The Hardest Question in AI Right Now: Is It Worth It?

Mitch Hey everybody, welcome back to Make Others Successful, a podcast where we share insights, stories, and strategies to help you build a better workplace. My name is Mitch. If we haven't had a chance to meet before, I'm joined with Mike and Matt. Say hello. Hello to the nice people. It's good to have you guys here. We're back in our studio. If you haven't had a chance to listen, we. We did sort of a cameo over at Tech Week in Grand Rapids where we got to do the podcast remote. Yeah, a different studio. So that was a little bit unique for us. Today we're going to be talking about drum roll AI again. Like I keep saying like, well, it's a little bit different than last time. It's sort of a continuation of last time. We've seen some shifts happen in the marketplace, and we've seen some things change according to how clients are looking at these things, how we're hearing, um, you know, thought leadership in the space. And so we want to, um, tack on to that and share our thoughts about where AI is on its journey and what leaders need to be doing differently than what they were doing before. It's not something that you should just throw at people or like, just give people the tool. It's something that you need to improve your processes with. But we didn't spend a ton of time talking about the process side of things. And I think that is what we want to spend some time about today. So does anyone else want to set the stage for today on your thoughts of what's been changing recently?

Mike I think one of the things that's going to feed part of the conversation is the introduction of consumption model or consumption based pricing. I think that scared a lot of people. We'll talk a little bit about that. I think that's worth, worth a listen, worth worth some thinking about.

Mitch Yeah. It's also growing to the point where people are spending money on it because they're excited about it and they see the potential in it, but they don't quite they're not quite able to attach it to ROI, something that results in an actual business outcome. So we want to paint that picture today. And one of the things that we wanted to call out was, uh, while there hasn't been necessarily an ROI articulated everywhere, McKinsey has a report that says eighty percent of people say that AI improved their workplace productivity in twenty twenty five. So with that said, and some of the models shifting, uh, their, their, their pricing model and things like that, it's, uh, an ever moving target. Matt. What are you seeing in the space right now as far as companies investing in AI?

Matt So people are, have really been progressing between a time period in which AI was cheap. Uh, and everybody was incentive to just give it a try, give it a go. Uh, and people were buying it and, uh, having people start to use it and. Like, as you mentioned in the McKinsey report, people feel like it's really helping. Right. Um, but as the models have changed and the way that things are being charged. Have changed, going to consumption or just doubling or tripling the price of, uh, of. Buying that all encompassing license. Uh, you really need to start attaching an ROI. Um, or organizations who've resisted AI now are being asked to spend money to adopt it. They also are asking, okay, how can I attach ROI? What, what's the return on my investment? How much am I getting back? How much faster? Yes. You feel like you got more work done. What? How does that help me? Right. Uh, and there's a lot of challenges with that, both from a practical perspective in actually doing it, but then also in making sure you're watching all of the, the areas that you really need to watch to make sure that you're understanding what that, what's really happening as your organization, uh, changes through the adoption of AI.

Mitch And I feel like one of the reasons why this is becoming more relevant is because the cost has been going up. Like you said, the sort of corollary I can draw is I remember when new streaming services come out, it's like five point nine nine a month and you can jump on and you get all this video to watch. And so everybody gets it. And then all of a sudden the shareholders start showing up saying, hey, we got to be turning a profit on this business. And so the price, the monthly subscription keeps ticking up and up. And I feel like where we are in that phase with AI, which it is now causing these questions of should we keep this or can we do without this? And sometimes that is not an easy question to answer.

Matt It's also true that AI is not just a standard software as a service. I'm going to put it in a low tier and get people to adopt, and then we'll raise up the price. They're constantly having to create new models that have different cost constraints and that have different size constraints. And so it's not only is it a, a, uh, a standard software as a service marketing business strategy. The truth of the matter is the more advanced the, the model, the more expensive it is. There's no two ways about it. Everybody. You can't you can't not be watching the news and see something about a data center going in and the cost of data centers and the cost of all of the electricity and all of the other things around these. The concept of AI solutions. So, um, it actually has a double whammy in that regard and related to the price, um, and how it keeps just going up.

