What happens when AI can do the work of an entire marketing team and a startup can clone your software for 90% less?
In this special AI Answers edition, Paul and Mike answer 15 questions from our AI for Marketing, AI for Sales and AI for Customer Success webinars: covering agents, job disruption, privacy, the SaaS apocalypse, and why your own pilot data beats any McKinsey report.
Listen or watch below and see the show notes and transcript that follow.
During the last few years, our free Intro to AI and Scaling AI classes have welcomed more than 40,000 professionals, sparking hundreds of real-world, tough, and practical questions from marketers, leaders, and learners alike.
AI Answers is a biweekly bonus series that curates and answers real questions from attendees of our live events. Each episode focuses on the key concerns, challenges, and curiosities facing professionals and teams trying to understand and apply AI in their organizations.
In this special edition, we tackle 15 of the biggest questions from our AI for Departments webinar series. Across these sessions, we shared fresh insights from our latest blueprints: AI for Marketing, AI for Sales, and AI for Customer Success. Paul answers every question in real time, unscripted, unfiltered, and just like he does live. This time, he’s joined by Mike.
Whether you're just getting started or scaling fast, these answers can benefit you and your team.
00:00:00 — Intro
00:05:18 — How should a CMO get started with AI?
00:09:57 — What is the difference between an AI agent and a regular prompt?
00:12:47 — Will AI labs fix their environmental impact?
00:17:04 — How can you convince skeptics that AI can help improve performance?
00:19:55 — How do you deal with AI sycophancy when using it as a thought partner?
00:22:06 — What efficiency gains are people seeing from generative AI in marketing?
00:25:42 — How can you track and measure time saved by AI?
00:27:47 — How do you manage information and prompts across multiple AI platforms?
00:33:59 — How do you balance AI adoption with data privacy and security?
00:36:17 — Which marketing roles do you think will be most disrupted by AI?
00:43:51 — Will sales calls made by AI chatbots just feel like spam robocalls?
00:46:29 — How can I reinvest time saved by AI into growth and innovation?
00:49:33 — When should you buy software versus building it yourself with AI?
00:54:35 — How do you protect yourself and your information when others use AI agents irresponsibly?
00:55:58 — How do you communicate effectively with company leaders that IT should not be the department driving AI adoption?
This episode is brought to you by Google Cloud:
Google Cloud is the new way to the cloud, providing AI, infrastructure, developer, data, security, and collaboration tools built for today and tomorrow. Google Cloud offers a powerful, fully integrated and optimized AI stack with its own planet-scale infrastructure, custom-built chips, generative AI models and development platform, as well as AI-powered applications, to help organizations transform. Customers in more than 200 countries and territories turn to Google Cloud as their trusted technology partner.
Learn more about Google Cloud here: https://cloud.google.com/
This episode is also brought to you by AI Academy by SmarterX.
AI Academy is your gateway to personalized AI learning for professionals and teams. Discover our new on-demand courses, live classes, certifications, and a smarter way to master AI. Learn more here.
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: These things are not reliable and in many cases they are not safe and people are definitely racing ahead and using them regardless. And we, we all could be collateral damage in that grand experiment.
[00:00:16] Mike Kaput: Welcome to AI Answers, a special q and a series from the Artificial
[00:00:20] Paul Roetzer: Intelligence Show. I'm Paul Roetzer, founder and CEO of SmarterX and Marketing AI Institute.
[00:00:26] Every time we host our live virtual events and online classes, we get dozens of great questions from business leaders and practitioners who are navigating this fast moving world of ai, but we never have enough time to get to all of them. So we created the AI Answers Series to address more of these questions and share real time insights into the top.
[00:00:46] And challenges professionals like you are facing, whether you're just starting your AI journey or already putting it to work in your organization. These are the practical insights, use cases, and strategies you need to grow smarter. Let's
[00:00:59] Mike Kaput: [00:01:00] explore AI together.
[00:01:06] Paul Roetzer: Welcome to episode 202 of the Artificial Intelligence Show. I'm your, host Paul Roetzer, along with my co-host Mike put, we're recording Wednesday, March 11th, 9:00 AM Eastern Time, in the middle of a thunderstorm in Cleveland. So hopefully this, we can do this straight through without any power loss. this is a special edition, so this is not our regular weekly, so don't get confused if you are a regular weekly recap listener on Tuesdays that we drop.
