A $6.5M Estimate in 23 Minutes: How AI Is Rewiring SMBs | David Flickinger Interview

A-ai-konsulttuun ja AI-expertsin, - mutta se on aikaan myös aiemmin, - että AI-konsulttuun, että tämä on yksi 1-10 miljoonaa - eikä perävästä. Tämä on yksi, joka on seuraavaksi, - joka on seurin monia, - että se on yksi ranninkampani, - ja on seuraavaksi AI-konsulttuun, - että seuraavaksi tuli, että seuraavaksi, - ja seuraavaksi. Ja minä oon, että minä oon, - ja minä oon, että minä oon, että David Flickinger, - David se on seuraavaksi, - ja minä oon, että AI-infrastuksia on seuraavaksi. Se on myös operatio, - johon ranninkampani on seuraavaksi, - ja seuraavaksi seuraavaksi, - johon, että seuraavaksi, - ja minä oon, että minä oon, että minä oon, - johon, että seuraavaksi 2017, - johon, että minä oon, että minä oon, että minä oon, - David, minä oon, että minä oon, että minä oon, että minä oon, että minä oon, - Thanks for having me, great to be here. -Now I want to listen, whether they're a business owner, - a CEO, a private equity professional, - search fund operator, - or simply a CEO, a acquire, - and once we finish this episode, - and I want to immediately forward this episode to their, - whether they're, it's their CEO, - of their portfolio companies, - or their group CEO, - or simply they're operating partners. So I want to make this episode so valuable for them. So let's get into this in a second, - but first I want to talk about you talk about your background a bit, - because I think it's very unique and very special. -Well, that'd be great, and I do think my background - is relevant here for the discussion in a couple instances. So right after I finished my undergrad, - I commissioned as an officer in the Marine Corps here in the United States. And in particular, - I had the opportunity to be an infantry officer. It's a pretty funny thing to be immediately thrust from college life - to being in charge of 50 infantry Marines about to go on a combat deployment. But you quickly realize that as an officer, - your two best weapons are a notebook in a radio, - because you really become an orchestrator. You're always talking to your squad leaders, - you're communicating back to your higher headquarters. You may be orchestrating fire support coverage - from artillery or air support, or snipers, - and it's really you as the orchestrator, - which I think is a relevant parallel to the AI orchestration - we'll talk about in a little bit. But after about five years of active duty service, - I transitioned into the private sector - and had an opportunity to work for a Fortune 50 company called Medtronic - which is the largest medical device company in the world. And got to do some really interesting things - from corporate development at the opportunity to work on some M&A deals - as well as pursue some business development initiatives there. They encouraged me to get my MBA, which I did. And I was very grateful for the opportunity to learn an industry. I really didn't have knowledge of prior to my time in the Marine Corps. So I've spent quite a bit of time post military service - in the medical device and healthcare sectors. I've bounced around a few different companies. And one in particular you referenced in 2017 - there was a really interesting technology called Medecrea - and that was the name of the business, but the technology was - using AI machine learning algorithms - to produce patient specific implants for spinal surgery. Which if you think about is a pretty crazy concept - to have an AI determine which type of implant a surgeon should put in a patient's spine. But it turns out that there's actually a very data driven - algorithm where you take an x-ray scan of a patient - that might have a spinal deformity, an adult or an adolescent - that has scoliosis. And you determine according to the medical literature - what is the appropriate curvature to return this spine to. And then the AI algorithm can determine what that is - and actually produce an implant specifically for that patient. That was the concept of this business. And it was tested over years of iteration against surgeons - who were just using their own creativity. And surgery is a little bit of an artistry. It still is, especially in spinal deformity surgery - where the surgeon looks at a spine curvature and says - "I think it ought to be here." But it actually turns out that the data informs a better surgical outcome. In other words, if you have an AI produce a patient's specific implant - according to its parameters, the patients had better short-term - and long-term outcomes as measured and follow-up studies over several years. So it was a really fun project in business to be a part of - it was acquired in 2020 by - incidentally by Metronik. And it was cool to be a part of that exit. And since then, I've diversified the industries - in which I'm focused, I've done some trade businesses, - I've done other peripheral healthcare or med-device startups - and invested a little bit myself. But I've continued to use AI based on that experience. And we really called it machine learning. The word AI has now become commonplace, especially with the launch of chat GPT - a few years ago. But I've continued to implement AI into portfolio companies - as a consultant and into my own businesses - to drive growth and extract efficiencies out of things that I've been doing. And then we transformed that into a business that's focused exclusively - on