AI That Delivers w/ Armando Viteri (Neubloc) | RevIntel [Archive]

Guys, welcome back. We have another episode today of the Connect Podcast Chief Executive Network. We've got Armando Viteri, CEO of Nublock, and we're going to hand it over right away. Armando, thank you for joining us today. You are the CEO of Nublock. Tell us a little bit more about who you are, what do you do, and why do you do it?

Well, let's see now. So, Nublox has been around for some— going on 20 years now. It's been a company that— that's the bald head— but it's been around for 20 years. I've been CEO for all that time. It is a company that focused on architecture and implementation work for mostly deep tech, deep tech products. Boy, it is early morning here in Phoenix, so it's going to take a little slow-mo, slow going. Anyway, so what we did is we, uh, we would work in areas like, for example, if you've ever taken a picture of a check and have it deposited into your bank account, the underlying technology for doing that came from us. Uh, the firmware that probably drives 50, 60%, maybe more of the global cloud The firmware inside those storage drives was developed by our teams, etc. So it tends— it tended to be heavy tech, deep tech kind of work, and where we were a specialty agency for providing high-level engineers with some combination, historically some combination of US engineers, project managers combined with Eastern European, initially Polish resources. And Poland wound up being— Eastern Europe in general, but Poland specifically wound up being a wonderful place to be able to acquire heavy-duty engineering, computer engineering talent. So that's been very successful for us. However, however, back in 2017, my CTO came to us, came to the partnership that owns Nublock and said, you know, we've been— he has— he holds his PhD from first-generation AI. In fact, I was a Lisp programmer back in the early '80s. And so we've certainly seen kind of the rise and fall of AI over multiple generations. And we've certainly lived through our fair share of AI winters. Once again, that's what explains the bald head. But back in 2017, our CTO, Dr. Nassim Moser, he came back to the partnership, said, geez, you know, this one feels different. This one feels like it's a fundamental shift. And that was actually prescient, right? Because this is clearly way before generative AI or any of that stuff. I mean, it was the year that attention is all you need paper that led to generative AI was published, but in the pile of thousands of research papers that came out, it didn't at the time seem something that was going to be earthshaking. However, what was earthshaking is, is a number of research institutions and both commercial and academic research institutions were having a lot of success, especially— you remember that was a time period that DeepMind was having a lot of success with things like AlphaGo, AlphaFold. Et cetera. So somehow this deep learning approach, these complex neural networks were yielding unexpectedly good and in fact superhuman capabilities. And it was based on that, that the CMAST partnership to go focus on AI. That wound up being again a powerful move for us because it allowed us to be able to invest in the infrastructure, gain the engineers and gain the traction, gain the processes internally to be able people do that. We had a great deal of success of getting departmental-level AI solutions in large healthcare companies, tech companies, etc. So a number of Global 1000 companies that wound up again implementing at the departmental level, director level, VP level, etc. Obviously, generative AI in the form of ChatGPT comes around and the world completely changed, right? So all of a sudden, all these conversations that were at the departmental level became C-suite level. And all of these customers that were pretty dialed in on what their use case was, what was the problem they were trying to solve, became instead a set of people that are being exposed to AI for the first time. Kind of, they knew how to spell AI. And intuitively, they kind of felt it's either an opportunity or a threat, not quite sure which. So in that process, we morphed into more of a consultative company.

Okay.

It's a pure implementation company. So what we do today primarily is, again, we focus on Global 1000 companies, is probably 75% of our business, where we go through a very careful pragmatic approach to, all right, these are the practical uses. This is where you get your ROI. Here's some implementation examples and here are the things to watch out for. And in that process, we tease out a roadmap for these companies to be able to evolve from traditional businesses to AI-enabled business to AI at the core businesses over a span of years. Right. But we do it in bite-sized chunks. Where we're really focusing on implementations that are high impact, low risk, and low cost to implement as we're moving through this successive wave steps of a roadmap. So yeah, the company today looks nothing like what it looked like even 5 years ago. We've kind of morphed into a totally different company. And in that process, we've also become very deeply embedded with a lot of digital transformation companies that are now morphing into AI transformation, etc. So it's, it's, it's an entire ecosystem that's gotten built up over the last couple of years as AI is becoming— going from hype stage into more practical reality stage.

So what is the most common use case that comes across your desk, and the client has an idea of this should deliver a high impact and reduce my cost. What are the most common use cases that you kind of come up against in, let's just say, in the past 6 months?

Well, you know, uh, once again, I'm trying to— so much hype around this stuff. Yeah, it's just too much. And, you know, let's boil it down. There's only 3 things you can use it for.

Okay.

