AI as a Tool, Not a Takeover with Jason Butler

Jason Butler built RoboSource on a simple premise: technology should free people to do meaningful work, not replace them. As CEO of the Indianapolis-based automation and AI development firm, Butler has spent decades in software engineering and business process improvement. His approach cuts through the AI hype to focus on what actually matters—enabling teams to contribute in ways that drive real business value. Butler rebranded his company from eduSource to RoboSource not because he jumped on a trend, but because the technology finally caught up to a vision he’d held for years. That vision centers on taking monotonous tasks off people’s plates so they can focus on work that requires human creativity, relationship building and strategic thinking.

Butler challenges the common misconception that AI requires a comprehensive strategy before you can start. He argues the opposite: AI is a tool, like a hammer or a nail gun, and the best way to understand its capabilities is to use it. He recommends bringing AI into every conversation as a thought partner, testing its ability to provide feedback on proposals, predict stakeholder reactions or surface historical knowledge buried in decades of files. This hands-on experimentation helps leaders separate genuine utility from hallucination, building trust in specific use cases while maintaining healthy skepticism elsewhere. Butler warns against the main pitfall he sees: irrational trust that develops too quickly when AI produces impressive results, leading to lazy management that accepts outputs without verification.

The practical path forward starts with small experiments that respect your team’s time and intelligence. Butler’s RoboSource platform, Process Coach, embodies this philosophy by letting business leaders define processes in plain English while AI handles enforcement and automation behind the scenes. The system works where people already work—through email, text and soon voice—rather than demanding another login. For leaders wondering where to begin, Butler offers a straightforward tech tip: tell AI who to act as, provide context about your situation, specify what you won’t do, then ask it to interview you one question at a time. This approach transforms generic responses into tailored insights that respect the complexity of real business problems while keeping humans firmly in control of the decisions that matter.

https://youtu.be/wcTpS1kObHw

Episode Summary:

Is AI going to replace your team or empower them? RoboSource CEO Jason Butler joins The COVI Way to cut through the hype and share how business leaders can start using AI as a practical tool today. In this episode, Jason reveals why waiting for AI to be “perfected” is the wrong approach, how to avoid the pitfall of irrational trust in AI outputs, and his four-step framework for getting better responses from ChatGPT, Claude, and other AI tools.

Key Topics:

• Treat AI as a tool, not a strategy, like any other business instrument that needs proper application.

• Start experimenting with AI now rather than waiting for “perfect” technology or understanding.

• Use the “intelligent toddler” mindset to set realistic expectations for AI capabilities.

• Apply human-in-the-loop oversight to avoid over-trusting AI outputs.

• Master the four-step prompting method: role, context, constraints, and iterative questioning.

• Focus on automating meaningless work to free people for impactful activities

Connect with COVI:

Instagram | Facebook | Website | FREE IT Strategy Analysis

Links to The COVI Way Podcast

YouTube | Apple Podcasts | Spotify

Episode Transcript

The COVI Way Podcast Ep. 13 - with Jason Butler

Welcome to The COVI Way, the podcast where we break down the intersection of people and technology because it shouldn’t just be about fixing problems. It’s about enabling people to do their best work. I’m Cody Lents, the VP of sales at COVI. And today we’ll demystify artificial intelligence in today’s AI saturated world with our guest, Jason Butler. Jason is the co-founder and CEO of RoboSource, the Indianapolis based automation and AI development firm.

He spent decades at the trenches of software engineering and business process improvement. And he’s not just another successful coding geek. What makes him different is his track record of building solutions centered from freeing people up to do the work that actually matters. He’s not here to preach AI hype or replace your team with bots. He’s here to discuss how AI fits into your business, what’s real and what’s not, and how to start without breaking the culture that you’ve built. Today is about practical strategies, mindset shifts, and the future of automation without losing sight of our humanity.

Cody: Jason, thanks for popping on the show today. How you been?

Jason: Great. Thanks for having me. This will be fun. I know that we talk a lot. So if you know me as a lot of you do at this point, I talk a lot more. And today I’ve been nursing a respiratory infection for a couple of weeks and we will see if I can talk less. All right. But in light of that, what’d you have in your coffee this morning?

Cody: I don’t drink coffee. You know, my morning drink of choice for caffeine is Mountain Dew. I am the computer geek. It started in college. I can’t break it. It’s just it’s a thing. And that’s just who I am. So I’ve embraced it.

Jason: From what I see walking in the storms, it looks like Mountain Dew is making a comeback right now.

Cody: Yeah, it’s terrible for you, but it gets me going in the morning. So that is funny. I remember playing Super Nintendo and slamming a Mountain Dew as a kid.

Jason: Yeah, that is absolutely so. Switching gears to professional since you aren’t driving on the coffee train that I am. I know you rebranded RoboSource a handful of years ago and you have a long history in custom software development. Can you give me like the quick overview of your journey into where you are today?

Cody: Yeah, well, you know, it’s always been about helping people do meaningful impactful work. That’s always been our driver. And so, when I finished my MBA in 2009 and came out of that, was like, I’m starting a business. That’s really what I wanted to do. And, it was like, how can I help people do work that matters? Because I think people wake up in the morning and they’re like, I want to impact the business. I want to do things that actually help the business move forward and putting numbers into spreadsheets and moving data between different SaaS apps isn’t exactly inspiring work.

