The AI Advantage You Can’t Buy with Adrian Rosebrock

[00:00:00] [00:00:05] All right, guys, welcome back to The Daily Mastermind. [00:00:10] George Wright III with your daily dose of inspiration, motivation, and education. And I'm excited to be [00:00:15] on the podcast today with Adrian Rosebrock. Adrian, how you doing? Adrian Rosebrock: I'm doing wonderful, George. [00:00:20] Thank you so much for having me George Wright III: Yeah, I'm glad we could line up. Guys, if you are looking at AI, [00:00:25] trying to use AI, you're dealing with it, whether you're a small, medium, or large business, we're gonna be talking about some [00:00:30] really good stuff today because there's a lot of hype out there in the industry when it comes to AI, [00:00:35] and you need to know what to do, when to use it, where the ROI is and what to think about if you're [00:00:40] trying to decide whether to buy solutions or create solutions. So let me give you a quick background on [00:00:45] Adrian. Computer scientist, entrepreneur, AI architect, investor. He's [00:00:50] the CEO at Chief of AI, the company. But he built Pie Image Search into a [00:00:55] leading computer vision and deep learning education platform that was acquired [00:01:00] recently a few years back. But now he helps organizations move past this AI [00:01:05] experimentation mode into actually getting an ROI and using it. And so this topic is [00:01:10] one that I think I've really vetted individuals that know what they're doing, and that's why, Adrian, I wanted to [00:01:15] bring you on the podcast. So I really appreciate you being here and I would love it if you just, before [00:01:20] we got into AI, told me a little bit of the background you had with your acquisition and with Pie Image, [00:01:25] 'cause it it helps people to understand where you came from and, why your knowledge is what it is Adrian Rosebrock: [00:01:30] Totally. Yeah. I always tell people I started as an entrepreneur when I was 19 years old, and I [00:01:35] built a couple apps in in college. The first one made enough to make my car [00:01:40] payment, the second one made rent, and the third one is the one that really took off. So during my final [00:01:45] year of grad school, I started PyImageSearch which is effectively an, an info product, [00:01:50] but in a very specific niche of computer vision and deep learning. So imagine [00:01:55] writing software that could see what is in an image, like face recognition, self-driving [00:02:00] cars license plate recognition. This is all stuff I did for my PhD, so I understood the [00:02:05] theory behind it, but when I was just getting started, I really struggled with the implementation, with the [00:02:10] experimentation, writing the code to actually do the thing. And I wasn't alone. Other [00:02:15] students, other researchers, other developers in, in the enterprise world were also struggling with this. [00:02:20] So I set out to, to fix that problem. I wrote nine books, two [00:02:25] courses, 500 and some tutorials, before generative AI was a thing. So all of this was written by hand [00:02:30] and grew PyImageSearch into a, a very large website top 20,000 website in the world [00:02:35] when it was acquired about four years ago, half a million people on the email list. So it was pretty [00:02:40] substantial for what it was. And then after it after it was acquired, I, started teaching myself [00:02:45] some some other topics just for fun, did some self-improvement, but ultimately came back to AI 'cause [00:02:50] I, I could see entrepreneurs, I could see business owners struggling, finding the signal through the noise, trying [00:02:55] to recognize what's true and what's overblown hype. George Wright III: Yeah. It's [00:03:00] interesting. You seem to have followed areas that you're passionate about, which is the journey of a lot of [00:03:05] successful entrepreneurs. And of course, AI is pretty big right now, but what [00:03:10] really made you decide to go down the path and really get deep in, in AI and create [00:03:15] this company around Chief of AI? Adrian Rosebrock: It was honestly working with my friends. I [00:03:20] belong to a number of entrepreneur membership communities, and just due to my background, people would [00:03:25] come to me and ask me questions about AI, what they shouldn't do, ask me to vet solutions [00:03:30] or look at kind of roadmaps of where they were going. And this was even before I had started Chief [00:03:35] of AI. As soon as Chat GPT came out this started happening where these business owners, they were [00:03:40] recognizing, "Hey, we need to get AI into our business." They could see the importance of it, but they had no idea [00:03:45] how to actually go about doing it. So it just started off naturally, me helping out my [00:03:50] friends, wanting to see my friends succeed. And then over time it