How do you layer AI onto a people-heavy service business?

Jason Kirby (00:00.217) Hey everyone, welcome back to Fundraising Demystified. Today, we're going back in time a little bit with a very close friend of mine, a friend of mine that pretty much introduced the word entrepreneur to me back when I was 18 in college. John, it's great to have you on the show. Finally, you know, I'm so excited to have you on the show. So John Tucker, founder of Help Flow, bootstrapper since day one of all of his businesses ever since I've known him. I'm excited to have you on the show, John. Jon Tucker (06:41.388) Thanks, man. I appreciate you having me, man. It makes me feel old as we... As you said, I don't know if you said it in this intro, but like we met when we were 18. We're not 18 anymore. So time goes by fast. Jason Kirby (06:51.201) Yeah, it's been a long time. But John, the reason why I wanted you on the show is you and I have been riffing offline about integrating AI into our companies, how we're using it, how you're using it. You're on a 200 plus person organization and it's heavily built on human labor. And you're being so proactive and integrating AI that I thought it'd be important to have you on the show. So that founders of startups and other companies could understand how, you might not be an AI company, but how you can use AI to improve your efficiency and build better products faster, quicker, better. So John, just to kick things off, can you give everyone a little bit of background on yourself and what you're doing at Help Flow? Jon Tucker (07:37.294) Yeah, so Help Flow started as basically a customer service agency in 2014. So it's been a little over 10 years. And by CS agency, I essentially mean doing customer support for other e-commerce stores. So we have agents, we staff those agents, we train them, and they do customer service for our clients. And so started that in 2014, have always been like really tech minded. I remember hacking together our initial technical systems and How do we learn the client's business and give the agent the right resources and like all of those things. So at our core, like we're a people business, right? Like we do customer service, at least for the first era of the company. But we've always been technical. We've always been systems minded. And so a lot has happened in those 10 years. I'm sure we'll get into, but now we're essentially at the point where we provide virtual assistance, like human VA's for any business, for any role, not just customer service, not just Ecom. but we layer AI onto the entire thing. So it's an AI-empowered human virtual assistant. And we've been able to accomplish a ton with AI, so it's been super, super exciting the last couple of years. But that's the journey in a nutshell. Jason Kirby (08:46.669) And so what's interesting about this and kind of what I think every founder needs to be thinking about is, we always look at like growing headcount and, know, growing the team to be able to handle more capacity and hyper scaling. And that's what the venture world is all about. And, you know, you have a 200 plus person team, but when you think about all the AI tools that you're doing, like, how does that change your staffing requirements and head count requirements? Jon Tucker (09:14.222) Yeah, I think I would almost answer it in like two ways. one of the things we've noticed with customer service over the years is, it's very systematic. And so I started to realize a couple of years into the business is like technology is going to replace this. And it happened like very fast over the past couple of years, but like in terms of our clients and what we do for clients and like how many customer service agents do you need? It's significantly less now because AI can handle a lot of the tier one questions, right? And so just in like our core business, it makes it so you need probably half the customer service agents that you used to. And I think that'll accelerate really quickly. Internally as a team, I look at it as basically an efficiency game where basically like a team member on our team in sales, for example, can handle two to three X the deal flow that another competing agency might be able to handle. Cause we're using AI for a ton of the steps and all the meeting preparations and all the email outreach and all that stuff, like we're layering AI and all of it. So I think of it, not so much as being able to downsize our team, but more so being able to increase the efficiency of the team. And I look at that as a competitive thing where it's like, okay, I know some of our competitors very, very well. They're using chat GPT maybe, right? And like that's very far off from where we're at. And so I look at it as almost, like a knife to a gunfight type of situation. I think we're in an interesting spot in AI right now where you can just run circles around people if you know, not even the advanced stuff, but just like, you know, more than the basics and you use it every single day, you can accomplish a ton with it. So I look at it as like a booster essentially to people. Jason Kirby (10:55.043) Well, what comes to mind is also just competitive rating, pricing. When you think about having AI supported, you know, staff, you're not ha you could probably lower your price while keeping the same staff and keeping the same profitability, to be able to compete with maybe, you know, full AI automation solutions, but still have the human oversight that still frankly, a lot of people want. so let's talk about how you think about building these systems. You know, we talk about you having kind of a algorithmic mindset when it comes to building systems and operations. you know, I have a little bit of that overlap myself. So go with that way too many rabbit holes in