Chat GPT will just do self-reinforcement, you know, and Claude will do the same. Now, these larger AI providers, like Anthropic and Claude, are trying to do better in regards to, like, social responsibility of putting guardrails in place to not reinforce any kind of, like, self-harm or anything like that, but you know, you fight hard enough, though, and we'll tell you what you want to hear. Yeah, so, yeah, that's dangerous. And you said, you know, there's a difference between knowledge, I think it's knowledge and experience, right? So, I've got 20 plus years of experience in technology and startup space, and so you know, I've deleted production databases, I've made the right, I've made the right hires, the wrong hires, I've done the wrong product direction, the right product direction, you know, I've had hard launches, soft launches, you know, I've taken my lumps, and that's what allows me to be more effective with AI than I see you know someone who just came out of high school or just out of college and has not had that real world experience and doesn't know when to trust the machine and when not to trust the machine.

Scott Groves  9:14  
It's funny, a buddy of mine, John Berghoff, shout out John Bergoff, was a coach of mine for a while, and like I think a decade ago he gave a TED talk where he talked about this very thing, not necessarily referencing AI, but how the path of technology, for basically the first time in human history, is outpacing our ability to learn, and definitely just, just with the advent of Google, has replaced our like relevancy of memorization, and so he gave this TED talk about how future genius will be predicated on how well we can ask questions, and the great example he gave, and this is way pre Chat GBT, he's like, you know, just these glasses are a frame or a lens through which I see the world, he's like a good question is kind of the frame through which you see the world, just by. Asking a good question, you're already changing your mindset of like the potential to get a good answer, and it's so relevant, right? Because, like, in the future, it won't matter if I can memorize the periodic chart of elements, periodic table of elements. It's like, can I ask the right question to get the right output, and that's just been accelerated at a crazy pace over the last two years with these large language models,

Taylor Dondich  10:22  
it's interesting, because if you open up Chat GPT now, or Clod, or whatever AI assistant that you use, more often than not, now the conversation after the AI will answer, the AI will say, 'Would you like me to do x as a follow up, or we'll try to ask more questions, and that's an interesting kind of change now, right? With the introduction to reasoning models and things like that, we're seeing the LLM come back with some questions, but are they the right questions? Not necessarily. They might have you go down a rabbit hole of wasted effort, you know, but again, the LMS are there just to please, and that's the crux there, without the experience, and so I don't know if you've seen the movie Idiocracy, yeah, from my great movie, right? And it's almost profit, profit, it, it's go, it's looking like the future, right, like we're putting so much trust into technology and things like that, you know, I work in technology, you know, and I've overseen and talked to other CTOs, other VP of engineering, who are leading, you know, hundreds of engineers, and everyone's adopting AI, adopt AI, adopt AI, and in fact, you know, my teams also adopt AI, and we do it in a way that the senior engineers are highly effective again, because they have the experience, and you know how to drive the AI to produce great results, and you have kind of like this stall in hiring junior engineers, because the senior engineers are doing so much velocity that other companies are saying, oh, we don't need to hire junior engineers, you know, we've got more output from our senior engineers and things like that, but that's pretty problematic, you know. And that's

Scott Groves  12:04  
how, so

Taylor Dondich  12:05  
I would say that's a bad deal, because sooner or later you and I are going to retire. I'm not going to do this for the rest of my life. I'm hoping not to do this in a couple more years, right. And so I'm taking my experience with me. Who's going to fill in the gap, right? You are now all of a sudden going to have this gap of experienced people who have left your organization, and now you're forced to fill in those seats, because AI still needs somebody to drive it, and I don't see that changing anytime soon. So, you bring on these engineers who had, haven't had the experience, and they're putting too much trust in the machine, and we're now going to have a problem where in the next five years the AI, the large language models, are going to be forced to train on data that was generated by AI, and where there have been studies that have shown that if AI consumes the content that AI produces and you keep doing that, you start seeing a degradation of signal,

Scott Groves  12:56  
yeah, like a negative feedback loop, yeah,

Taylor Dondich  12:58  
it starts reducing in quality, right, again, because it's a prediction machine, it's not true intelligence, right? It doesn't have the experience to kind of steer what's good information, what's bad information. So, yeah, I think it's going to be a very interesting social experiment in 1020 years when everybody who has taken their lumps, is retired, and now you've got this workforce that doesn't have the experience that is trusting the machine to make the right decisions. So, when you know business owners approach me and say, what's the right way to apply AI, I say, well, AI should not be replacing your workforce, it should not be the answer to, oh, I don't have to hire anymore, I don't have to hire junior engineers, I don't have to hire anymore, hire great people, continue hiring great people. AI is a fantastic screwdriver, it is a tool you want AI to 10x your workforce, not replace your workforce, and that is the key difference. There, you want to invest in AI, but more importantly, you want to invest in educating your workforce on how to use AI, and that is critical. So, I urge any business owner who has the knee-jerk reaction of, like, 'Oh, AI is going to say, I don't have to hire all these people and things like that, or I'm a solo business owner or entrepreneur, and now I don't have to find a co-founder or work with other people. That's a lie. Build your team force, use AI alongside them, and you're gonna see great output.

