Total Disruption Podcast

Input = Output: Why Your AI Results Suck (and How to Fix It) | S3E21

26 min
Apr 23, 2026about 1 month ago
Listen to Episode
Summary

Jake Denning and Mike Labersi discuss how to maximize AI effectiveness by understanding the principle of 'input equals output'—the quality of prompts directly determines result quality. They share practical applications of AI tools across their business operations, from content creation to app development, and introduce their new communication app launching May 13th that uses AI for conversation practice and role-play training.

Insights
  • AI effectiveness is fundamentally dependent on prompt quality and iterative refinement; users must actively train AI through feedback rather than accepting initial outputs
  • Organizations can deploy AI as specialized 'employees' with defined roles (ideation officer, research director, CFO) while maintaining human oversight and governance
  • The gap between AI capability and user expectation stems from insufficient time investment in learning and experimentation—10 minutes daily can yield significant productivity gains
  • AI auditing (running multiple AI models against each other) validates information accuracy and reduces hallucination risks in business-critical applications
  • Effective communication training via AI role-play addresses a critical leadership gap by building confidence without requiring hundreds of real-world practice conversations
Trends
Enterprise adoption of closed-source AI tools (Otter AI, Zoom) for risk-averse organizations concerned about proprietary data exposureShift from AI as automation tool to AI as augmentation layer for human expertise and decision-makingGrowing use of AI for internal R&D, content adaptation, and personalization at scale without dedicated technical staffEmergence of multi-model AI workflows where different specialized models validate and improve each other's outputsAI-powered communication and leadership training as alternative to traditional scripting and high-volume practice methodsNo-code/low-code app development enabled by AI coding assistants reducing dependency on specialized developersBrand voice consistency challenges as organizations integrate AI-generated content into customer-facing communicationsEthical communication frameworks emerging as counterpoint to NLP and psychological manipulation concerns in AI sales tools
Topics
AI Prompt Engineering and Iterative RefinementInput-Output Quality Correlation in AI SystemsAI Role-Based Deployment in OrganizationsMulti-Model AI Auditing and ValidationNo-Code App Development with AI AssistantsAI-Powered Communication TrainingBrand Voice Consistency in AI-Generated ContentClosed-Source vs Open-Source AI Trade-offsAI for Content Creation and AdaptationLeadership Communication and Difficult ConversationsSales Communication Without ScriptsAI Governance and Human OversightEthical AI Implementation in BusinessAI Tools for Research and IdeationMeeting Transcription and Summarization Tools
Companies
OpenAI
ChatGPT discussed as research tool and prompt engineering platform for various business applications
Google
Gemini AI tool mentioned as preferred research and validation tool owned by Google
Otter AI
Meeting transcription and note-taking tool discussed as closed-source alternative for risk-averse enterprises
Zoom
Zoom's native meeting transcription and note-taking features discussed as alternative to third-party tools
Opus Clip
AI tool used to extract quotes from episodes, create clips, and schedule social media postings automatically
GenSpark
AI ideation platform used for whiteboarding, slide deck creation, keynote development, and logo design
Replit
Coding platform used in multi-model AI auditing workflows to validate and compare AI outputs
People
Jake Denning
Co-host discussing practical AI implementation strategies and communication app development
Michael Labersi
Co-host discussing AI adoption challenges and communication frameworks for business leaders
Quotes
"Input equals output. If you put a junk prompt into basic AI and you get junk results, it's not AI's fault. You didn't ask it the right question."
Jake Denning~22:00
"Bad questions get bad answers. You've got to take time to train the AI because if you just take its initial result and you're like 'ah that's good enough' copy and paste and forward, you just train the AI that that's what you're willing to settle for."
Jake Denning~28:00
"I can build SaaS integrated applications almost faster than the coders can because I've got multiple AIs that I'm using."
Jake Denning~35:00
"Everything boils down to effective communication and everything boils down to practicing your communication."
Michael Labersi~55:00
"There's a huge gap in leadership, customer service, HR, and sales—people don't remember scripts. It doesn't work and it feels salesy because you have to be in a ton of sales engagements to actually master what you're saying."