this stuff. But how do you think about building the systems? it like Frankensteining a bunch of Zapier integrations? Is it building custom code? How do you do it? Jon Tucker (11:41.292) Yeah. So that whole concept of like algorithmic thinking, I've always considered myself like very systems minded is how I used to word it. And I wish I could remember who it was. I know we started the conversation on Twitter and then we moved it to a Zoom meeting, but I don't know who it was. If you're watching this, please let me know. It's cause I'd like to credit you. But this person said to me, he's like, John, you're an algorithmic thinker. Your mind like thinks in algorithms and process. And I was like, yeah, like that's a great way to say it. And so this whole... systems focus or algorithmic thinking is essentially being able to see the pieces of a process, right? So like when something happens, like what are the steps that get it to that point? And even one of our other friends, Greg from TicketKick back in the day, I remember an email he sent me before one of our meetings. He said, John, you are like the most process minded like person when you prepare for a meeting. Like I got everything here. Like I don't have to prepare everything. I know how the meeting is gonna go. And all of that, is really important with AI. Because AI can do anything, essentially. It can do a lot of stuff, maybe not anything, but it can do way more than most people are pushing it with. But you have to be very specific. And you could call that prompt engineering or there's a ton of terminology for it, but basically be clear on what you want done and give the context of what you want done. And so that whole algorithmic thinking mindset has caused me to really... be able to communicate effectively to AI to get really good results back from AI. And because I'm like that, our team is also like that, right? Like for good or for worse, our team culture is kind of based on the founder at a certain point, right? I know it expands beyond that, but we're very systems-minded probably because I was very systems-minded when we started the business. And so that's created an army of 200 people that know how to communicate to robots, because we communicate robotically sometimes internally. And that's created a super exciting time with AI because I realized like when we communicate in that way, we're able to have it do really complex things for us. And so that's kind of like the line of thinking of how we've gotten to this point in terms of how I think about communicating with AI in terms of the tools and like how, how we put it together. Definitely started off Frankenstein for sure. That was probably like. Jon Tucker (14:01.038) early 2023, somewhere around that time. And then, know, late 2023, I started to get clear on, you know, the open AI suite of APIs and doing a little bit of development work myself. I started to understand like how to move data around between these systems. We're now at a point, if we fast forward, where we're doing a lot of like agentic AI stuff, we're essentially like we're able to build these teams of humans almost is how I explain it to people. basically like through an API, we have a team of AIs that will go out and do stuff, know, research a topic, research a prospect, compile a report spot, like quality check and do all this stuff. And it'll come back with the work done. Right. And so from that perspective, we're essentially able to get a lot of leverage on ourselves as a team by essentially thinking of AI, like a human or thinking of AI, like a team of humans and kind of building the tools. So that's that's kind how we're thinking about the algorithmic mindset as part of it and then also understanding like what is AI in relation to your whole company and what your company does and how the team operates Jason Kirby (15:07.417) And see what I think is important for a lot of founders to realize, because everyone, know, most of the audience out here trying to raise money and, you know, be the, we are this, but with AI, and something along those lines, trying to get the big, big money, raised by saying they're an AI company. The reality is you can build great businesses, with very few people. If you know how to leverage these tools and build out these tools effectively and sure AI can. write the code yourself, but again, like you're saying, you have to really think like the, way the AI needs to be told what to do in a very strategic systematic way to get the output. one of the biggest problems I run into is the amount of tinkering required of just the constant tweaking and, you know, manipulating of the, prompt or the data that you're inputting. How do you go about keeping track of all that tinkering? Jon Tucker (16:01.566) EVALS. EVALS are, if you build an AI, you know what EVALS are. Maybe you're into it, maybe you're not. But EVALS are so important. EVALS are essentially quality control. Did the AI produce what you wanted? And doing that in an objective way to say, here's what I wanted, here's how I would measure what I want. And then you're essentially evaluating, did that happen? I think it was one of the YC guys, or at least it was on a podcast. They basically said like the two ingredients of building with AI is essentially workflow and evals, right? So like workflow is like, what do want done? What is the sequence of how it gets done, which is algorithmic thinking, right? So I've got a cooler name for it, I think. And then evals, like did that happen? That's basically like what you need to build with AI. And you could say you need training data and you need all this other stuff, which I guess is technically true. It depends what you're building. But with workflow and evals, you can get to the point where it's producing