Number one, it is, uh, AI, some combination of narrow or classical AI plus generative AI, can be used in very creative ways. So it is an enabler for innovation, and you see that more than any other place. You see that this in the life sciences. Life sciences. So what AlphaFold did some number of years ago, Google more recently took the next step with that line of products. And it's, it's the use, for example, in pharmaceuticals. The predictions are it's going to reduce discovery costs for pharmaceuticals by probably two orders of magnitude. Okay. So that's Okay, is number one, right? And we're involved in all kinds of— in fact, I said 75% of our business is Global 1000. 25% of our business tends to be where we get approached by VCs that have portfolio companies that are more traditional software, and they're trying to enable them to become at least AI-enabled, moving toward an AI at the core offering. Because in almost all cases, whatever you can do in traditional algorithmic approaches, you can do better in some combination of traditional algorithmic, narrow AI, and generative AI. And as a result, all of these software companies are really trying to get there as fast as they can to be viable over the long run. A lot of those companies tend to be in that innovation category. So life sciences companies are trying to create tools to be able to accelerate things like drug discovery as a result of being able to, to be able to better match, better match pharmaceutical approaches. Now, given that AI can make better predictions about what things are more likely to have positive outcomes, that kind of, that kind of stay. The second use case is, um, agentic enablement. And what that is—

enablement, I have not heard that before. So is that like agent and then IC enablement?

Yes.

Okay.

Yes. And what it is, very simple, you take any white-collar job that exists in a company Okay, you create an AI agent to assist them, and you can get, depending on the job category, you could get somewhere between a, on the low end, 20-30% boost in productivity, and in some cases up to a 5 or 10 times boost in productivity for that person at the higher end. Certainly marketing content creation, software development, those kinds of things. As a result, PwC came out with a very recent study that they showed— what they actually did, which was clever, is that they did is taking an agent model and enabling all of this broad class of employees and workers inside an organization. What kind of boost up are you going to get to margin within that company by industry type. And in the numbers in general, in almost every category, the improvement was in the teens, as high as 19%. Surprisingly, one of the highest ones was in luxury goods.

Wow.

So you have potentially, because you're boosting productivity of your employee base by so much, you're able to do all kinds of things that you weren't able to do for the same investment dollars. So, for companies that are using that agentic approach over time, over a roadmap, they'll wind up with 30, 40, 50, you know, depends on how many job categories they have, right? You're going to wind up with dozens of these agents, each of which are specifically tuned to supporting that one worker being better. And not only— and keep in mind, not only does it boost productivity, it does two things. Number one, it boosts productivity at work. Number two, it turns an average worker into a rock star. So all of a sudden, wow, you are getting levels, quality of production, because it's going to be catching errors. It's going to work as a coach with him. Well, have you tried this? Have you tried that? It will work through it like a partner., right, that he works with every day and over time gets to know better and better what that person wants, needs, what his tasks are, etc. So the very last step of that is you then start connecting the workflows between those various agents. So you evolve toward this model where you have a singular core of company knowledge. Some people call, you know, what used to be called a data lake, now people are calling it a lake house. Architecture, right, of which you have these dozens of agents that are pulling on that data to be able to help the individual to enable these individual workers. And then again, you're connecting the workflows, eventually connecting the workflows to me to give them an extra little turbo boost in productivity again in kind of in the later stages. So there, the only issue is which of these of the dozens of agents that you could go create Which do you start with first? So that's a lot of our process is figuring out, all right, let's say you're an airline. Okay, probably the things that you're going to want to— a bus company or airline, something like that. Probably the things you're going to want to focus on is agents to help you with route optimization, fuel economy, things like that. Right? That's going to be high impact for an airline. In a clothing design company, shoe design company, Probably the highest impact agent you could develop would be an agent that actually constantly is scanning the entire internet, looking at fashion trends as they move. You know, typically fashion trends do tend to move from certain centers outward globally, right? And you can almost track over time that path as those trends move. Well, now you can, you have a little magic wand that allows you to say, okay, this, this, the, this is the fashion trend in Milan. I know it's hitting, it's going to hit Houston, Texas in about 15 months. So I can actually be ahead of the game in terms of design. Right. So for design, for design company, a shoe company, etc., totally changes the game from the, the current process and much allows them to be much more on trend. So it depends on the nature of the company, which agent do you actually develop first. But that's where you wind up with a roadmap of these agents, right? The third one, and back to your original question, the most popular, the one that people are implementing first, uh, but I think a good margin is, uh, sales and customer support enablement.

Beautiful. I love that you said that. Okay, in what way?

In Well, you know, in this, let's, let's use the example of my United Airlines 3-bump method. Okay, so I have a couple million miles, I have a couple million miles on United. That means I have a special phone number that I call. It's a higher level of service. Okay, call up that United Airlines number and you're going to get somebody on the other end of the phone that probably is getting paid $20 a day, has been in the job for about 3 days, has no idea even what you're asking. So we've all done it. May I speak to your manager? Right now you're going to get somebody that has maybe been with the company 3 to 5 years. You can handle the simplest of questions, but probably the simplest of questions you could have just gone online on the United site and done yourself. So now you ask a bump again. You're now going to get somewhere between that second bump and the third bump. You're going to get somebody that's been working for United Airlines for 30 years, probably a retired flight attendant. And this person knows every single trick in the entire book. It was like, oh, you're trying to get there to there. And, uh, oh, I see. I see. Yeah, I gotcha. All of a sudden, oh, you, you, you hear this flurry of click, click, click, click, click, because they know exactly how to work that. $60 seconds later, it's like, I got you. And oh, by the way, I got it cheaper than you're going to get able to get any other way. Yeah. That's who you wanted to talk to in the first place, right?

