The Creative Agency Success Show: Episode 146
You don’t need more hands—you need better tools. For service-based small business owners, AI can lift the burden of repetitive tasks, helping your team do more of what moves the needle.
In this episode of The Creative Agency Success Show, we sit down with Ed Krystosik, Co-Founder of RAC Projects AI, to unpack the real-world strategies for integrating AI into small businesses. From streamlining workflows to navigating compliance risks, Ed shares how AI can support your people, sharpen your processes, and fuel sustainable growth.
Key Takeaways:
- Support Over Substitution: AI should enhance, not replace, your team.
- Start Small, Think ROI: Target low-friction tasks first.
- Fix the Process First: Map your workflows, then automate what makes sense.
- Win Team Buy-In: Make AI adoption a shared mission, not a top-down command.
- Play It Safe with Data: Avoid hallucinated outputs with smart model settings.
- Learn how to use AI to power smarter decisions.
Watch the episode of ▶️ Amplifying Human Potential In Your Business with AI Support.
Jamie Nau: [00:00:01] Okay, recording has started. So we will start in three, two, one. Hello everybody, it’s Jamie Na. It’s been a bit a while since we recorded, so we’re excited to jump back into this and have a really good guest today that I think the topic is gonna be something that everybody’s talking about. I know it’s something I’ve heard at the last couple of conferences I’ve been to. As always, I’m joined by Jody Grundin. So welcome to the show, Jody. [00:00:23][22.1]
Jody Grunden: [00:00:24] Yeah, thanks, Jamie. [00:00:24][0.5]
Jamie Nau: [00:00:25] And our guest today is Ed Kristosic. He’s the owner at ROC Projects, and he is really about AI, AI all over the board. He’s gonna talk a lot of topics today. So, Ed, why don’t you tell us a little bit about yourself. [00:00:36][11.4]
Ed Krystosik: [00:00:38] Yeah, thanks for having me on. I appreciate the opportunity. So I’m the founder of RAC projects. We’re an AI automation agency. And our primary two focuses are helping business owners to cut through a lot of the clutter in the AI space because there’s a lot of misinformation and a lot of hype out there and really focus in on business process automation that delivers an ROI. [00:01:00][22.6]
Jamie Nau: [00:01:04] Awesome, so let’s jump right into that. So when it comes to ROI, there’s a lot of different avenues where you can get an ROI. You can get it through marketing, you can it through processes. So can you talk a little bit about the first step you would do when talking to someone about using AI for that return on investment? [00:01:19][15.2]
Ed Krystosik: [00:01:20] Yeah, so when it comes to one of the biggest challenges that we’re seeing is hiring. And it’s very difficult to hire great people, right? So to give the people that you already have and you really have put a lot of time and effort into the tools to sort of amplify their workflow we found has been a great opportunity for the use of AI. So if you’re Sorry, I lost my train of thought there. It’s okay. Just start from the beginning, maybe. Yeah, yeah. So one of the opportunities that we’re seeing is the difficulty in hiring has been a real problem. And so if you’re using AI to amplify the people that you already have, or quite possibly replace the need for some new hires, that’s giving a tremendous amount of return on investment. And it’s really not about trying to replace an individual person entirely, right? That’s very difficult to do. What it’s best at is taking what you’re already doing and streamlining. Making it more efficient, cutting out a lot of manual steps. So that’s where we’re finding the most return on investment is finding small niches where you can use AI in a way that provides, boy, oh gosh, I am. No, you’re fine, you are fine. You’re good, you doing good. All right, so- That’s the areas I would focus on, right? Is not replacing people, but amplifying their current work. [00:02:53][92.5]
Jody Grunden: [00:02:55] So kind of looking at the agency world there, there’s a lot of things going on. Obviously, there’s lot of fear out there, I guess. And you mentioned that, is AI gonna replace the agencies themselves? And my view is no, but I like what your view is. [00:03:13][17.2]
Ed Krystosik: [00:03:15] Yes, there’s always gonna be a place for certainly people in any space, right? So I know personally, I don’t always want to deal with an AI, I wanna have a person, right. Maybe you don’t need a hundred people in your agency, maybe it’s 80, right, so it’s not gonna replace people. You still need ideas, you still need strategy, you still needs implementation, you need that relationship building that is really the core of any kind of business. So it’s certainly. Not something that I see coming where you just go and hire an AI agency and the AI does all the work. Now that’s possible in a few years from now, but I really don’t see it being the case where a person wants to go hire a computer. They want to hire someone and they can have a relationship. [00:04:02][47.5]
Speaker 4: [00:04:04] I’m not quite sure. [00:04:04][0.7]
Jamie Nau: [00:04:04] Yeah, I think one of the things that I’ve… I’ve seen a lot at a lot of a couple of conferences I’ve been to is there’s some websites out there, there’s an AI out there that can build websites for you. So that I think that freaks out some of the agency worlds. But at the end of the day, like if I, as a non designer, and have someone who hasn’t been doing this for the last 20 years, tries to create a website, it’s not gonna look nearly as good as someone who really understands like what color schemes work well, and what companies look for and what things what things they can use to, to really draw that attention. So I think that’s just one example is yeah, like maybe your team can that website or use that app on the AI app to create the website but they’re gonna know a lot better so like that may be working on behind the scenes working with their clients working with customers but they are still really understand the big picture of what things need to look like because yeah if I just try to do it because I start a business it’s gonna look like someone who started a business create a website still [00:04:53][48.8]
