This episode features a discussion between Todd, the podcast host, and David, CEO of Tekyz, about the critical elements for startup success. The discussion focuses on customer validation, developing Minimum Viable Products (MVPs), and integrating Artificial Intelligence (AI) in modern business practices. The conversation emphasizes the importance of early and thorough customer discovery to achieve product-market fit and explores how new technologies are changing the landscape for startups.
David’s Background and the Genesis of Launch First
David recounts his extensive background in technology, starting with enterprise sales and project management at companies like Intel and Texas Instruments. He then transitioned into entrepreneurship, launching his first startup in the early 90s, which he successfully sold in 2000. This experience and subsequent work with numerous startups at Tekyz led him to develop Launch First, a methodology focused on validating product-market fit before extensive development. He observed that most startups fail because they delay testing the market and securing revenue, often spending up to two years building a product nobody wants to buy.
The Importance of Customer Discovery and Niche Identification
David stresses that the primary job of a startup founder is to identify their early adopter niche and understand their top two or three root-level problems. These problems should have a high perceived impact and cost to the customer. He advises founders to adopt a clinical, hypothesis-driven approach, shedding the “black robe” of unwavering belief in their vision and instead wearing the “white coat” of a scientist seeking to disprove their assumptions. He explains that the Launch First methodology involves identifying potential niches, teasing out root-level problem statements, and mapping each problem’s impact and cost to each niche. This process helps to determine the most promising niches to target.
Pre-Selling Validation and the “Mom Test”
The discussion shifts to pre-selling validation to prove product-market fit before development. David highlights the limitations of beta agreements, noting that people often make commitments they don’t follow through on. He recommends the book “The Mom Test” as essential reading for founders, emphasizing the importance of focusing on the customer’s problems rather than pitching solutions. Launch First uses Figma prototypes to simulate a fully functional product, which is then demoed to potential customers with the understanding that it’s a prototype and the actual product will take months to develop. The goal is to secure advanced sales from committed customers, generate revenue to fund development, and validate the market. “If you’ve found the right early adopter where they’ve got a — the cost of this problem is big enough to them that they know they’re going to get this product when it comes out and they feel threatened by this so you can easily reach that stakeholder and you offer them a big enough opportunity up front…is that if you have those two things in place, then you can sell enough of your product in advance…to prove that you’ve got product market fit and to generate revenue to help you fund development.”
Leveraging No-Code Tools and AI for MVP Development
Todd points out that no-code tools and AI have made it easier and cheaper to build lightweight MVPs. He asks if David is seeing more clients who have already validated their concepts with these tools and are seeking to develop robust, scalable versions. David acknowledges this trend and notes that Tekyz is retooling its approach to incorporate these tools into the Launch First process. Todd shares his experience using Flutter Flow, BuildShift, and Supabase to create reusable code components, shortening the time to market. David agrees, mentioning tools like Bolt and Replit, and discusses Tekyz’s internal efforts to build AI models for microservices and UI generation.
The Future of App Development: AI Agents and Workflow Automation
The conversation turns to the future of app development, focusing on AI agents and workflow automation. David predicts everything will be built around agentic workflows and Retrieval-Augmented Generation (RAG) models. He shares that Tekyz is exploring the creation of AI agents for microservices building and workflow orchestration. Todd emphasizes the importance of workflow automation. He shares an example of how he automated a friend’s time-consuming business process, reducing the preparation time for a weekly meeting from seven hours to 15 minutes. David discusses his approach to creating playbooks for marketing activities, using AI to document and automate the processes.
Adapting to Rapid Technological Change
Todd commends David’s ability to adapt to rapid technological changes, noting that many people get stuck and fail to innovate. David responds that it’s essential to maintain a childlike curiosity and not be afraid to experiment. He acknowledges the difficulty of predicting the future of technology but emphasizes the importance of continuous learning and adaptation.
Key Takeaways
The discussion underscores the importance of rigorous customer discovery and validation before investing heavily in product development. It highlights the value of pre-selling to secure early revenue and validate product-market fit. The conversation also explores how no-code tools and AI transform startups worldwide, enabling faster and cheaper MVP development. Finally, the speakers emphasize the importance of workflow automation and continuous adaptation to technological change.
[0:12–0:17] David: welcome david to the podcast i appreciate you making some time yeah todd thank you for having me on the podcast i’m looking forward to the conversation well good well um i’m kind of
[0:17–0:35] Todd: interested to hear a little bit about your background as a starting point i mean uh i was doing a little research and uh you’ve got kind of a pretty impressive background over the last 30 years from, you know, like multiple, I think you started off in physics, if I’m correct, and then ended up in sales and then multiple, you know, different software companies.
[0:35–0:45] Todd: And now you’re a current CEO of a company. So, you know, I think there’s a lot to unpack in this and I’m pretty excited about it. So why don’t we just start with a little bit about your background and how did you end up where you’re at today?
[0:46–1:08] David: Yeah, sure. So I started out over 35 years ago, actually, in enterprise, in sales initially, in then in doing project management and software development for I was Intel, Motorola, Allied Signal, Arizona Public Service, and I was employed with Texas Instruments, and that’s when I started my first startup.
[1:08–1:26] David: I did it in conjunction with a partner who was also at Texas Instruments, and that was in the early 90s. And despite every effort on both our parts, we ended up growing it anyway, 800 customers in 22 countries and sold it to a publicly traded firm out of Toronto in 2000.
[1:27–1:42] David: So, of course, I thought I knew what I was doing for startups and what made that successful, which I really did. I did a lot of the right things, but more by accident, and I didn’t realize that until I had worked with a lot of other startups later on.
[1:42–1:52] David: I was VP of products for the acquiring company for three years, then cast out, again, on my own left there, and started Tekyz, my current company, in 2007.
[1:53–2:06] David: So I’m coming up on 18 years when I started that company. And we started out with Web 2.0 development. This is custom software for startups and mid-sized companies.
[2:08–2:11] David: Can you hang on a second? I didn’t think they were going to knock on my door to bring in the food.