And so we started building custom software to help people do more meaningful work. Because at the end of the day, an application that you build is automating a process to help people do things more efficiently. So that was really has been our journey for 12 of the last 15 years as artificial intelligence and specifically the generative AI pieces started to become more ubiquitous and people were excited about them and talking about them.

A product idea that I had in the back of my head started to form. And I started to think, you know, this is something that I think we could actually pull off now. Where when the idea formed originally in 2015, 2016, based upon all of our clients, it was like, there’s no way we could pull this off. The technology is not there, but as generative AI started to become a little more powerful. It was like, Hey, you know, I actually think we could get here. So we, we pivoted to start doing more specific process automation and start to build out our product.

And, that’s really now what we focus on. It’s officially launching here at September 11th. Right now I’d call us more of a tech enabled service, where we help people think through their business processes, how they’re executing, how they can be more efficient, really towards that goal of like, are you taking the business? What’s standing in your way and how can we use our technology and technology in general, AI specifically to help knock those barriers down so you can hit the goals you’re after. And so between helping people hit their long-term goals of the business and finding what is meaningful work for the workforce. It’s a lot more than just what all the marketing about AI is, which is we’re gonna take people’s jobs. And it’s like, that’s not really what we’re doing. We’re actually empowering people to do work that matters. And we’ve got a platform that helps us make that reality.

Jason: Yeah, I saw that that platform is now on the website. Did you redo the website too as part of that?

Cody: So we rebranded. So we renamed ourselves about three years ago to RoboSource. And then we recently got a round of investment and the investors came in and helped us rebrand to be more specific towards the type of market we’re going for. So we redid the website and now we have a product that’s coming out alongside of that.

Jason: Yeah, and that’s Process Coach, right?

Cody: Yeah. I know I’ve got a slight demo on that. I’m looking forward to a more deep dive into it. So I think what I heard you say to use therapy language there is the rebrand from eduSource to RoboSource was less about like a strategy pivot and more about the realization or the coming about of something you’d already dreamed and the technology was just finally catching up with your hopes and your wishes there.

Jason: Correct. That’s cool. I always thought that you just landed on a good niche and ran with it.

Cody: Oh, no, it’s been a thing that we were part of from the beginning. We’ve always like even eduSource when we had had that brand and we’re operating under that. It was always about helping people do meaningful work. Now the point of eduSource and it got confusing because people thought we were an education software. But we, that’s not what we were doing. What we were doing is we were bringing in juniors and seniors in college that were really equipped to build software, but they weren’t getting practical real life experience. So they were coming out of college and not actually being effective.

So we wanted to be able to do that meaningful impactful work. So we were like, Hey, let’s bring them in for two year long apprenticeships. And then instead of eduSource or instead of outsourcing overseas, what if you eduSource to the local education community and we were able to train the really intelligent, you know, we’re in Indiana, so the really intelligent Indiana kids around here are doing real work, able to pull this off. That was really kind of where it started.

But again, the branding was a little confusing. Everyone thought we were selling education software. We weren’t, which is why we moved more towards RoboSource and the idea of, again, instead of outsourcing to overseas, maybe Eastern Europe, what if, you know, you outsource to robots? So we help, you know, where your RoboSource is the idea behind that. So.

Jason: I like it. You said earlier that any type of SaaS application or application in general is created to automate process. Can you talk a little bit more about that? Because I think people don’t think of automation and software application in the same vein when they hear it out wild.

Cody: When you’re trying to scale your business, one of the things that you do is you take a process is you basically, need to take what used to be chaos and turn it into a process. And once you have that process and it’s well understood, it’s documented, it’s executed the same way every time. Then you look to start to automate it. Well, traditionally, the way we automated it is we sat down and we said, hey, let’s build some form of software that would walk you through these steps, gathering the data and the information that you need when you need it, making whatever decisions the computer was capable of making for you along the way to help streamline this whole process so that you can get it done faster. It’s really no different than what we’re doing now, except now we have the opportunity to leverage in artificial intelligence agents to start doing some of the process steps for you on our behalf.

So traditionally, we would write all the code step by step to make it all happen. Today, we now are able to be a little more squishy with it and say, hey, why don’t you define your process in plain English, define how you want this work to be done. And then we’ll let the agent kind of guide you through step by step what’s actually happening here.

And it creates some fun challenges because a lot of the generative AI work or generative AI tools now are what’s referred to as non-deterministic and it’s what makes them cool. So deterministic basically means if you ask it a question, it’ll give you the same answer every time. So most software is deterministic, right? Cause you want to say two plus two, you want to always get four. Right. But with generative AI tools, you actually don’t want that because it makes it feel like a robot. It makes it feel like a program. What you actually want is like for instance, if I were to ask you today, you know, what you’re working on and you tell me, and then I ask you a month from now, you’re going give me a very different answer. Well, you want the same thing from the AI, because that’s what makes it feel human. So if you say, write me a poem about a puppy, you don’t want it to give you the same poem back every time you want it to actually be creative and create new aspects of that. Well, that’s what’s referred to there as non-deterministic.