Ed Krystosik: [00:04:54] Yeah, that’s really good insight and the tools you’re talking about are like lovable, Replet, Bolt and we use those tools, right? But the most effective way we’ve found to use them is very fast iteration on new ideas, right. Get it out there, fail quickly or succeed quickly and then once you’ve kind of proven out the concept then we would take that to a developer or a website designer and say, okay, this idea works, right, so rather than spending 10, 20, 50, $100,000 on an idea, before you know it works, you’re able to really quickly iterate ideas, find something that has a true market fit, and then you can take that and hand it off to a full development agency, right? So for example, we use a lot of vibe code. I’m not sure if your audience is familiar with vibe coding, it’s basically using AI to build software. And it’s extremely impressive. Right up to the point where it completely fails. [00:05:52][58.0]
Speaker 4: [00:05:53] And it could be. [00:05:54][0.7]
Ed Krystosik: [00:05:54] It completely fails when you get into that complicated sort of multi-process, complicated workflows. So it’s very good for single page, maybe two page, but beyond that, it gets completely lost. So again, we’re able to demo to clients, show a proof of concept and say, is this really what you wanted? And they’re able to say, no, that’s not at all what we wanted. Okay, great. We come back with a second. Yep, that’s exactly what we wanted. So that kind of gives us a leg up when we go to full development and we know we’re building what the client wants. And so I think that’s a huge opportunity for any agency is not to replace you, but to show, hey, quickly in one hour, you can spin up a demo, sell that demo, and then build out the full product. [00:06:38][44.3]
Jamie Nau: [00:06:40] Yeah, it’s funny. It’s almost similar to what, to what we do in a way, right? So like a lot of people think they can do their own accounting, right. And they start there with their smaller business and then they, they, okay, we’re going to be doing our own accounting. We’re doing our books and they do it until they hit a certain point. And then like, okay. Now I need to hire a CFO and accountants. Then we go and look at their books. We’re like, oh, wow. We need to go back and, and fix two years of what you’ve done. And that’s, that’s kind of what I’m guessing will happen in the agency pace. People are like, yeah, I’ve been running my business for two years. I created my own website. The website keeps breaking down. And it’s because it’s hard to maintain a website. And it’s hard to build a website that can be maintained. And so I think that’s another area that they’re gonna really have to go into is be like. Yeah, what you did is good for when you’re this size, but now that you’re bigger, guess what? There’s a lot of pieces here that aren’t really working and the landing pages and all this stuff that we know as people who work with marketers are really important that a small business doesn’t know us. I think they may have to jump a lot more into that, that realm where they’re gonna be fixers versus like creators in a lot, a lot a ways. [00:07:31][51.7]
Ed Krystosik: [00:07:32] Yeah, so AI is really good at writing what we call spaghetti code, where it just sort of wanders and winds around. And if you get it right, if you go to try to change anything, you’re really going to run into a huge problem because it’s going to break something three files over that you have no idea what it’s doing because you didn’t write it. So yeah, that’s exactly right. And like we said, with the accounting, you need strategy, right? You need to say, we need to look forward and say, let’s spend some money here strategically. Let’s go ahead and identify different account types or different investment types that are gonna be tailored to you, not here’s general information about accounting. That’s not really that useful when you reach a certain point, sure. [00:08:12][40.1]
Jody Grunden: [00:08:16] Yeah. And with, um, you know, kind of jumping back a little bit, um with, you know, a lot of what you write about, I’ve talked about, it seems like it’s around project management, uh, legal, uh that sort of thing. If, if you’re looking at a, you know, if, if your talking to a project manager, because we use project managers all the time within our company, you know, how, how would AI, how’s AI going to help them out? And what, what would you say? You know, Hey, here’s the number one thing you should look for in project management. The agency we’re out there, we all have project managers, accountants have project manager, law firms, they should. So how is that gonna help their job? Again, the fear is that’s gonna replace their job. I’m completely saying, no, it’s not. But how’s it gonna help and aid their job, make it a lot easier to use? [00:09:02][45.3]
Ed Krystosik: [00:09:03] Yeah, so I actually came from the project management space before starting this firm. And when I moved into, started talking about AI, what I knew best was project management. So when I was talking about that, it was trying to find those use cases. And I found that it actually helps project managers become more organized because there’s kind of two ways to look at it. One is the AI is gonna go find all the answers and bring them back to you. And that’s not the case, because we know all of your information is spread out, different files, different folders, different teams, different chats, different meetings. So one, if you really think about how you work in a day and how your workflow works, it gives you the opportunity to organize your own project management style into a way that is useful for AI, which we also find is useful for a human, because it’s now more organized. And two, once you’ve connected all those systems together and all those data sources together, interact with that information in real time and sort of be able to see patterns across different sets of data. And I think that’s when the biggest thing is saying, okay, well, we have this client request over here. We have this project going on over here, we have this deadline here. What is the most important next step? And AI is really good at identifying, synthesizing, and bringing back a couple of ideas where you can say, Okay, that’s what we’re gonna focus on and then moving forward. [00:10:29][86.2]