Jake Denning~48:00
Full Transcript
Thank you. Dr. Clendang, how are you today? Good, V. Michael Labersi. How are you today? Outstanding. I'm still feeling spry. You guys picked us up on our last episode, which, by the way, you're back on Total Disruption. Welcome, everyone. We have a great episode lined up today. Jake's looking good, feeling good. I'm always looking good and faking like I'm feeling good. So we're here. And I want to talk, I want to dive in today to the world of AI and for practical and personal use, right? Like business a little bit, but I feel as if people in their every age group right now, there's probably 8% of the population that, and you're not in the same percent, 8% of population that's kind of using AI as an extension of like a Google or a search engine, which is fine. It's totally fine. But let's you're doing some cool things and you did a presentation about a month ago on a boat and where will be? Long Beach. California. And for a bunch of TV show hosts we were out there educating them on just how we do things, disruption. And AI is a great lever. And merging the world of AI, using it, having a broader spectrum of the tool, whether you decide to go into the far reaches of the tool or not, it's up to you. But as communicators and educators, it's our job, Jake and Mike, and Total Disruption, it's our job to bring people current information, ethical information, right? And make people better communicators. So let's go, man. Let's talk about how you got on the path of really getting into AI and using it. We'll call it AI 2.0 expanded, not just as a search engine. How did you bridge that gap? Well, I guess for me, it happened a little more organically over time as I had need of certain things. Um, so I started, you know, and I, I would love to say I was totally purposeful and I mapped this all out and here's my plan. That's not how it went. Um, it, it started with, yeah. Was I using chat GPT, um, the basic AI as a search engine? Yes, I was. I was totally doing that as everybody was when AI first came out because it was a little bit more, a little bit different driver of what it could find. And then as we built it out, we started, and Mike and I both use this, but I started asking it some different questions, which then prompted some different questions from me. And I paid attention to that. So now, you know, jump ahead a while. I mean, when AI first came out and we touched on this in a previous episode, Mike and I tried to build an AI voice, basically a voice agent that would do outbound calling. because that's what we all thought AI was going to be the best thing for was turn it into a person and it could call other people and it can convert sales. At the time, it was not that thing at all. And honestly, I still don't think it's that thing. I think it can be used for a lot of other stuff. But that's where it started in that particular spot. Right. Yeah. Oh, can I add something? Yeah. No, absolutely. For those of you that don't market on the Internet, right, This is the kind of formula typically really cool internet marketers use. You can put an ad somewhere, social media, magazine, billboard, anywhere. It prompts a trigger to go to the next step. The next step is go to another webinar, a page, something or other, exchange information, get something. When you exchange information and you get something, someone then gives you a call. and we were trying to utilize the marketing for that piece to get people to the next step, right? So we were trying to just do a thousand cold calls using AI to try to get people to talk to someone live. So where we thought leverage-wise, we were going to be able to pull a big button. Back in the day, you had to rent a list and then sit in front of a phone, hire call centers, hire closers. It was a whole big thing. We thought we really could reach a bigger and broader audience because we're honest people with ethics. We just needed to get past the, hey, let's get to know these things. We're going to leverage the AI that way. And that's why we thought, wow, this is really genius. So go on, Jake. Yeah, 100%. And we totally thought that. We threw a ton of time and a ton of money at it to crash and burn it because the technology was not ready for what we wanted it to do at the time. I mean, we were trying to force something that wasn't quite there yet. Might be now, but we're just on a different path now. We've marched down a different way. So I've used AI for a bunch of different stuff. I used it as a editor's research assistant. At one point when I was writing my new book, I was using it to go back and research some dates, pull some different quotes for me to use. And then I used it as an editor. When I'd finished the writing, I'd let it go in there. hey, I want you to grammar check this against it and see if it came up with some different sentence structure stuff. Worked out great for that, right? And that's some basic, basic things that you can do with it. They're not still using it as force. Whereas now we've actually built it out such that we have AI as employees. Literally, they have a role within the company that these particular different AIs function as. So like I use, we don't get paid by any of these guys. let be really really clear they just tools that we currently using that are working cool we use opus clip opus clip is a cool one because it take our episode it AI search some cool quotes that we have it cut them into shorter clips and it schedule the social media postings for us So that works out beautifully We don have to play with it. And we just have it directed very specifically of what we want and what we want it to do. I use GenSpark largely as an ideation format. Hey, I want to build this thing. tell me what you think. What if we did this? What if we did that? And I'll use it as a whiteboard just to get stuff out of my head and move forward on some different things. Recently, we built an app that used to require coding. You had to have