00:04:13 · David Flickinger
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 artistry. It still is, especially in spinal deformity surgery where the surgeon looks at a spine curvature and says, 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 specific implant according to its parameters, the patients had better short term and long term outcomes as measured in follow-up studies over several years. So it was really fun project in business to be a part of. It was acquired in 2020 by... Incidentally by Medtronic, 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 health care 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 ChatGPT 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 Vellm and what I'm doing today. So a long 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.
00:06:48 · Mikk
[00:06:48] S0: 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 is a typical ChatGPT, like just putting all sorts of documents there and just asking to to summarize few things or just the numbers or or things like that. So I wanna go smoothly and ask from you here, like, can you maybe walk us through the difference between asking ChatGPT a question versus having an AI system that actually lives inside the business?
00:07:31 · David Flickinger
[00:07:31] S1: Yeah. It's a it's a great question. And I think at this point, we've all used ChatGPT or Claude or Gemini or some chatbot 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 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 chatbot 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 a expert in x y z 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 chatbots at best are maybe 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, but it's delivered in such a confident way that they might make a bad decision. And that that would be the worst case scenario.
00:10:06 · David Flickinger, Mikk
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. [00:10:24] S0: 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'll 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 it's pretty dangerous in a good way. So please, could you please share the story?
00:10:57 · David Flickinger
[00:10:57] S1: Sure. I think this is a 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 to 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 a 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.
00:13:32 · David Flickinger
And, ultimately, that orchestrator agent is applying the rules and logic of the business to generate a proposal. That's really what the estimator is 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 I'm about to give you was sort of the ultimate test of, hey. Does this work? So the estimator, this... 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 gonna do it on their own, and this AI may not be ready for prime time. And, indeed, we hadn't actually deployed it live where it was being used every day. We were still iterating a bit, but we were pretty sure that we had it right. So we said, well, we'll take all the documents. We'll have it. We're not gonna interact. It'll be a blind test. And so we ran our simulation with the AI, and it took about twenty three 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 $6,500,000 estimate, they were within $400 of each other. And even us working and having designed this 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 and why shouldn't it? And so that was the moment, and there's a great story about that 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 use of these tools and these architectures in their business. But it's a really fun example because the size of the project was so large and the outcome was so close. Twenty three minutes for the AI, two out...