Jody Grunden: [00:10:31] Are we talking primarily chat or copilot or what are we talking about? It would be the primary tool that you’d see a project manager use or should be using. [00:10:39][8.4]
Ed Krystosik: [00:10:40] Yeah, so depends on your tech stack that you’re using within your organization. So if you’re a Microsoft shop, then certainly copilot would be great because then you can connect email files, calendars, teams. If you’re. The Google suite, you can obviously use Gemini, which is great to integrate all of your different sources of data. But if you are in between different suites, or if you want to build a custom application, you could use things like an eight N or make. To do a low code automation to basically build your own workflow and then go out and gather data from different repositories. So that’s a little more technical, but it’s certainly if you’re technically minded and like to tinker, there’s great opportunities to build out some custom workflows, which is something that we focus on as well. [00:11:27][46.9]
Jody Grunden: [00:11:27] Gotcha. So how would you come in and help like a, a team that had a bunch of project managers? How would you, how would, what would engagement kind of look like? [00:11:34][6.6]
Ed Krystosik: [00:11:35] Yeah, to start off with, we would go through identifying what your current workflow is. Because I think one of the biggest things that I see in the AI space is you’re always bringing a solution to a problem. And people say, I’ve got this solution and we’re gonna sort of crowbar it into your organization and it’s gonna fix all your problems. And that is not the case, right? That’s a recipe for disaster. So really identifying what you do now that works, What you do now that doesn’t work. And being able to have a solid manual process. And whatever you’re doing as a manual process, then you can start to automate, but not the whole thing, right? You wanna start with portions of it, right, because if you start turning it over the entire workflow to AI, it’s gonna hallucinate, it’s going to be confusing, and you’re not gonna get the outcome you’re looking for. You’re gonna be like, why did it do that? And the simple answer is, is because AI is a black box. We don’t know why it did that. So you have sort of deterministic processes, which are your standard automations of many years ago, A, B, C, D, E, which in the finance or accounting space is probably more beneficial because you don’t want the AI deciding a new tax rule on its own. You need it to like stick to the rules, right? But in the creative space, right, you’re looking for, you give it colors, you give sort of a creative direction and let it sort of explore the space. And that’s a non-deterministic workflow. And so I think the biggest thing is making sure that you know exactly how your business runs manually before you start trying to automate it. [00:13:12][97.0]
Jamie Nau: [00:13:13] So I think the interesting part of this is I’m gonna shift the conversation a little bit away from AI, because it sounds like what you would do is you would go in, and let’s just say we have a team of four people, and they’re doing all these things together, right? And then you go through and you look at their processes and you say, okay, 25% of what you can do can be moved to AI. 25% what you do can be move to AI, so you’re basically taking part of everybody’s tasks, putting it here, but then that means it’s all four people’s job duties can be a little different. So how do you work with those people and saying that, okay you were doing X, Now, AI is doing part of X and so now your job description has changed a little bit. So how do you get their buy-in on what they’re doing is going to be completely different because AI is taking certain elements of a bunch of different people’s roles. [00:13:53][39.4]
Ed Krystosik: [00:13:55] That’s a really tough psychological people question, right? [00:13:58][2.8]
Speaker 4: [00:13:58] So. [00:13:58][0.0]
Ed Krystosik: [00:14:00] It’s going to scare a lot of people, especially those who are maybe, you know, more… Settled in their career where this is what I do and here’s how I do it and it has always been done right and you’re always no matter what kind of organizational change you make that’s going to fluster some people right so what we found works best is to whenever we’re going into a new organization obviously we’ve got buy-in from the owner they’re ready to use AI we have them identify sort of a chief AI officer role within their organization someone who is super interested in AI, someone who’s super interested in innovation and technology. And we sort of partner with them internally to sort of be the liaison with the sort of the staff. And that sort of takes the pressure off of where the consultancy coming in to take your job. And once you’ve got that singular person who can then sort of help you say, okay, well, here’s how this department operates. Here’s a meeting I can set up with these people. And that allows us to have a lot more. Sort of friendly relationship with your job is changing. And here’s how, because we have that internal sort of buy in from that point of contact. Now I’d say the second part of that is when you’re identifying processes to replace, it’s everyone says they’re stressed, everyone says they’re overworked, everyone said they’re underpaid. Well, leverage that and say, well, here’s an opportunity for you to be opened up to do the thing that you think is very valuable that you want more time to do. And sort of give them their time back in a way that says, okay, what do you think the organization can do better if we gave you 25% of your time back and sort of put it on them to be a partner in it? [00:15:44][104.4]