Mitch Yeah, yeah. So hopefully that kind of sets the stage for our conversation. We're focused on two things. One is integrating AI into your workforce and what's the best way to do that? And then also, how do you think about this in a world where it's not just some flat fee, it's consumption based and there's a variable cost involved that you need to be weighing if it's worth it all the time. So let's dig into our first section about what's the strategy of implementing these things. So most of the time we've seen in the past, a lot of people look at this situation as, okay, we decided we're going to move forward with this AI model, this AI tool. This company, let's buy everybody licenses and let's do some training on those licenses. But there's so much beyond that. Do you want to uncover some of that?

Matt Yeah. I mean, the first thing that people need to recognize when they're rolling out AI is what type of AI that you're rolling out. Um, is this AI that's going to fit into a person's, uh, existing workflow? Uh, and they're going to integrate that II in themselves. So, uh, this is what most people think of when they think about ChatGPT copilot, uh, Claude, uh, etc. the actual applications where it's very free form. Uh, I have a problem. I'm going to use AI to solve that problem. So that's one, one way that people need to be thinking about it. And the other is integrating it into systems that people are already using. Right? So this is way more nuanced and you're not really buying a license of copilot or buying a license of Claude. You're actually integrating that into an existing business process or business solution. And these are very different approaches that are only appropriate for very specific use cases. Right. Uh, and why this really matters is, uh, one of the reasons is it's very, very difficult to capture and understand the ROI for the first, if I'm buying everybody a Claude seat, for example, how do you understand how long this would have taken if they had done it themselves? It's incredibly difficult, especially when you think, and I'm sure most people who have used AI, uh, will understand what I'm saying here. Maybe it would have taken you an hour and a half to produce something. But when you use Claude or Copilot or another AI tool, it took you half an hour, but it also produced ten times more content. How do you calculate that? How do you how do you understand the value of that? Right. It's incredibly difficult to understand that. Whereas on the other side, if I'm adding this to an existing business process or trying to improve sales, like a very common one for AI is the sales process or the invoicing process or like those types of things, you know, what you were doing before. You know, how many people and how much time it took you to do this before. Most people have some sort of metrics that you can watch. And so it's a little bit easier on that side to say, this is what our spend was. This is what we're doing. So first of all, what type, what are you trying to implement? What what are you doing?

Mitch Mhm. It reminds me a little bit of our dilemma that we faced still face is we have chosen this method of content authority of of thought leadership in the space. And the one great side effect of that is it helps people, right? Like that's our our mission make others successful, etc.. But there's a secondary one, which is people say, hey, I want your help. Can you help me with this thing? And it produces work for us. But mapping that to this video, we spent this amount of time on and it was part of this ecosystem. Is that directly attributable to that sale and connecting all those dots? Like it's like it's the content people's nightmare is that attribution and making sure that something has that full picture, but it's not in, it's not not valuable to try to figure some of that stuff out.

Matt Yeah. And a way that some organizations can look to solve this problem is by saying, we're going to standardize on a particular model as a baseline, right? Knowledge workers, workers who are doing this type of work are asking questions, trying to work in this space. That's a little bit nebulous, a little bit, you know, not just fitting into a particular process. Uh, we're going to get them all the same thing. And that thing isn't going to be consumption based. It isn't going to be, you know, I'm not if all of a sudden they use twice as much one month, I'm not going to pay a crazy amount more. This is just the baseline. We're going to buy this for everybody, whatever that might be. Um, and by doing that, you're basically saying we recognize AI as valuable. Uh, but we're not going to just make it available to everyone and we're not going to make it available to everyone at the all the same level, right. Uh, which is one of the things about copilot that's interesting because I think everybody in an organization can use free copilot and get value one hundred percent, right? I think there's a much smaller group of people that really can get value out of premium copilot. I think there's an even smaller group where like Co-work or some of these consumption based, uh, usages, they really are going to get a tremendous value. And so you should be thinking about it in that tiered approach, in deciding who it is that you are going to be giving these tools to. It's not a blanket thing. You don't want to just give everyone all the fancy, uh, AI tools and say, figure it out. Right. Um, because that's the other piece of it is you need to be helping people understand where they can get value out of it. Uh, that is where you're going to have the biggest impact in this style of AI that, that type of AI where you're giving this general knowledge. Again, GPT copilot, etc. that's where you're going to get the most value or you're going to help your employees get the most value. How you capture that, how you understand the ROI. I think that's a lot harder.