an app developer and you had to understand coding. Now with AI, I have one of my AIs that's CodeWriter, that literally that's what it does. And then I can use an app building program and I can use the two together, tell the what i wanted to build drop it over into the app say that didn't work this is what i want it to look like back and forth and we built this very cool communication app um which by the way short plug may 13th we have a training that we're doing go to total-disruption.com sign up for the training you're gonna get promo pricing on the app there i said it come check us out may 13th um so i mean i use it for a lot of design and ideation we use it for a lot of research basically we have an r d department that is run by AI. We go to that particular page. When I'm doing speaking proposals, because we do a lot of speaking, I mean, we get in front of people in either podcasts or on stages and different things. When I need to do a speaker proposal, I can take the standard formatted keynote speech that I've written and I can plug it into that and say, I want this adapted and I want the slides adapted to be able to be proprietary to this particular event. And it'll change some of the language a little bit to speak to that particular whatever it is like because you can't you can't talk to health care people the same way you talk to human resources people so you've got to change some of the language in there and I just ask it to do that so I've got a I have an editor that is AI that does some of that um and just a bunch of other stuff and we've got but I mean I did the presentation on it of here's who this is like for the most part gen you know I have a director of operations that's a live person. I want to be really, really clear. She's an amazing individual. And she oversees a lot of our AI that I'm working with. And we've got, you know, Jen Spark is our chief ideation officer. And we have a chief financial officer that's running AI. And we have a chief research officer. And we have all these different roles that are being fulfilled by AI, but we still have humans managing it to make sure, you know, we talked about this, and this is one of my favorite things that's come out of this recently, input equals output. You know, and Mike picked up on this, and there's just something that I said offhand because I thought it was kind of cool, and Mike brought it back, and he's like, no, man, that's the thing. That's totally what you've been doing. And it was awesome that he caught that, and that's been a big piece of this. If you put a junk prompt into basic AI and you get junk results, it's not AI's fault. You didn't ask it the right question. Bad questions get bad answers. Well, bad questions get bad answers. As you roll through the evolution, let's say you're somebody that wants to start using AI, and you want to start to expand the capabilities of AI. I mean, I suggest you give yourself probably a good 10 minutes a day to start working yourself through these questions, right? Because sometimes you might spend two and a half hours doing things, Even though that two and a half hours would have taken two and a half months or weeks anywhere else. Okay, this is what I want to do. Send it off. Get it coded. Go. Crap. It doesn't work. Send it back. Get it coded. Three, four, five days, right? Not two, three hours. Might sound like a lot of time. But when you're in the business world and when you're juggling between I don't have a true person I'm paying 100 grand a year down the hall or at a phone tip saying, hey, I need this by the end of the day. And it's a coder, a programmer. And you're trying to get things done, deploy things and launch and sell and do everything else. You need that, right? You need an AI. You need some tools and you need to spend some time with it. So if I'm hearing you correctly, the evolution came from, oh, we want AI to do as a business tool. You went from search engines and now trying to then figure out how to program it more to do things. And as you went down the path of trying to program it, there came out all these new models. Right? Is that kind of how it happened? Yeah, 100%. And I will tell you, a lot of this came from, not a lot of it. I mean, some of it came from, I coached some different large organizations that are, quite honestly, I'll tell you, they're terrified of AI. They're just scared of it. And they're largely SaaS organizations. And they don't want to use open source AI because they're concerned that open source, and I'll give them this, they're concerned that open source is going to take their proprietary information and make it available to the world because they're trying to cling to that particular information. All right, fine. I don't agree with it, but that's just what it is. And their ideology of using AI is it's got to be closed store stuff, like Otter AI, which is basically a note taker in Zoom when it comes right down to it. It's going to summarize an email. It's going to summarize a meeting and send out some notes. And it's a fine tool for that, right? I like that in a personal note. It drives me nuts. I personally don't like it either. But it's, you know, that Otter AI is not, it's not a true open source AI. It's great for a role of basically somebody taking meeting minutes. That would be like the old school version of somebody sitting in the corner doing dictation and then sending it out to everybody. That's its purpose. Doesn't Zoom have that now, though? I can't remember which one it has. It has one note taker or the other. I think Zoom because sometimes I on a lot of Zoom Sometimes I have some weird setting where I get reading assets and notes And it was like Michael did this and Michael did that And you know person whoever I all with Jake did this Jake did that And it gives me, like, paragraphs and lists. I'm like, that's pretty cool. Yeah. And again, it has its purpose. And it depends on what your goal is. If your goal is to function as a more