what you want. And so the constant tinkering is just part of the process of building. But I think when you have proper eval set up for what are you trying to do and what is the measure of good, essentially, it enables you to have an endpoint for that tinkering. And so one of the things that we do is, and this is more so over, Early 2024, I probably started studying evals and like how all that stuff works. Now we're at the point where for the things we're building, we have eval set up to say, okay, is this hitting the mark? And then once it gets to a certain point in terms of the eval and we say, okay, that is done being built and we will stop tinkering on it. We still revisit it. So like if we make big changes in our system, you got to run all the evals again, kind of see how it's going, make sure nothing broke because it always breaks. But once you measure it as like, this is what good is, it kind of frees you up to say, okay, I'm not going to take her on that one anymore. Cause with, think this has always been true for software. Like you could always tinker further. That's always been true. But I think because AI is a little bit of a black box, it's not like if then statements and code, the tinkering could go forever. Like you could just take forever. Jason Kirby (18:05.433) Yeah. And that's usually what I run into. I guess I don't really call it evals, but that's what we do with our team. So we build out our own algorithms and AI for, for Thunder in terms of the free tools that we provide, as well as a lot of the backend tools that we use for our clients when it comes to what we call like a capital strategy assessment. There's like sometimes like 14 different sections of data that we analyze companies and the market and all these different things on. And we realized that we can, we can. create a prompt to pull out the data that we need with like O1 or O1 Pro on Chatchabee T to get exactly like probably like 70 % of the work. You know, still gonna have a lot of human touch, but like the fact that 70 % could be done within like an hour of what would typically take an investment bank like weeks or like have some poor associate working 80 hours on a particular project and hating their life and. Jon Tucker (19:01.602) Yeah. Jason Kirby (19:04.697) you know, to having a partner do the output in a few hours. It's pretty transformational in terms of what could be done and the tools and whatnot. But when we look at the startup landscape and the transactions happening, we're going to throw something up here on the screen of how small teams are the future. Now granted, you have a 200 person team, but you're now getting more. Out of that team, you're not having to do layoffs or anything like that, but you're able to squeeze out more and more and more out of your team. And we're starting to see that across the startup landscape of companies being able to output more faster, quicker earlier. And the bar to accomplish, um, not accomplished, like, you know, get to the certain milestones that VCs want to see or potential acquirers want to see. You know, can do it with less and less people and less and less money, which I think is a fascinating. you know, scenario and you look at the list of companies we put up like cursor zero to a hundred million ARR and 21 bonds with 20 people, a hundred million with just 20 people and like lovable zero to 10 million, 15 people. then, um, you know, was the, other one? Yeah. Like magnificent, like 10 million in ARR with just two people. Like it's just wild what companies are able to do these days. Um, and so when you are looking at, at companies and and talking to various different startups or potential customers of yours, it's not your business. How do you look at these companies that you're talking to and advising them on AI integration or bringing AI into their business or automation into their business? Jon Tucker (20:45.294) Yeah, I think everybody wants to bring AI into their business. And then when you get into like planning it, like, what does that mean? Like, no, like many people are not clear. I don't want to say nobody, but everyone's excited. Few are clear on how to implement it. And so I think that mindset of understanding the process of your business, the workflow and the eval, I guess you could call it that. Like when you understand that. and you understand what's possible with AI and like what tools would be involved and roughly how would it be done, then it becomes clear to you how to integrate AI into a business. And so like the deal research process that you mentioned, you called it something else, but you know, the deal research process, going out and doing all this stuff that, you know, investment bank or what typically do, that is a process you know how to do. And I'm sure you guys have SOPs internally to do that stuff, but because you know it so deeply and you're doing stuff with AI, you can start to accomplish a bunch of it with AI. Right. And so I think, for businesses that are wanting to not disappear over the next couple of years, I think you have to understand how your business operates, which most businesses do. don't know if they could articulate it, but they need to know how it operates. Right. So take the time to like get really clear, but make sure you're paying attention to like what AI tools are out there and like, you know, use chat GPT, one, right. Cause that's like the best model now. I guess you could also say, three is better at certain things, but like use the most recent stuff. spend time researching how things are going and push through to the point of like actually understanding what's possible with these things. Because then your mind starts to like bring it all together and say, like for our deal research process, we could use, know, crew