AI & Cybersecurity: How Kevin Carlson is Guiding Leaders Through the Hype and the Reality

August 14, 2025•31 min read

Artificial Intelligence is everywhere. But with all the headlines, product launches, and bold predictions, how do you separate the real opportunities from the noise? That’s exactly what we unpacked on the latest episode of AI with Bry, where I sat down with technology executive and AI leader Kevin Carlson.

From helping organizations secure their systems to guiding executives through the complexities of AI, Kevin has a front-row seat to the intersection of innovation and risk. And as we learned, leading with AI isn’t just about chasing the latest tools—it’s about navigating with clarity and purpose.

Meet Kevin Carlson: Technologist, AI Innovator & Security Expert

Kevin’s résumé reads like a tech leader’s dream. A four-time Chief Technology Officer and current Chief Information Security Officer at Tech CXO, Kevin also spearheads the firm's AI practice—helping businesses adopt AI responsibly while keeping their data and infrastructure secure.

With deep expertise in big data, predictive analytics, machine learning, and even holding two AI patents developed for the media and entertainment industry, Kevin knows what it takes to lead through the fast-evolving digital landscape.

But beyond the technical credentials, Kevin brings something just as important to the table: level-headed perspective.

Cutting Through the Hype: The Real AI Conversation

When asked what he’s been learning about AI lately, Kevin didn’t hesitate: “There’s tons of practical application—but probably even more hype.”

It’s a sentiment many leaders feel but few say out loud. In an environment where every new AI tool promises to revolutionize business, Kevin’s advice is clear: take a step back, understand the true capabilities, and resist overcommitting to costly, overhyped solutions.

The Paradigm Shift: AI as the Interface

As we discussed recent developments from OpenAI, Gemini, and other tech giants, one thing became obvious—the way we interact with AI is changing rapidly.

“I think the paradigm is about to flip,” I shared with Kevin. “Instead of AI tools being hidden behind the scenes of apps, we’ll soon be interacting directly with AI platforms like OpenAI or Gemini—and those platforms will connect to everything else.”

For leaders, that means understanding AI isn’t optional—it’s becoming the gateway to how businesses, teams, and individuals engage with technology itself.

Leading Through the Complexity

With his background in both AI and cybersecurity, Kevin knows the risks—and the rewards—that come with emerging technology. His message to leaders is grounded in both optimism and caution.

“There's so much happening in AI. It's always more than you think,” he explained. But with the right approach—focused on education, clear use cases, and responsible implementation—organizations can tap into AI's potential without falling victim to the hype cycle.

Final Thoughts: How Leaders Can Learn, Leverage, and Lead with AI

The world of AI isn’t slowing down. But as Kevin reminded us, success isn’t about chasing every new shiny tool—it’s about thoughtful leadership.

Want to lead with AI? Start here:
✔️ Educate yourself and your teams beyond the headlines.
✔️ Evaluate tools based on real business value, not hype.
✔️ Keep security top of mind as AI becomes integrated into operations.
✔️ Stay grounded—the AI revolution rewards clarity, not panic.

🎧 Listen to the full conversation with Kevin Carlson on AI with Bry. And if this sparked new questions or ideas for you, share the episode with a colleague or connect with us—we’re building a community of curious, responsible AI leaders.

Episode's Transcript

Please understand that a transcription service provided the transcript below. It undoubtedly contains errors that invariably take place in voice transcriptions.

00:00:04.720 --> 00:00:06.860

Bryan Dennstedt: Alright. Here we go.