providing managed AI services, helping businesses deploy AI - in their own organizations. And I guess the last thing I'll say is we wanted to do that - in a way that was sort of an operator-first mentality - my own background, having lived that experience - both as an executive and as an owner operator. We wanted to do something that would not only improve efficiency - help a business grow, but ultimately foundationally make it more valuable. And so that's the approach we've taken with vellum - and what I'm doing today. So along and winding road to what we're doing now - but each step along the way has informed the way that we think about - this conversation which is deploying AI into an operating business. I knew that we have this episode coming up, so I did a bit of research - and I called around, I had like 15 calls - with traditional business owners here in Northern Europe - and 11 of them are actually using AI. But the way they're using it is a typical chat GPT - like just putting all sorts of documents there and just asking - to summarize few things or just the numbers or things like that. So I want to go smoothly and ask from you here - can you maybe walk us through the difference between asking chat GPT - a question versus having an AI system that actually lives inside the business? Yeah, it's a great question. And I think at this point we've all used chat GPT or cloud or Gemini - or some chat bot now. And I think what you find when you use those things - is they first and foremost typically give a very confident answer. And so one can maybe be persuaded by the confidence of the answer - but these answers when you use them in the format most people do - they typically lack context. And so what you find yourself doing is adding a bunch of paragraphs - maybe you're typing it in, maybe you're even uploading a couple files - so that the chat bot has some context about you, your business, your background. Hey, answer this as if I'm a sophisticated finance professional. Answer this as if I'm an expert in XYZ industry. And so you start to get some better context. But I think ultimately the difference is - having context about your business from a prompt - is very different from having context about your business as an infrastructure. In other words, the AI is connected into your CRM, into your ERP, into your databases, into your proposals folders, into your financial documents - and it has all the context about your business. It knows what your employees are doing. So it not only has context about what you're doing - but what your peers, your boss, your subordinates are doing in the business. All that is in this foundational memory layer of your business. And so when you ask a question of the AI then, you get a much different answer, one that's very relevant to your business. And one that I think gives you the confidence then to take that answer and go - apply it and use your own human judgment - where I think a lot of folks who are using these chat bots - at best are maybe wasting a little time. Because if they are really the expert, they could often arrive at an answer faster - than prompting with an insane amount of individual context every time. But at worst, they're getting an answer that may be different from the right one - because it's devoid of context about their particular business and situation - mutta se on yksi kohdalla. in a definite way that they might make a bad decision. And that would be the worst case scenario. Ultimately, I think you're seeing a lot of people use AI and in many instances being a little discouraged or confused about how do I do this in an enterprise setting versus just me individually trying to supplement my own skills or work streams with AI. - For me, a true eye opener was the example you gave me when we were doing those pre-calls. And I've learned a lot already from you when it comes to AI, but one particular example, and it was the roofing estimator. And you told me the story where I let you tell the story yourself, but this was really like where I, I've read and studied quite a lot about AI, but this particular example you gave and shared was like, okay, for those who know how to do it, it's pretty dangerous in a good way. So please, could you please share the story? - Sure, I think this is a quintessential example of an agentic AI architecture in a business. And so the example is a roofing business that is constantly bidding on commercial projects. And so they do a lot of new construction, which if you think about it is a bit like a math problem because the building doesn't exist yet, except on plans and in specification documents. And so what a good estimator does is they will read those documents, they'll use their homegrown spreadsheets and whatever tools they have, maybe they're using some third party software tools, but ultimately they're putting their own logic in, their own experience, their own materials pricing, their own assumptions about labor, and they're putting forth a proposal based on their best guess or estimate. And it was very clear in this particular business, we could use an AI orchestrator to task AI sub agents to do similar things. And so in this example, we have an orchestrator agent that tasks a plan reader agent that goes and reads these architectural plans, which are quite different actually from a basic PDF with just words. These have numbers, these have measurements. So you need to almost go pixel by pixel and