AI or some other, you know, agentic system. It doesn't mean you have to build it and code it, but you need to understand what's possible. And so, When I'm talking to businesses, I try to get like really down to like, what is the actual thing you need done in your business? And then let's figure out how AI fits into that. cause that's all that we've done internally is figured out, okay, can AI answer a customer service ticket? Absolutely. Right. for us, one of the things we've built that's really powerful is our onboarding process, which I think is very similar to the research process you mentioned. Jon Tucker (23:01.474) But like our onboarding process, like build a knowledge base about the client's business and build up all the FAQs and like all these things that our team needs that used to take probably six days, five to six days, two or three team members, probably all in, you know, 40 hours total across them. Not a ridiculous amount of hours, but more than one hour, right? Now we have it literally down to about 90 minutes and it's probably three times, four times better than the original one was. And cause we've, we've spent 10 years building up that onboarding process. So like, we've always been very meticulous about it, but then we started layering AI on to part of it. And now we're at the point where a client basically does like a 10 minute call with our AI system. don't even talk to a human anymore. They talk to the AI system that interviews them. And then out the backend of that 90 minutes later pops a super detailed knowledge base that my team then reviews and like, kind of annotates additional stuff that's needed. So there still is a human review, but it's. night and day different than what it used to be. The takeaway, I think for the listener is the reason why we were able to build that is we spent 10 years understanding what's the best way to onboard a customer. So we know our process, but I spent three years going really, really deep into AI to really understand like what is possible? How does this stuff work? What, what does a Gentic mean? Like is crew AI the right tool set to use? I got lost too over the years, but it all started to come together eventually. I went. like we could just do this with our onboarding process and now it's 90 minutes to get what used to take a week. Jason Kirby (24:33.463) Yeah. And insights that I have as I'm out in market, looking at &A deals and private equity firms and stuff like that, one of the common things, most common things are founders that have sold their companies looking at what to do next. And they're seeing this wave of AI and all these legacy agencies, heavy labor intensive businesses that, know, white collar, like computer labor intensive businesses, and they're just going on. buying these businesses and layering on AI and max it either laying off a ton of people and becoming instantly profitable or taking the existing staff and being able to 10X their capacity. it's, it's either you play the game and you catch up or you get eaten alive and either get bought for pennies in terms of the valuation that you could potentially get. And these are not AI companies, not selling an AI software. but they're taking advantage of the latest, greatest technology out there to maximize efficiency. And so I'd be curious when you look at similar structured businesses as you were kind of talking about your competition, like what opportunities come to mind? Do you see it as like, let's out-compete them or let's see if we can go buy them or do we make them irrelevant? Jon Tucker (25:55.406) It's so funny you bring that up. I'm literally looking at a deal right now for a VA business that I think we could replicate over tons of VA businesses. The main premise is basically what you said. We can be way more efficient and effective than other VA businesses, right? And I think that's because of how we think of the actual virtual assistant. The typical VA business or really any business with employees is You train those employees, they build up skill sets and then they can do really good work, right? The thing that's unique with AI is you can separate that work, the value of the employee's work. You can separate it from tactical knowledge of how something should be done and doing it efficiently versus what should be done, the strategic decisions. like tactical, get the stuff done. And then strategic, like what should be done? What should be the decision here? Right. And what's happening now is the tactical what to do can be built into a process, which has always been true. but now you can have AI do that process. And so then it becomes, okay, like the key value of the employees is the strategic decisions, right? Also called reasoning, right? Now, as of last year, the reasoning models are starting to come out. So now the reasoning can be done too, not all of it, but some of it. And so. When you think of the value of the employees, you can use AI to supercharge that. And I think that the VA business is a place where you could very quickly grow by going out and essentially acquiring businesses, rolling them up and then running them in your way. are without going like super deep into it. I'd love to nerd out with you later on like business strategy stuff. I think we have two options. think one is to go out and basically buy these companies, roll them into our model and do it better. The challenge with that is we would need to bring their VAs into our model, right? Which is not hard. Like we still have a process to do that, but that would, that would be one challenge of that model. The other way is for us to figure out how we can do what we do with their VAs. So essentially do what we're doing, but do it in a software driven way. Jon Tucker (28:12.706) where we don't