Scott Groves  14:39  
Hey guys, just a quick add from our sponsors here that make this podcast and our newsletter possible. Want to give a shout out to Ethereal Med Spa, a gym for your skin, so doesn't matter whether you're doing anything proactive all the way up to spot removal all the way up to beautification, right, or vanity. Ethereal Med Spa, they got. You covered Scott Wright over at Wright Autobody, love them. So, anything from a scratch and dent to a big insurance job all the way up to a custom rebuild of a classic car, they got you covered. And shout out to my buddy Casey Halstead, owner of 10th Planet Jiu Jitsu, which now has three locations. They just opened new location on the west side, so they got our Henderson location, they got a mid-city location, and they've got a west side location. So, anywhere in the valley where you're looking to train jiu jitsu, check out 10th Planet Las Vegas. Now, back to the show. It's funny you mentioned the screwdriver. I wish I could remember where I read this, because I'd love to give them credit for it, but something I read the other day was like exactly what you're saying, don't think of AI replacing the contractor on your team, or the guy that, like, you know, drives the nail, you got to think of AI as, like, you're upgrading from a screwdriver to a drill, you know, they'll, they'll be, they'll be 80 times more efficient, they can work faster, they can drywall faster with a, you know, an automatic drill versus a versus a standard screwdriver, but you can't replace the guy hanging the drywall, or you're just going to have a bunch of inputs with no outputs. Oh,

Taylor Dondich  16:06  
yeah, totally. And I mean, that same argument is for people who are very anti-AI. They're like, you're there was this argument given to me saying, you know, it's taking the joy out of the journey of learning, and I agree. I agree there to a sense, there is a lot of value in walking the path of learning by doing, right, but you're not going to use a handsaw forever in your life as a carpenter, right. Eventually, you're going to go to a band saw, you're going to find tools that will help accelerate your velocity, but ideally you're taking the time to learn, and with AI, you know, the screwdriver is a great analogy, you know, but a screwdriver can still only be used by one person at a time, and a lot of organizations are still using AI systems, whether it's their Chat GPT or Claude, you know, you've got a developer who's using Claude code locally on their desktop, they're using it as a screwdriver, as a singular tool for a singular person. Now, they might try and find weird, awkward ways to pass a screwdriver off to their peer, but you're losing all the AI context of that session of what they were working on, and things like that. And so, the next thing, the next big phase that we're looking into is how can we have very specific AI constructs, agents, I call them fellows, and I'll get into that a little bit later, that are not a singular screwdriver for a singular person, but works alongside a team, you know, because when you hire a person, you don't put them in a closed box, right? And, sir, you're going to do this one thing. It always always been really well to your neighbor. Don't talk to your neighbor. You're only allowed to talk to me. You know, if you hire somebody in your organization, that's a no-go, right? Ideally, you are hiring collaborators, right? You know that are working in teams of five to 10 people. Maybe you're hiring somebody in your marketing group, or in your technology, or support group, or whatever, but the expectation is that person is a resource for everybody. Everybody is helping each other out, right? Because you have multiple people working on the same initiative, project, and things like that, and so it's becoming more and more important for shared AI agents, fellows, whatever you want to call them, to be shared across a team, and how do you do that effectively, right? Because not everybody can, you know, stand around a desktop and open up the cloud for desktop or Chat GPT session and try and talk, and things like that. And Anthropic and Open AI are trying to figure that out a little bit. They're releasing some toolings there. There's personal AI assistants that have a little bit of that functionality, like Open Claw, and things like that, but that's that's thing is that that's the next iteration. So, like, I started a new project, a new company called Fellow Hire, and that is the goal there, where you actually have an AI fellow with a name with a face that's sitting alongside you and your Microsoft Teams or your Slack, and is working alongside your team. They're the shared screwdriver, they're the one that is doing the 10x multiplier to the rest of your team. Interesting. And I think that's the next evolution. And the other thing is, you know, with Claude, like, if you're trying to do, like, a shared Claude resource with your organization or something like that, you're trying to have it do everything you have, you're having to do marketing tasks, coding tasks, business tasks, research tasks, all these tasks, sales tasks, and I don't know about you, but if you've ever hired somebody who said I need you to do absolutely everything in my business,

Scott Groves  19:34  
they do nothing,

Taylor Dondich  19:35  
you get crappy output,

Scott Groves  19:37  
right?

Taylor Dondich  19:38  
Right, and that is exactly the same thing with large language models, right? You hear this thing about tokens and context window, right? It's the same thing as if you were dealing with a person, right? If you were like coming over and said, "Here's a folder with all of our business processes, learn them, right? You're right, the

Scott Groves  19:53  
executive assistant, the marketing assistant, the producer, the, you know, the salesperson. It's like, "Good luck.

Taylor Dondich  19:58  
Yeah, yeah, and. And you might get good output for like the first project you give them, and then you work on the next project, and you say, "Hey, do you remember that first project we worked on? That's when you start to see, like, Claude and Chat GPT start to suffer, right, because there is a minimum context window, just like you would have with a human being, right. So you got to kind of treat it like that, and so that's the other thing that I see is, you know, Claude, for example, has support for Claude in Slack, right? So everybody can, you know, talk to the Claude app in Slack, but you run into that problem, right? You've got your sales people talking to Claude, you've got your marketing people talking to Claude, you've got your tech people working to Claude, and now Claude's like the marketing person wants it this way thoroughly, so it's important that we start adopting AI technologies, which are a lot more role specific, so that they can be more efficient. You know, you hire a marketing person, you hire an outbound sales rep, you hire a support engineer, right? And you have clearly defined, when you are hiring somebody, don't you have a job description?

Scott Groves  21:02  
Right?

Taylor Dondich  21:03  
Right, this is what you're going to be responsible for. These are the tools you're going to use, and here's the acceptance criteria where you're going to be successful in your role. And also, here are the resources we're going to provide you, the people you're going to work with, you know, the tools that you're going to have to work with, like that's in the job description, right?

Scott Groves  21:18  
I love that.