make sure that you've extracted every number correctly because you pick a number wrong out of a spec document and it could be wildly different outcome in your proposal. So that's one work stream. Second is, okay, now that we've identified the type of building, can we have an agent go reference all similar projects that this company has done that are like this? Hey, this is an apartment complex building with five large HVAC units on the roof. How many similar buildings have we done like this with this sort of design? Let me go reference those and understand what those are. Another agent is going pulling the latest materials pricing. Another agent is pulling the latest labor pricing. Another agent is trying to understand risks in the job, whether that's geographical risks because it's 100 miles from the headquarters and so there's travel or we need cranes to lift materials that are unusual because of the dimensions of the building or how close it is to other things. So you have all these agents that are doing tasks. They're routing this up to the orchestrator agent and ultimately that orchestrator agent is applying the rules and logic of the business to generate a proposal. That's really what the estimator's doing, what the human's doing. He's doing all that in his head. And so as we started to iterate this in the business, this example of about to give you a sort of the ultimate test of hey does this work? So the estimator, the roofing business had really big job. 47 individual buildings in this very large complex in a large metro area in South Florida. And the estimator spent a couple weeks not using AI. Just hey this is too big of a project. They're going to do it on their own and this AI may not be ready for prime time. And indeed we hadn't actually deployed it live what was being used every day. We were still iterating a bit, but we were pretty sure that we had it right. So we said what we'll take all the documents, we'll have it, we're not going to interact, it'll be a blind test. And so we ran our simulation with the AI. And it took about 23 minutes to go through this massive pool of documents, 47 plans, specification sheets and generate a number. It took the human estimator about two weeks to generate a number. And on a $6.5 million estimate, they were within $400 of each other. And even us working and having designed the system, we were pretty blown away, but it was just proof and validation that this architecture works when the agent has access to the same logic that the business owner or key person in the business has, it can make nearly the same decision and why shouldn't it. And so that was the aha moment. And there's a great story about that roofing business in the sense that they have leveraged this architecture to grow tremendously over the last several months without having to hire additional people. Now they are replacing people with AI, they're making their estimators more valuable and productive through the supplementation, and use of these tools and these architectures in their business. But it was a really fun example because the size of the project was so large and the outcome was so close, 23 minutes for the AI, two out, two weeks for the human estimator, arriving at the same number and just the cost savings and productivity gain when you think about that applied across a year's worth of work is really, really cool to think about. And it's a lot of fun to be on the implementation side and see these success stories. - My question was and still is, we're talking about the senior estimator who has decades of experience doing this particular thing. Can you explain the logic behind the AI, how they're able to learn this, all this experience in such a short period of time? - That is really the key question. And I think in order to answer that question, you have to understand a little bit about the architecture. And so what the AI will do is it will start to build this memory and context layer, which becomes its foundation. So it uses access to all the documents. In this instance, there were over 3,700 proposals that were used to fill this initial memory bank of estimating logic. And so the AI doesn't get tired, doesn't get bored. It goes and extracts all the logic out of 3,700 spreadsheets and distills that out into rules. Similarly, it pulls the language of proposals and extracts that out. It looks at jobs that were lost and jobs that were won. And it looks at not only one jobs, but the job costing after the fact, hey, we estimated a 16% margin on this job, but it turned out it was 12%. Why? Well, the labor was underestimated. So it can fix that in a subsequent iteration. And so it builds this layer of contextual memories as well as rules, logic gates, things like this that are really the human decision making points that have been distilled out into technological gates or rules. And what you try to do is closely mimic the way that the human makes a decision. And of course, not every decision that a human makes can be distilled out into an AI rule book. But it turns out that a large majority of these reproducible tasks, whether it's creating proposals or responding to customer inquiries or billing and collections, can be written into rules. And as you start to think about this, and maybe there's something we'll get into later in the podcast, is how that knowledge, which we call tribal knowledge or typically lives in the minds of the key people of the business, if you