00:00:07.960 --> 00:00:32.480

Bryan Dennstedt: welcome back everybody to AI with Bry, the podcast where we explore how to learn leverage and lead with artificial intelligence. I'm your host, Bryan Dennstedt. And today we're diving into the critical critical intersection of AI and cybersecurity. Our guest is Kevin Carlson. He is a veteran technology Executive with a wealth of experience as a four-time chief technology officer

00:00:32.570 --> 00:00:57.410

Bryan Dennstedt: and a Chief information Security officer at TechCXO. Kevin leads our product and Technology group's artificial intelligence practice helping organizations navigate the complexities of emerging technologies and security challenges. Kevin's expertise spans big data predictive analytics machine learning and AI model development. And I think you're an inventor of 2 AI patents

00:00:57.410 --> 00:01:12.599

Bryan Dennstedt: that you've developed for a client in the media and entertainment sector in the past. So beyond his technical acumen. Kevin is recognized as an executive coach and guiding technology leaders through the evolving digital landscape. So welcome to the show, Kevin. Thanks for being here.

00:01:12.600 --> 00:01:15.969

Kevin Carlson: Hey? Thanks for the Intro Bryan! It's my pleasure. Happy to be here.

00:01:16.700 --> 00:01:30.429

Bryan Dennstedt: Well, let's start with that. Learn segment like with your extensive background, you know, and the rapidly changing world of AI that changes every day. I feel like what is something new or surprising that you've been learning about AI recently.

00:01:30.800 --> 00:01:45.559

Kevin Carlson: Oh, gosh! Well, there's plenty to learn out there, right. But I think the biggest thing that I that I learned is we're we're writing a hype wave right? There's there's a lot of hype out there. There's tons of practical application for AI, but there's probably more hype

00:01:45.560 --> 00:02:06.502

Kevin Carlson: right? So for me. It's it's really digging in and learning about what what capabilities are, what are great use cases, what are good applications of AI. And how can you approach them in a way that you know. Doesn't you know, over commit, you know, a client or a company to something that's super expensive, and, you know may not hit the mark. So

00:02:06.900 --> 00:02:16.689

Kevin Carlson: And I think the other thing that I'm learning about AI, that's just surprising is just how much there is going on with AI. Not surprising, but you know it's always more than you think.

00:02:17.160 --> 00:02:39.120

Bryan Dennstedt: Yeah, absolutely. I was watching something over the weekend. That kind of blew my mind. I want to get your take on. It is right now I feel like all these AI startups are secretly using Gemini, or Openai, or Deepseek, or whatever behind the scenes, and they're coming out with some amazing tools. But with Openai's announcement and some of the new tool sets I've seen Gemini roll out.

00:02:39.120 --> 00:02:47.730

Bryan Dennstedt: I think the paradigm is going to rapidly shift to the opposite. I think we're going to be going to Openai and Gemini as the interface

00:02:47.730 --> 00:03:03.520

Bryan Dennstedt: and those tools will be connected to everything else that you want to interface with and ask questions about. What is your thoughts on this rapidly evolving landscape of who's got the tools and the interfaces? And that's what I'm learning about, too. It's so exciting.

00:03:03.520 --> 00:03:17.579

Kevin Carlson: Yeah. So you know that that could happen for sure. But one of the challenges with with building a large language models. It just takes a tremendous amount of resources. So that's naturally going to lead to a situation where

00:03:17.960 --> 00:03:40.199

Kevin Carlson: large Llms are sort of concentrated in big tech. And you know, potentially government in the future. But just companies with tons of resources, and the rest of us are going to be using foundational models like Meta's Llama and others that are available out there and building on top of those, because you just don't have the. Not. Every company has the resources to, you know. Train a model like this from scratch

00:03:40.440 --> 00:04:03.399

Kevin Carlson: in terms of, you know, connectors and being able to do different things. You know, there's lots of cool technology out there right now to sort of standardizing the connectivity between things. So if you go back to what I consider sort of the beginnings of that with rag retrieval, augmented generation and providing context to these models, like in the moment. And then the ability to, you know, retrieve data from 3rd party services.

00:04:03.400 --> 00:04:19.599

Kevin Carlson: You know, I think that that sort of landscape grows. I do think it'll be possible, for, you know, non, big tech non-government entities to build similar models using foundational models and then implementing sort of those context additions themselves.