[00:01:32] this is a special AI Answers edition, so AI Answers is our 15th episode of this series. AI Answers is presented in partnership with Google Cloud. This is a series that we do based on questions we get from our monthly intro to AI and scaling AI classes. So if you haven't been to those every month, I teach a free intro class and a free scaling AI class, and we usually get dozens of questions during those live sessions.
[00:01:58] And so we use these AI answers [00:02:00] episodes to answer the ones we couldn't get to, but this is a even more special version of our special AI answers series. This one is actually based on questions we got during our AI four Departments week, which was also presented in partnership with Google Cloud. So AI for departments February 24th to the 26th this year.
[00:02:20] We released, three blueprints, AI for marketing, AI for sales, and AI for customer success. And we did that with a webinar each day. Those webinars had thousands of people registered for them. And so the questions we got were amazing and we could not get to all of them during the live sessions. So we decided we'll do a special AI answers edition that answers questions that came from our audience during those three webinars.
[00:02:45] So Mike has curated this. He's gonna go through and. Pick some, you pick some questions or marketing from sales, from customer success. I have not looked at them. I prefer to do these the same way I do it in the live environment where I don't see the questions until Mike asks 'em. so you [00:03:00] can, learn more about both of these, the BO webinars and the blueprints.
[00:03:03] You can go to SmarterX.ai/webinars. All three of those webinars are available on demand now. And then you can go to SmarterX dot ai slash blueprints and you can download on gated, each of those blueprints or whichever one is most relevant to your work. So again, thanks to Google Cloud for their partnership to bring these webinars and blueprints to life.
[00:03:25] You can learn more about Google Cloud at cloud.google.com. And I think Mike, that covers everything. Or I, am I missing anything here? No,
[00:03:34] Mike Kaput: no, that's it. That's some good context into the blueprint webinars, and like you said, we had such a huge audience for those. They were super popular.
[00:03:41] Paul Roetzer: Yep. Alright. And then this episode is also brought to us by AI Academy.
[00:03:45] By SmarterX AI Academy helps thousands of individuals and businesses accelerate their AI literacy and transformation through personalized learning journeys and an AI powered learning platform. There are currently 13 professional certificate course series [00:04:00] available on demand for mastery members and individual purchase with more being added each month.
[00:04:05] We just released our newest course series AI for financial services and AI for finance covers, real world applications of cross banking, insurance, wealth management, more that's the financial services one. So you can start applying AI strategically in your organization today. So the ifi, financial AI four financial services as part of our industry series.
[00:04:25] And AI for finance is part of our department series. So we have Mike, let me see if I get this right. We have marketing, sales, customer success, hr,
[00:04:34] Mike Kaput: HR, and finance at the moment. Okay. And we are about to do operations it and I believe legal to round out kind of the initial functions.
[00:04:43] Paul Roetzer: Yeah. So in the next couple months, we will have the vast majority of departments within an organization covered.
[00:04:49] And so they're a great starting point for people in your organization who maybe haven't figured all this out. It's a great kind of 1 0 1 to 2 0 1 level. It really gives practical knowledge about it, and that's how all of the certificate [00:05:00] series are structured across departments and industries. and then our foundations collection, which has fundamentals piloting and scaling in it.
[00:05:07] So great stuff there. Go check it all out at Academy.smarterx.ai. Alright, Mike, I think we've got what, 15 questions it looks like. Yep. Slotted. We'll see if we can get through all these in about an hour.
[00:05:18] Mike Kaput: All right, Paul, so first up. Someone asked, what is the best way to get started with learning all we need to know and implement about ai?
[00:05:28] Specifically, I believe this person is asking as a CMO, so leaders, how do I get started wrapping my head around all this.
[00:05:36] Paul Roetzer: All right, so this was not a planned plug, but do we not have an AI for CMOs webinar coming up, Mike? Is
[00:05:41] Mike Kaput: that We do indeed. Yes, indeed, indeed. Is that
[00:05:43] Paul Roetzer: been announced or am I announcing that
[00:05:44] Mike Kaput: now that No, that has been announced that, yes.
[00:05:46] So don't worry. That's no secret.
[00:05:48] Paul Roetzer: Yes, I would say, hold tight. We do, do you remember when that's coming up?
[00:05:53] Mike Kaput: Yeah, we've got it. It is March 26th at 12:00 PM Eastern. Perfect. It'll also be made available on demand, so it's again, at [00:06:00] SmarterX dot ai slash webinars. You'll see it right there.