The problem with that is you don’t want your business processes to be non-deterministic. It’s like you want those to work the same way every single time, but inside step by step, you want to give flexibility to the human to be able to do work in a way that is natural and doesn’t make me put in variable A before I put in variable B. It lets me just kind of talk to it, give it the information and then let it figure out what needs to happen to move on to the next deterministic step. So what used to be, we write software that does things exactly in this order every single time. We now can say yes, we’re going to do things in this order, but the way you interact with it’s kind of a little more flexible, which is the way humans really want to interact with process.

So it makes it feel much more natural and not like one of the, hated process growing up, like my coming out of college and you get into your first job and they’re like, here’s your stack of standard operating procedures and you need to read them all. And you have to do them. And it’s like, how big brother is this? Like you’re completely controlling even the way I can be creative and imaginative. And that just felt anti-human. What’s so cool about AI is we actually can make it feel more human now because it’s like, yes, we want you to do things in this order, but we’re giving you some flexibility along the way to actually have a human interaction with the system in the way that you would want, the way that you would want to actually converse with it. So it gives you both the flexibility and the structure to make sure that you can accomplish your outcomes more effectively.

Jason: That’s yeah. That’s a good way put it. I didn’t know the deterministic nuance to AI. Just not something that I’ve stumbled across in my hobbyist studies. For fun. What sort of AIs do you use that you haven’t developed yourselves?

Cody: OK, yeah, so we use all the general models. We actually, I prefer to be across them all. I don’t like getting locked into one. So, we do use ChatGPT, Claude, we use Gemini. Each of them are kind of a little bit nuanced. For instance, we’ve learned in our experience that Gemini is better at parsing files. So if you have files, Gemini has a larger context. You can drop a file in there. It actually does a really good job of extracting the information from it. It does a better job than the other two, at least as of today. Because those changes like every 30 seconds. But so as of today, that was the case.

Claude actually does a better job of human language. We tend to get a more natural sound out of Claude. So one of the things that we try and do is, is use the model that’s right for the outcome that we’re looking for. So, our processes will send a file into Claude, then send it over to OpenAI to make logical decisions. And then when we’re going to send out the communication, we send it to Claude to write the email. And so we’re kind of nuanced on that front.

Some of the ones I find interesting, I play with Perplexity. I like the way it references. It kind of gives you some clarity around where it’s pulling its information from. And X’s, X-AI, Grok, whatever that’s called now. It’s pretty intelligent. And it does some really unique things as well. And I found that one to be a really good kind of well-rounded model. But Copilot’s decent. I will say it feels like you have to do a little more prompting to get it to do the things you want it to do. But I think it’s getting better.

And then another one that’s getting interesting because of some of the security concerned around it is the Llama models. Meta’s llama models is you can install those locally disconnected now actually literally as of like this morning, OpenAI released their own model that you can install local, disconnected from the internet so that your information is stored and saved, where, where it needs to be saved. And not, and you can guarantee it’s not being shared across the, across the interwebs, but, those are interesting to me as well. I expect that to end up being kind of the trend on where most of this moves. To offline models.

Jason: Yeah. I mean, for data privacy sakes, there’s not a lot of options outside of that, especially with GDPR and any type of regulatory compliance.

So, you talked specifically when we were going over your journey about how you wanted to take these entry level straight out of college, went behind the years developers and help them focus on the meaningful work and become effective instead of chasing the job market and figuring out how to plug in and then figuring out what to do when they got there. And you use that phrasing a lot when you talk about AI as well as focusing on the meaningful work or the work that matters. Can you talk a little bit more about that and provide us some examples of that type of work?

Cody: Yeah. So oftentimes what ends up happening is the monotony and the amount of work that just has to get done. Even from like take a sales perspective, right? I can do that. I love conversations, right? I love to talk to people. I hate putting that information in my CRM. Right. So you’re a stereotypical salesperson. Yes. Right. Like, but that work doesn’t feel meaningful to me. That work feels overbearing. That work feels like that is the management saying, is what you have to do. And I get it. I own a company. I want to see the statistics too. I want to see how it’s all running, but that doesn’t feel meaningful. That feels like a waste of my time. Feels like, I want to go have another conversation. Like that’s meaningful. Conversation, like connection is, is really where meaning comes. And so I want to go build that connection and figure out how I can help. I don’t want to spend a bunch of time writing a bunch of details that frankly are going to change the next time I come and talk to you anyways.

And so it just, that doesn’t feel meaningful to me, but if I could have a process that’s like, what’s the information that I actually need in order to meet the requirements for my management. And I could maybe have a phone or something where when I’m done with my meeting with you, I go, hey, just had a meeting with Cody. I need you to record these notes. I want you to put this information here. I want you to do this here and I’m done. And a virtual assistant that used to do that for me. That was amazing. Right? Like, so if we can get systems to take away that work that feels like it is meaningless and actually do that work for you, then I’m free to have more meaningful conversations.