Mitch Yeah. And like you said earlier, there's a there's a balance to this of like giving people access to the tool for their regular work. But then there's the process side of things. We've seen both sides. One that kind of sticks out that I remember, which maybe you can talk about, is they were dead set on the process side. They were like, we have this way to use this amount of tokens and get it really cheap and integrate it directly into our process. But then they missed the, the knowledge worker side and they weren't wrong. They were just missing something. When is it right to go from that world where you're saying everybody should have this license to make them more effective to now we need to put it in a business process.

Matt I mean, I think.

Mitch Or and should they happen in that order?

Matt No, I think a lot of it is very specific to your organization. I think the one I think the main thing I would want to leave someone with is most or all organizations need both. Like, I don't think there's a single organization that can't get value out of a general AI knowledge worker. You know, copilot, GPT, whatever. Uh, and also make sure it's integrated into their business process now. They were doing custom development. They were like really integrating this into because their whole business model was around having custom business processes and workflow that they built software to support. I'm not saying every company needs to do that, right. There's tons of vendor provided solutions that have AI components, right? You have to start using them. You have to want to use them. You need to train on them. You need to activate them. But in some cases, you need to pay for a license to get that AI, those AI features. But, uh, you should really be looking at both sides and you should have both sides depending on your business, depending on where you're seeing the most issues and challenges. We'll kind of dictate where you should start, but I think at this point, most organizations have started already in one or the other at some level. Yeah. Um.

Mitch Well, how do you how do you distinguish between like the best place to start? Because when with AI being a non-deterministic thing, which if you don't, if you're not familiar with that term, it means you're not going to get the same exact thing twice. It is unique every time. When you're injecting that into a process, usually that process should be relatively controlled and very deterministic of how it happens. And so when we say, oh, put AI in that, it's sort of like Mrs. like it doesn't even make sense because it's, it could go different ways. And so there's this whole method of steering and, and. increasing the chance that it will do things correctly. But when I think of how we use it, it's things like, someone gave me this long summary of what they need help with. I need to come up with a deal title. What? What should we name this opportunity? And that I don't really care if it's deterministic or the same exact every time. I just need a title. Versus I want to AI to march it through this process that I like. I want strict control over how do you distinguish that?

Matt Um, it's not easy to explain. What I would tell you is that AI can play so many roles in so many different business processes, it's hard to distill it down into a single thing. Uh, what I would say as an example would be, uh, if you're having an, you have an intake process that brings content into your system, uh, and the user is entering that data, right? Using AI to help you understand what's missing, understand, transform that into something that you can use can be very powerful and doesn't need that much steering, right? It understands language already. It understands what understands what's going on. Uh, being able to say, hey, uh, I need to look at the data that we have right now and identify anomalies or differences or challenges, right? It's really good at looking at data, right? Uh, there are so many things that you can use AI for that will, uh, help and improve and drive efficiencies across business processes. It really depends on your business process and where you're seeing pain, where the problems are. I know, Mike, you've done some work with several of our customers to try to, uh, make these things better. And you've, I know you've used a number of different tools, but, uh, do you have any other scenarios where it's like, hey, this business process, this is a good this is a sign that AI could could really provide value.

Mike I think the, the intake one is, is valuable. We have a customer that's trying to process like volumes of documents, um, and glean information and instruction and figuring out how to do the right thing with the information from those documents. And that's, that's, that one in particular is proving to be a bit of a challenge because there's a bit of like, I can read the whole document, I can tell you what's in there, I can summarize it. And in that particular case, we're having to, uh, train the agent to only pay attention to the things that we want it to care about, right? Or to ignore specific scenarios or to highlight, um, you know, key things that, that are important for that particular scenario. Um, and so you got to spend more time with it. Like we've played with, you know, a couple of different methods. There's the old AI builder, uh, method that's been around for a long time, and you can feed it a lot of like documents and train the model and you get structured data output basically on the other side. Um, and that works well for like large amounts of data, but when it comes to less structured data, maybe interact with a, with a human and like, get some more information, um, generative AI and just chat seems to work a lot better than, than something like that. Um, when it comes, comes to processing that, the other thing that like we've heard from customers, um, recently is frustration just on the agent side that the agent is either hallucinating or going too far or going into a space that they don't want it to go into. So how do we tame that? How do we contain that? Um, and I was at a recent event and the presenter talked about agents being like third graders with PhDs, right. They've got access to tons of information, but when you tell them to clean their room, they still just shove the stuff under the bed and shut the closet door, right? And so there are certain things like when you talk about like the chat capabilities and some of the capabilities, there's like a baseline of those things and like meeting transcription and that kind of stuff is like, yeah, everybody should have that. Anybody who's a knowledge worker having access to that tool, it's going to make them more effective right away. And it's a pretty low lift, right? It's a low barrier to entry. Where it suffers is you have to be able to trust it to understand your business process, whatever that is, how your business works, and then apply that to live data. Let's say in a transactional system, like bringing those things together, I think is where a lot of the struggle is right now. Um, and that's, that's more it's agent level work. Um, but that's what we're seeing.