Yeah, absolutely.

And why does that person? Okay. Right? In fact, we're working with one major transportation company. And I love what one of the executives said there when we were kind of talking through developing this this customer support and ticketing agent for them? He goes, yeah, a very simple model. What I want to provide is I want to provide better than American Express Black Card level of service to every one of my millions of customers. That is impossible to do any other way than using AI. It's impossible because number one, you couldn't hire enough people. Number 2, you couldn't hire enough knowledgeable people, right?

You're absolutely right.

Even if it wasn't a matter of how much you paid them, you couldn't do it, right? You couldn't train enough people to do that, right? For them, you know, investment, you know, investment on the order of, you know, one of these agents, depending on the complexity of the company, is going to be in the range of maybe $300,000 to $500,000 to it. At the high end, maybe $2 million, right? It's a really big data set. You need a lot of GPUs to train on all this data. It's going to tend toward the higher end, right? But in general, you can get ROIs that are unbelievable with some of this stuff, right? One AI that we did for one Fortune 500 company, They invested, it was less than $500,000, $600,000 they invested total. It threw $10 million EBITDA to the bottom line with that one annual for that one investment.

Wow. That's insane.

So back to the sales and customer support. So obviously customer support, you're going to wind up with a conversational primary customer support winds up becoming a conversational agent. That's highly expert. That's, you know, the equivalent of the United Airlines 3-bump person.

That's incredible. So question on that, right? So question on the airline example. I think that's a perfect example. So many times where I don't have the premium President's Club airline package yet, but we're on our way there.

But when I do call, Yeah, you see the 3-bump method works for you too. It makes it to 4 bumps.

So I'm completely familiar with it. Now I have a question. So one of the clients that I work with, they are developing conversational AI right now. And it'd be interesting to hear your thoughts because the focus, and it leads actually into one of the questions that we're going to go over today, how should sales and marketing teams be using AI, which is in exactly that method. So the conversational AI right now comes from, um, models that are developed off of historical data from the client that we're working with. And then that historical data from— and this is where the question will come in— that historical data that happens over 10, 20 years depending on the, the life cycle of the company is then developed into a model that becomes your AI agent that is able to have conversations with prospects in your target market. And then through the development of more conversations, it learns, and then it's able to have more fluid dialogue that ends up in an outcome that you define, whether it's, let's say, for one of the clients that they have right now, it's debt recovery. So the ability to recover more debt, handle more phone calls, and pass on more qualified phone calls to human agents that can actually recover debt. Then you have pipeline augmentation. So AI, AI-infused pipeline augmentation where you assign 10, 20, 30, 40, 50, 100 AI agents onto an inbound qualified lead list, whether it's events, webinars, downloads, and then it can just communicate with those. So I see that.

Those are all very— this idea of using AI as a primary sales and marketing either assistant or direct interface that is being used in so many industries right now for so many different use cases.

And I see that it looks like it's becoming not saturated, but it looks like it's becoming a lot of noise. And then how do you kind of differentiate the impact AI, well-developed, well-trained AI can have versus what a lot of these faceless companies have where they upload a script and expect the AI to be, again, the 3-bump method versus using historical data, case studies, content, playbooks. How do you kind of differentiate then at face value which ones are highly developed and which ones are kind of snake oil?