have to sell you a VA, we sell you the way to work with a VA with the tools using our software. And so I think we're considering both, both might be a good option, but I think in every industry, I think you and I have probably had discussions like this. And again, going back to like that competitive mindset, for some reason, I'm just fascinated with the opportunity to like go into an unsexy business. And just run circles around competitors. Like we used to talk about this with digital marketing, where it's like you go into an industry where like, nobody knows how to do digital marketing and you're the best marketer. Like you could have some really good wins with that. I think AI creates a huge new opportunity across probably most industries to come in and say, if we are the best at integrating AI into a business to provide the value that the business provides, make it more valuable, like more valuable to their end customer. Right. Cause you're like, you're able to do it faster or better with AI. and more efficiently, so the business value is up. I think there's gonna be a ton of those opportunities over the next couple of years. And I think that's one path to capitalize on them is to basically buy businesses and improve the value by doing that. Another one is to embrace the fact that I think there are customers of every business, they're all gonna start to be looking at like, should we implement AI? So if a competitor of ours is not implementing AI effectively, we may be able to just sweep up their customers by having a better way to do things. And so it becomes a decision of like, do I buy that revenue or do I take that revenue? And the pros and cons to both, right? But I think there's opportunities across the board because of the disruption. Jason Kirby (29:45.753) Yeah, I think it's gonna be very interesting. I see a lot of companies in the, you know, kind of millions of revenue, but not venture scale or anything like that, where, you know, this type of process is so, you feasible of like, okay, do we, do we implement AI to be faster, quicker, stronger, and then go do business development and try to steal business? Or do we go in and buy our competition knowing that we can perform at a better, you know, gross margin or evident margin? than our counterparts. And I think that's going to be a major transformational shift over the next 10 years across all industries. And something that I think will be pretty powerful to witness as, you say younger talent, you know, comes to market and be like, why do you guys do it this way? Or what we're seeing with Doge, with Elon Musk and the US government of like them having to take an elevator down a mineshaft or like file like Jon Tucker (30:33.282) Yep. Jason Kirby (30:41.657) retirement papers for US government employees, you're like, what is wrong with our world today? Like how much opportunity gap is there to close by removing those levels of inefficiencies? Jon Tucker (30:47.246) It's crazy. Jon Tucker (30:54.382) Yeah, I think AI, I mean, on that note, I think AI took these inefficiencies. There's always inefficiencies in business, but it took these inefficiencies and like just blew them up very quickly. So like what's possible now in 2025, you know, we're recording this in February is vastly different than what was possible in like February of 2022, right? And that's only three years, right? It's not a long time business, but it's like completely different what's possible. And you could argue that like February of 2024, one year, Like what's possible now is very different than what was possible then. I don't remember when GPT-01 came out, but like the concept of a reasoning model makes it so that the strategic decisions can be made by AI too. And a year from now, it's gonna be as big of a leap, I think. Jason Kirby (31:42.881) I know. It's like people make the comparison of like when computers, you know, got mainstream call it the eighties and nineties when computers started to be adopted and stuff like that. But like, I think it's even more rapid adoption. Like that's almost probably 10 X faster than computer adoption, just because one already, everyone already has computers now and two, everyone already has access to chat. GPT. There's like no barrier to entry to have access to this kind of stuff. It's really more of, do you have someone dedicated to this and focused on this? Jon Tucker (31:49.378) Yeah. Jason Kirby (32:12.185) now, either the CEO who's focusing on it or bringing in someone that's going to be dedicated on it and what systems can be automated, whether it's hiring agency or bringing in like a chief of AI. I'm pretty sure we'll start to see that across a lot of the, you know, fortune 500s is the chief AI role hiring like 25 year olds and 30 year olds. Jon Tucker (32:24.376) Yeah. Jon Tucker (32:33.717) Yeah, I got something to add on that too. I've been doing a lot of reflection like at the end of the year last year. And I think what's unique about the journey that we went through that I went through as the CEO of the company is I realized like five years ago that like what we do is going to be completely replaced. And it was probably early and I probably thought it was going to happen faster. But I kind of got to a point where I said, I think what we do today is going to be completely replaced by technology. And so that caused me to start to like look externally of like, what technology and how does it work and how could we adopt it and all these things. So I got into the technology and into the disruption, um, early on. And so I had time to really like understand it. And then I think what happened is there was a healthy mix of terror and opportunity. And it was this dance over probably like, two year period of just knowing like, holy shit, I