[2:12–2:14] Todd: Okay, no problem. We can cut this out. Oh, thank you.
[2:16–2:16] : Thank you.
[2:17–2:43] David: Okay, I got it. No problem. Sit there and wait until we’re done. Anyway, so, you know, and it’s always been important to me from a software development perspective to always push the envelope in terms of being in modern technology, but also in terms of how we build our systems, accountability in the teams.
[2:43–2:56] David: And there’s all kinds of things we do from what I call an exceptional perspective that a lot of software development companies just don’t do because it’s hard work and it takes a long time to evolve the systems and the culture and everything else.
[2:56–3:11] David: So part of that is staying in front of technology. So as a result, of course, we keep moving and moving and moving and moving as technology shifts, trying to always provide modern platforms for any of our clients.
[3:11–3:28] David: and now everything’s all AI on the, you know, development front. But during that time, regardless of what technology stack we used, regardless of how we did it, we worked with a lot of startups during that period from the time we started Tekyz until now.
[3:29–3:46] David: And it was a real lesson for me how not to start a company. Over 80 startups during that period. and a few of them were really successful, but the vast majority of them failed.
[3:46–4:03] David: And they all failed, in my opinion, for the same reason. They lacked product market fit. They didn’t, and the reason they lacked product market fit, well, there are several reasons, but the primary thing is they waited way too long to test the market by asking their customers to give them money.
[4:03–4:18] David: they would do free trials and extended free trials and build out their products. And, you know, very often it’s about two years before they actually are in a position to start to market their product and try to get revenue, and they find out that nobody wants to buy it.
[4:19–4:35] David: They can’t sell enough of it for a high enough closing ratio that they can get to what I call the viability barrier, which is a three-to-one ratio between lifetime value of a customer versus the cost of that getting that customer.
[4:36–4:57] David: And they usually weren’t even close. So then they’d have to pivot, and pivoting is super costly when you’ve already built your product. And it was a frantic process because they then thought they knew, but, of course, they didn’t go back and do the real hard work, which is what I call there’s a hard, easy way to do this and an easy, hard way to do this.
[4:57–5:12] David: The hard, easy way is you do that customer discovery and you do it the right way because there’s a lot of ways to do customer discovery, and there’s a right way and a wrong way. There are more wrong ways than there are right ways.
[5:12–5:23] David: But once you really do that customer discovery, then you go and test your assumptions and you validate that your customer discovery was correct. So anyway, so that’s my background.
[5:23–5:38] David: This is kind of what led me to develop Launch First, that plus how we do our software prototyping during design. Those two things are the reason why Launch First was something that just naturally evolved in this process.
[5:40–5:53] Todd: So that’s interesting, and I think it lines with kind of what we see, too. I think that lots of people don’t — they think they do customer discovery, and they really don’t get down to brass tacks to the point where — I think they ask a lot of leading questions.
[5:53–6:06] Todd: They don’t really understand kind of what is going to change in the behavior. Are people really going to pay you for it? Maybe not understanding competitive analysis, too, like where is their white space? Is there a good beachhead niche you can actually start with?
[6:06–6:21] Todd: And so maybe help me unpack some of that, because I think that’s probably the starting point of this, is if you get this right, then lots of things get a lot easier. But like you said, if you get this wrong, you build and you spend a bunch of time, that’s a pretty expensive thing to go back and redo.
[6:22–6:32] David: Right. So this gets into the philosophy of what your perspective should be if you’re starting, if you’re doing a startup.
[6:33–6:56] David: So most of the people that come to me and want to build their apps if they’re startups, and we do a lot of startups also for existing companies that are looking to take some workflow that they have and turn it into a product, or they’re testing a new market, And those startups have usually a much higher chance of being successful because they do a lot of testing of their market first.
[6:56–7:07] David: And I’ll explain what I mean by testing the market. But a lot of startups come to me, and they’ve got the black robe on. And they have this vision, which you need to have vision if you’re a startup, right?
[7:07–7:30] David: But they believe in their vision, which is the kiss of death for a startup founder, is the belief system that they build around this vision. And so I coach them to take the black robe off and throw it in the fireplace and put the white coat on and turn into a clinician where you have hypothesis and you’re trying to prove your hypothesis wrong along the way.
[7:30–7:44] David: So that you become very clinical and analytical in the process of determining what the value of what the problems are, who the niches that you should be focused on, and how you arrive at those conclusions.
[7:45–8:19] David: So I always say there’s one job a software founder has or any startup founder has when they are starting their company before they do anything else. One job, and that is figure out who is your early adopter niche and what are the top two or three problems at most at a root level problem, not a high level generic problem, but at a root level that they need solved that have two components to it, That problem niche intersection has two things.
[8:20–8:30] David: It has a high perception of this problem needs to be solved, right? That’s not a real thing, but it’s a perception from the stakeholder. They perceive this as a problem.
[8:30–8:42] David: They feel it’s going to affect their career or whatever it is, but it’s a survival-level kind of root-level problem, and it costs the most because those two numbers are independent of each other.
[8:42–8:54] David: Somebody can perceive that there’s a big problem because they’re getting downward pressure from management or for whatever reason and they need the problem solved, but the actual cost of that problem is low.
[8:56–9:12] David: Whereas you can have problems that are really expensive, but the perception for the need to fix that problem is not high. They don’t feel it impacts them because maybe they’re in an industry where everybody struggles with this problem and it’s endemic, and so, you know, we’re just like everybody else.
[9:13–9:28] David: So what you need is to find that niche that has that top one or two or maybe three of those root problems that you are solving with what you’re planning on doing that have both high impact, perception impact, and high cost.
[9:28–9:39] David: That’s the number one job that founders have, but they don’t know that, and nobody tells them that, and it’s not obvious that that’s what their job is. And so that’s what launch first.
[9:39–9:46] David: The first thing we do is basically focus on discovering that, not discovering but uncovering that.
[9:48–9:55] Todd: And would you maybe walk me through how hard it is for founders to, A, wrap their head around it and then execute on it?