Jamie Nau: [00:15:45] And I think we did a while ago, Joe, I think it was the working genius one, where we kind of had our job broken into four quadrants and we said, okay, what are the things you’re doing that you love and that you’re good at? And that’s, that’s where you want to spend the majority of your time. And I can help people move from them from those other quadrants or the things they’re not good at that they don’t love, you know, kind of move from that quadrant up to that quadrant. So I think that’s how I’ve tried to use AI is like, what are things that I hate? What are the thing that just that I have to do or a part of my job? And I’m like, okay. This is almost an hour a day doing this, but it’s just, necessary evil as a part of my job, how can I automate that and put myself more into that space of okay now I’m spending 90% of my time doing what I love and what I’m good at and that’s really where we want our team to spend their time as well. [00:16:24][38.8]
Ed Krystosik: [00:16:25] Yeah, that’s huge. Now, the project management space, for example, I don’t know anybody who loves writing status reports, writing team updates, taking notes on meetings. So right there, there’s sort of three things that AI is really good at, listening intently, synthesizing into a summary, and then being able to distribute that. So right off the bat, that’s one of the biggest wins you could have is record all of your meetings, have an AI note taker and summarizer, put together the action items and the notes. And then distribute those. I mean, that’s a huge win right there. [00:16:57][32.3]
Jody Grunden: [00:16:59] 100% agree. So now we’ve got the team adopting the issue, we’ve got AI coming in really kind of helping. And then you got the big bad legal team that steps in and says, Hey, hold on a minute here. We can’t be doing this. How do you handle that pushback from compliance teams like that? Because I’m sure you get that and rightfully so you should get it. What’s the fear that we’ve gotta overcome or the issues that we’ve gotten overcome to make them satisfied? [00:17:27][27.4]
Ed Krystosik: [00:17:29] Yeah, so the biggest fear is gonna be that your data is gonna taken by the AI companies and then put into the models and then that information could leak out. That’s one big area of concern. Two is that the proprietary information that you have NDAs that you’ve signed with your clients and that information could again leak out into the wider ecosystem and then now you have a legal concern. And then the third and working with law firms especially is that the AI hallucinates. And it makes a law up or it makes an accounting rule up and then it applies that and you run out of the office, you’re too busy, you grab it off the printer and you put it down on the judge’s desk and it’s wrong. [00:18:09][40.7]
Jody Grunden: [00:18:10] Which happened. Which happened, right? [00:18:12][1.8]
Ed Krystosik: [00:18:14] So those are kind of like the three areas of most concern. So I would address the first one of the data leaking out into the wider ecosystem. There’s a lot of, so if you just get a chat GPT account or free account, yes, your data is going to be put into their training model and there’s a potential possibility that it gets leaked out. Now, there are steps you can take where you can change your data privacy settings where you tell OpenAI or Anthropic, which is Claude. Don’t use my data in training. Don’t retain my chats. Don’t send my information back to the servers. There’s some things you can set up sort of as a cursory level. But if you’re going into a larger organization enterprise, then there are teams you can work with at these major companies that can get you a sanitized version of JetGPT of different language models. And that’s your own version that’s walled off from the rest of the world. And then you can be certain that your information is not getting put into the larger model. The second thing about. [00:19:20][65.3]
Jody Grunden: [00:19:20] Can I back up real quick? When it gets put into the larger model, what does that mean? Does that mean that at Anders, I put in all my financial statements for all my clients and it gets in, this isn’t true, but let’s say I did that. I put all my financial statements on all my client’s and then you pop in and you’d say, you know, hey, what is client’s XYZ’s profit for this month? It could pop that up and you’re saying it would know that information. [00:19:44][23.9]
Ed Krystosik: [00:19:45] It could. Yeah. So if one of the use cases is doing like market research and if you’re a large enough company, right, and you have a lot of public information out there, you know, if you are a publicly traded company, your information will have been online and all of your financial statements. And if you ask the AI, you what is the profit margin for this company? It will return an answer and likely very, very close to being true. Now it’s unlikely that any one individual company. Will be able to have their information pulled at will, but it’s a very real concern, sure. So certainly any kind of private financial data, any type of trade secrets, proprietary information, we’ve worked with engineering firms, so drawings, blueprints, saying, build out a Q and A from this blueprint, which it was very good at. But the concern is, is that those drawings are proprietary and they could be. Could be, you know, leaked onto the larger internet. [00:20:46][61.6]
Jody Grunden: [00:20:51] Okay, and then your second concern then. [00:20:53][2.5]
Ed Krystosik: [00:20:54] Yeah, the second concern is that the AI hallucinates or it just frankly makes things up. And so the way that large language models work is that they’re doing a predictive next word or next chunk of words, right? So if it gets itself into a corner and it doesn’t have an answer, the AI wants to please the user. And so it will make up an answer to give you something. It won’t say, I don’t know. It will typically say, I know, and here’s the answer, whether it’s true or not, right? So that’s, that’s a very real concern. When a couple of ways we’ve dealt with this is having adversarial models. So you have a number of AI models in a row and one does the research. Another one does its own research, checks the other models work, and then feeds that into a third, which checks both of their work, right. You have this adversarial network. Of fact checkers that are AI in themselves. So it’s unlikely that all three of them will make up the same fact. So we’re finding- [00:22:00][65.6]