distill that into an architecture that lives and persists and that AI agents can interact with, does that make the business more valuable? And without doing a deep dive into software and coding and databases, what's happening is an AI agent is deploying swarms of sub agents to go look into all these areas of the business that it has knowledge about. And that agent, that orchestrator agent, only give you an answer as good as the systems it has access to. So the answer is, if you hadn't connected it to anything, it couldn't give you the same answer as the chief estimator because it would just use its generic logic. But if you had connected it into the systems, into the CRMs with thousands and thousands of customer records, into the supplier portals that has up to date pricing, into your labor and payroll data. So you know if you're putting a seven man crew on this, job you know what that's going to cost with a high degree of predictability, it can give you a very good answer. And so in short, it can give you these answers if it's connected into the systems. And that is where we really spend the majority of our focus in an engagement is first and foremost connecting an agent into your systems so that it has the context so that it can start to iterate and learn. So as a business owner, operator, from which moment will you be sure about this work, the AI was just doing when it comes to this particular offer, from which moment are you so sure that you are willing to send this to a client and making sure that it's all correct because people make mistakes, I guess AI is making mistakes. And I've been flying recently quite a lot and like I like to read like planes and things and they have those pilots double checking, triple checking, checking things like this. How do you make sure that this what you just sent to a customer is correct and you can actually sleep well on it? So we design a workflow that has human checkpoints at key instances along the way and that system, the goal would be that you over time remove those human checkpoints as you validate them. So you start with a lot of human checkpoints is the answer. You don't want to have AI going and emailing proposals to customers without you having checked it out of course. And you want to design the systems such that you feel very confident at each step along the way that you are either approving the next step, the next phase or you've approved it enough times you're confident that you only need to approve it here. But even in those instances where you say, hey, full approval, actually go email the customer. Do the proposal and email it. Let's say you're there. We don't have any clients that are just doing that yet by the way. We do have clients who are reviewing a final proposal in their draft emails, clicking the PDF and saying, well that looks good, send. So of course they could, but what it will do is identify anomalies and red flags as well. So if there's a confidence interval below a certain threshold, it will say, this is ready for your review. But please let me point you to here, here and here because this is worthy of additional investigation. This is an anomaly. I haven't seen this before. And so the goal is not to take the human out of the loop for these critical things. The goal is to make them more efficient and hopefully direct their attention to things humans are best at. I think ultimately that's one of the promises of AI more broadly stepping out and zooming out of what we've done is directing the human's attention to things we are best at and letting AI do things it is best at. Whether that's repetitive tasks, data analysis, data extraction. I got a 30 page contract yesterday and even if I was an attorney and of course, forward that to my attorney, I wouldn't even really know where to look to sit down and read a 30 page document. But this is an example anyone can do is you can put that document into chat GPT and say, hey, where are the most critical areas I should look? What are the red flags? What seems unusual in this contract and can direct you? So that's a micro example of how when you have this architecture embedded in a business at enterprise scale, the AI should be directing the humans to the most critical items, thus allowing them to be more productive, more efficient and automating work streams that should be automated, but make the human judgment element that much better in those businesses. And that I think is a competitive advantage. If you can make your human element of judgment better because you're directing your humans faster to those areas which need their attention, that's a win using AI. That I think is a great AI implementation. It's not always just let me automate everything. In fact, I think that that could prove to be a mistake over time. You see even this silly little thing of automated emails where you clearly get an email, we've all gotten them now. It's totally written by AI. There's not a human touch in it. And you're just like, throw this away. This was written by an AI that gave, whoever sent this gave no individual human thought to what they were sending. So again, these micro examples we start to see, but I think at a 30,000 foot view thematically, what we want to do is, again, start to streamline what can be automated and what should be delivered to a human so that you can run a business better. Because at the end of the day, a lot of these businesses are still people businesses. It's