00:04:20.490 --> 00:04:46.999

Bryan Dennstedt: I love it, have you? I mean, I primarily am using Gpt all the time for just chats and deep research. I'm totally in love with deep research. I don't like. I have to wait 15 min, but if I ask my assistant to go research it anyway, they'd take 2 h to do it. So it's impressive on that front like, what? What about the video audio like? I don't know. Talk to us about some of the tools that you're dipping your toes in and learning about.

00:04:47.210 --> 00:04:48.256

Kevin Carlson: Yeah. So

00:04:49.160 --> 00:05:13.810

Kevin Carlson: well, video is really, really cool. And it's come a long way in the last year. It's pretty amazing if you look back at things, you know, mid 2024 to now the quality difference is just staggering. The other thing that's staggering is the cost. For, like, if you want to do a say, a feature length film using AI, you still plan it out the same way, using a shot list and whatever, and generate things, but the cost is pretty high.

00:05:14.229 --> 00:05:21.569

Kevin Carlson: Not as high as actually, you know, hiring a crew and making a fully feature film. But it's it's up there. Yeah, for

00:05:21.720 --> 00:05:24.709

Kevin Carlson: for audio. This is kind of a

00:05:24.920 --> 00:05:29.630

Kevin Carlson: this is kind of a spot near and dear to my heart, because.

00:05:30.620 --> 00:05:33.970

Kevin Carlson: as you probably know I dabble in music on the side, and.

00:05:34.460 --> 00:05:38.549

Kevin Carlson: Have been have been a songwriter for years, and what I see

00:05:39.000 --> 00:06:04.459

Kevin Carlson: copyrighted material being used to train these models. That that concerns me right? So there's that side of audio like music, which I think is there's some dangers to that, and some concerns to that where actually, I think, can rob us of some creativity and make it sound very middle of the road mainstream across the board. If people really get into generating their own music. But the other side of audio is like

00:06:04.866 --> 00:06:11.389

Kevin Carlson: you know, speech to text processing, being able to generate audio like 11 labs, has got absolutely amazing product for.

00:06:11.390 --> 00:06:11.740

Bryan Dennstedt: Hmm.

00:06:11.740 --> 00:06:18.749

Kevin Carlson: Voice, you know, generation, can, you know, make it sound just like me or you? If we provide the right samples, and it can be very useful.

00:06:18.980 --> 00:06:40.999

Kevin Carlson: The dark side of that is, you know, if you use a bank that that has like voice identification, it's very easy. Now, if you get a voice sample to override voice identification as as a form of like multi factor, authentication. So like most technology, there's a there's a great upside, and a and a you know accordingly.

00:06:41.500 --> 00:06:43.579

Kevin Carlson: not so great downside.

00:06:43.580 --> 00:06:51.939

Bryan Dennstedt: Yeah, that's for sure. That's for sure. I mean, it's such an interesting perspective. I haven't found a

00:06:51.980 --> 00:07:18.269

Bryan Dennstedt: competitor to 11 labs. I like, you know the general models, you know. There is the Big 3 4 5 out there, I think, and they all neck and neck each other every every month. It seems like the voice side. It seems like 11 labs is the clear winner. The video side is just mind blowing again. There's 3 or 4 strong competitors there, I think. But yeah, it's definitely I'm so excited, but so scared at the same time on certain aspects of this, for sure.

00:07:18.320 --> 00:07:20.040

Bryan Dennstedt: Did you have like.

00:07:20.490 --> 00:07:40.180

Bryan Dennstedt: you know, as we're trying to to learn, you know about AI. Have you stumbled across? Some of the misconceptions leaders have about AI's capabilities, and and even like, as you talk about securing these digital assets that we're trying to. We've spent all this time creating, and we want to protect them.

00:07:40.300 --> 00:08:00.090

Kevin Carlson: Yeah, let me. So I'll use software engineers as the example. This is the one that you always hear about. And I think it's because to a lot of people in business, software engineering is, you know, kind of an unknown. They haven't done it right. That's changing, I think, with the younger generation. More and more people are coming out of school, even high school, with some idea of what it's like to program.