[9:56–10:14] David: Okay, right. It’s easy for them to wrap their head around this idea when I show them. Like, oh, you’re right. We absolutely need to do that. It’s really hard for them to execute on it because it’s tedious, and it takes grit to get through it and to do it deeply enough.
[10:16–10:26] David: The way we do it, the way that Launch First does it, is we come up with all the various niches. We tease out all the niches. I have a methodology for how you define a niche, and we tease them all out.
[10:26–10:38] David: and there’s usually anywhere from 10 to 30 niches that you could potentially target with your product over time. And then we tease out all those root-level problem statements. So they may say we’re solving these two or three problems.
[10:38–10:50] David: Usually those are generic problem statements that they think they’re solving. But when you stand in the shoes of that niche, what about, you know, why does that problem matter to that stakeholder in that niche?
[10:50–11:03] David: And you do that why two or three or four times. you get deep enough where you get to the root level problem for that stakeholder in that niche where they feel threatened by that problem. And that’s how you have to articulate the problem statements.
[11:03–11:16] David: And so when you do that and those problem statements start to distill out to more like 10 or 15 problem statements, not two or three. And then we create a grid with this, and then we map every cell in the grid.
[11:16–11:29] David: We say, okay, what’s the impact of this problem statement to this niche from a scoring perspective? And then we do the same thing from cost. What’s the cost that this particular stakeholder has with this problem statement?
[11:29–11:45] David: And then we bubble chart the whole thing. And when we do this, it makes it really clear that there’s two or three or four niches that float up into the upper right where their top two or three problems have the highest impact and the highest cost.
[11:45–11:59] David: And then we focus our deep dive exercise on just those three or four niches. to figure out which ones do we promote to the top, which ones do we demote for things like it’s costly to reach stakeholders.
[12:00–12:23] David: They don’t have decision-making. The ones that decide on the product don’t have budget and they have to get it approved. The sales cycle is too long things like that right Sure Or you know because you want somebody who can make a decision and purchase product because LaunchForce is all about doing pre sales as a way of proving product market before we even start developing the software.
[12:24–12:36] Todd: Well, that’s good. I mean, we have a similar approach, and I think kind of maybe where you’re going is we actually make the founders get a beta agreement, so they usually either use Flutter Flow or Figma or something to actually prototype what they’re actually going to do.
[12:36–12:49] Todd: So as a clickable demo, they’ll go back to those stakeholders that they’ve identified in their niche and basically get them to sign a beta agreement that says, at the very least, you’re basically going to use this product and beta test it for us and give us feedback.
[12:50–13:01] Todd: And if they can’t get more than 50% or 60% of the original people that they interviewed to sign that, we know that we haven’t gotten the value proposition strong enough for them to even do a beta test.
[13:01–13:15] Todd: And so before we even write a lick of code, we’re kind of on the same boat, which kind of sounds like maybe talk to me about your kind of pre-selling validation because I’m sure it’s got a similar concept where you’re really trying to make sure there’s a demand for what you build before you do it.
[13:16–13:27] David: Well, yes, and I think the beta agreements, if you get a high enough percentage of people willing to beta, that gives you a really strong signal that there’s demand for this product.
[13:28–13:40] David: The problem with betas is people will say a lot of things that they are not actually making a financial commitment to something, that they won’t follow through. There’s a wonderful book.
[13:40–13:53] David: It’s my favorite business book of all time. It’s called The Mom Test. Oh, yeah. You’re familiar with it, right? I am. And every single founder should read this at the very beginning of their journey.
[13:55–14:07] David: You were going to say something? Nope. Nope. Continue. Oh, okay. You froze for a second. That’s why I was asking. where basically the idea is that you’re focused on the problem, right?
[14:07–14:29] David: You’re not focused on the solution, but you just talk about the problem, and you talk with customers about the problems they’re struggling with and what they’ve done historically about those problems and how much does that problem cost you and why is that problem a problem and what other problems had you not thought about that they struggle with that are related to the problems that you’re talking about and on and on and on, right?
[14:29–14:49] David: And that way you’re getting truth from a customer because they are talking about themselves. They’re not talking about your thing. They’re talking about their thing. So what we do is focus on the problem, build a prototype using Figma if Figma will satisfy the requirement.
[14:49–15:07] David: And what I mean by that is the prototype has to be fully animated in such a way so it looks like you’ve actually built the product. in terms of on-screen behavior and workflows and pop-up messages because you’re going to demo that like you’re demoing a real product.
[15:07–15:18] David: You never lie to a customer. You always tell them this is a prototype. The first version won’t have all these features, and it won’t be out for three to four months if you’re doing a classic pre-launch sale.
[15:19–15:44] David: Or you say, we’re going to provide you with this service, but the software won’t be available for three to four months, but you can start using the service immediately, and you’ll cobble together some workflows and, you know, WordPress and Google Sheets and automate some workflows with App Scripts or whatever you need to do so that you can start to manage the business of a certain number of customers, but you get them to buy it in advance.
[15:45–16:20] David: So if — and the idea is if you’ve found the right early adopter where they’ve got a — the cost of this problem is big enough to them that they know they’re going to get this product when it comes out and they feel threatened by this so you can easily reach that stakeholder and you offer them a big enough opportunity up front in terms of a lifetime license or free implementation like in the case of clinical trial software where implementation costs cost more than the software typically or whatever the high-value opportunity is that triggers that client to feel like
[16:20–16:41] David: they don’t want to miss out on this opportunity. is that if you have those two things in place, then you can sell enough of your product in advance and be generating revenue from people you know are going to be very committed, invested customers to prove that you’ve got product market fit and to generate revenue to help you fund development.
[16:42–16:56] David: And you can usually, if you’re successful in the pre-launch, generate as much money or more from pre-launch sales from an intestinal fractional percent of your market versus giving away 10%, 15%, 20% to a seed investor.