Well, you know, well, I'll start with my own prejudices. For the enterprise customer, it makes absolutely no sense to use a commercial model. Okay. So if they're trying to work with Microsoft or OpenAI or Google, They're nuts because at the end of the day, at the end of the day, their primary value as a corporation will be that proprietary data that they own. Yeah. And they should protect that beyond any— that is, that is the crown jewels. And to allow a commercial model anywhere near that where the incentives are not aligned with this idea that this is proprietary data. Okay, I'm not saying that the IP leaks or doesn't leak with those. I would not take the risk because incentives are not aligned. And somehow when incentives are not aligned, you wind up with Facebook, right? That they use personal data in inappropriate ways. Right? Absolutely. So what we wound up doing is we know two things that's interesting. Interestingly, Microsoft Research wound up conclusively proving that if you take a smaller model, but you deeply train it in a very specific use case, it could outperform the largest models. Okay? So there is no really good reason to use a large commercial model. For those kind of specific agent kinds of uses. I'm not talking about the general purpose. I mean, I, I use ChatGPT like everybody else, right? But if I was developing for a guy doing route optimization for an airline, that's certainly not something that you're using a commercial model for. You're better off creating a bespoke model. So in all cases, we encourage It's up to the customer, of course, but in all cases, we encourage them to use public domain models, both for long-term cost reasons, but also because of protecting internal IP. So as a result, you have to think about these things in a different way. I think we are used to, we're all used to that, the idea that software is something where you have a whole batch of software developers that code up stuff and you wind up with some useful thing, right? That's not how AI works. You already hit on it. AI works. Yes, there's some amount, there's some user interface work, there's some development, there's customization work. There are software development pieces that need to be that need to get done. However, the performance is largely defined by the training, the fine-tuning. So as a result, it's a very different world from traditional software. So there, so going back to your question, is for enterprise customers that can afford to do bespoke, right? By all means, do bespoke implementation that's specifically tuned. I'll give you a couple more examples in the sales and marketing area. We own one company that is in the automotive industry. Okay. And for the sake of not disclosing without permission, what they do is that they act as a broker to large car companies. Okay. And they have a whole batch of sales guys that wind up being the brokers in the— they're a middleman, they're a broker in the transaction. Each of those middlemen, each of those brokers has a custom-developed bespoke AI that assists them in the sales process. Oh, that's a game changer. As soon as we did that, the sales productivity of those sales guys went up by 60%. That was in first generation. We have a lot of runway still on what we could get out of that AI to enable that guy. So, it's still in the sales marketing realm. But this one is an agentic thing, right? It's really supporting a person in his sales process versus the customer support, that example that I gave you where I'm on WhatsApp and I'm having a conversation with, with an airline, right, to book my ticket, right? Yeah. There's no, there's no human in the loop in that, in that conversation. This is, it's an enablement platform for those individual sales guys. Another example of that for a large tech company that I know very well, um, what they did is that they have, they have tens of thousands of SKUs. Of all these different parts, right? So, and these parts get baked into all kinds of products. In fact, there's probably from this, this company is probably hundreds of their parts that are sitting in your house or in your car today. Right. Well, company goes to purchase some parts from this company. And of course, they're going to ask some specific technical questions. Your data sheet says this. What happens You know, as your data sheet says, that's the performance up to 25 degrees. What happens at 40 degrees? You know, that kind of thing. Sales guy doesn't know the answer. What's he do? He fires off an email to the principal engineer. Principal engineer, not his main job to answer that question. He'll get around to it, but eventually he answers back. Customer gets his answer maybe a couple of weeks later. Right? Customer's unhappy because they probably delayed a product launch waiting on this answer. Sales guy's not happy because he did not make his sales. Company's not happy. They didn't get the order. Right? Now, what they did is they developed a system, an AI, that what you do is that customer asks the question. And I think initially, it probably could get about 80% of the answers because it had been trained on every stitch of technical data that existed within this company, right? So it probably got about 80% of the answers. Awesome. That's pretty good performance. But just as you said, what happens now is if it doesn't know the answer, it fires off an email to the principal engineer of not just that question, but all the related questions. And when that engineer answers, it's now trained. So now I I think they're up to like 92% answers. Wow, that's insane. Customer's happy, got his answer right away, doesn't delay his product. Sales guy's happy, he's got his commission. Company's happy, they got their revenue, right? Off of a system that— and this is a Fortune 500 company, you know, this is a system that they probably— they spent they certainly spent a low single 7-digit to be able to create this entire system. A lot of that was, was training costs, right? Because they had a lot of repository of data that they had to train on, right? So the kind of ROI you get, especially at enterprise customers, is outstanding. The tricky part is more of the middle market, the SMBs, because an SMB, if you're a $10 million a year company, you're not going to start having a half million dollar a year budget for developing AI agents.

That's very true, right?

And that's where the real— and I think a lot of the investing that's happened got really confused about this because all of a sudden they were developing these horizontal tools targeting enterprise customers. It's like, The enterprise customer is not going to, to use tools that they could potentially have IP leakage with, right? They're going to develop their own, tune to their own processes for that. It's not even a question of IP leak. It has to be tuned to their internal processes, right? But the SMB is not going to have that luxury. So there is a whole crop of startup companies that I do really like that are not the widget and gadget companies. They're not yet another AI presentation maker where we all know that, you know, Microsoft Copilot PowerPoint is going to be better anyway, right? It's not that class of startup that all kinds of money got thrown at, those kinds of widget companies. Instead, they're verticalized solutions, be it in the real estate business, be it in, you know, pick a vertical market segment in the SMB space. And you have a whole crop of these companies that are coming up with complete verticalized solutions. And now it's back to a SaaS model. So they're not, they're not spending half a million, million dollars. The, the real estate brokerage is not spending half a million, million dollars to go develop their own proprietary bespoke large language model. They are paying a, a software company, SaaS-like monthly payments per seat, whatever it happens to be. Those kind of companies have, have enormous potential, and they will revolutionize whole vertical sectors.