Taylor Dondich  21:19  
So, you got to treat it like that, and you're going to get really, really great output like that. So, you know, like for in fellow hire, you know, we deploy very specific role AI agents that are trained on your organization's context, like what do you do as a business, but also the organizational unit they're going to work in, right? Are they going to work in sales or marketing, and what are the business processes specific to that, and then here are the team members you're going to work with, and as you start working on these projects with people, start remembering their names, start remembering what they do, started start remembering how they communicate, start remembering the big projects you work on, and in that, because it's so role specific and custom trained, these AI fellows are much more efficient, and I see that as the key to really strong AI adoption, right? Being a lot more specific, don't stop treating, you know, the.. I mean, if you've had a pocket knife, right, like a Swiss Army knife, right, it's got all these attachments, it can do so much, not really good at them, right, and are going to build a house with it, right. You're going to have a plethora of specialized tools to get the job done.

Scott Groves  22:27  
Well, it makes so much sense, right? Because, like, the more specific you can be with the inputs, the better the output. I mean, I would never go to Google and be like, "So, what should I do today? Yeah, it's like, "No, I have to say I'm interested in athletic outdoor activity. What are the options in Henderson, Nevada? Right now, I'm gonna get four or five suggestions that are like, oh, good, that's great. Right? I have a question, for, like, you know, I'm thinking of the average business owner who doesn't have a tech business, right, and they don't know, they don't even know we're talking about agents and fellows and tokens and all that stuff. What I have found, for kind of like your solo entrepreneur or your, you know, small business owner, is to me, if you can set it up right, the best option, or the best optionality in any of these AI models, is like as a thought partner, right? So, like, I run the coaching business, I coach sales professionals, loan officers, financial planners, realtors, and some businesses that need lead generation. I don't have a lot of need for, like, the technical side of, okay, Claude, go into my email and do this and come up with this Facebook marketing campaign and whatnot, but what I use it a lot for is a thought partner. I'm like, hey, I'm thinking about this. Hey, my coaching client has this problem that I don't quite know how to solve. What are some resources or whatnot? So, for that business owner that maybe owns a jiu jitsu gym or a bakery or whatever, and they're like, I kind of just want to use chat or Claude to think through some ideas. What are some guard rails that they can give the engine in order to not just get fluff, and we're trying to make you feel good and telling you what you want to hear, like, like if you were to send something out from scratch. I'm a business, I own a jiu jitsu studio, and I want to grow, but I don't want Claude to just tell me what I want to hear, and be like, well, just put a banner up outside, and you'll get more students, because we all know that doesn't work, but like, what guard rails would you start to tell the system, hey, remember this, be specific about this, poke holes in this, like, how would you start to train for, I guess that's the correct terminology, your AI to give you good output, instead of just telling you what you want to hear.

Taylor Dondich  24:19  
There are a lot of great courses, you're going to hear the term or the phrase prompt engineering, right? Prompt engineering, and everyone gets a little worried when they hear the word engineering, like, oh, it's technical. No, it is a very specific type of skill now that I think everyone needs to be educated on. They should be teaching it in high school, in colleges on effective prompt engineering, and you know there's tons of books out there, you know, at the bookstore on how to ask the right question, and literally it's the same thing, right? It's literally the same thing. How can you write? How can you ask very targeted questions, you know, um. You know, ideally you're not asking questions that are leading the other person, right? You're asking more open questions. The questions are very limited in scope, right? Like, let's talk about this, or you ask one question and then wait for an answer, right? I've seen a lot of people use Chat GPT, and they're saying they have, like, they're just vomiting questions before hitting enter, right? They've got 20 questions in there. Hey, can you ask this, and then do this, and then can you do this, and then tell me a little bit about this, and then they hit enter, and they wait, like, you know, a minute or two for a response, and again the LLM agent's gonna try its best, right, but it's going to be the same with you asking the same person, right. You and I are having a dialog, normally it goes, we ask a question, we wait for a response, we ask, we then ask for a more detailed question, right. We are not asking like 510, questions all in a row.

Scott Groves  25:58  
Dude, imagine how boring this podcast would be if, like, in the first five minutes, I just went through 35 questions, and then sat back to enjoy my beer, was like, 'Go,

Taylor Dondich  26:05  
yeah. And then that is a great way to say, because then you'd say, 'Hey, do you remember question one? I'd be like, 'Hell no, are you kidding me? I remember maybe the last question, right? And it's the same thing with AI agents, right, with large language models, you know, the more that you give them upfront, the harder it is for them to think about it. You know, when you're using a long session with, say, Chat GPT or Claude, or any AI assistant, or something like that, the longer your session, you start to realize that it's taking a little bit longer for them to respond, right? Right, and if you're technically aware, that's because you have really thrown a lot of tokens at it, right? Every time you give it a sentence or a question or anything like that, the LLM needs to identify every word, every punctuation mark as a token, and the more tokens there are, obviously, there's more data, but of course it's going to take longer to consume that data, and LLMs are stateless. They're stateless. They don't remember, they don't actually remember what you said last. Just the conversation carries over into the next prompt.

Scott Groves  27:19  
Can I? I wondered about this. Can I ask Claude to remember a conversation

Taylor Dondich  27:25  
with Claude for desktop? And yeah, yeah, if you have like Claude, a Claude subscription and stuff like that. Claude has this concept of like projects and things like that. And what they do is they have a flat file called memory.md and I don't know if you're familiar with like Markdown or it's like a text file, it's a Tegra file, right, and it's kind of like an organized text file, where it says this is a header and this is a paragraph that it's a very weak structured text format, but it works very good for larger language models, because really it is this token parameter text, and as you ask or have a conversation with, like, Chat GP or Claude. I'll use Claude. I love Claude.