evolved as [00:03:55] AI evolved into more of this small-medium business enterprise [00:04:00] solution business where we're actually helping people from the ground up actually architect AI [00:04:05] solutions. And not just around building workflows or giving you prompts, but understanding the [00:04:10] privacy concerns, the policy concerns, how you actually architect the data, how do you build data sets for your [00:04:15] company? How do you train custom models that gives you an actual, real proprietary [00:04:20] lift in your business? George Wright III: Yeah, there really is so much information out there, but I [00:04:25] think one of the challenges a lot of businesses face is that it comes along with this idea that [00:04:30] because it's AI, you can figure it out yourself. And so it's a pro and a con. [00:04:35] And so with all the hype around AI, I wanted to ask you, where do you believe business [00:04:40] leaders really should be focusing, and where do they start? Because you talk [00:04:45] about, I know that there's, architecting solutions versus, buying [00:04:50] products and subscribing to, systems and software products with AI. Where do you get most [00:04:55] businesses to start, and where should they be focusing? Adrian Rosebrock: It's a great question. I explain [00:05:00] this to people as looking at where AI adoption currently is and where it's [00:05:05] going. There are three waves if you look at this. The first wave AI access is like [00:05:10] 2022 to about 2026, where the primary adoption metric has [00:05:15] been, how many people are using AI within your organization? Do you have, ChatGPT accounts? Do you [00:05:20] have Claude accounts? Maybe you have some some power users on your team have Claude [00:05:25] Code or Claude CoWork set up and they have some custom skills. Maybe you're sharing some skills back and forth [00:05:30] betwe- between your team. But the primary success metric here is adoption. How many [00:05:35] people within your organization are using AI? The issue here though, especially moving [00:05:40] forward, is that there's no moat surrounding your business whatsoever. Anybody with 20 [00:05:45] bucks a month spend can get access to Claude or ChatGPT, have access to the exact same [00:05:50] models that you have. Now we're starting to shift more into wave two, which is about the [00:05:55] economics, which is about 2025 to 2028, which isn't honestly that, that far [00:06:00] away. And now we're starting to understand like the actual cost. What's the ROI? Are [00:06:05] we getting real value out of that? And also what's the governance and the policy surrounding these things? [00:06:10] And starting to look at the real unit economics of this. So in April [00:06:15] 2026, OpenAI was showing some financials that demonstrated they had about [00:06:20] 3.7 billion in revenue, but cost them about 12.5 billion in expenses to [00:06:25] actually run and operate. Obviously, that's not a sustainable business, so you have to look at [00:06:30] the, the token economics of how this is even gonna work out at scale. And, [00:06:35] Claude, ChatGPT, these companies can eight to 10X your token price over- [00:06:40] overnight, which can completely blow your entire business model if you're, have a lot of token [00:06:45] spend. So that's where George Wright III: if you build your business around it, right? If you're building your business around it now thinking it's [00:06:50] working, and by the way, not even looking at the economics and they start to scope creep on you, [00:06:55] and then all of a sudden something changes, it's like building your whole business around social media and having them pull your page, [00:07:00] right? Adrian Rosebrock: Exactly. Exactly right. And that's why tokenomics is this brand new [00:07:05] idea and thing to consider where businesses are literally purchasing token futures so they can lock [00:07:10] in their token costs for the next few years. Just as you, if you're running like a, a- an industrial [00:07:15] operation, you might wanna purchase oil futures so you know what your true oil costs are gonna be.[00:07:20] Tokenomics are also becoming very much a thing to consider and with that, [00:07:25] where we're going and I think is the most important part to focus on is this wave three or what I [00:07:30] call the, the differentiation period. And that's, I estimate 2028 to 2020-- 2032. [00:07:35] And here we focus exclusively on the proprietary capabilities the competitive [00:07:40] advantage of actually using AI within your business. And my favorite example of this [00:07:45] is the hedge fund Bridgewater. Bridgewater is known for having [00:07:50] recorded like all of their internal conversations, every investment they've made, they have a detailed paper [00:07:55] trail of why they made certain decisions. So when, large language [00:08:00] models became prominent and useful, Bridgewater had this treasure trove of dataset of a dataset [00:08:05] already available to them. So they built a dataset that the goal was to help to create an AI [00:08:10] model to help their team members make smart decisions for the company. So [00:08:15] they focused exclusively at least initially on their investments, and they trained or [00:08:20] they they just gave gave the model to their team members and just started...