still human to human interaction. And of course, people and will build systems for these agent to agent interactions. But the human human interaction element, if you can use AI to make that better, I think it's really a tremendous force multiplier. Now in the roofing business, what about junior estimators as the senior ones are so more efficient now? What's going to happen with the junior ones? So they get up to speed faster. That's for sure. They can do more complex estimates earlier on than they otherwise would have. And so if you want to use them in that way, that's great. I could tell you one area where I think we've seen that instance you just described and my prior answer align really well is maybe the junior estimators are able to spend more time with the customer with the decision maker because they're not spending so many hours as an inexperienced junior estimator in front of their computer screen. They're spending more time with the customer. They're spending more time on a roof. They're spending more time doing the things that help them win jobs because the answer that really was the answer to a math problem has largely been automated. Their review was directed to these specific areas. Now there's a high degree of confidence that they're getting to the right answer. The chief estimator, while he may have had to spend a lot of time reviewing a junior estimator's proposal before he approved it to be sent, now maybe has to spend a minute or two because that person has the confidence as well that this proposal was generated with sound logic. It was using the context of the business, the correct rules, pricing, rates for materials and labor, etc. And that the junior estimator was focused in the right areas. And then the senior review of that is again much more efficient. So I think the promise is twofold one that you are just more efficient with the workforce that you have. With junior employees, what can they do faster and above and beyond their current skill set with less training and less time in the role? But two, can those junior employees focus again their time and effort on things that win business, win jobs, help customer retention, help improve a customer experience versus spending more time in front of a screen? Okay, and last one here, are you now able to make and send more proposals to potential clients as previously for the senior estimator, it took it two weeks to create it. Now it's like 20 to 30 minutes and obviously some additional time to review and double check. So are you now able to just simply send more offers to win more jobs? The short answer is yes. And this example is a lot of fun because this engagement was predicated on a request from the roofing business as customers that they respond to more proposals. In other words, they work with these general contractors that have a wide net that they cast in terms of the jobs that they're trying to win. And so they want their roofing subcontractor who they really like working with to bid all these. But of course in the absence of AI that's almost impossible without hiring a big team which doesn't make sense ahead of winning that work. And so the idea was, hey, can you send out more proposals? Thus when more jobs but also satisfy a customer request by using AI. And I would tell you absolutely they've been able to send out more proposals but beyond just the volume of proposals out the door, the amount of one business has been tremendous. This company has doubled the size of their commercial contract, win rate. And this business will be doubled the size next year. And I think you directly attribute that to an AI tool that was well-architected and well implemented and they haven't hired any additional people. And that's really, I think, one of the promises of AI too is not that you replace a workforce but that you're able to grow productivity in a way that doesn't bloat your overhead. I think that's what so many small business owners, mid-sized business owners are looking for because as I, aside from AI, when I talk to business owners and in my own businesses, you always reach these inflection points where you really believe and you're right, you could grow here, but you reach this inflection point where, okay, to grow there, I've got to hire all this additional overhead infrastructure staff. And so you end up staying and topping out at a certain size and that is different for everybody based on their individual market, their vertical, but almost every business runs into this problem. And I think the beauty of these AI implementations is that you can blow through that ceiling, that previously to blow through. You would have had higher a bunch of people ahead of doing that. You can now drive that growth without additional overhead. Well, you never have to hire another person, of course not. But as you grow, now you can justify it. And I think growing from a place of efficiency where you're an AI first business growing in to a new revenue category or threshold is very different from where you see a lot of businesses that are at this massive threshold. They've already got a lot of overhead. And now they're implementing AI as it means to reduce headcount. That that story exists out there. We hear it in the news, but the much more interesting story that I think we'll start to see in this lower middle market space is companies grow with an AI first infrastructure because they had the AI layer that empowered their