[00:06:03] Paul Roetzer: And that's gonna come with a blueprint as well, right?
[00:06:05] Mike Kaput: Yes. Yep.
[00:06:06] Paul Roetzer: There we go. All right. So the answer to your question is join us March 26th for a webinar that actually explains all of this, as well as a download blueprint. Now that being said, the way I've been thinking a lot about AI adoption in organizations, whether it's a team department or organizational level lately, is the need for leaders to have a higher degree of AI literacy competency.
[00:06:30] what I mean by that is the CMO is gonna be the one that's gonna have to push the team to figure out how to apply AI for efficiency, productivity, creativity, innovation. They're gonna be the person who's gonna have to deal with. The employees who don't wanna learn ai don't wanna do it. I mean, the CMO is often overseeing the creative within an organization.
[00:06:50] There are a lot of creatives, whether it's writers, designers, video producers who see AI as a threat to what they do. So, you know, for the [00:07:00] CMO, it comes with, I think, at the starting point, just a deep understanding of what AI currently is capable of doing, what these different models are capable of doing across all their, you know, again, not just text in, text out, but audio, video.
[00:07:12] code, design, you know, in terms of image production, video production. So you, you have to have a deep understanding. Reasoning would be a really important one when it comes to strategy. You have to understand all that, and then you have to be modeling use of these tools for your employees. So you don't have to be playing around with all the image generation of video generation tools and things like that, but you need to be using it every day.
[00:07:36] So I guess for a CMO, the starting point is understand it deeply. And like our foundation's collection on AI Academy is, is that like you could come to our free intro to AI class if, if this is all new as a CMO or you have a friend who's a Cmmo and it's all new, go to the free intro to AI class. But if you're part of our AI Academy, take the Foundation's collection and I mean, literally 95% of what you need to know at a t a decent [00:08:00] confidence level, you will get from those three, course series alone within the foundation's collection.
[00:08:06] So, or just come to the free March 26th event and do the AI for CMOs, webinar, download the blueprint. So that, that would be my, my kind of quick advice. Anything to add there, Mike? Anything you're thinking of?
[00:08:17] Mike Kaput: No, it's just a really good reinforcement of something that we mentioned on the actual webinar is one of the experts we interviewed at Google.
[00:08:26] Emma Dell Roetzer, who's a manager of AI transformation there, she mentioned basically the best. Organizations doing this that she sees have leaders that model this stuff, that talk about this stuff. And that's something, one of many things I very much appreciate about your approach is like you're always telling us all sorts of stuff about how you're using AI and it's a massive inspiration and motivation for the rest of us.
[00:08:50] Paul Roetzer: Well I think, you know, not just me. I mean you do the same thing with the podcast for sure. And even internally I think there's things you're doing all the time and just sharing with the team. But that is the key. And I [00:09:00] think a lot of times, even with the podcast. Yeah, there's things I'm doing where I'm like, I don't know.
[00:09:04] This is pretty basic. Like, I don't know that I should share this, and then I'll share something. That to me is just sort of second nature at this point.
[00:09:11] Mike Kaput: Yeah.
[00:09:11] Paul Roetzer: And then I'll go do a speaking gig and someone will come and say, Hey, this is a podcast. That example you gave was amazing. I actually went and built something.
[00:09:16] I went and used that and I taught somebody on my team something. And so I think that there's a lesson in there for all of us that don't take for granted the knowledge you have and the capabilities you have and the things you're doing with AI that you might think is basic. To a lot of people that might change the way they think about ai.
[00:09:32] So, you know, definitely think about your own use of it and then think about sharing and modeling for others how you're doing that. Whether it's through LinkedIn posts, internal Slack messages, you know, running lunch and learns, whatever it is. And you know, I think CMOs being a leader in that is a great way to think about it.
[00:09:48] And then if you transfer that to other departments, you know, obviously we're talking specifically about CMOs 'cause that's the question, but this goes for the leaders of any department.
[00:09:55] Mike Kaput: Hmm. All right.
[00:09:57] Mike Kaput: Question number two. When we're using this term agent, you know, referring to AI agents, what exactly is that?
[00:10:03] Can you help explain the difference between an agent and just a regularly, regularly used task or prompt that you're repeating with ai?