got to figure this out. Otherwise I'm going to be displaced. And I've got actually some family background story around that, that might be interesting. Like why I reacted like that. But that's, that's what happened. I felt fear of being made obsolete essentially. And I'm way into tech and way into technology and very systems focused. Right. And so it caused me to be able to dig into it and actually enjoy doing it and love doing it and be interested in it. And I think that with any tech change, it's important to actually be interested in it and enjoy the process. You could hire a chief of AI, yes, but I think also like every leader in every business has to understand at its core, like how does AI function so that they can identify opportunities of where it fits. Cause that way you can go through that transformation effectively. Everybody will become obsolete because of AI if they don't embrace and use AI effectively. I don't know the timetable. It's faster for customer service than other industries. It's slower for certain industries, but everybody will become obsolete if they don't adapt to it. Right. And so I think it's really important as founders and business leaders to like actually be interested in it and enjoy it. cause it causes you to, to go into the rabbit hole and spend time. And eventually you start to see all the dots connect. Jason Kirby (34:45.689) So, you before we hopped on, you were kind of talking about this workflow that you were using in terms of creating internal tools and products. I'd be curious for you to kind of share some of your insights. Like we have your pet project of yours right now that you think is taking the latest and greatest AI and kind of sharing how you're thinking about it what problem you're trying to solve. Jon Tucker (35:10.094) I could go in a couple directions with this. I know we talked about the coding one. We could go into a little bit of that. Or just in general of how we're taking our current process of layering AI onto the team and taking it to another level. What do you think would be most valuable for people listening? Just knowing your audience. Jason Kirby (35:30.999) I think the coding example being that you're not a coder, but you know how the process works and systems work to be able to conduct your team appropriately and effectively, I think would be an interesting example. Jon Tucker (35:34.604) Yep. Jon Tucker (35:42.146) Yeah. So I think, I think for people that are running software businesses, I think it's, critical to like go very, very deep into, you know, AI assisted coding. you know, one of the guys, that I think is really pushing the envelope on this is, I think his last name is right. McKay, right. We can link it up in the show notes, but these workflows you can build with like chat, GPT zero one. Claude cursor and like knowing how to work between those tools and how do you give your entire code base to zero one since there's not or zero one process. There's not the API at least as of now. Like when you try to figure out a workflow for how all these things can help in your software development process, you can get a lot more done. And so I think for every, you know, software company, I think it's really important to make sure that your team is starting to use all that stuff. But even for non-software companies, which I would put us in that bucket, at least for now, I've spent a lot of time specifically over the last 18 months learning how to code essentially, like code in this new world. So understanding front-end, back-end, frameworks, what's the pros and cons of different frameworks, to use the same framework between the front and back-end, database stuff, users, payments, all these things. There's a lot of aspects of software development. Because I've gone deep into how software development works and how do you architect a piece of software, we're starting to realize how we could take what we do and turn it into a software product and do it fairly quickly using AI coding in a solid team. And so I think the reason why I bring that up is I think it's important to start to understand how software is built using AI. because then it becomes way more feasible for you to integrate AI into your business, right? Integrating AI is not as straightforward as like just go out and buy, know, Salesforce is doing some cool AI stuff. Like maybe we should move to Salesforce, right? Or there's a bunch of help desks that are doing, you know, some cool AI stuff for customer service. Maybe we should just integrate them, right? Like that might be step zero, but like the most effective step to take or the most important step to take is to understand. Jon Tucker (37:56.35) What are they doing and how are they doing it? So you can start to apply it yourself. and I think studying how software is built and studying how all this AI stuff works, is the path to get there. I would wrap that up by. You got to watch for this blind spot, I think, which is really important. When you start to see AI tools out there, a lot of them are doing some powerful stuff, but a lot of them are just, you know, nine to. nine to 12 months ahead of maybe you or maybe their competitors. And so they come out with this product and say, look, like you can automate a customer service ticket. And they wrap it in a nice interface and do all these things in marketing, but it's automating customer service ticket. If you know nothing about AI, that's insane to see, it's magical. But if you understand even just a little bit about AI, you go, aren't they just passing the ticket up to OpenAI and then getting the response? Sort of. Like what else are they doing? Like once you understand a little more about AI, you can understand what these tools do and then you can work with your team to build