[16:57–17:17] David: And the work you do to try to get that money from a seed investor does zero to move your business forward, whereas everything you’re doing in Launch First is moving your business forward in terms of growing your customers and building a sales and marketing engine and figuring out how to speak to customers so that they’re your closing business.
[17:17–17:24] Todd: What percentage of your clients actually get to that point where they’re basically generating that amount of money before you write a lick of code?
[17:27–17:41] David: So we had, well, the clients that get all the way through this, the niche analysis piece, there’s a high percentage of them that will get there. Or they fail fast and fast.
[17:41–17:43] Todd: Yeah, which is perfect, too. Right. Exactly what you want.
[17:43–17:55] David: Right. We do several. Exactly. Right. So I try to convince my clients if they’re not getting there, let’s pivot two or three times because we’re dealing with a high-fidelity prototype and a marketing stack, right?
[17:55–18:11] David: And those are easy things to pivot as compared to software. And if they still can’t get past it after two or three months and they’re not getting close to what I call that viability barrier, that three-to-one ratio, then I try to say, look, it’s a reasonable thing for you to consider.
[18:11–18:22] David: You fail fast and cheap. We don’t want to invest hundreds of thousands or millions of dollars in building this big, sophisticated system if you don’t have a market that’s going to buy it. Right.
[18:23–18:38] David: So, but, yeah, a decent percentage of them. I don’t have actual metrics, so I can’t tell you exactly. That’s interesting. But when we originally tested this, we did this with four different clients when we tested this model a few years back.
[18:39–18:45] David: And three of them were successful. and got to that point really pretty easily. Board 1 didn’t.
[18:46–19:01] Todd: So what’s interesting to me is there’s kind of an other trend on this, too, and I think both kind of the no-code and then some of the AI tools have made building a lightweight, non-scalable version of this so much easier and so much cheaper.
[19:01–19:22] Todd: So less than, you know, in some cases $5,000, $10,000. You basically can piece something together to prove out your concept, go to an early adopter, maybe get your first, if it’s a B2C, maybe get your first couple thousand people on it, realize you have a business, and then you go back to basically rewriting an enterprise or a more scalable version of it.
[19:23–19:40] Todd: And I’m curious, do people come to you with some of that already starting to, they’ve figured some of that out, and they’ve trialed that super cheap, and now they’re coming to you and saying, build me a robust, scalable version of this, because I found the niche and I think, you know, and I’ve spent less than $10,000 to do it.
[19:41–20:01] David: Yeah. And so, yeah, and this is possible now, right? There were a couple tools you could do this to some degree and not necessarily create something that would really be all that appealing and attractive but would work and you could at least see that the workflow had enough value that you could start to sell it.
[20:02–20:12] David: But now all of a sudden, really, in the last three or four months, really in the last month or two, this has really opened up a lot. In fact, this is something we’re completely retooling.
[20:12–20:23] David: I haven’t had people do that on their own and then come back to me and say, let’s build a robust version because everything we’ve built in the past was always scalable day one.
[20:23–20:35] David: It’s just sort of built into our DNA. You know, but that’s going to change for this, specifically for this proof of product market fit, right?
[20:35–20:49] David: Because with a product, you’ll be able to basically get a product on the market that can satisfy a certain number of customers, right? And satisfy a certain level of sophistication in the workflows and things like that.
[20:50–21:07] David: Really, they’re just like you said. So it’s funny because I’m wrestling with the right way to approach that right now, exactly what you just said. What are the right tools to use, how to leverage it, and to make this a part of the launch first offer instead of just doing the design prototype.
[21:09–21:40] Todd: Yeah, and so I’ll just tell you what we’ve been thinking about in trying to pilot with some of our customers. you know we’ve been trying to use Flutter Flow and saying instead of using Figma can you actually use Flutter Flow and actually you know write the code and the entrepreneur in general most of our entrepreneurs that come to us they’re not technical they’re not sales guys they’re generally product people so they’re in general they’re fine like doing some design work and then we’ve been looking at build shift for a lot of the business logic so basically doing a clean segmentation
[21:40–21:59] Todd: to make sure the business logic is not put into the UI and is a clean layer, and then Supabase on the database side, Postgres. And basically what we’re finding is once they kind of upgrade through that, there is a fair amount of reusable code that basically the next party can take a look at and really leverage to not start again, but basically leverage it.
[21:59–22:16] Todd: The second piece to it that we’ve been kind of saying is how do we basically reuse a number of those things so that you’re not, you know, authentication, password reset, Stripe integration, A lot of that stuff almost every one of these guys has, or we have Marketplace, or we have B2C, or we have screens for B2B.
[22:16–22:26] Todd: Like, there’s a bunch of reusable components even in that strategy that you can go do that basically shortens this time frame to get them to market. And I don’t know if any of that resonates with you.
[22:26–22:36] David: Oh, I completely, exactly the same sort of thing. In fact, there’s also tools like Bolt and Replit. I don’t know if you’ve played with either of those.
[22:36–22:37] Todd: Yep, I’ve played with both of those.
[22:37–23:04] David: And they’re improving very quickly in terms of what they could do a month ago versus what they can do now. So, in fact, I’m using Replit right now to build a referral portal for our own internal use because we have a lot of referral partners, and we need a portal for them to be able to use, and we need a portal to be able to track it and all that.
[23:04–23:13] David: And so, you know, building it in days using a tool like that. And it looks nice, too, on top of all that.
[23:14–23:23] Todd: How do you think about reusable code from that standpoint, though? Do you think that a lot of these tools, that if you build them in there, there’s some leverage as you build the scalable version?
[23:24–23:36] David: I think the UI, for sure. We’re also, like, things that we’re doing internally is building a, you know, We have several AI models that we’re building internally for our own use right now.
[23:36–23:51] David: One of them is it’s still at early stage, but this is something we feel really strong about. It’s an AI model for building microservices, which, of course, with all the communications layers and everything in place, because we do everything in microservices.