Matt But that's, so that brings us back to the ROI. Like in that particular case, the one that you talked about initially, uh, there, uh, the return on investment has to do with today. They have to spend a large amount of time training humans to do this work, or they have to trust and, and the pressure on the humans is very, very high because they need to get this right every single time. And so the value for them is pretty high to invest in, in that, in trying to refine that and make that better. Because even if you drop that by a few percentage points, that's going to provide tremendous value to their business.

Mike Yeah. And they're like, that is true. But their particular challenge is they cannot afford risk in that scenario either. So they can't just feed it through a bot and expect what comes out the other end to be correct. So they have to have a set of human eyes on it regardless. And so it's an interesting balance.

Mitch Yeah. So you have to have one last human set of eyes check. Yeah. The example I was thinking of was in the copilot studio. There's it's like the newer workflow automation type tool. There's an action called classify where you give it this email. Incoming email could be one of three different things depending on which one it is, steer it one way versus the other. And there is no guarantee that it chooses the right one. And so you need to operate almost as if should this go down the wrong path, what's my eject that I can pull so that it doesn't actually cause damage or do something destructive? So I know I'm painting it in this like super negative light, but I think we need to be aware that it is not foolproof. It's just really smart and does the right thing. A lot of the time.

Matt We talk a lot about the treating AI as your assistant and providing IT instruction. And it's the same case whether or not you're in the mode where you're integrating it in your business process, or whether or not you're in a mode where you're adding it to your knowledge workers. In both cases, you need to treat it like you're handing this off to someone else and you need to check it. You need to verify it. But that makes it even harder to understand the re right back to that conversation. Like, uh, if I have to have somebody that is smart and knows all the details anyways, why do I need AI? Right? And the answer to that question is if they can do one hundred of those tasks instead of ten in a day or an hour, uh, that provides value.

Mitch And you actually have a webinar coming up about treating like part of the content is about treating AI like, so if you want to learn more about digital events, we'll have it listed there. That goes into that a lot deeper. But yes, Mike, what was your.

Mike I was just, Matt just mentioned, like, if you have somebody that's really smart that you rely on and, and how AI can kind of feel like it can fill that gap, I immediately thought I was like, I use a variety of AI tools now instead of asking Matt Dressel.

Mitch And that's your determining like a sign of progress.

Matt Uh, so it provides all the value because now you can get rid of me.

Mike No, it's not about that. It's, uh, before, like, we always had search engines before, right? But then even then you'd get stumped. And then, like I always had Matt and I could be like, hey, you help me think about this, right? And give me the angle that I'm missing. And I could always get that. And now I don't need to do that all the time, so I don't have to tie him up.

Mitch Yeah. I've reflected on the same thing, which, uh, I've Matt is one of them. I've also had other previous coworkers where it's like, man, it was really good to have their, their brain on this thing, otherwise I would have been lost. But now I'm like, I can just ask a smart AI model and it could probably do a pretty good job and probably just do it for me. And it's a weird.

Matt How much value is that?

Mitch Yeah, you. You tell me, Mr. ROI. I don't know. It's.

Matt I don't know. That's that's the challenge.

Mitch It's worth what we're paying for it. I know that.

Matt Well, you mentioned about that is interesting too, because we have gone through. Now now that we're doing, we have a number of things that are consumption based. And even all three of us have had the conversation about exactly what we've been talking about in the podcast, about businesses choosing how much to spend and how to do it, and what's the value of it. Uh, and, you know, you get to the end of the month and you go, oh, we spent two thousand dollars on credits or whatever that might be. Uh, and I think in one of the meetings I came in was like, are we sure? Are we okay with that? And Mike, your response was, I mean, Mitch was able look at what Mitch was able to accomplish that or look at what he was able to accomplish. Like it would have.

Mitch Taken.