Scott Groves  28:04  
Hey, too. I recently abandoned Chat GPT and went to Claude.

Taylor Dondich  28:08  
Yeah, I mean, for a lot of reasons, right? But if you're using, like, Claude for desktop or something like that, you could say, 'Hey, Claude, remember this. This is this is a phrase that I use constantly as I'm working with my AI fellows in Slack, I'll say, remember this key piece of information for this session and future sessions. When you say something like that to an LLM, and if your AI agent, you know it's an AI agent, is more than the LLM. There's a lot of tooling around an AI agent, right? Part of it is memory. Part of it is, how do we keep track of things long term? And you know, there's a lot of AI agent platforms out there that do it in a lot of different ways. Fellow Hire does it in a very different way as well, but that is a key thing for the for the AI agent to go, okay, I have to write this now into memory.

Scott Groves  28:57  
Okay, cool. I did that for the first time the other day. I was sick of uploading my style guide for consolidated coaching, because they like couldn't remember what my colors were, and then I tried to do, like, you know, a Word document with my logo on there. So finally I told, I said, like, Claude, why can you not remember my style guide? And it's like, oh, you don't have the memory button turned on. So I had to go to settings, turn on the memory, and then I upload. I'm like, for all future conversations around consolidated coaching, remember this style guide and my logos and whatnot, because it is kind of infuriating. It's like talking to somebody with Alzheimer's, yeah, where it's like, no, no, I told you that like 17 seconds ago, but a lot of the LLMs will not remember stuff.

Taylor Dondich  29:36  
Yeah, I think I think it's very important for us to maybe step away from the term LLM, because now you're getting into AI agent territory, and LLM is just the model, it's the large language model, tokens in, tokens out, right? An AI agent includes a lot of the environment are wrapped around LLM to provide things like memory.

Scott Groves  29:57  
Okay,

Taylor Dondich  29:57  
and let's talk about memory real quick, because you said, hey. You had to flip on memory, and now, oh, great, Claude's now remembering it, but maybe my responses are just a little bit slower. That's because, again, AI agents are sessionless, they have to take in your memory file every time, every time. So, if you have a memory file that starts to fill up, because again, you know, you tell Claude, hey, remember this, remember that, remember this, remember that there is this memory MD file that the entire contents have to be put into every conversation, every session context, right, and as the conversation goes on, Claude is constantly trying to optimize the tokens that it has to send every time, so it loses a little bit of the detail every time, every time, every time, so you know, if you've been in an hour long conversation with Claude, you start to see little artifacts of that degradation, because it starts to forget things that maybe you talked about at the beginning of the hour, right, or it's like, oh yeah, I had to look that back up. It's because it's had to like reread the memory file and things like that. So, you, your conversations to be effective should be targeted and short-lived. Okay, okay. So, if you say, "Hey, do research analysis on this competitor, it goes off to the web, and it does that, and everything like that, and it presents to you know, big blurb of text of what it found. Then you say, okay, write the key findings into a file, and refer to this file in your memory whenever we have to refer to this competitor again. Great, done. Start a new session, okay? Because all the key findings are now in a file, right? You start a new session, all your tokens have basically been cleared, clean slate. Now, if you ever want to work with that kind of data again, you say, 'Hey, Claude, read this file about this competitor. Now those, now that context is back in, and now you do that. So that's that's what I say: start multiple sessions, you know, multiple times be very targeted in those sessions. Remember, what if there's anything you need Claude to remember, or any agent to remember for that session and future sessions, state that right? Some agents will be easy, you could just say, "Remember this. Some agents, you might be like, "Hey, write to this file. Okay, you know, but do that. Do that that way your future sessions again can be very targeted to what you're trying to achieve, and then you can pull in data as necessary.

Scott Groves  32:27  
So, let's talk about the business owner that's maybe okay. I'll spend like I do 20 bucks a month on Claude, right, for the for the basic functionality, and then we'll move to like the super advanced stuff that you do, which I'm guessing you're not cheap, but we'll get to that. So, let's talk about the $20 model. I'll give you a couple examples of like what I've used Claude for in helping with business tasks and whatnot, and it's usually like analysis and thought partners. So, for example, on the Henderson HQ, I told Claude, "Hey, here's my Instagram address. Yeah, I want you to analyze all the analytics and tell me what content is performing the best, and it went and then it told me, hey, it seems like you're getting a lot of feedback when there's dogs and restaurants, so more dogs, more restaurants, amen, that will help your exactly, amen, that will help your Instagram virality, and then, and then I was able to ask some deeper questions about, like, how should we be tagging stuff, how should we be maximizing for audience participation, so we can grow only a Henderson audience. That was great. So that was one use case. Another use case for my business partner in the mortgage space, which I handed off to Leah. Shout out, Leah, James, if you need a mortgage. I said, "Hey, I want you to analyze our Yelp page and give me the top three reasons people work with the Scott Groves lending team, and it went through, and it actually gave me some ideas that I was like, 'Oh, I did not realize those were the things that clients were keen on as like a good client experience. So those were kind of like very basic level one use cases that business owners can kind of implement AI in order to like improve their business. Can you give me a few more on like the basic any business owner could do this with the $20 version of Claude. And then we'll get into the super fellows and all theirs and stuff.