Jamie Nau: [00:22:01] that’s an effective way. It’s so tricky, right? Cause it, like you said, AI is convinced it’s right, right. Like, so we were recently doing our, our team retreat and I was looking for a tool, like an interactive, like tool that people could like, you know, the slides would be up there. People would be answering questions. It would show it live and I wanted it to be free. And I asked AI like, can you give me three tools I can use for this? And they came back. It’s like, yeah, these are the three tools you can use. This is how much they cost, this is how they’ll work. Here’s links. Like sweet. I clicked every one of those links and none of them Cause I all have like one of the things that didn’t work. And so it’s like so confident that it’s It really is tricky, so I love the idea of taking that answer, running it through a different AI. I’m like, is this answer going to work? And they’re like, no, it wouldn’t work because instead of me spending two hours finding out that all three of those websites didn’t work, another AI might be able to tell you that no, those are not going to because they don’t have your qualifications that you said they did. [00:22:49][48.4]
Jody Grunden: [00:22:50] So is it a separate AI or is it like three versions of using the same chat and you’re just putting it into another chat? Is it completely a separate software that you’re using or? [00:23:01][10.9]
Ed Krystosik: [00:23:01] Yeah, so typically it’d be three different models, right? So for example, in a workflow would be have something like perplexity, which is sort of a AI search engine, go out and find information on this client, find information, on this case, find information whatever the question is, and then bring back that information for processing. And then the second model would be something like open AI, chat GPT, right. So go ahead and say, do your own research. And come back and bring your research. And then a third model, like say Claude, for example, or Gemini from Google, you say, okay, look at these two summaries that were written by the previous two. Do they align? Are there facts that seem to be different? And then also, you know, just go to it, and you can have it do its own research too. So there’s a lot of different ways you can structure it. It’s gonna be your own individual use case, but. Yeah, we’ve used three different models to fact check for mission critical data. [00:24:04][63.2]
Jody Grunden: [00:24:06] Oh, I love that. Love that idea completely. After [00:24:08][2.5]
Jamie Nau: [00:24:09] What about your third risk there? So let’s go down that one. [00:24:12][2.8]
Ed Krystosik: [00:24:13] Uh, the third risk is, uh, we talked about proprietary information. Well, we’ve talked about data and we talked about hallucinations. So I think those were the, the kind of the three areas of most concern. But the hallucinations, I think are the scariest when it comes to work product, because you are putting your name on it when you put it, you know, you know on the, on the desk or put it on the internet or out the door. And that could be a very scary thing, right? Your reputation is on the line. And certainly if One, if you’re wrong and there’s a problem, that’s bad. But two, if your wrong and they say, why were you wrong? And you say, well, it’s because I was using AI. And then the question becomes, well if you using AI, couldn’t I just use AI and fire you? Yeah. So you’ve got to be very careful with your outputs and use it as a tool, right? You use an accounting software. You don’t pretend to say, I do all this, but pencil and pen, right, you use software, right. But you’re putting your stamp of approval on the outcome of that software. So the same goes for AI. You can’t just single shot and publish it and say, this is right, you got to fact check, you’ve got to edit, you’ve you’ve to make sure that you’re okay with publishing or delivering that work product. [00:25:26][72.5]
Jamie Nau: [00:25:27] Yeah, and it’s, it’s crazy how fast that it’s taken over, right? I heard recently that more than like 50% of all written information on the internet is AI generated. It’s like, how long has AI been around? And the fact that like more than half of information on the internet is already AI generated is crazy to me. So we’ve reached the point. [00:25:44][16.4]
Ed Krystosik: [00:25:44] Real quick, something interesting there. There is a fundamental problem with AI and having the internet full of AI-generated content because AI needs data to train itself to become better. So as the human information is all vacuumed up and now has basically been ingested, it needs new sources. And so it goes back out to the internet to find more data and it finds its own data. And they’ve done testing where after I don’t know the exact number of iterations, but after a few generations of it doing that, the model itself collapses. It just starts to hallucinate itself into a fever dream of just gobbledygook. Because it’s now feeding itself its own nonsense. And that’s a very real question of how much better can AI get over the next period of time, because you need synthetic data, because there’s no more real human data out they’re doing, Jess. [00:26:44][59.4]
Jody Grunden: [00:26:45] That kind of brings me to a question, Jamie, before we go into the final, is that, you know, like we, we’ve traditionally in, you know, we we’ve advertised through, um, the internet, through media, uh, through SEO, uh we’ve done podcast webinars, we’re done a lot of different things to generate opportunities for, you knows, sales calls. Um, you know, how, what’s the most effective way now? Is it, is it writing? Is, is doing podcast? Is it being on somebody’s podcast? Webinar, you know, because it sounds like AI is looking for new information and I think new information comes from what? Voices and videos and stuff like that. Am I going the right path on that or am I completely off? [00:27:27][42.2]