growth, that empowered productivity and efficiency in their business. They were able to jump to a new revenue threshold that otherwise wouldn't have been possible without bloating the overhead. And so to do that with an AI first mentality and infrastructure, you can then hire in a very smart, efficient way that of course supports that new revenue threshold or tier that you're in. But you haven't had to blow your overhead in a way that well now you have to lay off any year or something like that. But it's really interesting. We're starting to see that happen in businesses with whom we're engaged. And it's really fun to see them grow just to these total new revenue categories that even they didn't think was possible. But now they're doing without adding a headcount or maybe they're adding one or two, but it's after they've already grown 40, 50, 60% year over year. We have a lot of folks listening and obviously the previous guests as well, they just they buy a lot of businesses and the very obvious problem in most every SMB and what private equity and overall business buyers have spent decades to reduce is the key man risk. And now listening from your end that you can just put all this data into those into AI and educate this AI agent and what you do pretty much is get all this knowledge and experience from the business owners head and put it into AI. It I can see it makes a big difference. Obviously you you can't replace the in person relationships which someone has built over over decades. But when it comes to business multiples and the overall key man risk, maybe you can talk what you have seen there. I think it's it's a great point. One of the things of course you're right. You're not going to totally eliminate a key man risk in a business. Of course these human to human relationships are so critical as I've said. But I think when you can distill what's in that owner's mind over time into this memory layer such that you better understand how they make decisions, how they evaluate risks in the business, how they see opportunities. You can start to value and think about that business differently. And I think a lot of the discount that a buyer gives to a business is just a lack of I don't understand what drives risk or decision making. I don't understand what drives success that guy does. So I need him not only because he has a relationship but he or she understands what is going to be a successful project or job or whatever in this business. And I'm not sure I understand that for this particular business. Maybe understand it in a category but for this little business in this little market, you're not sure you understand it. So you apply a discount. And I very much believe that a well deployed AI architecture can distill and extract that knowledge into a layer that a buyer can look at can interact with can use their own AI agent as they do diligence to connect into that layer and really start to understand the nuances and the specifics of that individual business like people all these businesses are individuals and they're different. And they have their own nuances and they have their own risks. And so there's often not a one size fits all approach but to the extent you can more deeply understand what makes this business successful, what makes it risky, what makes it a tremendous opportunity if you could only do X, Y or Z, you can more fairly value that business. And that's a win for the buyer and the seller. And so I think the opportunity to deploy these architectures particularly in the lower middle market where you see these massive discounts on purchase prices, I think benefits both parties because you're getting a better deal. The buyer more deeply and more confidently understands what they're buying. And the seller has set up their business in such a way that it's more likely to succeed in the hands of another buyer of another owner. And I think that part maybe has lost a little bit too. It's not just, well, how do I get an extra turn on the sale of my business? So I want to do this. Typically you have some maybe role equity as an example and you want to make sure that you maximize that. And so you care deeply as well as a seller about making your business more valuable in the hands of another owner. And so why wouldn't you want an architecture that gives you more predictability with respect to how to run your business? And I think that's where a lot of our customers and requests are really focused, especially these individually owned businesses is I want to make sure I get the most value for my business when I sell it. But I also want to make sure that it's really successful in the hands of another owner. Let's say someone is listening, thinking of selling maybe one of their portfolio companies within the next 12 to 24 months and what they hear is like, okay, it makes a total sense. But where do I start? What knowledge or data should I be putting into AI to make sure in 12 to 24 months my business would be a bit more valuable, for example? Well, I can't believe that people are just using an off the shelf tool and calling it AI in their business. So if you're doing that, great. It's better than doing nothing. But that certainly doesn't provide any real value enhancement to your business. What I think folks like that ought to be thinking about is how do I have a real AI architecture in my business? In other words, how do I have my own memory layer, not somebody else's memory layer, somebody