[00:08:25] They said, "Start prompting this thing, and we'll measure the accuracy and the quality of the [00:08:30] results that we get." And they got about 50% accuracy, about a coin flip not very good. [00:08:35] Then they handed this over to their engineers and said, and their investors, and they say, "Tune these and [00:08:40] write custom prompts for us, like expert written prompts. Let's see what type of accuracy we can get [00:08:45] off of that." Accuracy jumps to 60, 70%. And then they go into their AI [00:08:50] team and they say, "Okay, take this dataset and just [00:08:55] fine-tune a large language model, custom train it on our dataset." And when they did that, they [00:09:00] measured the accuracy over 80%. And what's even more interesting is that their [00:09:05] token cost dropped by 14X because they were able to self-host their model [00:09:10] independent of ChatGPT, independent of Anthropic and the Gemini, the other players. So they [00:09:15] have this self-hosted sovereign model that doesn't update Claude and ChatGPT [00:09:20] will at a whim. Like that they control the token cost on, and it's beating, [00:09:25] just copy pasting prompts and skills into into a large language model. Now [00:09:30] Bridgewater has a true proprietary lift because n- very few hedge funds, [00:09:35] especially not to the degree that Bridgewater does have access to these types of datasets, and now they've [00:09:40] trained the custom model on top of that. That is where AI is gonna be going in the [00:09:45] next five years. Because again, there's no moat around a $20 a month account. You [00:09:50] have to be looking internally, what's your competitive advantage? What are you doing internally that [00:09:55] no one else can? And start engineering your AI efforts surrounding that. George Wright III: Yeah, you [00:10:00] actually said some things that really resonated with me, which is a moat around your business. Obviously when you [00:10:05] do that, but when you combine it with a competitive advantage, because most people are utilizing some of the same [00:10:10] stuff, and they will, but when you have proprietary information and you build it the right way, [00:10:15] it not only creates a moat around your business, but you get a distinct competitive advantage. And like we're starting to [00:10:20] use more agentic sites with some of our stuff that we're doing. We're not really necessarily [00:10:25] containing that, and I think our token cost needs to be looked at as well. But as you start [00:10:30] to do that in a in a system that can [00:10:35] self-evolve, the accuracy and the ROI grows [00:10:40] substantially versus just constantly using the tool, right? So we've had the same thing happen where, [00:10:45] individuals in our company, we want them to be using AI, but when we start to collaborate around AI [00:10:50] operating systems, they get even more productive because now the learning is collaborative. But then [00:10:55] I could see what you're saying when a business starts to build an infrastructure or moat or [00:11:00] even operating system around their system. You gotta build that, right? So I guess the [00:11:05] question for me would be, should you build something or should you buy [00:11:10] solutions? And you talk a little bit about this, so tell me what your philosophy is on that. Adrian Rosebrock: Yeah, [00:11:15] so whenever we talk to a potential client or a client, we're trying to help them out and [00:11:20] evaluate where they should apply AI into their business, which by the way, is the number one question we [00:11:25] get. Where in my business should AI go? We tend to just show them this, this [00:11:30] two-by-two grid where on the X axis we'll say, "What's the degree of differentiation to [00:11:35] your business?" Does every business run this like a CRM or, employee payments? Or is this [00:11:40] something highly differentiated to your business specifically? And on the X axis, we'll [00:11:45] say, "What's the degree of consequence or the risk?" If we, if you get this right, if you get this [00:11:50] wrong, what are the consequences? And if you plot this out, what you'll see is most of [00:11:55] the time you're gonna buy something off the shelf. If there's no [00:12:00] differentiation and there's very little consequence why would you build that? That's not a good use of money. You're gonna buy it off the shelf. [00:12:05] If