high-functioning executive or leader or whatever is in your role, use AI to help you. There are tools for that. I mean, there's different things. I am not a SAS coder. I'm not. When I say SAS, it's S-A-A-S. I'm not that person. I didn't go to college for that. I went for leadership in organizational development. I didn't go as a computer technologist. But I can build, at this point, I can build SAS integrated applications almost faster than the coders can because I've got multiple AIs that I'm using. And one of my favorite, this is one of my favorite things to do with AI. I absolutely love this. This is my favorite thing to do with AI. Have AI check the other AI. I love running AI audits. It's one of my most favorite things is to take, you know, I'm just going to use some examples. Like I'll take Replit, GenSpark, ChatGPT, and Gemini and basically put them in an octagon and have them fight. right you know i want i want you guys to go in there and and and i'll pull one copy and paste stuff right for the most part i'll give them website links and go what are you seeing here and then i'll plug it over into the other one say hey gemini saw this what do you think about this and i'll i'll have them basically work on and against each other to be able to get to a common solution and it's a great way to audit what one is doing to where you can validate the information you can look it up you can be able to find it you know that's that's what i use gemini largely for it's owned by google gemini is a great research tool love it for that chat gpt is a good research tool you know the paid version is even better than that um gen spark again it's an i it's a idea machine is what i do it it builds a lot of slide decks for me and i have it do a lot of keynote work do some ideation around logo and you know different different results against spark I gotta say like over I checked a little clunky when you're trying to do graphic stuff I know that it works like for meetings and you know when you're kind of sending around your friends but when when you put out a product from Jen spark it really looks like you spent like a few thousand dollars on it and it's really slick. I'm massively impressed and maybe one day, maybe you and I just have to have a course on one of our little heads in the bottom corner and we're just walking people through how to use some AI. I think it might be a nice freebie for some of our audience. Yeah, it can be fun and there's so much of what the output equals the input, right? Input equals output and we've said this before of if we want it to look cool. Like we did not have AI design our original logo. Like we had a human being design our original logo. Mike's wearing it right there. And yet what's great about that is that is the brand. It's a play button and a Havintosh at the same time. Yeah. So that particular trademarked piece, we plug that in and say, here's the brand colors. Here's the brand flow. Here's the imagery. This is what I want it to look like. Give me an example. And then you build it from there. I like this i don't like that do this do that and you've got to build it it takes time to train the ai because if you just take its initial result if it spits it out and you're like ah that's good enough copy and paste and forward you just train the ai that that's what you're willing to settle for you've got to take time to train it in every one of those plain and simple so that's you know much like just we just finished the phase i guess phase two of the app is out now um and may 13th you're going to roll it out. You guys are going to get some special promo pricing. It's in final beta testing right now, which is super cool. But it helped me to finish out a communication model I've been working on for years to get done related to what's called the communication compass. And this is a super awesome tool because every single person out there has had the notion in their head, hey, I got to have this really difficult conversation today with this person, be it a salesperson, be it an employee, be it a leader. And I mean, I need to know how to start. Where do I start with this conversation? And we actually have a AI tool inside the app that you can plug all that information out and it'll give you suggestions on where to start. And then it will actually build out the role play session for you. So you can practice on the AI, not practice on the hard conversation. And this is a massive gap that I have found in Michael as well, that we found in leadership, that we found in customer service management, that we found in HR, that we found in sales, is that there's a huge gap. And it used to be the old school mentality was, well, go out and learn all these scripts so that they can say the thing and you can say the thing and they can say the thing and you can say the thing. Nobody remembers all the scripts. It doesn't work. And then it feels salesy, 100%. And everybody, I've heard, how many times have you heard that i don't want to sound salesy yeah it sounds salesy because you have to be in a ton of sales engagements to actually master the stuff you're saying right like it's not native it's beaten into you because of volume it's just like anything else right when you're in a position of anything right let's say you're uh carving out circles in a piece of wood for a living There's the mechanical efficiency that happens after your hundredth car. Then there's another one and another one and another one every hundred times. The same thing works with these scripts and reps. What we're trying to do is, without being in a sales environment, trying to train the brain and the person to be able to communicate so they're not getting tripped up with self-guilt. and they're not getting tripped up with, hey, wait a minute, I'm trying to get past barriers so that I can introduce my point of view, topic, brand, product to somebody else and create a positive result. And the positive result may lead to a sale it may lead to somebody