00:08:00.590 --> 00:08:26.889

Kevin Carlson: And there's no doubt that AI is going to make programming different than it has been historically. But the misconception that I hear from a lot of folks is I'm going to replace all my software engineers with AI, and you know, my advice is, Yeah, go try that. And let's see how it works out. Because, look, the fact of the matter is, it takes care of things it knows how to do. AI is, you know, right now, is primarily

00:08:26.970 --> 00:08:52.000

Kevin Carlson: pattern recognition and regeneration. So for things like basic crud operations, create, read update delete operations on a database or authentication, or things that have to be done for every, say Sas application integration with payment systems like stripe and things like that. Those have been done hundreds of thousands of times, and there's lots and lots of code out there for them, and it can generate those things much more quickly than a developer.

00:08:52.730 --> 00:08:54.159

Kevin Carlson: great application.

00:08:54.560 --> 00:09:07.729

Kevin Carlson: But you know AI can primarily recreate or maybe synthesize from a couple of ideas, but it it still struggles today to do something that's truly innovative, or to integrate with a new service. It's never

00:09:08.330 --> 00:09:31.010

Kevin Carlson: seen or heard about or hasn't been trained on right? So when I when I think about the level of hype around replacing software engineers, I look at Microsoft for an example. You know they're big investors in Openai. Obviously, they're rolling out AI across all their products. And they came out with a statement a couple of weeks ago saying that 30% of their code is now generated by AI,

00:09:31.060 --> 00:09:59.860

Kevin Carlson: all right, that's and that's a good target. 30% is a great target. So if you're creating an app and say, 90% of it is stuff that's been written before. Is it really a unique app? I'm not so sure. But if you can get 30%. That's great. Right? That's definitely more speed. But this is what I was talking about earlier about the hype, the hype out there is, you hear people saying we're going to replace our entire software development team with AI.

00:09:59.860 --> 00:10:00.550

Bryan Dennstedt: Yeah.

00:10:00.550 --> 00:10:08.265

Kevin Carlson: A few. I think it's a couple of months ago I read something on Linkedin about a company that got rid of a handful of developers and then was hiring them back because didn't really work.

00:10:09.890 --> 00:10:19.560

Kevin Carlson: you still need people that understand the code and can look at the I generated code because it does hallucinate right, and it does do things wrong, and you have to tell it to correct itself.

00:10:20.740 --> 00:10:45.469

Kevin Carlson: But advances come every day. I was recently looking in a new Vs code Plugin, called Klein, CLIN. E. That integrates with a bunch of different models, including locally deployed models, and it does a pretty amazing job of just writing code, creating files, correcting errors in the files. So certainly we'll get a lot of efficiencies there. But full replacement, I think, is probably a little further off.

00:10:46.000 --> 00:10:58.140

Bryan Dennstedt: Yeah, I would have to agree with you. I mean, I think you you hit the nail on the head. In my opinion on 2 major things people humanity has to come to terms with is the creativity that

00:10:58.140 --> 00:11:14.209

Bryan Dennstedt: authors and musicians and everybody out there has got to be solved like we've got to still reward those people for their contributions that on creativity. I think AI can usher in a new era of creativity. But we need to reward the creators. And the second thing is.

00:11:14.210 --> 00:11:38.480

Bryan Dennstedt: I mean I've tried it myself, you know, regular coding versus the vibe coding with AI and the vibe coding gets me in these loops where it just doesn't know how to do it. And then I'm just like I should have just wrote it for myself. I would have saved an hour of my time, because I know how to do it. But you hit the nail on the head. I think 30% is the right target for AI to help me with the Login page. Help me with the forgot password page. Help me with the stripe integrator. Those are

00:11:38.550 --> 00:12:06.970

Bryan Dennstedt: perfect perfect use cases for AI, and it will save 30%. But that 70% that's left is the hard part of the programming where you probably still need that creative human in the middle. There, I mean, I think that's a perfect segue into the leveraging component of the podcast here, you know, how are you applying AI in the organizations that you're a part of like, how are you getting these people to leverage AI day to day and get to that? 30%.