Taylor Dondich  34:07  
Yeah, yeah, I mean, you know, you kind of nailed it, right? Like, if, if you're doing like competitor research, or you're saying, 'Hey, why is my page? Give me an idea on like why my page is performing well, or you know, take a look at my socials, tell me, you know, anything that you can find about, like, my followers or anything like that, anything that you could do yourself with, like, an open browser, AI can do pretty well, right, to a degree. A lot of tools out there now are getting a little bit more protective, like, for example, you can't tell an AI agent, "Hey, go on Reddit and start automating my posts, or or start, or even gather intel on your

Scott Groves  34:46  
automation. Most, most business owners, like solo entrepreneurs, they won't even get to the automation. It's

Taylor Dondich  34:52  
not even automation, it's just like, hey, go find the top 10 communities on Reddit to to promote my idea and. Then most AI agents now will say, hey, Reddit's not letting me, right, right, like even Reddit's like, you know, blocking automated AI traffic and things like that, because again, it's a computer program, and it identifies as a computer program, and there's a lot of platforms out there that are trying to block that, right, but there's ways around it, but right, you know, generally anything that is like, hey, you know, I'm going to take a look at my Instagram, and I'm going to take a look at my followers, and try and figure out some information, and that might be like a four hour job yourself, right? Because you can go down a rabbit hole, but you tell that to an AI agent, an AI agent can do that in a matter of like couple minutes, right? So those types of tasks are great, right? It starts to get a little bit tricky the moment you say, "Hey, I need you to log into this service. Got it? Right, so

Scott Groves  35:45  
it's like, if you want, if you want AI to look at something, evaluate something, that's a great business use case at $20 a month on Claude or Chat GBT, or whatever. But once you, once you want to start doing stuff, that's where it gets more complex and expensive.

Taylor Dondich  36:00  
And here's another very important thing to know: is that LLMs or AI agents are not aware of the current present.

Scott Groves  36:08  
Okay, like, let that let that soak in for a moment. That was really philosophical.

Taylor Dondich  36:13  
Large language models have to be trained on data, right? And so, ChatGPT, Claude, they have to be trained on data, and there's a cutoff date on that data, because they have to process all that data, you know, like the most recent Chat GPT models really have only been trade up to data ending in December of last year, so got it, you know, the last six months LLMs don't know anything,

Scott Groves  36:37  
so Chat GPT has no idea why the straight of her moose,

Taylor Dondich  36:41  
yeah, yeah, they don't know the political landscape of the last six months, which has been a doozy, right, right, they don't know what your competitor did in the last six months, you know, right away,

Scott Groves  36:52  
right,

Taylor Dondich  36:52  
that's why there is now again AI agents, we're not talking about LMS, AI agents, and AI agents have tools like fetching things from the web, got it right. So, when you say, tell me about Company X, an AI agent, if, especially if you're using a lower model, right, like Haiku or Sonnet, instead of like Opus and Claude, or an earlier version of Chat GPT, it might just tell you the data that it knows that it was trained on right, so you might need to be a little bit more specific and say go out on the web and identify the most recent news in the last month, three months, six months of Company X. See how that's a much more targeted question versus tell me, tell me the news about Company X,

Scott Groves  37:40  
better answer better

Taylor Dondich  37:42  
questions, better answers. So, if you are interested in things that have been happening in the last six months or so, first off, if you're going to use a model for market research or for competitor analysis or anything like that, know when the training data was cut off, okay? And also recognize that it's only going to tell you, even if you say go out on the web and do this, it's only going to tell you things that it found publicly, right. And so it might look at Crunchbase and things like that, but there's a lot of great data platforms out there, which require an account, but will get you much richer information, right? Like, if you're a sales outreach, right, you could say, hey, go find me some emails for this company, and it might return you like the emails that are on the public website and stuff like that, but it might not be able to tell you all the key decision makers in the organization and their contact information, because you do need some sort of data provider to give you that access, they have all that rich data, so you have to understand the limitations of what AI agents can do, right? They, you know, if you are a person using that $20 cloud, that $20 cloud is probably going to have some rudimentary tools that go out into the public web, that's that's it, right? If you're using like cloud desktop and things like that, you can easily connect it to like very, very common services like Google Workspace for Gmail, and things like that. So now you can say, "Hey, go look at my inbox, prioritize my emails for me, or highlight the ones that require immediate responses. Again, if you say something like that, that's vague, that's that's LM right there. But if you say, "Identify any sales-oriented emails that require feedback from me or identify any emails that are coming from customers that are reporting an issue. Those are very, very targeted questions, but that's something a small business owner can easily do.

Scott Groves  39:34  
I'm thinking we used to have a rule on my team when I was running a very big mortgage team, you know, we like my partner and I were doing like 100 and $60 million in production, and then we had 13 people in our branch, and we had 40 people in our region. We were managing all this stuff. I mean, 500 emails a day was a light day, and so we had to create this rule where it was like, if my name is in the to line of the email, I will address it. If it's in CC or BCC, my virtual assistant is going to auto delete it. Or move it to the file folder for that particular loan, and I'm just thinking that would normally take her hours a day of work. Now, if I could just set up an agent to be like, hey, same rule, if the subject, or if the line is to me, or if the subject says, you know, reaching out for the first time, whatever, I want you to leave that in my inbox, everything else gets archived or deleted, or whatnot, like I mean, that can happen almost, I imagine, instantaneously. Could have made my VA so much more effective five years ago, 10 years ago, when I was running a big team.