some of the same functionality in your system for like eight to 10 X cheaper than some of these companies are charging. So I think it's really important to not just adopt like AI tools, but again, to understand how they work. So you can build some of that into your own, your own functionality. Cause otherwise you become, you become somebody else's margin. Like you don't really get the margin benefits. by adopting this stuff. Jason Kirby (39:26.585) Don't get them out. You're someone else's margin. I like that. That's a quote. Jon Tucker (39:29.774) is a Jeff Bezos quote, right? Your margin is my opportunity, I think. Yeah. Jason Kirby (39:32.218) Yeah, your margin is my opportunity, but your someone else's margin, I think is a different more like a Jon Tucker (39:37.058) Yeah. So open AI wins the margin battle. Like they're going to eat all the labor of many, many industries, but in the interim, there's going be a ton of companies that take the open AI margin and then market up, you know, six X or something somewhere around that. And they'll have a good business model for two, two to five years. Some number, I don't know how long. I don't know how long it'll be. A lot of margin of many companies will do very, very well. But there's a lot of companies out there that like the tech they're doing is not that complex. If. Jason Kirby (39:54.809) Two to five years. Jon Tucker (40:06.442) if you're interested in learning it. But if you're not, you just pay that. That's still fine. It's still a good win. Jason Kirby (40:11.565) Yeah, I think at the end of the day, people still for as long as our generations around and maybe even Gen Z still like there will still be a desire for human engagement and human interaction and doing business with humans. But if you can make humans faster, stronger, better, it's like, you know, all the cheesy. Marvel movies and other movies out there like the AI soldiers and stuff like that, or like the robotic soldiers and make them super enhanced. Like it's always like a common theme, but like at the end of the day, people would love to have that functionality, but still have the human oversight because well, humans trust humans as well for the most part. But I think, um, this concept of you have a short window to capture a margin opportunity while things are still difficult. Things are still hard to figure out. You can build the business that does it for people for a while, but at some point. Jon Tucker (40:32.974) Yeah. Jason Kirby (41:00.729) You as you were going to say, like that, you might be able to just do it yourself. a lot, I see a lot of companies talking about ripping out legacy ERP systems and, legacy software says internal tools and just building exactly what they need only for what they need, uh, faster, cheaper. like I was at Walmart and we had, it said billions on software and then they spent billions on people replacing the billions in software that they were spending. because it was just such a scale, like paying $10 per seat for 3 million people, know, it gets a little pricey. And so they were like, well, we can build a team of 150 people and build that tool ourselves. Now it's like they can build a team of five people and build themselves. So it's gonna be a very, interesting future ahead. know, John, this has been really fun talking AI, future of AI, how companies should be leveraging AI and their startups. But for people that don't know you, John, they want to learn more, what's the best way for people to learn more about you? Jon Tucker (42:04.044) Yeah, I think helpflow.com is probably best. You can learn more about our approach and what we're doing. We're doing some cool stuff with virtual assistants. And then the main thing I would do, whether you're interested in working with a VA or not, if you go to the site, you'll see a call to action where you can literally talk to the AI on the phone and just experience it. One of the big things that we do with VA is we make it so you don't have to over explain everything to your VA and type out all these emails or SOPs. You basically just talk to an AI on the phone and it will interview you about the work you want to under the process. And then it creates all the SOPs for the VA. But you can literally do that on the site and just talk to the AI and like really experience it. And that tends to like kind of open people's minds quite a bit to like what it's like to actually work with AI. And so I think that would probably be the best bet. Just go to helpflow.com, check out kind of our approach of what we're doing and then talk to the AI so you can really see how it works. Jason Kirby (43:00.505) You know, I was going to wrap up, but now I to talk about Borty. I don't know if you've come across Borty AI, but they blew up the scene with their AI for like networking kind of thing. they used, supposedly Borty raised $8 million around that they recently closed. And I called it. I was just like, I got to give it a try. It's an Australian accent, which is brilliant. You know, it's just like, so. Jon Tucker (43:05.592) Yeah. Jason Kirby (43:28.013) Hilarious and it's got, you know, it's a witty personality and all that kind of stuff. You're like, okay. he, anybody, like what they did where it's like, this is AI. There's no faking you or fooling you. You know, this is AI and this is your, kind of, was my first experience having like, what was considered to be a natural conversation with AI. Results, you know, as far as the output of the recommended connections, kept recommending people like specifically said, I don't want to meet, but, you know, interesting nonetheless. Jon Tucker (43:35.128) Yeah. Jon Tucker (43:53.718) Yeah. Workflow workflow and eval workflow and eval. I'm sure they did, but that's an example of like, it's