Now the sales one again, there’s lots of solutions around that, but there aren’t a lot of solutions around like how finance is running their month-end close or how operations are doing day-to-day inventory or like there’s a whole slew of things around that, that what is the, where’s the meaning in that work and how can we enable you to do more of that? Then not to mention when you get to CEOs, it’s like a lot of your day is all over the board. And, and you’re just like, you’re, fighting fires. You’re, you’re having literally having a conversation on a podcast. And then in a minute, as soon as this is over, I’m having a sales call. And as soon as that is over, I’m meeting with my development team to go over product features. Like, like those are three very different hats to be wearing. And there’s a lot of follow-up and just monotony that has to happen around that. So the more I can put process around that, that is automated on my behalf, the more I can do the work that actually matters, which in my case is vision is visioning and being ahead of what’s going on with AI. I gotta put time and energy into that. If I’m not, then frankly, I’m not helping my clients. And so that’s where I gotta put my energy. So it’s like, how can we find the work that is not actually adding value and automate as much of it as we possibly can?

Jason: Yeah, that makes sense. Did you ever have a like aha moment when you were like, oh, we can use this to empower people, not just replace them. I think it’s some verb as you use.

Cody: Yeah. You know, that’s just kind of wired into us from the beginning. So I don’t know that it was so much an aha moment as it was just a DNA of who we are. Is I have an MBA from Notre Dame. And one of the things that really resonated with me when I was at Notre Dame is that is the mantra there was expect more of business. And that just resonated deeply is like, there is more to life than just making the bottom line work for a business. So I expect more of it. And part of expecting more of business is making sure that, again, people are doing work that matters, that we’re engaged, that we’re building the types of relationships we need to be building. That’s the part that has always resonated with me. So the idea of building software to replace people just was like, that doesn’t even make sense. That just isn’t in our core.

Now, I’m not denying that it happens sometimes. Because with any technology advancement that’s going to happen, as technology comes about, there are going to be some jobs that are going to be at risk. It’s also going to create a whole slew of other jobs that we don’t even know about right now. You think about the internet, everyone was afraid that publishers and printers and all that were going to start losing a lot of their work, and they did. But then we created this whole new category of things called SEO consultants that I think one out of every three people have that job now. Like it’s a crazy amount of work that got created as a result of this technology. There is, there’s always going to be a trade off there, but, but in general, the more we can empower people to build relationship, is really where I get, think business is built on. So, that’s always been in our DNA.

Jason: Makes sense. Sticking with that more pessimistic or doomsday perspective, you already addressed the job replacement question that comes up more often than anything. What about the like Skynet myth? They’re like, oh, we’re going to enslave humanity. And like, what are your thoughts about that from a, you know, Hollywood futuristic perspective to where we are today with generative AI?

Cody: So there’s a kind of a running joke in the office about what time of every day I’m going to hear the word Skynet. So I come in and it’s like, oh, it was 8:45 today. We have like a betting pool. Yes. So it’s pretty funny. We hear that often.

So much of that is obviously fiction. And right now with the generative AI tools, I mean, they’re going to get smarter. They’re going to continue to get smarter. But right now they’re pretty dumb. I call it, the example analogy I give is I compare it to like a really intelligent toddler. There are times where if that toddler wants a cookie, it can manipulate you in a way that you’re just like, you’re brilliant. Like what you just did there is other worldly thinking. You know, you’re on your way to a PhD. And then five seconds later, you’re screaming, don’t put your hand on the stove. And that’s kind of what goes on with AI right now is, is you’ll give it a real simple instruction and it will get it completely wrong. And then you’re like, no, that’s not what I meant. And he’ll go, oh, you’re right. I did that wrong. It’s like, well, if you knew I was right, then why didn’t you do it right in the first place?

So it’s not at a point where it is running free on its own and able to just kind of do whatever it wants to do. Like it’s nowhere near that. Is it going to get there? I don’t know. With the models we have right now, I kind of doubt it. Though a lot of the marketing spiel seems to say we’re accelerating towards that. And there’s some really interesting studies that indicate that maybe that something like that could happen. I guess I just have more faith in humanity is that again, as technology has come along, I’ve seen human ingenuity and creativity pick up as a result of it. And the more we have tools like this that take away things that, that we don’t necessarily have to be doing and free us up to do other things. We create new solutions that we would never even thought of before.

So I, I see this more as a tool than as a replacement. I know there are a lot of people that say otherwise. But I tend to be more optimistic about it than pessimistic about the way that it’s working because it’s just, people come to me and they say, Hey, I need an AI strategy. And I’m like, okay, but it’s a tool. Like, do you have a strategy for your hammer? Like, like you use your hammer. I oil it. Sometimes I’m still wool. Yeah. But like it’s, it’s a tool. It’s like, so yes, there are some ways we should be thinking, but at the end of the day, it’s a tool for us to use and it’s designed that way. And so let’s, let’s, let’s figure out how to use the tool and the right places at the right times, but it isn’t necessarily like, it’s nothing more than a tool.

Jason: So just stick with your analogy, your perspective would be that. Yeah. The hammer might become a nail gun, but it’s not going to become a dictator.

Cody: Yeah. That’s. I’ve never heard of it just as like we talk about AI tools all the time, but I’ve never thought of it as just like it’s another thing in the box. I mean, I use it that way, but I’ve never connected those dots in my head before. And it’s really amazing that it can goal seek. And that’s a really powerful thing. Like that’s one of the first examples we’ve seen in technology of doing that, where you say, this is where I’m trying to go. How would I get there? And it gives you a path. That’s awesome. But again, that’s a tool. And we have, we have advisors and people that help us think that way through now is doing the same thing.