Jesus Christ. So from where you sit today and where you believe the market is going, aside from the airline example, pipeline generation, I think What you said and our CMO said is completely aligned, but have— it's like empower your sales rep with its own AI copilot. And it's like your AI assistant that is able to help you deliver on the task that you're supposed to do on a daily basis and increase productivity. But if you were to be more specific, closely related to tech, AI, SaaS, SaaS, uh, software development even to a certain extent, maybe an element of pharmaceutical and healthcare. How should those verticals and the sales and marketing teams in those verticals be using AI to the highest efficiencies today?

Well, again, you know, um, let's, let's focus it on the enterprise customer. For the enterprise customer, they need to they need to develop their roadmap, right? And what unfortunately what's happening right now is, especially for the enterprise customer, is that there is not a deep level understanding what AI can and can't do for them. They just have this intuitive feel that it's either an opportunity or a threat, right? So by default, they will go to large management consulting firms. Firms, right? We all know who they are, who have absolutely no domain knowledge at all, right? But are awesome at generating really pretty PowerPoints, right? And they will waste millions of dollars, and even more importantly, a lot of time with these engagements where there's actually meaningful content being generated for them. And then at the end of that process, what are those large management consulting companies going to do? They're going to say, 'Oh, now that we're completed with this project, here's the business card for, uh, one of the large tech companies that are happy to do a consulting engagement for you.' What do these large tech companies do? Well, they need to have a large win. So what do you wind up? You wind up with these $50 million plus, plus, plus AI development agreements. And I, once again, I have the bald head to prove it. I have seen these kind of boil the ocean kinds of agreements. My entire decade, and the likelihood of success of them is near zero. They're too big, they're too complex, they're too whatever, right? So what if you're an enterprise? What I would do is engage in a company that has A, has domain capability, and B, is not trying to sell you into the idea of here's the roadmap of 30 agents we're going to develop for you, that'll be $50 million. Please, right? Over 5 years or whatever it is, right? But instead says, okay, your highest priority thing you should do— and more often times than not, it's in sales and marketing, right? Or it's in software development. Those, those two come up every single time, right? Go do, go do, uh, agents for those specific use cases. Then pick off the next priority, next priority, next priority. In bite-sized chunks. After you've achieved success with the first one, then you move into second, the third, the fourth, right? That approach to me is much more sensible, but unfortunately it's not what a lot of large companies are doing. For the smaller companies, even for the SMB, what the tricky part is, is And I think— I don't— I think we're too early, but I think eventually that we will develop kind of industry vertical experts that really understand the deployment of these kinds of bazillion different vertical market solutions that are being developed. Yeah, because it's very confusing right now. I mean, if you are a real estate brokerage and you're trying to go find the right combination of packages for you that's really going to turbo boost your brokerage? How do you even do the analysis, right? So I think that that is a cottage industry that's, that's developing of people that have— that are kind of AI for pick a vertical market, right? That can advise these sorts, these sorts of companies to pick the right collection of tools for them, right? So for, so for the enterprise, a little bit easier because I think there's, there's a more developed infrastructure of companies that, that look like NewBlock. And certainly NewBlock, you know, and there's, there's certainly a huge infrastructure of digital transformation companies that in general may or may not have AI capabilities, but they know for a fact that digital transformation is evolving into AI transformation as their primary business. So What you do is go back to your digital transformation partner and make sure that they have partnerships with somebody that looks like a Nublock that has a— and that partnership together, right, that digital transformation firm that already understands what the data infrastructure looks like inside of that enterprise customer, etc., coupled with a company that actually understands how to architect and implement AI.

Magical.

And notice that none of this, none of this And that's why we very, very proactively are looking to set up more and more partnerships with qualified, competent digital transformation firms. Because I think it's that ecosystem. It's actually beyond that. I mean, it's a digital transformation firm combined with the AI architecture and development firm combined with the change management and data governance kind of firms. So there's, there's cottage industries of these, these sorts of companies that are now doing joint ventures or just playing partnerships, etc., that wind up looking like a one-stop shop for the, for the enterprise customer. But you notice that none of that, none of that was big foundation model suppliers.

I agree. So that makes sense. Now So let's say there's a CEO watching this that fits the criteria of a digital transformation partner. They see an opportunity to partner with a company like Nublock. What is the most imminent problem or set of problems they should be— well, they experience or are experiencing to step into a conversation with you and that be the right fit?

Well, the problem is a lot of it is the hype around AI, like every other technological wave, is really geared around making it seem as complicated and as confusing as possible to make it mysterious. Yeah, right. And so in our particular case, we have what we call the discovery phase. That's the very first engagement. And it's a combination of education process, but also kind of mentored kind of coaching brainstorming at the departmental level to extract out what are the potential agents so that we can cobble together a roadmap, but principally to be able to get to the point and say, you know, of your entire organization, these are the 1, 2, or 3 agents I would start out with. Let's start with the smallest, smallest of these agents. The least expensive of these agents, just so you've been through the experience of deploying an AI agent. Because you're going to— your organization is going to have to go through some cultural shifts too, as you shift into a model where your workers have a symbiotic relationship with their individual tuned-in agent, right? So, So that discovery process, whether Nublock does it alone by itself, or oftentimes we'll do it with a digital transformation partner that already understands the organization, whether, whether in partnership or alone, that kind of discovery process is kind of a pragmatic, simple way to do that's very ROI-focused and not lingo and hype-focused.