Ed Krystosik: [00:27:28] No, you’re absolutely correct. So I would say the content is the number one area I think for creating really good high quality leads, right? Because the simple fact of the matter is with AI basically being the smartest individual ever to exist, the value of information or the cost of information has basically dropped to zero, right. If I wanna know some fact, I can get that almost instantaneously. What becomes valuable is is the human connection. What becomes valuable is the experiences that AI doesn’t have, right? We talked about this a little bit earlier. We all have life experiences to say, okay, we tried that before and it didn’t work. AI will say, hey, I read that this works and you should do it. And so that real life lived experience. And also, yeah, you’re correct with content, video especially written similarly and audio obviously was same as video. The AI treats that again as real human-generated content and it gives that a priority. And so that’s what we talked about a little bit earlier about getting AIO ranking, which is basically showing up in a chat as a possible company to work with, right? And so, that’s a huge opportunity there. And once we get to a point where AI video is good enough to fool people, where AI audio, which is, basically is, is good to fool enough people. And AI written obviously is close enough that it could fool people. What you have is your own personal brand or your own sort of company brand. And if you’ve established that now and you have a back catalog of 50 videos of you as a human, and then there’s another organization that is using AI and they’re brand new and it looks all slick and clean, but they really can’t prove that they’re sort of been in the trenches for this period of time, I think you’re going to come out on top. [00:29:23][115.1]
Jody Grunden: [00:29:23] Gotcha. So we’ve got a limited budget. You’re saying put it in podcasts, put it and YouTube, YouTube webinars, webinars, that sort of thing is the key because you’re thinking that’s going to really prove that hey, you truly exist, you’re truly a person, you’ve got original original material. Another question would be is, is AI looking for a consistent message over and over again, like, like let’s say I’m on 10 different podcasts and I I tell, we told earlier before we hopped up, before we started the conversation, we told why I’m wearing Hawaiian shirts all the time. If I were telling that to 10 different people, is that what AI is looking for? Or is AI looking for 10 different things from me as opposed to the same story? What’s the most effective way of getting AI to really embrace me? [00:30:17][53.4]
Ed Krystosik: [00:30:18] Yeah, so consistency, right? So the way AI operates is it’s a vectored language model. So every time that you wear a Hawaiian shirt in a video, that’s gonna show up in the training data as Jodi Hawaiian shirt. And if you think of it as a four dimensional sort of graph, and you’ve got Jodi and you got Hawaiian shirt, and every time it happens, those two get closer and more associated, right. But if your Hawaiian shirt. And Tuxedo and gym shorts, and you’re not gonna know which one of those to focus on, right? So the more data it has around a certain outcome or a certain fact, the more it’s gonna be confident to say, this is the answer. So if there’s, you know. You know two people and one is sort of all over the place and one is hyper focused and you ask hey I need somebody specialized in this it’s going to associate your name and that that specialty or or that trait more highly and that’s going to be more confident in giving that to the to the user. [00:31:24][65.6]
Jody Grunden: [00:31:24] So like, for instance, we talk about profit-focused accounting. That’s something that we’ve been, been evangelizing. I guess you could say over the last, what, 20 years, you know, the different models and here’s what to say consistency. Here’s what our CFOs teach cash production, financial pipeline, you know, sales outlook, that sort of thing, financial reporting. Um, if I, if, I’m the one saying it a lot, or should I have Jamie saying it and then Hannah saying it, and then Adam on there saying it should and multiple people. Talk about the same thing, or should it just simply be me talking about it to bring that there? I mean, what’s your thought on that? Because again, they’re hearing the same message coming from different people from the same organization. Is that what they’re looking for? [00:32:04][39.5]
Ed Krystosik: [00:32:04] Yes, and that will associate your organization with these people, with that topic, right? So one of the things you can do to, again, no one really knows exactly how the AIO or even SEO truly works, but creating frameworks, creating sort of deliverables that are like sort of step-by-step or process-oriented or strategy-oriented, giving them a unique name and then associating that with your organization or you as a person. Right, starts to build up that association in the AI’s vector database in its mind. And then when it’s looking for, hey, I’m looking for an accounting strategy that gives me X, Y, or Z. And it goes and finds in its database, well, we’ve got this one that seems to be really well laid out, really well thought out. And I can absolutely ingest it in a very simple way because it’s got bullet points or it’s step-by-steps or it has got breakdowns. And it’s gonna be more apt to serve that up than having to go and synthesize 10 different blog posts that are sort of disparate topics and say, well, here’s what I think that Jody means. [00:33:10][66.0]
Jody Grunden: [00:33:12] That’s cool. Cause we, we just put together a maturity model. Jamie was the, uh, the spearhead of that and it addresses those five areas. So I’m thinking for me personally, I’m, well, maybe we should be having many people talking about the maturity model, uh whether that’s on a video in person speaking events that are recorded, that are put on, on the internet, uh YouTube’s, you know, breaking it down, you know, piece by piece a little bit. But, you know, making it very educational so that people will listen to it and that they picks it up as if it’s so it’s not like a true, you’re not not unquote, a sales pitch type of thing. [00:33:47][35.2]