00:12:07.330 --> 00:12:20.740

Kevin Carlson: Well, the the 1st thing I always do when I'm talking to a company about AI is, it's it's education, right? It's what's real and and what's hype? And that's where I start. Because the hype is so

00:12:20.800 --> 00:12:43.840

Kevin Carlson: attention getting that. That's what a lot of people, you know, sort of glom onto. So you know, once I sort of give them examples of what's real and what their expectations should be. With regard to AI. I walk them through a deployment model. I call it aim AI implementation model. And it's essentially starting with commercially available models because

00:12:44.060 --> 00:12:54.369

Kevin Carlson: you can do proofs of concept there for specific use cases at low cost, but the downside of that is that you also have low differentiation, because everybody else is pretty much using the same model

00:12:54.830 --> 00:13:24.419

Kevin Carlson: moving up from there. If you need more privacy, or you have some specific data that you want to retrain a model on, or put layers on top of a model like Llama and train it. You can use foundational models for that. That's a little bit higher cost, but it gives you much higher differentiation, and really, for the greatest impact is also the greatest investment is developing a truly proprietary model from scratch. I see very, very few people doing the latter anymore, unless they have something really unique to do like you mentioned the patents earlier.

00:13:24.815 --> 00:13:33.530

Kevin Carlson: That was, you know, extremely unique. We developed a model for a client that could isolate the most interesting moments in a podcast among other things.

00:13:33.610 --> 00:13:34.225

Kevin Carlson: and

00:13:35.310 --> 00:13:47.509

Kevin Carlson: you know, got got. Some, had some really interesting experiments around that. And it took a couple of months to get to, and then refined over a period of like like a year and a half so much bigger investment.

00:13:47.620 --> 00:13:51.060

Kevin Carlson: So after we understand the different sort of implementation

00:13:51.440 --> 00:14:15.330

Kevin Carlson: types. For for AI, then we focus in on proofs of concept what's going to bring value to the company right now, and let's scope it to where it can fit. Maybe with a with a commercially available model. If there's privacy concerns, we'll just go straight to a foundational model and deploy that you know on Prem, or, you know, in the cloud, in a private, you know, Vpc. And they can use it, you know.

00:14:15.520 --> 00:14:29.830

Kevin Carlson: however, they wish without worrying about data getting out. So that's generally how I approach it. Like, let's start small. Let's focus. Let's start on quick returns with low cost, commercially available models. And then let's take it step by step. From there.

00:14:31.020 --> 00:14:58.300

Bryan Dennstedt: I'm curious. Yeah, I love the aim model. And like as we go in as ctos and cisos and stuff and help these companies lay out their AI roadmap. The the components that I see is a how do we get everybody to that same playing field? Do you know what a basic, prompt engineering concept is. And let's guide you through that. But it's also like, how do we just get you this general AI level as a starting point. And then it's almost like looking at it.

00:14:58.300 --> 00:15:08.199

Bryan Dennstedt: In my opinion, like on a department by department level of how can a specific tool help marketing or developers or security, etc. What do you see? Like?

00:15:08.200 --> 00:15:22.360

Bryan Dennstedt: I see a couple of different companies like exploring the light Llms or the I Forget the fox one or something that's out there, or versus just signing up with Gemini or Openai for the $20 per month per employee.

00:15:22.360 --> 00:15:39.089

Bryan Dennstedt: So a hundred person company, they all need the tool. And we need to lock that tool down and control our data, governance and things like that like, what's the what's the? How do you guide them through that 1st entry point piece and then get them into those specialized components you're talking about.

00:15:39.090 --> 00:15:43.851

Kevin Carlson: Yeah. So it's really understanding largely the companies.

00:15:45.930 --> 00:16:02.720

Kevin Carlson: data, I guess, for lack of a better better word. And maybe that's the right word, but it's it's really understanding that data and the privacy concerns around it. The regulatory concerns around it. Look, I mean, you can use Openai's Chat Gpt, and you can check the box and keep this chat private.