So is there, is there a procedure or a methodology when considering implementing AI? A lot of people, or a lot of business owners, I believe, would, would go right to let's start with the end in mind. What does the outcome look like? Uh, now if I have this picture in my head, can we go and look at an AI development firm to create this outcome. I believe those outcomes that they envisage in their head are sometimes just pure— it is purely imagination, and some of them is— they're not doable. So is it the right methodology to begin with the end in mind, or is there a different methodology that you've seen to be more fruitful when creating your roadmap for an AI process in your business?

Ball and tell you— could tell you what AI will look like in the year 2029, I could sell that crystal ball for a lot of money, which is, which is exactly the problem. Yeah, I think, I think at kind of at, at a, uh, not 100,000-foot level, but at an ISS level. We have an understanding that the enterprise is evolving into a completely different architecture. Happy to share our vision of what the enterprise IT architecture looks like at the end of implementing this stuff. There's a paper that we've written on that, that seems to have a lot of resonance for a lot of people. Happy to share that for you to share. Yes. With your folks. However, it is at such a high level because really what it's saying is that you're going— the data lake becomes even more important. You're going to wind up with some infrastructure that is the gatekeeper to this data lake. People are starting to call it the lake house architecture. And then you're going to wind up with these dozens of agentic enablement AIs over time, and then the processes will wind up being connected. So in an overly simplistic way, that looks like a totally different IT infrastructure than what exists today, and that is going to be an evolution. And the preferred evolution that kind of future-proofs it is do one agent, go spend a little bit of money, do one agent to enable your your most— the highest impact area inside of your organization. So you're on— you're getting an ROI right away for that thing. Then do your second one, third one, do your fourth one, do your fifth one. But right off the bat, you're getting an ROI from having implemented it instead of going down one of these, you know, $50 million boil-the-ocean kind of, I have this grand— I have this grand vision, give me a whiteboard, let me show you my grand vision of what it could be, that would be $50 million, please, where it's unclear where the ROI or where or when the ROI is actually ever going to happen, right?

Okay, so, so let me ask you this, right? There's a ton of businesses that I'm speaking to, my clients are speaking to, I'm sure you're speaking to, and they've got one toe in, one— well, one foot in, one foot out, and they're just on the cusp of, I'm going to go forward with this, to give them a kind of overview or a glimpse into the development implementation of these processes or the system around it. What can a business owner expect, or how would you describe what the process of building AI or integrating AI into somebody's company It's like, what would you describe that as? How would you describe that process?

Using the keep it simple method, right?

Um, you're—

what you're doing— enterprise goes through a discovery process in which they identify a handful of high-priority agents, right, based on the criteria: it's high impact, lowering cost, and lowering risk. Because some of these AI things can quickly turn into kind of more research than development, right? Yeah, yeah, yeah. So using those 3 criteria, you select a handful of these, you start implementing your first one. Okay, you decide, all right, this is the one I'm going to go do. Well, maybe it's, maybe it's a, uh, I don't know, it, it's, it's some sales enablement or customer service enablement app. You start going down that path. What's going to happen In our case, as we move into the second phase, which is called— which is the scoping phase, where we assign— typically we would assign a project manager and an AI architect and also full-stack architect. Those three people would act in partnership and they would wind up doing a lot of deep, deep, deep interviewing of the end users to extract what the real business requirements are of the system, but then a lot of deep deep interviewing of the existing IT infrastructure, right? The people in IT to be able to figure out what real data stores are, how do we need to augment that, how are we really capturing 100% of data that we're going to need in future to be able to continue, as you said, the continuous training that has to occur, etc. And at the end of that process, we will, we come back to the customer and say, all right, here are the requirements, Here's the architecture, here's the cost, time and dollar cost for being able to implement this. And at that point, that customer has everything they need to be able to hopefully then engage NewBlock, but could engage someone else too, right? Go do this, right? And then it's a straightforward development process. A whole lot more training involved, but a development process where you have the normal development and piloting and deployment and training and all the rest that goes into a successful deployment of a system.

And do you see that taking— is that like, so from the point of sale to the point of now you're integrated within the company, project managers are assigned, work has begun, I know it's very hard to tie a piece of string to this, but every project will be different. Do you see these, let's say, an agent, the first foundational model, a testing environment, checking that it's successful and can work within the business? What kind of timeframe is a business owner looking to have that kind of model installed in their business? What is it, 6 months, 12 months? They're looking more 24.