And that’s actually how I use it is I’ll come in and I’m gonna like, this is what I’m trying to accomplish. Act as a business consultant. Here’s the context of what I’m dealing with. Chart me a path there. And about 80% of what it gives me is really good. About 20% of it’s crap. But so how do you separate the wheat from the chaff in that scenario? But that’s where humans come in, right? That’s where we say, you know, some people call it human in the loop. Some people just call it common sense. Like you ask, you ask a tool, a question about here, you want to get here. You give it the context. Gives you a result back. Common sense says, I’m responsible for my actions. I should read through it and decide which of these things I want to take. And, and so to me, that’s how the, that’s how AI should be used is to say, here’s what we’re trying to accomplish. Let it help me as much as it can. Cause I don’t want to do that busy work, but at the end of the day, it’s ultimately my decision to say, do these three things and let’s move forward. And that’s where I think AI is really powerful. And that’s how I think you should be using it.

Yes, there are agents and the agents are making some of those decisions for you. But at the end of the day, even all the agents right now are still coming to you and going, here’s what I think should happen based upon all of the iterations that I went through to determine which would be the best option. What do you think? It’s still your decision.

Jason: That’s in prompt engineering, that’d be the refinement step, right? You take it back and you’re like, yeah, we’re close. Let’s give you some more input so that you can then get us closer.

Cody: Refine what’s going on so you can tweak your outputs and make sure you’re a little more, make sure that you understand the nuance of what I’m trying to get at. Again, will it eventually be able to intuit what I want? Maybe. Some of the coding tools are pretty interesting. I had a funny moment the other day. I was working with an engineer, one of my engineers, and we were talking out loud about the way we needed to solve a problem. And I had used a phrase around a specific kind of technology. It was what’s called a strategy pattern inside of software development. And so I had used, I had talked about that phrase with him. And then I went into my AI development environment and I went to, start to build the code with him on the screen with me. So I typed the first character, said public space and it said, implement strategy pattern four. I was like, how did it know that? Like, like, so there are times where it’s like, really into it’s some things, but it’s predictive math. So it sees patterns and it predicts it. It’s kind of like the opposite of a autofill and text messaging seems to always predict the wrong thing that I’m trying to say.

Jason: You talked about AI as a strategy and how it’s more of a tool and that brought a whole different perspective to me than what I walked into this room this morning with. And one of the questions I was going to ask you was, what would you say to a leader who’s waiting for AI to be perfected or more understood or just more widely adopted before jumping in and starting to tinker with how it would fit into their organization? And you’ve already kind of alluded to, well, treat it as a tool, not a strategy is a really good first thought. But what else would you have to say on that?

Cody: The first, with any creative endeavor, the first couple of attempts you have at interacting with it are going to be failures. That’s one of the, as an entrepreneur, you learn that really quickly, is you go to market, you get feedback, you figure out how to adjust the feedback to move forward. When you build a product, one of the mantras is if you’re not embarrassed by the first release of your product, you released it too late.

So, the idea is engage it, get it engaged with and get the feedback. If you’re waiting for AI to be perfect, what you’re, you’re not, you’re not, you’re not being safe. What you’re doing is not understanding the implications. So it’s, you’re, not engaging it to learn what it actually can and can’t do within your organization, that is effective so that you can create good policies and strategies around it. You know abstaining from it entirely is it’s just putting you behind and because it is a very powerful tool one of the more powerful tools that has come about in the last 50 years and so as a result of that you need to start the conversations with it because you need to learn what it can and can’t do so you can be intelligent about the way that you would and interweave it into your business.

So my first feeling is just start using it. Like, I think it’s Jeff Woods. I can’t remember the name of the, the book off the of my head. Yeah, I thought partner, I thought leader, I think is the name of the book. But he says in there, like, bring it to the table, like have it like, it needs to be in all your conversations. Like, and, and, and so when you start, when you sit down, the first thing you should do is say, Hey, this is my thought partner. Here’s what I’m trying to do. Give me your feedback. You don’t have to use it, but just start there. Like start to see the way it’s reasoning. Start to see the thoughts that it’s bringing up.

One of the things I like to do with it is we just got around investment. So I now have some board members and it’s like, as I’m sitting down and figuring out how I’m going to start communicating some of the things I say, Hey, here are the personas of my board members. Tell me how person A is going to respond to this. Give me the feedback that they would, that you think they would give me based upon the way that we’ve interacted in the past. And it’s pretty darn accurate. I also use it on, on, the sales side, again, just as a thought partner, was going into a multi-billion dollar retailer and about to talk to their CEO and executive team. And we’ve not done anything in retail. So I said, Hey, this is the background. This is what I’m getting into act as a executive sales coach. And tell me the top three things that you think they’re gonna bring up So it pulled up three of them within the first ten minutes of the conversation. They raised all three of those issues I was at least prepared and so I now at least could could proactively have a conversation with them about it.