there's little differentiation, but there's high risk and a good example like this could [00:12:10] be a CRM. Nobody wants to build and maintain their own custom CRM. That's a lot of work, but [00:12:15] it's very important to your business, and risk of getting it wrong, typically very high. So again, [00:12:20] you're gonna go buy off the shelf. Now, let's say like you're run- you're writing or running this [00:12:25] underwriting company where you offer loans to individuals or businesses, and you have a very specific [00:12:30] underwriting process. That underwriting process, that's a very intense degree [00:12:35] of differentiation. Very little people are going to be doing it the way you are. So now we're getting into a [00:12:40] little interesting area of the world where you're like do I buy off the shelf?" [00:12:45] Probably not, because nothing is really gonna serve what you need for your specific use case. But you should [00:12:50] first do your research, see what's out there. Can you find a solution that could get you 70, [00:12:55] 80% of the way there? And then we just have to modify a little bit. Maybe we write a little bit of code, write some [00:13:00] API integrations, try and get you most of the way there. It's only in this upper [00:13:05] right quadrant where something is extremely differentiating to your business and the risk of [00:13:10] getting it wrong is extremely high. That's the quadrant where you're gonna spend, the [00:13:15] effort, the time, the money to say, "Okay, we need to build this from scratch," because that is where [00:13:20] that true proprietary moat is gonna come. And again, like when we look forward to, [00:13:25] 2020 and beyond, companies that do the, the build to win approach for their high risk, [00:13:30] high differentiation, they're gonna be like Bridgewater. They're gonna have a actual real moat [00:13:35] surrounding their business. George Wright III: I love the way you said that and just for my purposes, maybe for the listeners [00:13:40] to clarify, you're talking about with the AI specifically. So if you have a high degree of [00:13:45] risk of getting it wrong and you really clearly wanna differentiate, this is where you [00:13:50] build, and this is where you build custom solutions basically. But the other areas would be, [00:13:55] can you buy a solution off the shelf, or can you buy a solution and modify it so that you can [00:14:00] definitely differentiate, but you really don't wanna get that, that process wrong? And I [00:14:05] would imagine that maybe some businesses, this would morph maybe over time or [00:14:10] as they develop more and more but this really does answer the question [00:14:15] of or at least the process of this, of where to deploy it, right? Because one of the things I was wondering, and I was gonna ask you, is [00:14:20] how should a company really identify the area to start with where [00:14:25] there's the most measurable ROI? Is it always to start with r- replacement value of [00:14:30] productivity, or is it to focus on the highest ROI first to [00:14:35] drive the metrics in the business and then come backfill it with productivity? What are your personal [00:14:40] thoughts on that? Adrian Rosebrock: That's a great question. I think focusing on productivity or hours [00:14:45] saved is a good place to start, but it's where everyone else starts, and I don't think [00:14:50] it will serve you in the long run. So my personal belief is that [00:14:55] the more and more AI gets integrated into our daily lives, the, the more [00:15:00] we see AI-generated text on Facebook, on LinkedIn, the more our brains [00:15:05] tune it out as just noise. It's like, we see it all the time on LinkedIn, low-effort posts [00:15:10] written with AI just makes you wanna just roll your eyes and move on. It's like this person couldn't [00:15:15] be, They, they-- Not only could they not be bothered to write it by hand, but they couldn't even take the effort to [00:15:20] use AI to make it interesting George Wright III: Yes. Adrian Rosebrock: So why... Let's zoom out for a second. [00:15:25] Why does that happen? A large language model is trained on effectively the entire [00:15:30] corpus of human information most of the internet. And when you train a model [00:15:35] in that way, n- it becomes an average of the homogeneity of this internet. [00:15:40] And then when you go and write a prompt and you get a response back, that's another average based off [00:15:45] of the weights of the model at runtime. So you get an average of an average, and for [00:15:50] your mathematically inclined or statistic-friendly people, listeners, what's an average of an average? It [00:15:55] reduces the variance. So your output actually shrinks, yeah, [00:16:00] significantly. So now everything is starting to sound the same. And again, I'm sure, Jordan you've [00:16:05] encountered this. Your listeners have encountered this. You're like, "Wow, the internet sounds like the same now. Everybody sounds [00:16:10] the same." It's because of everyone using large language models. [00:16:15] So It's my personal belief that if you want to use AI within [00:16:20] your business, use that more in the back office. Use it in ways that can [00:16:25] optimize your underlying processes. Use it in ways to increase your your [00:16:30] proprietary moat. What you shouldn't do is use AI excessively in [00:16:35] user-facing interactions, because those are gonna g- become more and more [00:16:40] rare and therefore more and more important and e- and then the most [00:16:45] differentiating factor with, within why someone would choose to go to your business versus a competitor. [00:16:50] It's because there is a real person on the phone. There is a real person at the end of the [00:16:55] chatbot. There is a real person who's typing that email to you. Maybe they're using AI [00:17:00] to assist, to pull knowledge from the knowledge base, but there is a real human being there. [00:17:05] So if the user has a problem, they will know there is a person there. So my [00:17:10] personal opinion is that if you wanna be successful with AI, don't use it to [00:17:15] replace all of your using-- user-facing interactions. I think that's actually how you hurt your business in the [00:17:20] long run. Instead, start looking at it as how can I start [00:17:25] optimizing and removing repetitive, boring tasks from my [00:17:30] employee's plate such that they can spend more time with the c- with our customers [00:17:35] to develop more face time with them, to develop actual, real relationships with them. [00:17:40] We live in the most technologically advanced time in our world, but yet [00:17:45] we all suffer from this, this deep loneliness. [00:17:50] So what's the solution moving forward? The companies that will be successful, the individuals that will be [00:17:55] successful will be the ones who learn how to nurture these relationships [00:18:00] while optimizing the other aspects of their business using artificial intelligence. George Wright III: that is so [00:18:05] critical. I'm a big believer in that, and I give you a very tangible example. It's like [00:18:10] I think a lot of people start with the intent of increasing their productivity, but then they think [00:18:15] geez, if I could do it in half the time, why don't I just do more of it?" Case in point, at The Daily Mastermind, [00:18:20] we receive hundreds of requests to be interviewed at any given time. [00:18:25] And, in the beginning, it would be like, "I listened to your episode, and I heard an amazing [00:18:30] topic, and, I've got a couple ideas, and I'd love to talk with you." And you're like, "Wow, this person really spent some time." [00:18:35] Now, because of AI, everyone says that, and they have extreme detail. [00:18:40] But what's happened is I get now more requests, and they're all personalized, and [00:18:45] it's basically cold email coming to me now that looks extremely personal, so I just tune [00:18:50] it all out. In fact, we send it through filters now of filters. And so I think people think, "Oh, it's gonna [00:18:55] increase my productivity," and they're like man, I could do 100X more of these," and what they realize is that [00:19:00] personalization is no longer personal, and so they're doing it wrong. Now, if you can increase your productivity [00:19:05] to spend more time in your unique talent doing what you do great with relationships, now you can spend more time [00:19:10] with relationships 'cause you don't have to do the busy work. That's an effective use of it. So I'm a [00:19:15] big believer in what you're saying, and I think when it comes to ROI, it is a bit of a mix because [00:19:20] I know that if I innovate with AI, it might create a large [00:19:25] ROI, and then I've got this little save me a little bit of time thing. And so I think it's a [00:19:30] balancing act, obviously. But you guys at your company Chief of AI i'm sure that when you [00:19:35] architect solutions for businesses, you have to look at all these levers, right? Adrian Rosebrock: [00:19:40] Absolutely. And it's worth noting, it's like we, we are an AI [00:19:45] company, but I joke with, the team members and our clients. It's like we're actually the, company. [00:19:50] We don't really wanna sell you a workflow or a set of prompts or to come into your business and sit [00:19:55] next to you and watch you work and figure out how to optimize everything. That's what AI was, but it's not [00:20:00] where it's going. We're much more interested in yeah, sure, understanding where the key [00:20:05] areas of your business are, but understanding where that true proprietary advantage is, understanding how [00:20:10] we can create AI solutions