[23:52–24:17] David: Like I said, it’s sort of built into our DNA. But it doesn’t make sense to always be building them over and over again or copying and pasting code. So this is something where we can generate them based on the business rules for this particular service and its need for being able to be provisioned and orchestrated in, like, a scalable environment like Lambda or Cloud Run or whatever.
[24:18–24:36] David: And using possibly you know I starting to consider maybe using Replit for generate building the UI for us because it creates all those UI components for you and it does it pretty fast, and it does a decent job of it, and then just plug that into our own stack.
[24:36–24:50] Todd: Yeah. Yeah. How do you, I guess, you know, one of the things that’s kind of hard to think about is, like, especially in your world, these things are changing so quickly. how do you think about building kind of modular components of this when you know maybe that’s not the tool?
[24:51–25:05] Todd: You know, like a lot of this is API-driven, so you can pick different models. You can basically specify different models. But even from an architecture standpoint, how do you think about leveraging these tools in a way that, like, six months, a year from now, you can swap these things out to whatever’s the latest and greatest for you?
[25:06–25:19] David: Yeah, and that’s tough, right? Because how do you predict the future? Yeah, right. I mean, right now, if you talk about a year out, that’s really hard to predict. There are certain things I think are predictable three to five years out.
[25:19–25:30] David: A few things are predictable, like the end of, I don’t know, I keep saying this, but the end of multi-level parking structures. That’s three to five years away. There will be a tipping point.
[25:30–25:46] David: They’ll be completely empty because of self-driving vehicles and Waymo and Tesla. and people will stop buying cars because it just costs too much and it’s too unsafe.
[25:46–25:47] Todd: It’s not a good return on the investment.
[25:47–26:05] David: Right, exactly. And there will be a tipping point where it’s not like Uber where you have to wait 5, 10, 15 minutes. It’ll be like you’ll have a car in a minute because it may be parked in somebody’s, you know, somebody in your neighborhood has a car that they own a car and they’re not using it and it leaves their parking lot when you need it, right?
[26:05–26:17] David: So, by the way, if you’re invested in multi-level parking structures, sell them now. Get the heck out. And people don’t believe these tipping points can happen that fast.
[26:18–26:32] David: I’ve seen it happen that fast. Massive tipping points that happen. Anyway, so some things, back to your question because I’m rambling now. Some things I think like that are predictable, at least I believe they are.
[26:32–26:44] David: Other things like how are we going to be building apps in a year from now are not as predictable, other than I know everything is going to be agentic workflows built around RAG models, right?
[26:44–26:57] David: So this is kind of all the — this is our orientation these days. And we’re thinking what agents should we be building right now? Like one of the agents is our microservices builder.
[26:57–27:15] David: what other agents should we be creating right now, maybe a workflow orchestrator that’s an agent as opposed to a workflow orchestrator software, where right now with workflows, you configure them by dragging different components and connecting them and putting in logic.
[27:16–27:28] David: There’s a tool, but now there’s AI tools that will build those for you, like with Make and Zapier, and you can say, this is what I want this workflow to do, and it kind of like wires it up and it gets it right sometimes.
[27:29–27:32] David: But in a month, it’ll get it right a lot more often.
[27:33–27:52] Todd: Do you think there’s a shift in your business from going from these horizontal LLMs to things that are a little bit more vertically focused, or do you think that’s still a ways away? I mean, they’re Swiss Army Knives, and I think what a lot of enterprises are finding is they’re good for a lot of general stuff, but the domain expertise in these businesses needs to go way deeper.
[27:52–28:06] Todd: And so dumping more domain expertise into these that are specific and then optimized for their usage, A, is much cheaper and then much more valuable to their business, and they get rid of a bunch of stuff that’s not valuable to their business.
[28:07–28:24] David: Right, right. And I think you look at something like notebook LLM, right, or notebook LLM, which is sort of that, right? Take your expertise and you upload it to notebook.lm and now you’ve literally embedded all this expertise into this tool that can do all kinds of stuff.
[28:24–28:44] David: It can create a discussion, which is wild. I remember when that came out, I was blown away by that. And an amazing number of podcasts started publishing these episodes, which, of course, I think they’ve stopped because they realize that they’re going to get dinged from a consumability perspective.
[28:44–28:55] Todd: I would even lean into that one further. I think my workflow for just producing these podcasts has probably dropped from something that’s probably four to six hours to down to a couple of hours, or maybe an hour now, right?
[28:55–29:07] Todd: Where it’s like I use the notebook LM. I take stuff from YouTube. I take the podcast. I take some articles that maybe you’ve written, some social media. I put it all in there and start asking some questions about, hey, this is what I’m looking for.
[29:07–29:21] Todd: What does David have that’s interesting for my clientele or the people that listen? and then help me build a podcast script that really kind of facilitates that. And, you know, like, again, that would have taken me hours to do, and now it’s taking me less than an hour.
[29:22–29:32] David: Yeah, and, in fact, the message that I got from you that talked about what you wanted to talk about was really impressive. So, you know, congratulations for figuring out how to use these tools.
[29:32–29:51] David: This is what everybody should be doing. This is what people that are not adopting AI right now just don’t realize how they think that it’s just a big learning curve or, again, whether we’re talking about startups or existing companies, we’ve talked about enterprise.
[29:51–30:08] David: But there’s just a lot of resistance to this, and they don’t realize that these tools will teach you how to use them. You don’t have to know how to use them. You just open it up and you ask it a question and it starts a conversation and pretty soon you’re starting to feel pretty comfortable with the tool.
[30:09–30:19] Todd: Yeah, I think it is a different engagement model, though, not just because of text or anything, but I think so many people that I’ve seen start using it like Google and then they say, oh, this is disappointing.
[30:20–30:34] Todd: Or they get their first hallucination and it’s not right and they’re like, oh, this is terrible. I think you really have to dedicate some time to trying to figure out ways to use it and really stretch the bounds of what it can’t do.
[30:35–30:46] Todd: And I think then it’s incredible, right? I remember I dropped a bunch of stuff in. We were doing a trip to southern Utah. My kids go to college in Salt Lake, and we were doing a trip afterwards, and I just listed, I don’t know, 15 different places we were going to go.