else's tool that I'm just using as if it were a SAS subscription? How do I have my own memory layer? How do I have my own architecture? How do I have an AI agent that connects in to all the systems that run my business such that when I ask it a question or an 18 months when a potential buyer asks it a question, it has all the relevant context answers and with a high degree of confidence can produce an output or work product that's very closely aligned with what my business does that's very specific to my business and is not some generic answer. That I think makes a business more valuable, subscribing to an AI product or maybe even building a little homegrown SAS product. Sure, maybe there's something to that. It's better than nothing, but I think having a real, a gentick architecture and you start to hear this buzz word more and more about a gentick AI, that's really what a gentick AI is is having your systems connected to sub agents that all connect up to a main orchestration point that you or your employees interact with that can give you answers, create work product and provide insights about your business that are real, relevant and context rich based on real things that your business has. It's all great. Business is risky, entrepreneurship is risky using AI seems like an operator. What are some of the downsides or risks when doing an implementing all this? The first one that comes to mind is cost. So cost is a huge one. And for a small subset of business, it's data security. I think data security is one that's a risk for everyone, but of course, on the spectrum of lower middle market companies, maybe someone is doing $5 million a year is less concerned with their customer data being sent over an API to open AI than a business that does $500 million a year. That business, no matter what it is, it might actually be concerned with customer queries that are unsanitized going across to Gemini, Claude and Thropic all in pursuit of well hopefully those companies now soon to be public have good security policies themselves but maybe you're making those businesses smarter or you're very least sending customer information that your customer thought was proprietary out into the world. So there is a data security layer and we deal a lot with healthcare businesses, my background is in healthcare so we've thought a lot about that we essentially have this anonymizer or sanitizer that we can plug into this architecture that will remove either patient information or sensitive customer information when you send an API request. So data security is a big one for the right size businesses. I'm not saying everybody needs a sanitizer on top of their agentic architecture but everybody should be concerned with cost and I think in the news there's been several articles now Microsoft canceling a bunch of cloud code subscriptions was probably the most notable one recently where I think in the absence of this context layer I described these agents are using the most powerful models to just bludgeon through and it's very inefficient and it's very very expensive and so I think one needs to think about how you deploy AI with the right model for the right task and I'll give you an example so let's take this roofing estimator as an example. You have this AI orchestrator maybe you want to use Mithos or Opus 4.8 or GPT 5.5 whatever the latest model is that you want to use you're really smart model as your brain but if you're reading PDF specs that just need summarization well maybe I don't need the latest and greatest for that maybe an open source model could do that most of these open source models are as good as the frontier models were six months ago so they're really good maybe I can use that for free maybe I have now spreadsheet extraction or I'm double checking spreadsheets I could probably use a Gemini model from two iterations ago that's really really good at math is excellent at Python knows spreadsheets and it will do a great job but for fractions of a penny whereas if I did that with the smartest model I would be racking up tons of costs I'll give you an example in one of the businesses that we engage or have an engagement with doing document generation based on a workflow it was a few cents to do document generation using the latest version of cloud and that jumped up to about $14 per iteration over the last few months so these companies open AI and Thropic X AI SpaceX there's soon to be public companies and they have shareholders now to answer to and they have to start making the economics work the only way to do that is to continue to release top models for which they can charge high per token or per usage fees and so if you as a business owner want to say well I'm just going to sign up with cloud or just going to sign up with with an Thropic or open AI and I'll just use that for everything that could be a very very expensive mistake we're already seeing companies come to us and say I got my bill from Anthropic and this is insane I don't know how people are using AI in their business well what they are doing is they're putting everything through the smartest model what you want to do is you want to have smart tasking like I described and to the extent possible you want to use what are called local models which don't go outside the walls of your building or maybe they live on a cloud server but they're not going across and so if you have sensitive customer information or you're concerned about