Matt Him three times as much. Like that's it's tremendous.

Mitch It was a proud moment.

Mike If it's well directed. Right. It can be highly useful. And like we're not just wasting our AI credits on like generating random images, right?

Matt Hello, world.

Mike Right. I wrote, well, I shouldn't say I wrote we wrote a bunch of code using AI, and I did a little bit of math on it, but it's like, you can then feed that back through AI and say, hey, how much? How long would this have taken a team of developers to, to build? Right. And it's ours versus or weeks versus a couple hours.

Matt But Mitch but Mitch, I gotta ask for real, real question here. Would you have would you have built it as robust as AI built it?

Mitch No, not a chance.

Matt So then it would have been less so. That's not a valid comparison.

Speaker 4 No no no no no no no no.

Mike It would like. But.

Mitch Oh so you're saying I could have gotten it done quicker dirtier.

Matt Uh yeah. It's it's very it's a difficult when, when AI produces, you know, four truckloads of stuff. And if you do the same thing, you would have done one truckload. The four truckload is way better, more secure, more whatever. Like all these like, yes, you can make all the better, better, better, better, better. But you would have never done that. So then what's the.

Mike Value.

Speaker 4 You.

Mike Know? Well, we would have said, oh, it'd be really good to do those things. And then somebody would have said, well what's that going to cost?

Matt No, I know, but it's, it's a cyclical. It's a, it's a, it's a, it's a snake eating its own tail.

Mike But in the, in the software world, all that's created is a scenario where it's like, wow, those product ideas that everybody had that got shot down was like, well, let's revive them.

Mitch This feels like a perfect policy that you should create that dictates what level of security and self-correction our AI built code should do requires. Yeah. So that we can truly.

Matt That sounds like a.

Mitch Good.

Matt Use of my time.

Mitch ROI. ROI. Let's talk more about the consumption based stuff because it's something that changed recently. And it's something that, uh, put a sour taste in my mouth. I'll say like, one of the closest examples that I can think of is Cowork when it was in, uh, preview or what's the, the frontier or whatever. Yeah, it was.

Matt Free. Do it, do it.

Mitch It was awesome. Like I was, I was having it. I think I said at the time, my favorite thing was every day I was having it. Look at all my priorities and plan my next week out ahead of me and preschedule my time and whatnot. I haven't done that in months. Cowork went to a consumption based model, and all of a sudden you started to face the reality of, oh, that query is going to cost you seven dollars every single time you run it. And it created this like dilemma in my brain of.

Matt Is it worth.

Mitch It? Can I, can I really justify that? So one of the metaphors that I've seen made, I think it was on LinkedIn. Sorry, I don't know who said it, but we're entering a world where AI is now the scene where Microsoft or whoever will sell you the razor for free, or they'll give you the razor for free, but you need to buy the blades, you need to buy the the replaceable thing that comes with it. And so if we think of it as like, yeah, you can get copilot for free or just thirty bucks for that initial slug, like that is valuable. But to get that nice clean shave that doesn't pull at you and feels really nice.

Matt I'm gonna, I'm gonna, I'm gonna take that metaphor to the next level. Do it. The real problem is that everybody kind of knows how much hair they're growing regularly. All right, so you know, kind of what to plan for, but this is like giving you a razor or giving you the razor, the, the, the handle and then buying blades, but you don't know whether or not the next thing you're going to shave is an alpaca or a like tiny dog. You don't know, like it could be dogs.

Mitch Is that is that. Well, I guess people Sarah just shaved her dog. She people do. Yeah.

Matt Like you just don't know. You don't know what you don't know what you're going to get next, right? Yeah. Uh, it is a, like it's a very interesting problem. And, but what's really interesting is this isn't new. Like, just like you were talking about with that previous customer, any developer who's been integrating this for any length of time, that's all they cared about. They've been talking token price for years and years since the day it came out. Yeah, that has been everything they've done. But then much like loop back around to the beginning of this conversation to get the market going, these big organizations, when they built, they abstracted all of that. They said, I'm going to give you a either an all you can eat or a quasi all you can eat, uh, like all you can eat for a particular day for this price a month, and it made it really easy for people to make a decision. And then we all got hooked on it. And now everything's turning into a little bit more, a little bit more, a little bit more money grubbing.