Ed Krystosik: [00:33:48] Right, right. So I did a test with my co-founder the other night, we were having a strategy meeting and he was showing me the new Gemini 2.5 model. And if you Google yourself to the AI, and it’s an experimental thing, you have to go to Google labs, it’s it’s a little bit more complicated to set up. But if you just say who is Ed Kristosik, I mean, it builds a CIA dossier on me with everything about me. And it was like, I was very flattered, but it actually even probably overstated my credentials. Because I had put out information talking a lot about software development. I am not a coder. But it’s a software engineer. And I’m like, Okay, well, I can’t, I got to figure out how to fix that. But but it built a dossier on me. That was very impressive, right. But if I had been talking about basketball and skiing. And whatever it may have been, because I did a test with a few other people that I knew who weren’t in the creative space and who weren’t making content. And it was like, we think it’s this, we think they’re that, we they’re into this. It didn’t really know. But boy, it was confident in my answer because I have a ton of content out there. Right, right. [00:34:55][67.5]
Jamie Nau: [00:34:57] Yeah, I think the maturity model is a nice path to go down to, I think, because it is a very specific topic, but it’s a wide topic, right? So kind of to Jody’s first question is, yeah, we can talk about Hawaiian sure and get those two points closer to each other. But with the maturity models, we can talk about five different things and get all five of those things closer to maturity model, closer to financial forecasting. So when people say, hey, I need help building a financial forecast, like, well, these guys talk about this all the time, this is the place to go. You know, so I think that’s a nice. A nice starting point for any organization is something that is very specific, but also very broad where every podcast we go on isn’t going to be exactly the same. I might go and spend an hour talking about cash forecasting, but still talking about maturity model and Jody might go to another podcast talking maturity model, but talking pipeline and sales pipeline. And the two, the topics are just merging towards each other. [00:35:45][48.0]
Ed Krystosik: [00:35:46] Yes, absolutely. And like I said, when you give it a framework, as you said earlier, you Googled, I need, or you didn’t Google, you asked ChatGPT, I need the software to do this thing. And it just made it up, right? Because maybe that doesn’t exist. And it confidently created something. If you give a framework and it finds it and ingests it and it hears it over and over again, if someone were to put in ChatGpT, I need a financial framework that focuses on cashflow. It’ll maybe return three ideas. And the top one might be yours and say, here’s this framework. And here’s who you can talk to about it. Yeah, absolutely. [00:36:22][36.0]
Jody Grunden: [00:36:24] And hope it doesn’t end with my name. [00:36:25][1.5]
Jamie Nau: [00:36:29] Joanie’s awesome! [00:36:29][0.7]
Speaker 4: [00:36:30] Exactly. [00:36:30][0.0]
Jamie Nau: [00:36:32] I can’t wait to see what George’s Gemini looks like. George’s Geminis would be very broad. There’s a lot of good stuff in there. All right. So obviously we could talk AI all day and I’m going to, I am going to still talk AI here because this is my, this is, my favorite question to ask. And we may just keep doing this one on the podcast is in the last, last month or last 60 days, what is something really cool? Personally, you’ve used AI for, because again, we’re moving out of the work realm here. What is something that in your personal life or in your everyday life that you’ve been using AI for to, um, to make your life easier, just for, just for fun. So, um Joe, I know you’re always in the, in the tool. So what’s something fun you’ve I’m using that as well. [00:37:06][34.4]
Jody Grunden: [00:37:07] I’d say the funnest thing is that I do a, every morning I put together basically three things, my to-do list, my three things that I’m going to do today, and if today was a perfect day, what would it look like? And then I also talk about, hey, here’s what I accomplished yesterday. So I put that into a framework that works really well, and it was a visionary form framework, basically. And so every morning I take a really long walk on the beach for about, it’s basically an hour, three mile walk with the dogs. And during that walk, you know, it gives me a chance to clear my mind and so forth. And then when I get done with it, I actually just talk into it. You know, here’s what happened yesterday. You know here are the three things I want to get accomplished today. And today is perfect. You know and here’s where it would look like. And then AI then puts it into, Chai GPT puts it in to the exact format, puts it into the words. And a lot of times I’ll have to massage a little bit. Hey, sounds too much chat, GPT-ish. You put it in my language again, and it does that. And it’s been, I’ve been doing that since last September, October. And the goal is, is so that chat knows a ton about me. And so that eventually it could be, you know, my ultimate goal here is that Jamie is like, hey, what would Jody say? And boom, it knows exactly what Jody would say based on everything that Jody’s had. You know, I can go back in and say, you know, hey, what did the month of December look like for me? And it pops in. Hey, here’s what you did. Here’s how it worked out. It’s like, oh, that’s pretty cool. So, it’s kind of like a journaling type of thing that I’ve been putting together. Hopefully, like I said, it’ll lead to something bigger, but I’m super excited about it. I enjoy doing it every single morning. It’s just one of those things I look forward to and, you now, that, so that’s kind of how I’ve using it just recently. [00:38:58][111.4]
Jamie Nau: [00:38:59] Cool. Yeah, I can’t wait to talk to JodyBot at some point. [00:39:02][3.2]
Jody Grunden: [00:39:03] Yeah, exactly. [00:39:03][0.6]
Jamie Nau: [00:39:04] This is very exciting. All right, Ed, what about you? [00:39:07][3.6]