00:16:03.530 --> 00:16:13.090

Kevin Carlson: But what if there's an error on their end? Right? And that's what you know. That's what I'm concerned about, like when I use ChatGPT, and I wanted to have it do something, and I've got some

00:16:13.190 --> 00:16:33.480

Kevin Carlson: something I want to feed into a prompt, then I make sure I completely anonymize it first, st just in case. And I do want to come back and address the point you made about prompt engineering. To me this is such a funny concept, because prompt engineering is to me nothing more than the ability to describe to the model

00:16:33.590 --> 00:16:34.730

Kevin Carlson: what you want.

00:16:34.920 --> 00:16:35.420

Kevin Carlson: That's right.

00:16:35.420 --> 00:16:35.890

Kevin Carlson: Great. Yeah.

00:16:35.980 --> 00:16:51.759

Kevin Carlson: talking to a person. And it just it kind of blows my mind that you know, we give all this thought around how we talk to an AI model, and we don't give the same consideration to somebody else. We have a conversation with them, you know. Imagine how much better it would be if if we you know, if we did that so.

00:16:52.023 --> 00:17:08.899

Kevin Carlson: and some learning curve right? It's you know. I don't try to give my clients canned prompts, because then that's what they'll use. I wanna just like, you know, here's a basic starting point. And then here's some directions. You can take it. And I encourage them to explore and try different things, and if they don't get what they want the 1st thing

00:17:08.900 --> 00:17:30.979

Kevin Carlson: the 1st time through change what they're asking for and try it again. There's no penalty for it, and you know it's just a it's a matter of trial and error, like if you just hired a new employee, and you threw them into the, you know. They threw them into the fire and said, Okay, you have to go do this. You're gonna have to tell them exactly how to do it if they're new. So you kind of have to look at at these Llms the same way.

00:17:31.930 --> 00:17:58.779

Bryan Dennstedt: I agree completely. I mean, they ask a simple question and expect this big, long answer. I do think you nailed it right on the head like you have to treat the AI prompt just like, you know the smartest person on earth, and be very, very explicit, and just walk them through that. So I typically say, the Ryzen framework role. And you know all that kind of stuff. But like, either way, it's sad that we have to teach people how to talk

00:17:58.780 --> 00:18:04.410

Bryan Dennstedt: on that front. But I think that's what we need to do as leaders is to help mentor and guide these people

00:18:04.410 --> 00:18:29.120

Bryan Dennstedt: to that next stage of it really, and guide them with how to best communicate. Like, just like, I remember one of my 1st jobs Ibm sent me on a 3 week training program to be indoctrinated on how to do expense reports how to run a project. All that kind of stuff. And I think that's lost. I mean, I know I prepped for probably at least 20 min for this, podcast to make sure, I had some good smart questions to ask you

00:18:29.120 --> 00:18:47.969

Bryan Dennstedt: and stuff like that. And we need to find that that time blocking to have that like heads down time instead of going from meeting to meeting to meeting, and and then being exhausted at the end of the day. So I don't know we're we're dipping into the the last segment here, which is really our our lead segment.

00:18:47.970 --> 00:19:12.500

Bryan Dennstedt: And that's what it's all about. I think that's the essence of why you would potentially hire me or you to help a company grow in the AI space on that front. That's why I wanted to put this podcast together because this is such an important topic of how to embrace our new digital employees that are coming in and getting the digital employees to work well with the real ones. So how are you finding leading

00:19:12.580 --> 00:19:21.389

Bryan Dennstedt: technology leaders and companies and educating them and leveraging it, but really leading these companies through on their AI journeys.

00:19:22.060 --> 00:19:30.595

Kevin Carlson: Yeah, I think, 1st of all, you have to understand the business right and understand the challenges that they have and what I like to get people passed. Is.

00:19:31.680 --> 00:19:49.729

Kevin Carlson: are you looking at AI because AI, or are you looking at AI to actually do something real? Yes, it's great. You can deploy or use Chat Gpt, or Gemini, or Claude, or any of these, to do one-off tasks. But if you really want to integrate it with your business.