that leverage that proprietary advantage. We're not [00:20:15] here to sell you some workflows and then leave after that, because I think there is a huge [00:20:20] disconnect between the founder, the C-suite, the, the stakeholder mindset and [00:20:25] your employee mindset. For people who have a very owner [00:20:30] mindset, they look at productivity and they'll be like, "Oh man I just got this task done in 10 [00:20:35] minutes. It used to take me two hours. Look how much more productive I am. Now I can do all these other tasks." [00:20:40] Whereas someone down, farther down the food chain in the, the business, they might not have [00:20:45] that ownership mindset. They may be a very good worker, but they're looking at it as "Wow, this [00:20:50] only took me 10 minutes. It used to take me two hours. Great. I'm can go hang out with my [00:20:55] family right now." And that's-- Right? So it's and there's nothing wrong with that mindset either. In fact, I'd [00:21:00] encourage that, that mindset. But just realize that, if you're a key stakeholder in the business, you're [00:21:05] more likely to want to optimize, optimize. Whereas, just your typical nine-to-five [00:21:10] employee, they just wanna do their job, do it well, get paid, and go home. [00:21:15] There's-- The incentive isn't there to optimize every ounce of the minute, every minute [00:21:20] of every day, which typically if you're an entrepreneur that's more of your mindset. George Wright III: Yeah, I think it's [00:21:25] the difference between dropping a really cool tool into the team and [00:21:30] architecting your company's vision and strategies around it so [00:21:35] that we're saying, "Okay, we're doing this with a purpose to optimize for more results," and then it grows [00:21:40] you, grows us, grows everybody. And I think you used a co- you used it a couple times, the term leverage. I really [00:21:45] like the idea because I've... Originally you start of it and think, it's like replacing a [00:21:50] task with a robot versus having a robotic arm or something. It's like [00:21:55] where you can say, "Now I'm augmented to where I can be super..." When I use that to augment my [00:22:00] own talent and skills, it's different than replacing talent and skills because now you're still [00:22:05] maintaining that individuality, but also just getting more productive and more ROI. So I really like [00:22:10] that. Do you see some... 'Cause you work with a lot of businesses, do you see what are maybe some of the biggest [00:22:15] mistakes that people are making with AI? And what comes to mind for me as you and [00:22:20] I talked for just a minute, because I have a pretty great relationship with a banking institution, [00:22:25] and we're doing a lot of media work for them and helping them to incorporate some things, and it brought up a an example [00:22:30] you gave of where there's some mistakes might have been made in banking. But give me some ideas that these [00:22:35] are topics around AI that you don't normally think about, but that you should start [00:22:40] thinking about and you're making mistakes with. Adrian Rosebrock: The, the biggest one is [00:22:45] governance and policy. And for good reason. It's governance and policy, it's so [00:22:50] boring. Like it's not a cool flashy demo of AI doing something amazing and [00:22:55] sexy for your business. It's not this great, product shot commercial that gets you [00:23:00] sold and amped up on wanting to go and do that thing. Like governance and policy is, that's admin work, that's [00:23:05] HR work. It's, there's so many things rolled up into one and it's just tedious and boring. [00:23:10] But the problem is that is exactly where if you're running a, a large [00:23:15] organization, where you need to spend that time because you can become liable for it. The, the example [00:23:20] we were talking about offline was this bank in the Midwest. They had an [00:23:25] employee who was, not doing anything nefarious, just wanting to get their job done. They uploaded [00:23:30] some sensitive information, some personal identifiable information to ChatGPT, [00:23:35] and obviously broke broke a lot of compliance rules in the process. The [00:23:40] bank had to report, self-report to the SEC regarding this violation. First time it's ever [00:23:45] happened in history due to this example. And I've seen studies that suggest that 50, [00:23:50] 60% of CEOs believe their employees have uploaded sensitive [00:23:55] information to these tools, either knowingly or not George Wright III: Yeah, and it's it's [00:24:00] interesting because you don't... It's one of those topics, it's like taxes. People just, [00:24:05] it's a boring topic, people don't wanna think about it, but if you're an active investor, you realize that taxes can [00:24:10] completely compound and change your ability to invest if you just put a little effort into it.