6 months is a good guess. I mean, it would have to be a really big, big model or agent to be developed to start going much beyond maybe 6 or 8 months. And also part of the time frame is defined by whether or not, you know, I describe a 3-phase process, right? Discovery, scoping, implementation. Those 3 phases. For a lot of customers that are a little bit further along the learning curve, they don't need a discovery process. I would say 50% of them do, 50% of them don't. For those that don't, you're, you're buying at least, you know, 30, 45 days of elapsed time by not going through that process. And we, you know, we, we are happy when a customer says, hey, the biggest problem I have is customer support, go do one for me. It's great for us, right? Because then we can immediately move into scoping and implementation and and skip the whole front-end part of it. So, that's fine. But yeah, I think the 6-month kind of timeframe beginning to end is about right for most, mostly any enterprise-class implementation.

Wow. So, backtracking a little bit now, I'm sure whoever's watching has a great understanding of how AI can be used, how it can be installed, integrated within the business. What kind of upside that you can experience, what kind of downside you can kind of get if you do it wrong. But for you, let's backtrack before you started Nublock. For those that want that little bit of inspiration, maybe there's a 20-year-old, 18, 19-year-old, 20-year-old kid saying, I want to build an AI company. I think I can do it better than this guy or just as good. Why did you get into entrepreneurship and building businesses in the first place? What was the very first thought or instance that you went through to have that, I'm going to go build a business?

Well, I knew it from the offing, you know, I'd wind up down an entrepreneurial career. I have MIT computer science, electrical engineering by academic background, worked for HP as my first job, and I got recruited into a tiny little startup by the name of Sun Microsystems. Wow, no way! I know Sun. So I was one of the, one of the earliest people at Sun, um, and in fact Sun on the East, East Coast of the US, that for years was at least 50% of Sun's global sales, was me, my boss, and a secretary. That was it. That was who, who we were. And that entrepreneurial ride was so fun. I was there for 10 years. So I was there from its earliest revenue days all the way through where it was a Fortune 500 company. That was such a fun ride that immediately after that, I did my very first startup. And then I continued down that path. So kind of early career, I did a series of startups, kind of mid-career. I was turnaround CEO in a couple of situations for some investor groups. And then kind of later in my career, I've become more focused on more on the mentoring investing side and in an operator of NewBlock, which NewBlock, I'll tell you, you know, NewBlock has been the most fun I've had in my career since those Sun— those early Sun Microsystems days. Wow. Quite frankly, I'm catching the second wave of my career, right? You know, the first wave, the rise of the internet, all that I was part of in the days of Sun. I was never really part of the mobile, social, all that. That was never really part of my career. So I totally missed that wave altogether. Not that I would have wanted to have been involved in the rise of social. But here's the— third major wave of my career and arguably the largest. And I'm getting to catch that wave in an interesting way. And in fact, in this particular case, what's happening is that I get, you know, in the morning I'm talking to a massive agricultural, industrial-scale agricultural farmer looking at how to use AI. Then in the afternoon, I'm talking to a major airline, how they're going to revamp opera. Then later on, I'm having drinks with a major music production house, right? Wow. One of the big movie production companies, or etc., etc. You get the idea. Yeah, yeah, yeah. Multi-billion dollar real estate, commercial and commercial retail and home residential real estate. Developer. That's like my typical week. So, I'm getting to see all these really incredible, interesting applications being implemented. They're revolutionizing whole new business. Well, whole existing businesses into a completely different stage in their evolution. So, it's kind of like of the three waves of my life, I missed one, I'm catching two of Well, this is the second one.

So for those that would like a little bit of social mentorship, what would you say are the top 2 or 3 habits that you've been very consistent with that have kind of ensured or enabled you to achieve the success you have today?

Well, I'll kind of slightly redirect that question. You know, for somebody that really wants to go do a startup company. Uh, the— I'm not gonna— this is not gonna be the most earth-shattering revelation. Almost all startups that fail never achieve product-market fit. That's the reason why they fail. Okay, there could be lots of ingredients to that, but I would say the primary ingredient is that The biggest problem is people fall in love. They get some idea, they think it's the most brilliant idea they've ever heard, they fall in love with it. They go to dinner, they buy roses for that idea. They are so enamored of that idea because it's the best idea ever, right? And they don't wind up doing primary market research. Okay. Wow. Great. Awesome idea. Okay. Who am I selling this thing to? Guy who would actually write the check for this. Yep. Do I know how to get to that person? This is great. If the primary market buyer is a CEO of a Fortune 50 company, probably not gonna be able to get to him. Number 3, is do I solve a problem for him that is something that he needs to solve or would like to solve? How really imperative is this? Now, clearly that's a model for kind of B2B startup. B2C startups are much more random than that. I mean, they catch— nobody can ever predict which one's gonna catch fire, which one won't. And God, you know, I don't know, it's voodoo which ones actually take off. But for B2B kinds of companies, it's much more pragmatic in the approach. So the advice I would give is that before you go out buying expensive dinners and roses for this idea that you just had, do the homework and really establish whether or not you actually have a real idea or not. I've made this mistake early in my career. In fact, if I could go back to the long hair version of me, back in the '80s, hey, hey, what are you doing? Go do some research. I would have saved myself years and so much heartache, and I probably would have hair still today.