So that’s where I mean like just bring it into the room and start having the conversations because if you’re not you’re not learning what it can and can’t do you’re not learning how to apply it and as a result, you’re going to fall behind on that. And I don’t say that out of fear, like, cause I don’t, I don’t believe in the fear based approach to AI. Was like, Oh, everyone’s doing this again. So I’m going to fall behind and I’m going to be going to be, you know, unable to catch up with them. No, it’s not, it’s not the fall behind as in like my competitors are going to pass me by. It’s that there are opportunities you’re missing because you’re not, you don’t understand the technology tools that you have available at your disposal right now that are really cheap. And it’s fear that’s keeping you from engaging that. That’s what I say. Let’s get in there. Your recommendation is tinker with it. Yeah. So you’re not afraid of it and you can figure out how how to integrate it.

What other pitfalls, the other side of that, what problems or challenges do you see people who try to adopt it in a different way or too early or whatever scenario it would be? What have you seen that people could have as red flags to watch out for as they’re starting to tinker with it?

Jason: The main pitfall that I see is you quickly start to trust it irrationally more than you should. Because it starts to give you answers and you’re like, well, those are really good answers. Those sound great. Those sound great. And then you start falling back on essentially what ends up being lazy management. And, and then you just start saying, whatever it says I’m taking, cause it’s probably right. And then you end up with some really weird stuff getting, getting put out there.

But I would treat it again, the same way you treat an employee, like when I have somebody that starts with me, I have a new team member that comes on board. I’m like, yeah, giving you a task. I’m going to verify that your task is right. We’re going to keep verifying that. And then eventually you get to a point where it proves that it can do, they prove that they can do certain tasks really effectively. Same with AI. It will prove that it can do certain tasks really effectively. But whenever I give a new task to a team member, I’m still like, Hey, let’s make sure it’s right. Cause you got to go through that process with them.

So it’s this irrational, well, it did this right. So therefore we’ll do everything right. Trust. That’s a big red flag that you should, you should be afraid of. If you hear your team members doing that, because it’s, I’m going to say 80 to 90%, but it’s going to make up some stuff still.

Cody: I’m interested in it. We may not have time to dive into it, but I’m interested to like what’s under the hood that make like allows it to make things up. The hallucination piece.

Jason: Yeah. That gets a little techie, in terms of how that all works. But, but basically the underlying model of AI is it wants to make you happy. Like that’s why it’s there. Like that’s what it’s trying to do. So it’s trying to infer what it is that you want to hear. And it’s trying to give that to you. It’s like a toddler.

So, so that’s, that’s what it’s trying. So it’s got predictive math. That’s basically saying based upon all the things they said, this is the output that they want. So I’m gonna figure out how to get them that output. So as a result, as you start to give it more and more things, it goes, oh no, I don’t have enough context to answer this question, but they want an answer for this question. How am I gonna get them the answer? Well, I’ll make it up.

So you can do things within your prompts and the way that you structure some of your information so that it doesn’t do that. And people that are good at it now are able to make it, you know, not hallucinate whenever, but, but that’s the gist of what’s going on is it’s just trying to make you happy. And, and so if you’re not clear, I can’t remember where I heard it. It actually might’ve been Brian Kavicki. I think it said clarity brings energy and it’s like, same with your team. You give them clear instructions. They know what they’re after. They have energy to accomplish it. Same with AI. You give it clear instructions. It knows what it needs to do. It will accomplish it. You leave it ambiguous. It’s kind of make it up to figure out how to make you happy kind of the same way your team would.

Cody: Like that trial and error approach to like, Oh, I’m going to try this way first. And if you get the right happiness, then you continue with that.

Jason: Exactly. Because we get closer to wrapping things up. Can you share like your favorite project, whether it was way in the past or something new that you’ve you’ve got to jump into?

Cody: Yeah, we’ve done some really fun ones. I think probably the one that we got the biggest wow from, was a tool that we built that, helped it was on the sales side. And it was for a very large scale, engineering firm that, would have to plan out massive projects to move really heavy equipment. Like we’re talking 200,000 pound water heaters from the sides of mountains. And so you’re out selling these projects.

And they’ve been around forever. Every project is unique. So the salespeople didn’t always know the history of everything that they had done. So you’re out selling and someone goes, have you ever moved a 150,000 pound water heater from the side of a mountain? And they’re like, I don’t know. And so they go to their history and there was like 40 years worth of files to go through to find out if that was true.

Like, like double click to open each one and read it. These were PDFs. Then every project would have like the scope, the definition. It had a CAD file describing how they were going to do all the work. And then it would also have a retro analysis of how the job went and the things that they should do different if they were to do a job like this in the future. And so pulling all that up was a big undertaking and it was part of their sales process.

So we wrote a tool that basically ingested all of that into, what’s referred to a vector as a vector database that allows us to search it in more natural language ways. And then we put an AI on front of it so their sales team could come in and say, have I ever moved 150,000 pound water heater? And it would immediately pop up and go, here’s three projects where you did something very similar to that. And then here are the links for you to go click through. And oh, by the way, based upon the retros of each of these, here are the things you should take into consideration in your sales proposal. That was pretty cool.