00:19:49.730 --> 00:20:05.159

Kevin Carlson: then you have to have, I think, a pretty solid understanding of where you can get efficiencies, or, you know, either through doing things faster or doing things in a more automated fashion. So like document summarization, or even some, you know.

00:20:05.810 --> 00:20:21.859

Kevin Carlson: like, if you want to generate a Powerpoint using chat, Gpt right and then put it into a you know, a company template or something. That's that's something pretty quick. You could certainly do it yourself, but it could be done in maybe 2 min instead of, you know, 30 min or an hour versus something that's completely automated using.

00:20:21.970 --> 00:20:32.960

Kevin Carlson: you know, agents, before you know, before you jump in start making decisions about, we want to use this. We want to use that because of news articles, you've read or podcasts, you've listened to, it's

00:20:34.350 --> 00:20:45.150

Kevin Carlson: important to understand the business and where you can really use it. And I think that's where folks like you and me come in is, you know, we've done this, we've been involved in AI for a while. And

00:20:45.290 --> 00:21:00.470

Kevin Carlson: we we typically know places in a business that where AI could be leveraged quickly and successfully. And then, as the executive team or the leadership learns more about it, and they start seeing those opportunities in different places and the discussion opens up. But

00:21:00.680 --> 00:21:08.429

Kevin Carlson: ultimately you have to pick a place and start somewhere where you're going to get, you know, or see a quick win. See? Some evidence. This is really going to work for the company.

00:21:09.777 --> 00:21:13.380

Bryan Dennstedt: Yeah, I mean, you're spot on. I think that's what separates

00:21:13.814 --> 00:21:34.600

Bryan Dennstedt: some people from the others is is having that business acumen. Having sat at the board table, having sat at the executive table enough to understand, you're like, look! We have got to generate sales. We've got to generate revenue to make the payroll and to do those things. And this is how technology and marketing and others are gonna work hand in hand.

00:21:34.600 --> 00:21:52.950

Bryan Dennstedt: I think it's interesting because everybody wants to jump into AI and go straight to the agentic automation to replace employees or something like that. But you really do have to do exactly what you said, crawl, walk, and then run into AI and don't expect to get there tomorrow, just because you saw

00:21:52.950 --> 00:22:00.769

Bryan Dennstedt: us talking about it on the podcast. I think it's also really important, like one of the threads I want to pull out of what you said is

00:22:00.770 --> 00:22:27.980

Bryan Dennstedt: is automation versus, like true agentic, like, I think, most of the time like, I hope most businesses have got these standard operating procedures in place. This is how we handle customer success. This is how we handle leads. This is how we handle deployments of code. Now, you have that standard operating procedure. Now, how do I take AI and automate that to

00:22:28.130 --> 00:22:44.800

Bryan Dennstedt: remove the human from that rote, repetitive thing. And I think that's a good place to start with the crawl. Let me help you automate this or help me. Let you automate that, and then you can get into these agentic things where you have this ability for the AI to think

00:22:45.060 --> 00:23:00.499

Bryan Dennstedt: and reason, and try and predict which path it should go to, or which other agent or automation it should follow. And I think people are confused by that. Do you want to help unpack? You know automation versus gentic, and maybe this crawl, walk, run, approach a little bit more for us.

00:23:01.129 --> 00:23:03.049

Kevin Carlson: I can try right? I think.

00:23:03.660 --> 00:23:14.170

Kevin Carlson: so I guess an example I would use it in the context of my day to day is when I need to write a proposal for a client. Right? I can.

00:23:14.610 --> 00:23:27.220

Kevin Carlson: Autumn, I guess, automates the wrong word, but I can record the call like, I have a call with a client about a potential project, and we talk through it. And we, you know, ask questions about the scope and the timeline and everything, and then I can use a tool like.

00:23:27.590 --> 00:23:54.720

Kevin Carlson: you know honor or fathom, or something similar to record that. And you know then I can ask it a question. Right? It says, you know, what are the requirements for this for this project? So that would be, I guess. In a sense, some level of of automation of that, or it's really more like efficiency gains, I would say, but true agentic would be like I would have the call, and 5 min later a proposal would pop out complete with S.