a broad, that's a very broad thing for them to try to do is like make recommendations on your network. Right. like I know the data is there, but that's a fairly complex thing to do. so. Jason Kirby (43:57.719) Yeah, yeah, they didn't run it either. Or maybe I have some kind of edge gate. Jason Kirby (44:12.377) But it's, but the thing is, data is not there. Like I do a lot of AI, we have our own matching AI and algorithms. Like we're having to fill the gaps with AI and make reasoning assumptions with O1 because there's just so few data points that are actually up to date and relevant and public on a lot of companies and firms and investors and stuff like that. And that's where it becomes really difficult. if you just go to someone's LinkedIn, it's like. Jon Tucker (44:30.851) Yeah. Jason Kirby (44:40.761) It says they're an investor in like 20 companies on their profile. then the fifth one, it's the actual real job they have at the firm. like for an AI to skim all that data on LinkedIn and have a contextual awareness of like, what's actually right. Um, I've seen it miss a lot. Um, and I'm sure they'll get better and better at it over time, but right now it's, totally misses the mark on, you know, accuracy. were, I think we're at like 70 or 80%. Um, and that's after. Jon Tucker (44:59.447) Yeah. Jason Kirby (45:10.817) a couple of months of tinkering in terms of kind of filling the gap on investor related data. Jon Tucker (45:12.482) Yeah. Jon Tucker (45:19.692) Yeah. I think that's a symptom of like, think of this as an interface and an infrastructure problem, like our interface and architecture really. Like when I say the data is there, like it's not there in one place. And that's an architecture problem of like, where do you get that data? and so I think that's, that's one thing that needs to be solved, in the coming years is like where, where is all the data and how do you make that all available to AI? And then the interface part is like, how do you communicate with AI when it's not sure, does it go up to you and then go back to AI or does it just escalate to you and say, this one's not good. Like this didn't pass. Right. and so I think like it's a whole different way of like building software, but I'm starting to see companies start to kind of think like that. And that's how we're thinking is, when you think of AI, like a human and you treat it as such, and you work with it back and forth, it becomes much clearer on how, how it would be built. doesn't mean it's easy to build it, but it becomes much clearer on how it would be done. So that's kind of how I'm thinking about that human. Jason Kirby (46:20.761) How I always treat it is like I treat it like an intern, you know, like I, I, I talk to it as if it's an intern and then I expect like an associate level response, you know? So it's like, it's where I grade and judge it against an associate level. So it's definitely don't want, like it was an intern output, you know, last year. Now it's like, it's getting closer to what I would consider like a B level associate and B level, meaning like you have a players. Jon Tucker (46:33.966) Hmm. Jason Kirby (46:49.411) Then have like your B players and your C players. It's usually between the outputs, usually between a C player and a B player. but if someone's slightly more experienced than who I'm having to explain it to, which is like how I typically will engage with, with AI, you know, on a regular basis, that's always what I tell people as far as my metaphor when, you know, when they don't know what to do. Like some people just don't have the right mindset when it comes into managing. they don't see it as managing a person or, know, they, or, they just. Jon Tucker (47:11.853) Yes. Jason Kirby (47:19.809) Do magic, make it happen. Jon Tucker (47:21.006) Yeah. Yeah, I think the way I think of that part is like, in prompting, you should think of it as a human and make sure it's like crystal clear what you want done, but also repeat, like, a lot, like repeat very awkwardly. Like if you're talking to a human, you said, I want you to do this because of this, blah, blah, blah. And like, here's what I want it to be. And it's really important that it's like, you basically are baking crystal clear, like there's no room for confusion of what you're asking it to do. That's the first piece and that's very awkward for people, think, to do. The second part is give it all the context. So like give it a lot of context and then at the bottom, if you're just doing it in the chat in GBT prompt, repeat again what you want done so it's crystal clear. That's super helpful. And then the other part is after you do all that, at the end, you say, you know, before you get started or before you provide this, like feel free to ask some questions if you need some clarifying, if you need some answers clarified, right? Again, very similar to a human. You wouldn't just send them off and not let them ask questions. When you do that, I think the results end up being a lot stronger, but it's definitely a constant game of tinkering to see what's the right way to do it. Jason Kirby (48:31.001) Well, all I can say is it's worth the tinkering. John, thank you so much for being on the show. Really appreciate your insights and just catching up and having a fun AI chat. So we'll get this episode out and share with our audience ASAP. Jon Tucker (48:33.805) Absolutely. Jon Tucker (48:46.508) So thank you so much, Jason. appreciate it, man. And for everyone listening, I hope it's been helpful. Get into AI. Get interested in it. Enjoy it. And be grateful for the time that we are in life right now. This is the biggest change that's ever happened. And when you zoom out and see that, it's pretty exciting. So I hope this has been helpful for you guys.