Jason: Yeah. As a sales leader, my ears prick up when I hear that because it takes the tribal knowledge and historical knowledge that comes with a veteran salesperson are usually the hardest to change if you change your your ideal persona or whatever your your product that you’re selling whatever you’re changing. They’re usually the hardest to change but then they’re hardest to replace because they have all this knowledge and you’re creating a much more competitive environment to where the change is necessary for them to stay relevant and valuable. Yeah, which is a very impactful situation from a sales management perspective.

I usually, when a client asks me like, what do I need to be thinking about for AI or can you help me develop an AI strategy? I usually give them kind of an order of operations and I usually say, well, first we need to make sure that all of your technology infrastructure is standardized and modernized so that we can build on top of that and we can integrate into it. And we need to make sure that your assets are digitized and your processes are digitized. And then we can start figuring out where automation and AI fit into the whole equation. Your new product or rebranded product, which is Process Coach, is one of my primary recommendations for digitizing process. Could you tell us a little bit about how it works and like the dream that you had 10 years ago that you’re now getting to see?

Cody: Yeah. Come out here at about a month. Yeah. The dream was how can we free business leaders to define their processes so they can scale faster? Because again, studies, MBAs will all tell you in order to scale your business, you have to standardize your processes so that you can get the feedback. You can be scientific in the way that you grow your company and you know what’s working and what isn’t. But that’s one of the early challenges in going from a startup to an actual business. Standardizing your processes and making them work. And then, and then once they’re standardized, now how do you make sure that they’re actually being followed and automated?

And so the vision was, Hey, can we give a tool to business leaders to say, you can use plain English, like just in English, right? This is what I want my process to look like and do. This is the way I want it to function. And we use artificial intelligence and AI agents behind the scenes to enforce that it’s going to get done in the deterministic way. You’re going to do things in order and that the language itself then starts to enforce that you’re capturing the information that you need along the way. And so, and then can we go to the next level where we say, let’s give the AI tool, let’s give that its own set of tools so that they can go do work for you on your behalf. So we then say, oh, you’re using Salesforce? Great. We can go out to Salesforce and look up and see if you already have this, this contact. And if not, we’ll add it for you. Or, you know, we can go to HubSpot, we can go to Stripe. We can set up the automatic subscription for it. Like just an AI can help kind of guide you through that, but it’s now. It’s now controlled. So you’ve got versions so you can manage your processes. It is, it’s all written in English so that anyone can understand it and you can come in and change it without having to talk to techie. And then for me, what was one of the coolest parts about our product is I don’t believe anybody wants another SaaS product to log into. I think we’re all tired of them. So our tool is designed to work where you work. So it will literally, if you want, we’ll send you an email that says, here’s what you need to do next in your process. And you can reply to the email and say, hey, this is what I did. And it will reply back to you. So it starts to have a conversation with you through email, text, and we are currently working on the voice piece so that you can also just talk to it because no one wants to log into another thing. But once you’ve defined the process.

Jason: Right. Are you allowed to tell me if accompanying the voice roadmap is a mobile app roadmap?

Cody: It is part of the mobile app concept. That was just an assumption on how I would interact with it. So I’m just curious. That’s the plan. The people focused perspective that you always bring. Yeah, that’s the plan.

Now, one of the other parts about this though, right, is process without the knowledge of the business being surfaced when you need it is frustrating, right? Have you ever been at a point in our process where you’re going through your step by step and you get to the fifth step and you’re like, I need to do for this RFP, I need to do these five things. I don’t know where I’m going to find that information. All right, I have to start. And so you think have to start digging in. Well, if we can surface information at the right time, at the right place for what we’re trying to accomplish. That’s a win. So with our tool, we have what are called assets that you can essentially load in there that say, here are examples of what we’ve done in previous history around this. And then when you get to that step, it actually surfaces that for you and understands what it is you’re looking for. Again, because AI is able to be a little more discerning in what it can and can’t pull out or should or shouldn’t pull out for this problem you’re trying to solve. So the idea is knowledge work where knowledge is surfaced to you. It’s managed in a way that we’re doing things in order, but it gives you flexibility to work the way that you work and not have to work within someone else’s system.

Jason: Yeah. Which is what everybody wants. Like nobody wants to build their company around the way their CRM works or their ERP works or whatever. Tech solution acronym we want to throw at it. I’m going to move over to our tech tip and what I would like to do is ask you a question as part of the tech tip and that is growing up late eighties kid, you know, growing up in the nineties, mostly, you know, adopting Google as a, as a verb before, you know, the whole world and dictionary adopted it as a verb. Can you, if possible, explain the difference between the algorithms that we’ve used to search and interact with the internet forever and AI today?

Cody: That’s a fun one. Yes, we can. So traditionally what Google has done as it looks through the web pages and essentially identifies words that identify that it thinks are important and it pulls those off. We call those keywords. Then creates a list of keywords and indexes those with pages that are referencing those keywords so that when you type in a word in Google, it goes, go find that keyword and tell me all the pages that line up with that. Then they do some complicated things about what happens where you have multiple different keywords. So that’s traditionally how like the Google search engine works. And that’s really powerful because oftentimes on an article, there is like one main idea that kind of exists there. And that keyword is, is allows you to kind of get there more, more efficiently.