[30:46–30:57] Todd: And I said, just give me an itinerary. I want to be gone for two and a half weeks. Tell me how long. I like hiking, biking. And it just blew it out for me. And, you know, how long that would have taken me to put together. And here’s the trails to go see.
[30:58–31:08] Todd: Here’s the biking to go see. Here’s a spot to go do. Spend two days here. This is longer. Oh, you don’t want to drive any more than 100 miles in a day? Fine. We’ll take care of that. And you’re like, holy smokes.
[31:09–31:30] David: Yeah, I know. It’s amazing. So when I first, when ChatGVT came out, right, a couple years ago, and I was, even before that, I was talking about OpenAI the year before and what it was capable of doing, you know, in the two version or whatever it was back then.
[31:32–31:50] David: And then ChatGPT comes out, and all of a sudden it’s so consumable. Yeah, it’s hallucinating like crazy, and, you know, there was a lot of stuff that really needed to be fixed, but it got fixed pretty fast from my perspective because it rarely hallucinates that I can tell now.
[31:52–32:03] David: But it came out. I was talking about this with my wife a lot, And she went, yeah, you know, it sounds really interesting. I don’t know that I’d ever use it. Anyway, so three months later, four months later, this is still pretty new, right?
[32:03–32:18] David: Chachi Pigi’s only been out for months. We’re in the backyard. We just bought a house. And she asked me, we’re talking about, because we had moved from Scottsdale to San Diego, and we’re looking in our backyard thinking she wants to do gardening because she loves to do vegetable gardening.
[32:18–32:31] David: She wants to grow all our own vegetables. She says, how many beds do you think I’m going to need for our backyard? four-by-four beds to feed us. We’re not vegetarians, but we eat a lot of vegetables. And then her sister called right then.
[32:31–32:42] David: And so she gets on a call with her sister for like 10 minutes, talking about the backyard and the move and all that. So I’m there. I pull out chat GPT on my phone, and I ask her that question.
[32:42–32:53] David: I say, we live in Vista, California, so considering the microclimate where we live, how many beds would we need considering we’re also not vegetarians? vegetarians were both in our early 60s.
[32:54–33:16] David: And so Chachapiti says, okay, given your age and all these things, it says you probably need 10 4×4 beds or whatever the number was. I don’t remember. I said, okay, so in each bed, how should we plan, what should we plant considering that we want to have companion vegetables in the plant, but certain things don’t like to grow in the same bed as an onion.
[33:17–33:29] David: Right? And so then it gives us a planting plan. And I said, okay, now do that for succession planting for all the seasons because when you harvest something, you have to make sure what you put in there is compatible with what was going in there before.
[33:29–33:44] David: And so then it does that. And I said, what companion flowers should we put in each? Because you want certain types of flowers in each bed because they draw away the insects that would go to those. But other flowers will draw insects in and they go, oh, I like that vegetable, and leave the flower, right?
[33:44–33:56] David: So you’d have to put the right flowers in it. So I’m asking it all these questions. And it was answering it all for me. I said, okay, create a table by bed by season for all these things.
[33:56–34:08] David: And so it did a beautiful job, right? She gets off the phone with her sister. It was like 10 minutes. And I said, what about this? I said, remember I’ve been talking about Chachipiti while I was talking with her about gardening, and this is what it came up with.
[34:09–34:15] David: And she went, she turned her head. She was looking at me, looking at me. She couldn’t believe it.
[34:16–34:20] Todd: How much time do you think that would take you to do, right, and to get it right? That’s crazy.
[34:20–34:29] David: Well, she even said, she said, it would have taken me days to come up with this, and everything I can see, it looks correct.
[34:29–34:32] Todd: And you were talking to your sister for 10 minutes, and we got it taken care of.
[34:32–34:45] David: Right. So that’s about as practical as it gets, right? And what I try to tell people is don’t ask a question that you want an answer to if you’re getting used to using this.
[34:45–34:48] David: ask it what questions should you ask.
[34:48–34:50] Todd: You should ask it. Right. To learn something.
[34:50–35:01] David: If you’re trying to get this kind of a result out of it, right? What questions should I be asking? And if it doesn’t give you the right questions, say, yeah, those aren’t the things I’m interested in.
[35:01–35:14] David: I’m more interested in this, so what questions should I? And then you’ll start a conversation. Then you can say, yeah, those are good questions. How about you start answering those questions? And then you can just start talking to it like anybody else.
[35:14–35:26] David: I ask please and thank you. I’ve been doing that from the beginning. I thought the please just because I feel like the value that I’m getting from it is so high. I wanted to be nice, not that I needed to be.
[35:27–35:39] David: And the thank you was because I wanted to give it confirmation when it was getting things right because I figured it could only help it. And I didn’t have any evidence of that before. It turns out now later on that it’s true.
[35:39–35:50] David: It actually gives you better responses. And plus, if AI ends up taking over the world, I wanted to remember him. You want to be nice to it. You’re overlords. No, you’re safe.
[35:50–35:52] David: I’ll protect you. The rest is ever gone.
[35:52–35:52] Todd: Exactly.
[35:53–35:53] David: You’re good.
[35:55–36:08] Todd: That’s good. Well, let’s continue in this line because this is one that I guess I’ve been kind of wrestling with a little bit is, you know, you’re starting to see some things with, like, computer vision and stuff, which Claude is doing.
[36:08–36:39] Todd: It seems pretty clunky. It’s not really ready for prime time At least in my opinion quite yet But like where do you see Some of these like it’s one thing to have these Agents and they kind of a lot of them Are kind of standalone right where it’s You ask it a question and it will give you an output But we haven’t really connected it very easily With workflow right whether it’s Belling equations if then if it meets This criteria do something else the conditional logic That I think is really what going to unlock this Who are players that you think like I mean Zapier and stuff is there but like I mean and I don know
[36:39–36:47] Todd: Make is a really good one. What is it? Make is really smart. Yeah, I’ve looked at it. Yeah. Okay. But is, like, Microsoft or anybody else really building, like, an enterprise?