Mitch No, I think that example of like, you don't know what you're going to get, what you're going to have to shave is helpful. It, uh, it introduces something that is highly variable that could balloon pretty quickly. That the thing when you were talking about developers thinking about tokens, I think it variable costs have existed for a long time, but thinking about something like server infrastructure where you need a certain amount of horsepower CPU to facilitate, say, like a software as a service to run and provide good performance to all of your users. Some might look at that and say, wow, that cost keeps going up and up and up. And I hate that from a financial standpoint. But that also implies more people are using it. That means more money is coming in. And so it just hopefully it's this they're diverging. They're, they're two lines that are both going up, but they're diverging like going farther and farther apart where the benefit is getting more and more while the cost is staying relatively. I don't know exactly how to think about that, but.

Matt That's one of the ways to think about it. But that's going back to the exact topic. How do you judge the value? Because I can tell you, uh, me, our real world, our situation internally, I, you try to use co-work for something and it didn't work three times. And I used almost all of my individual credits for the month and it never did what I wanted it to do. It could be my prompts fault. It could be coworkers fault, I don't know. But I got to the end of it and I was like, this is stupid. Like you spent thousands of credits to tell me you couldn't do it. What? Like I don't.

Mike You should have asked for a refund.

Matt Yeah. I mean, I did message to Microsoft, the product team, and went, this is stupid. Like, I get that it's really hard for them because they're like.

Mitch It's non-deterministic.

Matt Well, no, no, the bigger problem is I thought to myself exactly what he said. Give me a credit. Right? Like give like, give me those credits back. I'll just click that every time. Now you get into the model of like, like Amazon, where like, if you've, if you've done too many returns, it's like we're going to, you know, like there are ways you can do it differently, but it's just, it's so much more complicated. It just gets more and more and more complicated. But, uh, yeah, that was a really frustrating thing for me. This last, this last month, I was like, I was not only did it take a long time, like from a time suck perspective, but then I just had to use regular copilot to do it anyways.

Mike So I think like that event that I was at yesterday. Actually, there was a bit of a knee jerk reaction. Uh, and the company that was represented there basically took, and they built a model on basically a Mac studio.

Matt On premise.

Mike On prem. And, uh, so then they have to pay for the electricity and for that machine and for somebody to put the things together on that machine, but they don't have to, they don't have the variable cost issue necessarily, right?

Matt There's a lot of people talking about local LMS, building your own model, owning your own model, how much.

Mike How much pressure do you think something like that or, or a transition to something like that would will put on the big players to reduce cost? I know there's all kinds of other stuff going on right now where they're like, oh, AI is going to kill us all. We can better regulate it and all that kind of stuff. And I see a lot of that as a play to keep the little guys out of the market and to put some pressure that keeps the money going in for those consumption credits.

Matt It's a complicated problem. There's lots of reasons why local models can make sense. One of them being cost. But I would also say there is a significant cost to maintaining and keeping those things running. Uh, if you had to do those things custom anyways, and you were iterating on that custom stuff, you're going to be spending the money anyways. So you could probably make some justification. But if you're like, I just want to use an LLM to ask a question about a generalized topic. That seems crazy. Uh, there's like, unless you're like in a highly regulated, highly secretive situation, you know, the cost just like it just doesn't make any sense.

Mitch Is the future where someone, every household just has an appliance at their house that hosts their AI? Like I've seen a lot more consumer targeted where it's like this. Battery backup slash. Network attached. Storage slash AI brain for your house. Yeah. When I think of I was trying to think of some like similar situation where you wouldn't expect someone to always go to a restaurant to eat a meal. They can eat their meal at home and they have the, the stove and the microwave and all the things to do that at home. Is it wrong of them to bring that stuff back? And I know it's totally backwards. Everything went to the cloud. I don't think.

Matt Back. I don't, I don't, I don't think AI is commercialized enough for that yet. Right. Uh, when you think about those things, electricity, electricity comes out of a wall socket and everybody's home. All you're doing is putting a battery pack and management around it and putting it in your house. That makes sense. That's not AI, right? Like AI at its core, what most people use AI for is based on a centralized model based on everybody's data. That's what most people use it for. Why would I need that? Why would I want that in my home? Like, it doesn't make any sense, right? Uh, but, uh, have.

Mitch You ever been to the data hoarder subreddit?