Okay, so last question. We've reached it. I didn't— again, these pod— I, I love doing these podcasts because you never know who you're going to meet, and the conversations that come out of the other side of these chats are just— for someone like me, the way I've acquired and gained so much experience, I've run over 300 episodes, so I've got a chance to meet with 300, uh, CEOs in the past year and some change for an hour each time and just get busted up on theses that were untrue or badly taught. And then on the other side, just gaining a ton of experience. So how— for this is, I think, is one of the most important questions. How do business owners and CEOs stay afloat when times get tough?

For kind of the startup world or for established businesses?

I'm not— let's say your first 5 years. Let's say your first 5 to 10 10 years, and you can use it, you can double-edge it, you could use two, two answers for it.

Okay, no, no, I got, I got the path I'll go down, which is for— let's go down the startup path, um, path I'm much more familiar with, right? Um, I think, I think there are two things you can do that are really, really, really key beyond the primary market research is figuring out whether or not you really have something or not before you start dating seriously. There's two things you can do. Uh, number one, go get yourself absolutely kick-ass advisors. Okay, I'm not talking about accountants and lawyers. I'm talking about CEOs that have achieved success, that have gone through all the way through the startup funding through the exit, have been through that path maybe even more than once, right? So they've sat in the big chair. And in addition to that kind of person, people that are really respected members of a specific industry that's deeply networked, etc., go get yourself that, that kind of advisor. And generally speaking, very senior people. They've been through their career are more than happy to give back. They're more than happy to be advised. Maybe, yeah, you just throw them some equity or whatever it is, but they're happy to volunteer their time because they see them, the, the younger versions of themselves in these, these young startups, right? The second thing, and probably even more important than that, is listen to them. I would say Of the startups that I see that do pick up good advisory boards, have really good mentors, oftentimes it's a question of hubris that these startup CEOs don't wind up listening to the advice being given. That's not to say let them go run your company for you. That does mean that you give meaningful thought and weight to what you're being told, because that person that you're sitting across from probably has had maybe not exactly the same situation in their career, but something really related, and saw how the playbook plays out as a result of this action or that action. Right. So, so yeah, I would say forming up an advisory board, it's never too early to form up a good advisory board, even, even if you're doing seed or in pre-seed funding, having a world-class advisory board behind you gives you a tremendous amount of credibility with prospective investors. So it's never too early. Go get really good guys and make sure that they're people that are actually experienced. Again, no slight on accountants, no slight on lawyers, but they're being advisors for the sake of eventually getting a customer in a funded startup, right? That's their objective. So we're— a real advisor has a completely different lens of why they're doing this.

Mate, that's been an incredible hour for me. Um, we've hit the top of the hour, so that's a wrap for the— for today's show. We may have to add on a few new, uh, topics of conversation. What are— are there any lasting words for people who are watching this interested in AI, interested in implementing AI, what would you say to those people?

Wow. You saved the hardest question of all. That is a good question because I think, I hate to say this, but I think 99% of kind of content that's out out there for people that are interested in ramping up in AI tends to still be, you know, there's a whole bunch of gossipy channels, you know, who fired who at OpenAI this week, you know, they're very gossipy, not trying to, you know, it doesn't help you ramp up, right? Then you have a whole other ones that are really kind of deep tech kind of jargon-filled techie tech stuff, which really doesn't translate into kind business objectives and ROIs and all the rest. So I think, I think there is an emergent set of sources, this among them obviously, where it's more kind of a more pragmatic, okay, we're past the hype stage now, and now how, what is the real part of emerging technologies, right? And I think that there's, there's relatively few sources of those kinds of information in, in the form of podcasts or YouTube channels or, or anything else. But I think they are, they are starting to emerge. These more practical ones are starting to emerge. So if you do a little bit of hunting— of course, here I should be able to spout off 3 or 4 of them that I like doing. I can't off the top of my head, but they are, they are out there. But there's relatively few of them.

Awesome. And where can people find you if they want to contact you or get in touch and learn more?

Well, feel free to reach out to the Nublock website, but if you want to contact me directly, I'm always happy to engage. My email address is aviteri@nublock.com. So feel free to reach out to me directly in email and I'm happy to engage. Otherwise, look for me on LinkedIn.

Very easy to find.

And I certainly encourage you to kind of take a look at my LinkedIn channel because it has an awful lot of business content around the AI industry-specific as well as kind of more horizontal views. Feel free to find my name on LinkedIn and I'm happy to link up with you.

Absolutely amazing. Armando, thank you so much. It has been extremely informative. And until the next one, bye for now. That's a wrap.

Thank you so much.

Take care.