[36:49–36:55] David: Yeah, Microsoft. I’m trying to think of their something flow. What’s Microsoft’s called? They definitely have.
[36:55–37:02] Todd: I can’t remember. I just read an article that was talking a little bit about what they’re doing. They’re starting in this direction, but I was just curious if you’ve had any first-hand experiences with any of them.
[37:03–37:25] David: I mean, workflow automation to me is everything, right? I agree. Right. The biggest benefits I’m able to achieve for people that are struggling with a workflow, and I’ve got a lot of examples of this where I’ve automated workflows for people in the past, are dramatic in terms of the impact that they can have.
[37:25–37:36] David: Like a friend of mine runs a really big business network, and every Tuesday morning he has a meeting of about 70 different people, but he has 1,500 members.
[37:37–37:48] David: And it took him all day long. This was several years ago. It took him all day long, like seven hours, to get ready for his Tuesday meeting because he’s pulling the information from different places.
[37:48–37:59] David: 30% of the people are new. They’re guests in there. He’s got to bring the members up. He’s got to organize the order that he’s going to announce people because he announces everybody.
[37:59–38:12] David: Everybody speaks. And it took him like seven hours, and if he got any kind of distraction, partly because he’s getting calls while he’s trying to do all this. So I said, let me automate that for you.
[38:12–38:48] David: and when we when we did this and this was just sort of a hobby personal project I did for him because he’s a friend when we did this he said okay we said you can automate that and I said yeah I think so so let’s do it and he said okay and he says here’s we go through the whole workflow he shows me exactly what he does and I said okay give me a couple weeks as I’m running a software company and I’ll do this in my part time but give me a couple weeks I’ll use Google Sheets and write some app scripts in the background to automate this and connect in using the API from your ticketing system.
[38:49–38:59] David: Anyway, so he did. Come back two weeks later, and I showed it to him. He said, oh, that won’t work. And I said, why not? He said, because it doesn’t handle these two conditions. I said, well, you didn’t tell me about those two conditions. Exactly.
[39:00–39:02] David: Let’s go ahead. Right. This is just typical, right?
[39:03–39:05] Todd: Requirements. Requirements, requirements.
[39:05–39:19] David: Right, exactly. So we went back, and I thought I asked him every question, But this is like you were talking about corporate business intelligence inside your company, right, domain expertise.
[39:19–39:33] David: You know, this was just part of his DNA. He had been doing this now for many years every Monday. And so there was so many little nuances. Anyway, every two weeks they’d come back and they would then realize why it wouldn’t support his business because it was missing this case or that scenario.
[39:33–39:47] David: or, you know, you can’t put people in when their email is whatever it was. It was all these little nuances that he didn’t tell me about, and it was not something I could easily predict because it was so specialized to what he does.
[39:48–40:10] David: So after about three months, every two or three weeks, we’d get back together when I’d have a few hours to work on it here and there. And then finally we got it to where it looked like it was doing it all. So we ran it in parallel with his normal process for one week, and it ran it perfectly, and then he started using it himself and ran across a couple more of these little glitches over the next month or two, which I took care of pretty easily.
[40:11–40:25] David: And now he’s been using it every single Monday for the last four years, four, four and a half years now. And it takes him 15 minutes to do his Monday instead of seven hours. And whenever we talk, he says, you know, I’m still using that program you wrote.
[40:25–40:59] David: and he’s got a big and this is typical when you really dig into work automating a very manually intensive person intensive workflow and we’re about to do the same thing with his podcasting network because he has a radio station and a big podcasting network and I told him I’m going to be starting my own podcast soon and so that sort of conversation he said well I’m kind of not sure I want to tell you this but we have a few members that send me their podcasts when they do them, and I distribute it in my network.
[40:59–41:12] David: And I said, weren’t you sure you wanted to tell me? He says, yeah, it’s just a lot of work. And I wasn’t sure I was going to keep this going because of how much time it takes me. But, you know, these are really great members, and that’s why I’ve continued to do it.
[41:12–41:23] David: And I said, this sounds really familiar. Why don’t we get together? But now it’s not going to take me three months to do this. It’s going to take me hours probably to do it.
[41:23–41:41] David: spread across several days as we go back and forth, and he remembers what he didn’t remember each time, right? Maybe a week, week and a half instead of three months, and probably four or five hours of my time, and I’ll be able to automate the whole flow for all the distribution, publishing, and outreach.
[41:41–41:49] David: So I love workflow automation because this is where businesses really, really save money and make money.
[41:49–42:04] Todd: Yeah, and I think that’s where this starts, right? Just take out all the stuff that was bubble gum and tribal knowledge, or somebody like your friend basically has all this tribal knowledge from doing it, and if you can basically turn that into an automated workflow, then that’s great.
[42:04–42:14] Todd: That guy can go on vacation, or he can take us 15 minutes on a Monday to get it knocked out. And then if he’s not there and he turns it over to his son or daughter, then great. They’ve got a process to go do that.
[42:14–42:18] Todd: And so I do think that’s kind of the entry level where a lot of this starts to happen.
[42:18–42:43] David: I want to add one thing to this because this is something I’m grappling with right now. So we’re building playbooks for how we operate in our business, in particular on the marketing side now that I’m starting a podcast, in the podcasting side, in the marketing, the interviews, and when I go on an interview like this, that sort of thing, right, and the communications with people like you.
[42:43–42:54] David: And this is all, you know, I was doing it myself for a little while until I felt like I had kind of a consistent process. And then I got my VA to start to do this for me.
[42:54–43:09] David: And now she’s got it, so I’m going to have her document the playbook on all the nuances, basically like any workflow, right? There’s all the things you do regularly, and then there’s all the nuances that are the exceptions that you have to cover.
[43:10–43:20] David: And so she’s going to do that. But I want these playbooks to be — I want to use AI to then be able to document the playbooks as best I can.