willingly energetically doing a task it may lead to a deal it may lead to a relationship it may lead to something And it's an ethical communication tool where we're not doing an NLP, it's not psychological programming, it's not about the hocus-pocus. is basically working on somebody's self-confidence using an AI tool that has already been warmed up and already been programmed, therefore saving somebody thousands of hours and thousands of minutes of headache, heartache, and confusion. And it's like at the tip of the, you know, like at your fingertips. And we'll figure out, like, go to total-disruption.com if you want to get more information on this. we'll figure we'll let you know where to get this app and all that and we'll launch that on the 13th of may sweet man total dash disruption dot com i mean that's this ai stuff jake i gotta be honest with you if we may just take a little sidebar it's not so intimidating when you think about input equals output and it's it's it's a per it reduces ai down to a tool if you think input equals output ai is not so scary what i put in is what i get out whether it's open source closed source whatever so i want to put something in and that leads to better quality ask better quality input to get better quality output like i have done and when you use ai a lot you start asking broad and then it's wait i don't like that and you specific specific specific you may have to ask ai three and four times the same thing tighter and tighter we'll say oh yeah and this is this is where my brother-in-law was really on the forefront of this like he he mentioned this that ai had been out feels like a week like maybe even been out a day when he did this and we were sitting down having lunch the one day and he was playing with it at the time we all said that hey i'm playing with ai you know i'm playing with this tool and basically what it boiled down to was he said you You know, I didn't like what it gave me, so I kept questioning it. And I kept questioning it and challenging it to go back and find the answer I was actually looking for until it did. And that was, you know, he was very much on the forefront of this in the very beginning of, he didn't just take the first answer it gave him and copy and paste it and send it. You know, he made sure, and he's an engineer, he challenged it until it came up with a result that satisfied him both ethically, truthfully, honestly, authentically, got him to the end, and he's like, yes, that's what I was looking for. And it wasn't so much that he talked to it like a person, and we've talked about that too, as you kind of have to do that to give it your brand voice. Because how many of us type differently than we talk? 100%. I'm guilty of it too. I type very Elizabethan when it comes right down to it. And that was beaten into me by a British literature teacher in high school of how I needed to, how I needed to write versus how I talk. But you've got to talk to the AI. If you want your brand voice, you got to talk to it. Like you talk to it. You got to type like you talk because if it comes back with a website that looks, you know, like something, I don't even know how I would explain it. You know, our website looks very, very specific to how we talk and yet you didn't have to it could have come back as something very you know could look like kitten calendars when it comes right down to it if we did just let it do it but you got to speak with your brand voice well speaking of brand voice i i want my name to be associated with a behavior like elizabeth right like i want i want la barson it may or may not be a good thing. We'll see where that lands. It doesn't matter. All I want to do my ego is so fragile that I want to somehow have, you know, like, when people take a blast from me, like, take my name in vain as long as you're saying my name. You know what I mean? Like, let's go, right? What an impact. That's some old school writing stuff. Yeah, I may have dated myself just a little bit in that statement about some of the writings. How can I work my way through AI to get people to use me as an example of how to write and communicate? I don't know. But that's going to be something I'm going to be pursuing here in the future. So when we're talking about educating people, right, the biggest thing that you and I think we find in people is that learning curve and the adaptability of I have an idea, trying to get it out, trying to either make money on it, use it, make my life smoother, make my workday smoother, make everything more efficient, close more sales, communicate better to my kids and family, communicate better to prospects, communicate to my boss, boss down to my, you know, subordinates or my reports, direct report. Everything boils down to effective communication and everything boils down to practicing your communication. I myself have four different ways I communicate. I communicate in a professional manner, I communicate in a personal manner, a coaching manner, and then a friend manner. And I got very good at blending and separating the three because I don't want my messages to get misconstrued. Now, I know that when you listen to somebody like Tony Robbins, for example, I don't want to date myself, but Tony Robbins in this day and age was a big forefront in breaking open communication. He's a great marketer. He uses communication to change people's face. That's not quite what I'm talking about. I'm talking about communicating so that people can hear your information the way they're used to hearing it. Not everybody has to learn and adapt to your style. Sometimes you can quickly adjust your style to hear that. Now, I know we're running out of time. So I think we should pick this up next episode. What do you think, Jake? Totally, man. Love it. All right. So go back to disruption.com. My name's Mike LaVerti for my very super esteemed colleague, Dr. Jake Lynn Denning. You were just live on Total Disruption. We'll see you next time. Thanks for tuning in.