Taylor Dondich  40:30  
Here's another key thing, right? Like, you're talking about basically email filtering and triaging, right? Yeah, if you can't write a flow chart of what you want, then you haven't gone into enough detail with your AI agent, right? If you say, "Hey, Claude, go check my email and prioritize my email, you know what you mean, right? Right? Like, well, I prioritize sales over marketing, you know? I prioritize B2B over B2C outreach, and you have this mental model, but you haven't actually describe that raddle to the AI agent, so as an exercise, if you say, you know what, I'm going to ask Chat GPT or Claude to do this, whether it's a one-time thing or whether it's an automated process that you want to do, say, "Hey, at 3am every morning, do this that way, at 6am I have this on front of my plate, if you can't outline the exact steps the agent should do, then you, you're not ready to give that to the agent, because the moment you have the agent have some level autonomy, right, having giving access your AI agent access to your email is a pretty cruelty, trusty kind of process, right, because it's

Scott Groves  41:40  
only two years to get done with my virtual assistant. You're

Taylor Dondich  41:42  
not gonna, you can't, you can't give an AI agent read-only access to your email if you connect it to your Google or anything like that, right? So, the most dangerous thing you can do is just have that AI agent run off the rails because it thinks it knows what you want, right? So, you have to be very, very specific. We call those things guardrails, right? Keep this, keep this agent on guard rails, you know, so if you can define your workflow as a flow chart or a list of steps, you're in good shape, and you need to include that list of steps in your AI agent.

Scott Groves  42:11  
This gets kind of meta, but could I ask, we'll see with Claude, could I ask Claude the right question to have it ask me all the questions that it needs answered to do that correctly.

Taylor Dondich  42:23  
I do that all the time. I'm like, hey, all right, Claude, I need to build a sales outreach campaign, and I need all the steps there. I need to make sure that there is no room for assumptions by AI agent, things like that help me draft this battle plan out,

Scott Groves  42:44  
and so it'll ask you all the questions one by one. Yeah, I think you probably have to tell it to ask you one by one. That was my mistake one time. I was like, ask me all the questions you need to know, and 300 questions. I was like, okay, ask me these

Taylor Dondich  42:55  
one at a time. You know, one of my workflows is so I use Notion for, you know, my documentation space, right? A lot of people use Google Drive, things like that. I use Notion because I work with other people too, right? So it's very important that my AI agents or my fellows write to a document space that other people can see and provide feedback and things like that. But that allows me to have a living, breathing document of processes. So what's an interesting thing is, if I ask an AI agent to do something, and I know I'm going to need to refer to that process again in the future, I say, "All right, we're going to draft this process out in Notion, right? And let's draft it out step by step, and the AI agent will do that, it'll connect to Notion, it'll write it all out for me, which is great. I'll review it, and then I'll be like, hey, you know, let's think about this step or that step until it's refined, and then what's great about that is I can say, okay, at 3am refer to this Notion document, execute all the steps, perfect, that way in the future, if I need to modify those steps, I can modify it in Notion, and the next time that AI agent runs it runs through that that playbook, right.

Scott Groves  44:03  
So that

Taylor Dondich  44:03  
is one thing, is if you're going to start having your AI agent do repetitive business tasks, right, or you need to refer to some kind of process that you did before in the past, right. Find me 20 leads or whatever, build a playbook, build a collection of playbooks that the AI agent can easily refer to right, and is known as like the source of truth.

Scott Groves  44:25  
So, let's talk about you and your business specifically. Let's say there's a business owner in Henderson or Las Vegas who's like, I know there's some efficiencies to cut to pick up in my business. I know that again, whether I'm a banker or a tech guy or a jiu jitsu studio or whatever. I know AI could make my business more efficient, and they want to hire you to make that happen. How do they do that? How do you go about entering that relationship? What does it cost, which is not a fair question, because the use case could be crazy. The

Taylor Dondich  44:53  
use case can be crazy, but I mean, so I started a company called Fellow Hire, you know, it's Fellow hire.com Take a look at it, it follows a lot. Lot of the things that I talked about, you know, you want to bring on AI, autonomous AI agents that are working outside your team are very role-specific and are highly trained on the business processes and tools that your organization uses, right. So, this is not for the solo business owner that's using the $20 clod model, right. This is for someone who says, okay, I have a team of five to 10 people or 100 people, and I need to introduce AI, but a lot of my work is, you know, first off, nobody in our organization is technical, right? I'm a law firm, right, I'm a marketing agency, nobody knows how to connect my, the tools or the right thing, I just need something that my team can talk to and perform these ad hoc projects and things like that on demand and be and just elevate everybody, that's the proposition. So it's not for the solo business owner, but it is for a law firm, it is for a marketing agency, it is for a car dealership or something like that, where like we have a team of people, we need to find a way for AI to be easily approachable. Right, I can't have my team logging into another platform or looking at another tool and training on that. You know, my team is already on Microsoft Teams or Slack, or we're all talking on WhatsApp or Telegram or something like that. You know, we already have our communication tools. We need something that lives in there and also be experienced in all the tools that we use, and so this is where, like, Claude for desktops are still fall down, because they'll provide access to the most common tools, but there's always going to be a law firm that uses some crazy archaic system on an old Windows machine, and they're so under their office, there's industry-specific CRM, yeah, exactly, or ticket managing software, or something like that, that you know is cool, or maybe they're their family member made, and they're using that, or you know, something archaic, or you know, some medical device, or something like that, that does not have a common platform, and things like that, but you still need something that can help auto integrate with that, so with fellow hire, what we do is we do an onboarding session and say what's your goals, right? What are you, where are you falling a little short on? What do you want to optimize, and if you were to hire somebody, that's the key thing, if you were to hire somebody to try and solve this problem, what would that job description look like, so we actually start from a human kind of perspective. We say, what's the role, what's the job description, right? And that leads a very organic conversation of, like, well, they, they're going to work with these people, and we're going to expect this type of tasks to be done, and they have to know these tools, just like you would hire normal human Pearson, and then what we do is we actually launch our own customer infrastructure, it's all SOC ISO compliant for the customer and everything like that, and then we have, we launch one or more AI fellows that have been pre-trained on the organization, so we've gathered all the information we can about the company, what they do, their industry, you know, their history, everything, their org chart. Who works in there? What are they responsible for? We preload the AI agent with all of the integrations with the tools that they use. If there is some weird archaic system or something like that, we custom build the integration for that. Yeah, whether it's through like browser automation or custom connections or special VPN connections or anything like that, or we, we can build, we build all that, and then we plug them into their existing communications platform, like Teams or Slack, and this fellow shows up with a name, a picture, and you can just start talking to it.