Ed Krystosik: [00:39:08] Yeah, so it’s something very similar to Jody’s. The recently open AI sort of released their memory. So it does memorize all of your chats going back, like you said, over since basically inception. And I basically asked it, give me a psychological and business and relationship breakdown of me. And there’s some prompts you can find on Twitter and different places that kind of that are very effective at this. And it gave me a full psychological profile. And then I was able to interact with it and say, Okay, well, What are my shortfalls in business? What are the things that I focus too much on and things that could do better at? And it gave me some really good insights, right? It was like, you focus too on iterating and perfecting rather than executing and launching, okay? You seem to be very interested in this, but it seems to be more effective when you do that. And it was very, very insightful. And I had it write me sort of a 30-day plan to improve Um, you know, sort of how I operate in my day to day in business. So that was, that was very interesting. And then the second thing is something also to Jody, very insightful. Um, I’ve cataloged all of my back content in the last couple of weeks. So all the podcast transcripts, all my newsletters, all of my social posts, anything that sort of I have is a creative output. And I put that into a vector database where now with content is being written on my behalf. It not only takes the idea I’ve given it, it goes to my back catalog, searches through all of that and finds relevant stories, anecdotes, things that make it more human and then uses that in the new content. So it’s always building upon itself. [00:40:52][104.1]
Jody Grunden: [00:40:54] I love you, Ed. I love y [00:40:55][1.2]
Ed Krystosik: [00:40:59] These are simple things you don’t think about. You’re like, I have a bunch of content that I’ve written. Why don’t I give that to the AI as it’s training? [00:41:06][6.5]
Jamie Nau: [00:41:07] He might be one step closer to this Jody bot than he thought. [00:41:09][2.3]
Ed Krystosik: [00:41:14] Yeah, but it’s knowing how to effectively use that, right? If you just give it like, there’s a context window with LLMs and if you give it a hundred pages of your back content, it’ll cut it off at like page three. But if you are able to vectorize that into a database where it can search over it with a confidence score, then you’re gonna find a much better output. So there’s different ways of going about it, but it certainly not rocket science. Yeah. Awesome. [00:41:40][26.1]
Jamie Nau: [00:41:42] My item is a little. [00:41:43][1.0]
Jody Grunden: [00:41:43] Out of dribble basketball, Jamie, is that what it is? [00:41:45][1.9]
Jamie Nau: [00:41:45] Yeah, I know how to do that. As well as I do use it for a ton of sports stuff too. But the one that I’ve been using it for is I actually have two projects open in chat GPT that I have been basically loading books in. Right? So I started off the first post was, okay, here’s a list of 10 books I love and five books I did not like reading. And I have one for personal and one for like business personal development type stuff. So one like entertainment books and one like development books. So I started off with that. But now every time I read a book, I go in there and I’m like, I loved this book. This is why I loved it. And then I added to that catalog or I hated this book and this is why it was too slow. It was too educational. It was to, it sounded like a textbook versus a personal book. And so I’ve been doing that. So I think I probably have like 50 books in each of those catalogs now. So every time I’m looking for a new read, I go and ask them like what I’m looking for, a new book. This is kind of what I am thinking. Like today I want to read a mystery. What mystery would you recommend? Or today I wanna read a book about like AI processes, which one would kind of fit my criteria. And so it’s I’m using it for book recommendations with the catalog already built up of what I like and what I don’t like. And then again, I finished that next book, I go in and log it. Like this one, this is why. Didn’t like this one. This is why, so that’s one thing that I’ve been using it for, which has been super helpful and I’m finding a lot more. I’m hitting a lot higher percentage on my book reads lately. [00:43:00][74.8]
Jody Grunden: [00:43:02] That’s awesome. [00:43:02][0.3]
Jamie Nau: [00:43:03] Phenomenal yeah [00:43:03][0.7]
Jody Grunden: [00:43:03] So wrapping things up here, what’s new for you and how can people get a hold of you? [00:43:07][3.8]
Ed Krystosik: [00:43:08] Yeah, absolutely. So the most effective way is to subscribe to my newsletter and you can go to the CAIOnewsletter.com. There you can get subscribed. You get information about working with us on projects. And in that new weekly newsletter, I share a lot of deep strategies. So it’s not AI news. It’s not tips and tricks. It’s how do you actually effectively in a strategic way implement AI into your business. [00:43:33][24.6]
Jody Grunden: [00:43:35] Well, you got a new subscriber for me, that’s for sure. [00:43:37][1.6]
Jamie Nau: [00:43:37] Yeah, that’s awesome. Yeah, we appreciate you coming on the show. This was this great episode. And again, a topic that we know we’ll get a lot of listeners. So appreciate you bringing all your knowledge to this. [00:43:47][10.0]
Ed Krystosik: [00:43:48] Absolutely glad to be here. I really appreciate the opportunity [00:43:50][1.7]
Jody Grunden: [00:43:51] Yeah, thanks, Ed. [00:43:51][0.5]
Jamie Nau: [00:43:59] Okay, starting in three, two, one. Hey Jody, welcome back to the podcasting world. We’ve been a little busy for a while so I think our podcast has been on hold but it’s excited to be back in here with you and recording and of course, talking about a fun topic as always. [00:44:14][15.4]
Jody Grunden: [00:44:15] Oh yeah, yeah, AI is one of those topics that I definitely enjoy talking and listening about and hearing more about it. And Ed’s a great speaker, so I’m looking forward to hearing everything Ed’s gonna say. [00:44:26][11.4]
Jamie Nau: [00:44:28] I know Ed, the nice thing about Ed, we pre-show with them and recorded with them and Ed can really talk about any topic when it comes to AI. We could have asked him anything and he had a lot of insight and I think there are a lot of really good topics here that I think all of our listeners will really enjoy.