00:23:54.720 --> 00:23:55.220

Bryan Dennstedt: Good.

100

00:23:55.220 --> 00:24:22.849

Kevin Carlson: Right. So the danger in that right now is that, do you believe the models are correct? Right? So back to the coding example, where you've got, you know, 30% of stuff that can be done pretty reliably. Do you want to take a chance on AI creating a proposal that's going to be for a non, you know an unprofitable, you know project for for your company, right? Probably not. Right? So there's still that human element in there.

101

00:24:23.570 --> 00:24:26.290

Kevin Carlson: The best use of agentic that I've personally seen

102

00:24:26.520 --> 00:24:55.069

Kevin Carlson: is actually in software development where I can, you know, write a prompt, you know, right inside of my code editor using Klein and have it, you know, generate files and code and test it out. And you do all kind of stuff right? And it's it's not. I would say, end to end automated because you still have to approve things which is a good thing right now. But it's closer to agentic, you know software development than what I've seen before, which is typically generate the code, copy it and paste it into the editor.

103

00:24:55.670 --> 00:25:11.960

Bryan Dennstedt: That's right. That's right. Yeah, no. I mean, yeah. You. You described it very well. In my opinion, I really love the emphasis that you've put on, you know, the leadership, adaptability that we've, you know, subtly talked about throughout this conversation.

104

00:25:11.960 --> 00:25:27.239

Bryan Dennstedt: And you know, I think, being on the forefront of AI, like you and I are to a degree, it showcases how we have to be continuously learning and adapting in this really fast changing world.

105

00:25:27.240 --> 00:25:47.530

Bryan Dennstedt: I love also, like, you know, we have to just be guiding these teams through technological shifts. And it does take a little bit more than just some technical knowledge. So thank you, Kevin, we really, I really appreciate you jumping on the show and and hanging out with us, and departing some amazing insights into this for us, so.

106

00:25:48.100 --> 00:25:51.189

Kevin Carlson: Sure man, thanks for having me and enjoy the conversation.

107

00:25:51.770 --> 00:25:54.710

Bryan Dennstedt: I want everybody to know that.

108

00:25:55.480 --> 00:26:19.270

Bryan Dennstedt: our list. Well, I guess. Let's just say everybody that's listening out there. If you love this episode found it enlightening in some fashion. Please subscribe to AI with Bry, leave a review and share it with some colleagues that are interested in the future of AI. We're just getting started on this. Podcast and I'd love to have you along for the journey, make some suggestions on who else we should have on the show or let them know, and I'll be glad to get in touch.

109

00:26:19.330 --> 00:26:28.499

Bryan Dennstedt: You can connect with Kevin through Tech Cxo's website. And I think you have your thought leadership blog is that it's fractionalcto, dot blog. Is that right?

110

00:26:28.500 --> 00:26:33.569

Kevin Carlson: No, I've changed it. It's now it's it's kevincarlson.substack.com.

111

00:26:34.110 --> 00:26:52.710

Bryan Dennstedt: Perfect. So check out that substack. I've read several of your articles. I really like the one you did a couple of days ago on on how to speak up and just ask some great questions. It was a great article, and I think it can be helpful to some of the younger people that are just getting into this space. So thank you for your contributions on that.

112

00:26:52.880 --> 00:26:54.700

Bryan Dennstedt: Appreciate you being here, Kevin.

113

00:26:54.700 --> 00:26:56.140

Kevin Carlson: Thanks, Bryan. Appreciate it, man.

114

00:26:56.660 --> 00:27:02.570

Bryan Dennstedt: Until next time. Everybody. I'm Bryan Dennstedt, reminding you to learn leverage and lead with AI.

Bryan Dennstedt

Bryan Dennstedt is a Fractional CTO at TechCXO, helping startups and growing businesses optimize technology strategies for sustainable growth. Specializes in aligning tech operations with business goals to drive efficiency and innovation.