Matt My point is there you need to there's going to be specific scenarios that I think that may be valuable, but most people aren't storing their data locally anyways. Ninety nine percent of people are sorting their data in the cloud and couldn't be happier, right? When you start to see that change and you start to see that data live more local or be have purpose that is local to, to people's home and people's environment. Maybe. Um, but I think it's, I don't, I don't see that as being a, a big change across the market. I do think there's tons of people, you and I both have a lot of data that's local at our houses and I could see us.

Mitch You don't know that.

Matt Yes, I do. And I could see people like us for sure. Yeah. And that could be a large number of people. Like I think there's, there's a market for that. Uh, but it's, it's not going to be. Is that just.

Mitch A natural progression though, where something is super specialized, costs a lot of money to have your own. And then big businesses abstract it, provide it as a service for people where they can just subscribe to it, and then it balloons enough so that people want to pull that back because it, uh, the technology develops enough where it's easy. It's not a super specialized thing anymore. Yeah.

Matt I mean, that, that progression is absolutely natural and common. What's what the, the time in which it comes to be local, uh, is that the cloud, like there has to be a market force cost or security or some other market force that brings it back, forces it down people. If you have to choose between, uh, everyone could change their oil. Very few people change their oil in their car. Why is that? Because it's cheaper and easier for me to just have somebody less risky to just have a mechanic do it, right?

Mitch Like it's funny you use cheaper because like in it depends on how much you value your time.

Matt No, to a lot of people, they go, the risk is I don't, I'm not experienced. I'll snap something off or not do something right. And that's a very expensive vehicle. Now I broke the vehicle and right. Like there's lots of ways you can think about this, but that's like, it's the same like you talked about food, right? People can make the food at home. Most homes have stoves and like all those things, right? But do I always want to do that? Is that like, it's all the same? It's the market force conversation. And right now, I'm not sure that I see a scenario where the cost of it out in the wild being so much more expensive, uh, or so much less secure or, you know, I just don't I don't know that I see it. Me personally.

Mitch Yeah. I feel like a lot of the noise I've seen recently online is more powerful, less cost at the same time with AI models. And so as long as that stays relatively flat, it's probably okay. But I think there's some super technology somewhere that is obnoxiously expensive, but is not going to see the hands of consumers hopefully anytime soon. I don't know, we've diverted a little bit from our conversation, but it's it's sort of fun.

Matt It's all the same thing. ROI.

Mitch Yeah, so we need to move out of this space where we're just going to spend whatever we need to in order to get AI integrated. We need to intelligently pull it into our processes and.

Matt Give business scenarios for people that you think are valuable, right?

Mitch Yep. So maybe a question to ponder as you go is which process did AI actually change in this quarter? Not just speed up for one person that that's one aside, I feel like I keep hearing these big companies that spend tons of money on AI and they displace a bunch of jobs. That has to come from we have implemented AI into this process. And so somebody, it hasn't been us is implementing AI at such a high level, such a core level at these businesses. I think it's companies closer to like the, the actual AI providers. Um, I don't know how to think about that.

Matt I'm going to bring it home for you right now because I've thought about this pretty extensively. Uh, we are a services company. Project managers are really critical to our organization. Uh, we have had one project manager for a very long time. If we had not implemented AI by now, we would have hired at least one, if not two, more project managers for the number of projects that we have right now. There's no question in my mind, no question whatsoever. AI has prevented us from hiring at least one other person, for sure. And I could say the same thing about developers or people who are executing on our projects, right? When you think about a very large organization that's been doing this at a very large level for twenty years, they have all of those people already. Someone came in and implemented AI and made all those people faster and then shrunk. That's what they did. And that doesn't have to be AI within a particular business process. It probably is. Yeah, there probably is some of that as well. But in our case, we haven't really implemented AI in a custom like in a specific process. AI has just enabled all of our team to be more efficient.

Mitch Okay. Next episode we're going to cover. Are those businesses evil for doing that? Save save your thoughts for next time. I'm just kidding.

Matt That sounds like a. Chances are, the.

Mike Chances are we would have said they were evil for doing something else before that.

Mitch Okay, we're going to ponder that next month. Anyway, thanks for joining us today. It was good to talk about AI again, but like I said, it's a moving target. It's flying at us and we're doing our best to make our way through it. So that's all for today. We'll see you next time.

Matt Thanks, everybody.

Mike See ya.

Speaker 5 Hoo hoo hoo! Hoo hoo hoo hoo! Hoo hoo hoo hoo!

‍