Scott Groves  48:39  
So here's an interesting use case that's relevant to Henderson HQ, so we do a piece of content, you know, whether it's a clip from this podcast or whether it's my jiu jitsu coach, Val kids jiu jitsu coach, Val's awesome, best black belt female in Vegas, you should get her on UFC BJJ, she goes out and does content for us, she'll go to a restaurant or whatever, and she has a piece of content, then my virtual assistant in the Philippines has to edit that she has access to the Facebook Business Manager, so she can do basic posting on Instagram and Facebook, but there's a limitation on the Business Manager. You have to be in the app natively from your phone in order to tag somebody or invite them as a collaborator, and then we would like to also put that video on Patch, but patch.com because they want you to be local or Nextdoor, any of these things, you have to have a VPN that's in Vegas, and then we would like to post it to a Facebook group, but business manager doesn't allow you to post a Facebook group, and then this I call the daisy chain of fuckery, at some level, even though everybody's working very hard, it breaks down, and so each piece of content, instead of being properly up and tagged and labeled and everywhere in five or 10 different systems, maybe gets up accurately on three, and maybe kind of half-ass on three, and not at all over here, and so it's like just like having a marketing fellow to be like, okay, we have a piece of content, you go to work and you do all. These things, and then you know, you like, I don't want the content or the newsletter to come from AI, because the whole point is like build local trust with a real human being or a real set of human beings, but for like the busy work behind the scenes, I just want a button that I click and say every day at 4pm look to see if there's new content in the Google folder, and then do all this work. Could you train a fellow to do that?

Taylor Dondich  50:24  
Yeah, absolutely. Like, that's that's the whole point about AI.

Scott Groves  50:27  
How much are you gonna charge me?

Taylor Dondich  50:28  
So, I mean, so fellows like that, and you mentioned a lot of stuff in there, is like, oh, some, some of these apps only have like a mobile interface, and some of these apps require you to be local, so it requires a VPN, and stuff like that. We've thought about all that, so we actually have a residential proxy service that we automatically provision across our AI fellows. We have the ability to control native devices like phones and things like that. We've, we've, we've thought of that, and when we sit down with you, and again, this is like a white glove like treatment, we go, okay, How are we going to build the playbook of some of the business processes, because that's what you define, you define a business process at the well-defined already existing business process.

Scott Groves  51:05  
I'm good at that. Yeah, the implementation,

Taylor Dondich  51:07  
yeah, yeah. And the implementation, you know, you're pretty tech savvy, that's great. But, like, you know, an owner of a law firm, or, or, or a car dealership, or anything like that, they're not going to know, they don't know what a proxy is, or anything like that, so that's why we need to take, we take all that technical know-how out of that. We don't have anything running on a computer in their environment, or you don't have to install anything, or anything like that. We can do that, that's a great business process, but there's also things that are ad hoc requests, right? So you might have this podcast manager, or marketing manager fellow that you've brought on that have these well-defined things, but you also have a team of other humans that might need ad hoc requests at any time, and that's that's where it gets really special as well. He's like, "Hey, let's work on a new project together, and, "Hey, Kyle, come into this, you know, into this chat room with this AI fellow with me, and let's discuss it as a team, and start building out a new process ad hoc. Yeah, so this AI fellow is really treated like an employee. Our fellow hire AI fellows are basically on a two-tier kind of setup. There's your standard fellow, which doesn't use the most latest foundation frontier models. They're not going to be like software developers, but they're going to be amazing researchers. They're going to be amazing at doing exactly that business process you talked about, that's already automated, you know. They're going to do like competitive analysis and executive assistant bookkeeping and things like that. If you define a role in that space, and that's about $1,800 a month.

Scott Groves  52:35  
Okay,

Taylor Dondich  52:35  
yeah. If you are looking for a highly technical or a highly reasoning AI agent, so like if you wanted something that that would take a lot of the tasks off of a paralegal, right, you need something that is very, very fine-tuned, very skilled in that space, highly reasoning. We have senior fellows, which are at 4000 Oh, I got

Scott Groves  52:54  
it. You have to do a lot more training, upload a lot more data. There's a lot more

Taylor Dondich  52:57  
training, so we do, like, for example, the first week is white glove onboarding, right. We work with you to train the AI fellow, so we're gathering all of your documentation, all of your business processes. We're getting very intimate with how you run your business, what tools are available, what do we need to enable for the AI fellow to be effective on day one. And then after that week, we bring them into your virtual collaboration space, whether that's Teams or or Slack, or whatever you have to get going, and then we do weekly check-ins. You know, how is the fellow doing? You know, is there anything we need to change? Is there anything we need to train on? Is there anything new, things like that? And then after that, we do like monthly check-ins, but we always have like a media support line in case you're like, 'Hey, I need the fellow to do this, and it's not responding very well in this, let's, let's fix it up, so it's a very white glove, right. We need to remove that technical thing, so you know if another organization that is not very technical savvy says, "Oh, I'm going to bring AI in here, and I've downloaded Open Claw, it's all the rage, and I've got it running on my Mac, and I've given it full access to all of our business accounts, and it now has access to our file storage, and everything like that. Let's go