[00:24:15] But this is one of those things, compliance, governance, policies you don't think about, number [00:24:20] one, because it's boring, but number two, because you just don't think about. You don't know if your [00:24:25] employees are using this stuff, just trying to be more productive and help the business, but uploading sensitive [00:24:30] data, uploading proprietary information, and so there's a lot of that. How do you [00:24:35] correct that? Is that just a process of architecting governance and compliance into [00:24:40] your overall business policies? What do you do with companies to help with that? Adrian Rosebrock: It's a, [00:24:45] to start it's a three-step process. The first thing you do is work with your lawyer, [00:24:50] work with your stakeholders, and you come up with your policies and rules of what happens [00:24:55] and what the consequences are. And then you inform George Wright III: by the way, with AI never really existed before. So these are things you [00:25:00] have to think about, not like what are our terms and conditions. You have to like creatively think of this probably, right? Adrian Rosebrock: [00:25:05] Exactly. You need to sit down and really think of the use cases how people could be using AI within your [00:25:10] team and mark off what is allowed versus not. You need to circulate to your [00:25:15] employees. There needs to be a bit of education surrounding your employees over the dos and do nots [00:25:20] surrounding AI, and most importantly, the consequences of what happens if they break these rules. That's [00:25:25] like step one. You need to get that. That's very defensive for you, so if an employee does something and you [00:25:30] get sued over it, you can at least point to these and saying, "The employee's responsible. We terminated [00:25:35] them." And then that'll help, help... It doesn't eliminate your liability, but it can certainly help your [00:25:40] case George Wright III: Yeah, 'cause intent, bad in- non-compliance or non-policies can comp- [00:25:45] compound the consequences Adrian Rosebrock: Absolutely I can. Absolutely I can. [00:25:50] So that's step one. The next step from there is you, [00:25:55] trying to think how I want to word this properly. Next step is start your [00:26:00] internal team discussions and figure out what what's what you need to do to [00:26:05] interface with your IT department. So this is typically for your much larger [00:26:10] organizations, 100 members, 200 members plus. Start looking at your firewall, your [00:26:15] routing rules, and you can start installing tools that can monitor traffic [00:26:20] in and out to ChatGPT, in and out to Claude. You could set up rules for common file [00:26:25] names that could acci- accidentally be uploaded to to ChatGPT. That way you're scanning [00:26:30] your network traffic for these violations actually happening, so you could step in before [00:26:35] something, something truly bad happens. This actually happened to me, ironically. I I was [00:26:40] working with a financial advisor from a, a large organization and one of their [00:26:45] employees became victims of a phishing attack. And the reason- and the way they were able to figure it out [00:26:50] is because all team members have to VPN through this central server so they can monitor traffic [00:26:55] going in and out. They were able to detect, oh, there was this file name that was uploaded [00:27:00] to, to ChatGPT that shouldn't have been. It was due to this phishing attack. They were able [00:27:05] to pinpoint the exact employee, the exact time that this, this violation happened. [00:27:10] That's another way that you can start looking at this. 'Cause you, you can't tell-- Without doing something like that, you have no [00:27:15] idea George Wright III: You'd never know Adrian Rosebrock: You never know because, someone's on their home laptop, on their home [00:27:20] internet connection, they're at a cafe using the internet connection. They could download anything [00:27:25] and just upload it to upload it to one of these services. George Wright III: Yeah, and you don't know until it's too late, so 'cause you [00:27:30] don't know what you don't know, right? Yeah Adrian Rosebrock: And what's worse is depending on how your [00:27:35] organization's enterprise settings are or the individual's on their individual Claude or ChatGPT [00:27:40] accounts, if you're uploading something via the web interface, that can become training data for the next [00:27:45] model that is released by Anthropic or ChatGPT George Wright III: that's true. So step one would be kinda like [00:27:50] identifying it all. Step two would be like guardrails and systems for compliance and monitoring, right? Adrian Rosebrock: [00:27:55] Yep. And then from there, if you're a Google Drive or a Microsoft shop, I [00:28:00] recommend preventing employees from downloading files to local disk so everything just [00:28:05] goes into an appropriate Google Drive folder and is loaded from there. This helps... It does, again, it [00:28:10] doesn't eliminate data being uploaded to, into a large language model, but it makes it [00:28:15] significantly harder, and it allows you to control file permissions folder permissions at an [00:28:20] enterprise level George Wright III: Oh, I love it. I love it. No, this is really good, and this topic was very important for me to [00:28:25] mention because I really feel like there's a lot of uncertainty. Even though there's a lot of hype around [00:28:30] AI, there's a lot of uncertainty, and that can be really clearly coming [00:28:35] from compliance liability, governance, and so I think that's important. [00:28:40] Last question I had for you is, how do you see AI changing the roles of [00:28:45] founders, entrepreneurs, and business leaders over the next, few years? Adrian Rosebrock: That's a great [00:28:50] question. Honestly it ties back to what I said earlier. It l- it allows people to [00:28:55] experiment faster, build proof of concepts especially for entrepreneurs who typically [00:29:00] neurodivergent in some way, ADHD or autistic, we're bouncing around all over the place. It's nice to be able to [00:29:05] create a proof of concept, translate that to a demo, and then hand it off to a team member, be like, [00:29:10] "Okay, now go and build this." Versus trying to have explain yourself and invest all this time and effort. It allows [00:29:15] you to take-- go from idea to proof of concept really quickly. Now, the trick [00:29:20] there is don't think that your proof of concept is production ready. Don't buy, code it, push it [00:29:25] online. That creates a lot of security risks. But it does allow you to be a bit more [00:29:30] creative and explore that yourself. And the other most important area I would say that I see AI [00:29:35] changing our lives for business owners is it gives us more time to build the human capital [00:29:40] relationships. Human capital is always, at the end of the day, been the core of a [00:29:45] business. It's the relationships you, you build with others, the agreements of this is what we're exchanging in [00:29:50] value for money. It's gonna become even more important with AI because we'll be able to automate more and [00:29:55] more tasks, do things more efficiently. So the reason why you would do business with so- [00:30:00] with an organization versus another one actually doesn't-- has little to do with the organization. It has [00:30:05] to do with your personal relationship to one or more people over there. That is why you're going to [00:30:10] make that business decision. So as you look for ways to AI to improve your life, to [00:30:15] automate your life by all means explore that, be curious, but try and focus on ways [00:30:20] that you can actually get more relationship value, and real authentic [00:30:25] relationship value, not, posting on Facebook or Instagram, but real in-person [00:30:30] relationship capital goes so farther than anything digital these days. George Wright III: Yeah, I love it. When you [00:30:35] go hand in hand, I think then you're really truly getting the benefits of AI, not just productivity. Man, I [00:30:40] appreciate the, the insights. There's a million more questions I could ask you, but for the sake of time, Adrian, I appreciate [00:30:45] you being here. And what's the easiest way or best way for people to connect with you, check out the [00:30:50] company, learn more about Chief of AI or keep tabs on the kind of great content that you're putting out? Adrian Rosebrock: [00:30:55] Just head to chiefofai.com. We're happy to talk to anyone who has any questions about [00:31:00] AI. We have a form there to book a call with us, and we'd love to get to know you George Wright III: Cool. Great. [00:31:05] All right, guys. Listen, if you're listening to this, I hope this has opened your eyes to a few things, given you a few ideas and [00:31:10] strategies, and I wanna be able to hear from you, so hit me up on The Daily Mastermind. Let me know what you're working on, what you're struggling [00:31:15] with, the things you're learning about AI, and what you wanna hear more about. We really look forward to [00:31:20] interacting with you, and make sure you share the show. That helps us grow the community, the network, [00:31:25] and that's what we're all about. So keep in mind and as I always say, it's never too late to start creating the [00:31:30] life, the business, what it is that you truly were meant to live. And so I appreciate [00:31:35] you joining us today. Hope you have an amazing week. This is George Wright III and The Daily Mastermind. We'll talk with you [00:31:40] soon. ​[00:31:45] [00:31:50] [00:31:55]