cost routing tasks to a local model can be a great option because there's no usage fees for those the usage fees are baked in to whatever you spend to run that model whether it's a local server on premise or whether you use that in a cloud server you could use it as much as you want without any additional cost so I think cost will continue to be probably the top point of discussion and concern among both large and small businesses as they think about AI deployments David remember I said at the very beginning of this episode that I want this business owner investor whoever is listening to to really forward this episode to to their portfolio company or the co-founder to just keep these AI ideas going for them that being said I want to finish with you getting given maybe more practical let's say there is someone listening who runs one of those traditional businesses one to ten million of EBITDA and they would like to do something within the next maybe maybe 30 days so they have a bit of like a before and after what would be something you you you suggest them to do so the first thing I suggest every business do is connect an agent to your systems so that when you start to interact with AI you can have an answer that's relevant to your business that first step is easier than one might think so that's what we do as a first step engagement for a lot of businesses and then you can start to think about which work streams whether that's roofing estimating or medical device compliance or some sort of collections reporting that you want to automate as a particular work stream maybe as an agent loop you'll start to hear more about these agent loops which you essentially set goals and have an agent run within your architecture to continuously iterate that's step two three four or five the first thing is connecting your systems into an agent that you can interact with so you can start to ask questions of these powerful AI agents these powerful AI models that have context about your business so that'd be the first thing I would recommend every business owner do and there are a couple open source tools that folks can use folks have probably heard about this open claw idea that became really popular and then kind of fizzard fizzled out and I still use my open claw here and there but what you really find is at an enterprise or business level those don't really cut it they're probably not secure enough either but they're really not built for real enterprise or even small business they're built for individual use cases but you could certainly as an individual if you have a really really small business try to do that the problem with those is they could be technologically overwhelming to set up you almost have to be a developer in order to get those working properly and most folks I've talked to have tried to implement open claw successfully who are not developers have just spent ten times more effort and energy trying to get it to work than they have actually using it but my point would be connect your agents to your systems as the first thing you do before you try to do anything else with AI in your business what I learned today is if done well if AI integrated in a correct way you can make a simple business a lot more valuable by just having all the information in one place making better decisions sending out more and better proposals so that being said I know you're busy but maybe maybe someone would like to reach out what would be the best way for them to contact you yes I'm on X DW flickinger flicai and gbr or at vellum.ai you can go to our website we have a contact form there and you can schedule an intro call we don't do a pitch deck we just have a conversation about your business we are starting to deploy forward deployed engineers into larger businesses but for smaller businesses we typically engage folks remotely and so we're always happy to have a call a conversation I would recommend most folks go to vellum.ai and you can quickly and intuitively book a 30-minute call we'll have one of our folks chat with you about your business and see if there's a fit but I would continue and encourage folks to think about how however they use AI whether it's through us or through another provider that does similar things to us if your AI deployment is increasing the value of your business relative to competitors if you're just using the same off the shelf tools that everybody else is using are you really enhancing the value of your business are you really gaining a competitive advantage and I think that's the crux of the conversation that we like to have with folks is to explore ways in which we can help them do that so would welcome any outreach and always here to chat awesome this was 50 minutes well spent So thanks a lot David for sharing all those lessons. Yeah, thanks for making it was a pleasure to chat with you and thanks so much for having me. I hope you enjoyed this episode of Pires and Builders. Please remember to subscribe wherever you're listening and leave a review. It really helps the show reach more buyers and operators. And if you already haven't, go back and listen to some of the other episodes of Pires and Builders. This podcast is built for people who buy, build and hold great businesses long term. Every guest you hear on this show has real, hard-won experience, acquiring and operating companies and the goal is again very simple. To give you ideas you can actually use in your own journey.