Scott Groves  54:04  
be prepared to be scammed, be

Taylor Dondich  54:06  
prepared to be scammed, hacked, deal with, you know, upgrades, and you're still have to be that technical person to deal with it. So, you know, fellow hire, you know, we are fully SOC ISO compliant. We handle all the infrastructure, you don't have to host the thing, and we make sure everything is stable and reliable, just like you would want a reliable new hire, right? Yeah, and so it's

Scott Groves  54:29  
really funny because nobody would hire like an executive assistant and be like, 'Here's my bank passwords, here's the keys to kiddo at company, here's access to all of our client data. It's like you earn your way into that level of stuff, but people are doing that with, like, Open Claw,

Taylor Dondich  54:43  
and just giving them keys to the castle. And then you see the horror stories, and if you search for horror stories, you see a lot of them. Yeah, and they're like, "Oh. And then I put all this trust in this machine, and then I upgraded Open Claw, and nothing's working.

Scott Groves  54:53  
I was just reading a story the other day, and he posted the screenshots, and I know what the screenshot looks like, because I've looked at 1000s. Of bank statements over my mortgage career, some guy, he had wrote, he had written the perfect algorithm to day trade and gave Open Claw access to his Charles Schwab account, and I think within 12 hours he was down $380,000 or anything. Yeah, just like what's

Taylor Dondich  55:17  
in the machine, right? Whoops, no guard rails in place. And so you bring up a good thing, you know, with fellow hire, when we do that onboarding and stuff like that, we say, well, what's the success criteria for this fellow, what's what do they need to do right, and so we go just like with an appropriate hire, when you hire, do you have a successful 30 day plan, 60 day plan, 90 day plan for that important people,

Scott Groves  55:37  
no, but

Taylor Dondich  55:38  
no, but you should, right, and so we're kind of like training the client, say, "Hey, in the first 30 days, what do we want to see this AI fellow doing effectively? Great, and we drive that as the North Star, right? And as it goes on, and we build trust into the AI fellow, we give it more keys to the castle, right? We give it little by little, right? It's access by least privilege, you know, like you want to make sure that you're giving just enough to make sure that you're refining the processes that the AI fellow is doing exactly what it needs to do, just like you would do with a normal hire, right? Except that they're 24 or seven, they're never tired, they're running multiple conversations with multiple people, you know, and they have immediate access to do tasks that you know, somebody would have taken hours to do the key part there, though, is that they're working with real human beings. So this was a very interesting experiment, you know, when I, in an organization where I was working at, I brought in a bunch of AI tools for, and it was a technical company, right, and so you had like tools that would monitor their project management as soon as you assigned it to the AI agent. The AI agent would run through and do it, and everything like that. But it was very cold. It was very cold, and engagement wasn't very high with the staff. Same with, like, saying we're going to adopt AI. You hear a lot of business owners say, we're adopting AI, and you'd get a lot of pushback from the employees, you know, the rest of the staff is like, is this taking away our jobs, is you know, how do I even work with this thing, and stuff like that. And so the interesting I did at this organization is kind of like the starting point of fellow hire was I decided to build an AI agent that had a face and a name, so instead of Chat GPT or Claude, it was Morgan in marketing, and you pop them in there, and I said, you know, you know how I said, like, you know, every agent you kind of pre-train with fellow IR, we pre-train and say, okay, not only is this the organization, stuff like that, but I also say, here's your personality, here's your quirks, maybe you love waffles, you know, maybe, maybe you're overly excited whenever a new project comes on, something that just gives a little spice, right? And that, along with the name in the, in the picture, the avatar of a real person engagement went through the roof. Now you're seeing people actually talking to these AI fellows saying please, thank you, they're being mentioned in all hands as like a great employee, things like that, because it's removed kind of like that cold tool factor, and now they feel like they're working with a teammate versus a cold iron box, so that's been a really interesting kind of revelation there as well.

Scott Groves  58:21  
Well, we're gonna have a long conversation offline because Henderson HQ is one of, I think, 30 newsletters that our parent company, Hometown Legend, manages, and we have a, we have a lot of need for one of these AI agents immediately, so we might, we might be hiring a fellow from you. Awesome, but real quick, again, where can people find you so that they can reach out? If they're

Taylor Dondich  58:40  
absolutely, you know, the best place to reach me is on my LinkedIn, so just search for my name, Taylor Donnage on LinkedIn,

Scott Groves  58:46  
spell it,

Taylor Dondich  58:46  
Taylor T A Y L O R, then Donnage D O N D I C H, or you can just go to Fellow hire.com reach me there, taylor@fellowhire.com eager to talk to you. I love talking about AI, I love talking about technology, what works, what doesn't. You know, I've taken my lumps, I've gotten the experience, I've seen success, I've seen failure. So I'm eager to talk.

Scott Groves  59:07  
Well, we're gonna, we're gonna have you back on after we work together and talk about all the successes. So, thanks for being on, man.

Taylor Dondich  59:12  
Yeah, thank you.

Scott Groves  59:15  
Hey, it's Scott Groves with the Henderson HQ Podcast. I hope you got something out of that episode. If you enjoyed it, please don't forget to like, comment, and subscribe to the podcast. It really helps the show grow. And by the way, if you are a business owner, or you know a business owner who has an interesting product, service, or just an interesting backstory, please, please get in touch with us. Email us at the Henderson hq@gmail.com We would love to interview you, because that's what this show is all about. It's about building community, supporting local, individually owned businesses, and just making Henderson a great place to live. And don't forget, go to Henderson hq.com and make sure you sign up for our newsletter. We send out a once a week newsletter, no space. Am about the most interesting local businesses, hot spots, restaurants, community events. Thanks for watching the show. Really appreciate you.