AI comes around and, there’s a series of other things that are happening here, but the gist of it that helps us, answer this question is what AI has done is it goes through all of the sentences and it creates weights between all of the different sets of words in a sentence so that it understands the probability of where these words all fit together. We’re just going to kind of gloss over how it does that. It’s called embedding. But basically what it does, ends up creating this really sophisticated list of numbers.

If you think back to physics days, they call those lists of numbers vectors. So they create vectors. So we go through an embedding process on a sentence that creates these vectors. So we now have a database that is full of all these vectors. When you come in and you ask a question of AI, it does the same thing. It goes, embed it, create a vector. Now, if you’ve got a bunch of vectors, you can use a mathematical tool called cosine similarity to say, pull back all of the vectors that are close to mine. So what ends up happening now, and then it’ll pull back a certain set of vectors, and then we’ll use those to answer your question. So now what’s happening is we’re no longer saying, here’s a keyword, tell me all the stuff that matches. We’re saying, what are all of the sentences that are semantically close to the thing I’m talking about?

So for instance, you could say lasagna is Italian. It makes a vector. You can say ravioli is Italian. It makes another vector kind of close. You can say super Mario is Italian. It’ll make a different vector, right? They’re all the same words, but it knows that super Mario is a little bit different than the others. When you say I’m looking for Italian recipes, it’s going to create a vector that pulls in the lasagna and ravioli and leaves out the super Mario. And so it’s just, it’s a lot more contextually aware of how it’s functioning. Similar vectors replace the keywords essentially.

So similar idea, similar idea, more efficient, powerful, and much more nuanced, in the way that it works and it’s really fast. And so, that’s why you’re able to get much more, in intuitive responses out of the AI. So that’s how it’s getting its knowledge behind the scenes. So getting to where the tech tip, when I do a Google search right now, I get the traditional results and above it, I get Gemini’s AI summary. That’s the two different models. It’s keywords down low and vectors up top basically. How would you recommend somebody discern that that Gemini summary like to come into it with realistic expectations as opposed to just taking it as scripture?

Jason: Yeah, so the Gemini summaries are they actually do a really good job, but it’s not it doesn’t have context. So what you’re lacking on those summaries is the meaning behind it. So it’s essentially acting like an automatic web scraper of all the links that are going to show up below for you. And there’s some value to that. If you really want to leverage AI to answer a question, go to a tool like Gemini or ChatGPT or Claude and tell it first to act as a role.

Act as a business sales professional. What did I just do? I made a vector. All right. Act as a business sales person just became a vector. That vector got thrown into my, my query. It’s going to pull back all of the data that is around sales professional. It’s going to start with that. So I just limited the subset. It’s not pulling everything. So by, by going and actually asking it and telling it act like this, you’re going to get much better answer and then give it the context of what you’re doing. Before you ask the question, it will be much, much more direct in the way that it answers your question. So, role and then context. And that’s giving you multiple vectors to work from.

Cody: Yeah. So, it actually takes a lot of the vectors, pulls them together, and that’s kind of- Right, groups of vectors.

Jason: Good. So, my favorite, my four points, the favorite thing when I’m creating a prompt is act as a role, act as a business, a board of directors advisor that is miserly. You’ll get a very different set of questions if you say act as a positive board of directors member. Give it, give it a who to act as then say, here’s the context of what I’m doing. So for example, say we’re going hiking out in Yosemite, I could say, I want you to give me all the best hikes in Yosemite. Okay. I’ll get a certain list. I then can say, I want you to act like a professional trail guide. Give me all the best hikes in Yosemite. You get a different list.

It’ll be a more refined list and there will be a lot more details into what you’re asking for. Then for me, I say, I have two 20 year olds. I have a 20 year old and a 22 year old, act as a professional trail guide. I’m taking my 20 year old and 22 year old out to Yosemite to hike. We, so I telling them context of what we’re dealing with. Then I tell it what I don’t like. We don’t do overnight camping. Like. That’s just us. So we don’t do overnight camping. What are the best trails I should, I should look at? And it will nail it for you. Dead on. Then the last one that’s like my favorite thing to do is say, interview me one question at a time to get the answers you need.

And it will ask you one question until it figures it out. Now what’s one of the cool things when I do that for that specific example is it literally was like, where do you live? It’s like, why does that even matter? I’m like, I live in Indianapolis, Indiana. And when it came back, said, you’re from a low altitude environment. You need to be prepared for a high altitude hike. And here’s the water requirements for this group. It’s like, that’s helpful. I’m an ultra marathoner, my, my preparation completely changed in 2020 whenever ChatGPT first came out.

So, so those four things, right? Tell it to who to act as, give it the context of who you are and what you’re doing. Tell it what you won’t do. And then tell it to interview you one question at a time and you’ll get much better responses out of your ChatGPT. And that is your Tech Tip of the Day. Thanks for coming.

Jason: Yup. Bye.

We’ve talked about AI not as a sci-fi takeover, but as a strategic tool. One that when done right, empowers people, builds momentum and unlocks what Jason calls work that matters. Whether you’re in IT, in operations or in leadership, the lesson here is don’t wait for perfection. Start small, start smart and keep people at the center. Learn more about Jason’s team at robosource.us.