[43:21–43:39] David: To a certain degree, that’s a little difficult because you still have to describe all these cases, right? But then I want AI to then say how can we automate this process, the outreach and everything else, and then start to build the work — automate these workflows.
[43:41–44:09] David: This is kind of, I’m starting to kind of come up with my own internal methodology for how we take something that’s amorphous like this marketing activity that has regular steps and nuances for all these exceptions and turn that into playbooks and then take the playbooks and use AI to then automate the playbook for wherever it can be automated and do it in steps.
[44:09–44:21] David: Because you can’t do it all at once, right? You do it all at once, you end up with a mess. It’s got to be done in steps that are kind of controllable and that are testable, that you know that it’s being done all correctly.
[44:22–44:37] Todd: It’s funny. Like, it sounds like we think a little bit like, I mean, I always like using, like, something like Lucidchart or something just to outline the process, right? So if we can agree on kind of what the process looks like and there’s a flow and there’s decisions and then you’re like, okay, here are all the variations that I have, okay, that’s pretty cheap to figure that out.
[44:37–44:55] Todd: And then basically if you want to take the next steps into how do I automate some of this, at least you’ve got that framework to start with. And that’s even what we’re talking about is let’s make sure that’s all documented because the tool over time could change, but the workflow is probably not going to change because once you nail down the workflow, it’s just applying it to the next tool.
[44:55–45:12] David: Right, exactly. I totally agree. And that’s exactly what we’re doing. So you’re doing basically a workflow. I’m doing playbooks, which is the same thing. And it’ll probably have the workflows in it, you know, in terms of flow charts and all that in the playbook.
[45:12–45:22] David: I just want a place where somebody can go and I’ve got a new person coming because you need this to scale, right? So I want to add another person because maybe we’re doing twice as much.
[45:22–45:33] David: I’ve got more people doing podcasting and we need more support. And so I can just hire somebody, they give them the playbook, and then we have a playbook about how to train them, right?
[45:33–45:43] David: how to get them trained so that they’re doing these things. Or we’ve got what parts are automated, what parts have to be addressed by them individually, and then we can, you know, continue to scale and roll this out.
[45:45–46:08] Todd: Well, I might leave it there. I think, David, maybe, I mean, you’re a little bit older than I am, but not much. It’s just interesting the time, you know, like I probably got in computers early 80s, and you think about all the changes technology-wise that we see, and I’m appreciative of you basically trying to use all these different tools and stitch all this stuff together to leverage them and figure it out.
[46:08–46:19] Todd: I mean, I think we probably all knew a lot of peers that maybe got stuck somewhere and didn’t continue to poke and innovate and challenge. And it’s just cool to have a conversation and see where you’re going.
[46:20–46:29] Todd: The AI component of it wasn’t in our agenda, but it seemed like a good meat to the topic and something we both had some in common too. So I appreciate that.
[46:29–46:47] David: Yeah, and, you know, it used to be a lot easier to predict where the fuck was going. And now it’s not. It’s like impossible, especially in the technology world, you know, in terms of how the technology is going to affect technology.
[46:47–47:02] David: But as far as you’re, you know, I mean, you know, I always say to people, because they still come to me to help them work through PC-related issues, right, My friends and family and, you know, sorry to bug you, David, but I’m struggling.
[47:02–47:13] David: Got a question. That I’ve never seen before, right? But I — so I always tell them, and they ask me, how can you do — how can you figure that out? I said, well, you know, I channel my inner five-year-old.
[47:14–47:28] David: Exactly. That’s how I — because I just pretend — I don’t know what I don’t know, so I’m not afraid of any button on the screen. Failure. I just buy everything. Yeah. Right. Yeah. So if you don’t have that mindset, you’re in the wrong business with how fast technology is moving.
[47:29–47:40] Todd: Yeah, it’s exciting. It’s kind of a fun time. So I’ll wrap it up here, though. But, David, I really appreciate you sharing your insight. I think we have a lot of short kind of core values on how startups really need to get going.
[47:40–47:51] Todd: I mean, I think, like, just the customer research, the validation before you spend a ton of money, like that just really resonates with me. I think so many businesses fail because of that.
[47:51–48:06] Todd: And then it’s exciting to see how these tools to build kind of MVPs, you know, first versions of it are just getting cheaper and quicker to market to validate those assumptions. And then, you know, like companies like yours where they can basically then build for scale going forward.
[48:06–48:11] Todd: So it’s an interesting time, and I appreciate you spending some time to share your insights and wisdom.
[48:12–48:19] David: Yeah, and I really appreciate the questions and the conversation. We’re definitely aligned in a lot of these ways. So thank you for having me on your show.
[48:20–48:32] Todd: Yeah, well, thank you. If you enjoyed the podcast today, please just take a moment to rate it or comment on it for us.
[48:32–48:39] Todd: This feedback really helps us, and it helps us get the word out to helping other entrepreneurs and founders. Thank you.
David Hirschfeld founded Tekyz, a company dedicated to transforming business software development. With over 30 years of experience, his journey began with a physics degree from UCLA and a successful sales career at Computer Associates. After launching and selling his first software company in 2000, David found his passion for empowering entrepreneurs.
He developed the Launch 1st™ methodology, which focuses on generating revenue before coding begins. This helps startups gain traction while minimizing risks. With a commitment to innovation and collaboration, David leads Tekyz in providing AI-powered development and SaaS solutions, making a meaningful impact in the tech world.
Tekyz is set to launch two new AI applications: one for automating the Launch 1st Methodology Niche Analysis and Estimiz, an AI-based project estimation tool. Outside of work, David enjoys golfing and woodworking.
You can learn more about David Hirschfeld and Tekyz by following his LinkedIn profile — David Hirschfeld LinkedIn Profile.
For more information about Tekyz’s services and how they can help you harness the power of AI in healthcare, visit tekyz.com or contact the founder directly at [email protected].
The Hard-Easy Way: A Veteran Founder’s Guide to Startup Validation was originally published in Tekyz Blog on Medium, where people are continuing the conversation by highlighting and responding to this story.