[00:10:13] Paul Roetzer: So, phase one of generative AI was text in, text out. So that was, you know, ChatGPT When we first started using it in 2022, you would put a text prompt in and you would get some sort of text output.
[00:10:25] same would then apply to images and video you were just putting text in and then you would get an output from the machine. Agents are, AI systems that can take actions to achieve a goal. So they can, in some cases, develop a plan of what they're gonna do. They are sometimes given access to different tools like the internet to conduct searches.
[00:10:44] and so they can, you can say, okay, I wanna produce a research report, and then rather than just producing a report from its knowledge base. It goes and searches the web and then it curates information on the web and then it synthesizes that, and then it writes an [00:11:00] output, then it verifies its sources.
[00:11:01] So it's actually going through and doing a sequence of actions. And so you can imagine if you are, let's say you're gonna run a marketing campaign, 'cause it looks like this question came from the marketing one. you know, if, if you say, you go into Claude and say, okay, help me. Write a landing page.
[00:11:19] And it writes the landing page. That's just a simple chat bot, like just creating an output. Then you say, okay, now let's go through and let's build, an entire marketing campaign around this. Well, now it's gonna go and build a plan of how it's gonna do that. It might need to call on, you know, different knowledge base you've given it access to.
[00:11:39] So it's gonna go and start taking actions that might be 10, 15, 20, 50 things. And then it'll eventually create all the final outputs for you. But it's actually going and doing a bunch of things and taking these actions. And then you may actually have it set up where it's like, go ahead and do the thing.
[00:11:55] Send. Send the emails, do the advises. That's agents. Now [00:12:00] the confusion comes in with how autonomous these agents are. Many of the agents you would be using in your regular workflows at a, you know, an enterprise. Are gonna be pretty basic in terms of their autonomy. There's still maybe some rules built in there as like the humans really heavily in the loop.
[00:12:15] The stuff you hear about with like OpenClaw that we've been talking about, that's much more autonomous where people are trusting these things with access to a bunch of information to just do stuff. in some cases, leaving them running overnight. We'll talk about Andrej Karpathy had a tweet just this week.
[00:12:32] We'll talk about an episode 203 next week. Where people are on the frontiers, on the edges, really starting to push the agentic side, the autonomous part of the age, agentic side, and that's gonna create for some pretty interesting environments.
[00:12:47] Mike Kaput: So question three is something we hear a lot of variations on, not just in marketing though, that's where this one came from.
[00:12:55] Do you think AI Labs will actually fix the [00:13:00] negative impact their technology is having on the environment?
[00:13:04] Paul Roetzer: I think they think they will. I don't know that you and I Mike, have any specific insight information about how the Exactly. They'll execute this. But in essence, like, just for reference, if people aren't kind of familiar with this topic and context a few years back, a lot of the AI companies,
[00:13:22] it set out to be, carbon neutral. Like they, they, they wanted to, you know, keep their impact on the environment, neutral or actually a positive impact, actually give back energy. when AI exploded in 2022, that just got thrown out the window. it became, build data centers, consume energy, build as much intelligence as quickly as we can build.
[00:13:47] And so we have to delay this idea of the, you know, the neutral impact on the environment. Now many of them think that if they build more intelligent systems, these super intelligent systems, those intelligent systems will solve [00:14:00] for this. right now the way they're solving for it is they're making, every year or so the cost of compute drops like 10 x.
[00:14:09] So they're making more efficient algorithms that use. Less computing power per token of output, I guess is the way to think about this. So, you know, the energy it took to write, let's say a 10 page research paper a year ago, or a ten second video from, from soa, it, the amount of energy it would've taken to do either of those things has, has dropped 10 x probably in the last 12 months.
[00:14:35] Mike Kaput: Yeah.
[00:14:35] Paul Roetzer: So to do the same thing. Now the demand for those outputs is on an exponential, so net we are requiring way more energy. Yeah. Having a way greater impact on the environment because demand is rising, but they're doing, they're satisfying that demand more efficiently. So that is their current path to do it.
[00:14:56] But they are all looking at solutions, in terms of [00:15:00] different energy sources and how to get it more efficiently. you know, including off earth stuff where the data centers live in satellites, through like x AI and stuff. So I am, I am concerned about the environment. Like, like many people. I don't know that there's too much we can all do about it at this point.
[00:15:21] The couple things I've talked about is use the more efficient model and get really good at prompting. Like, those are two actual things we can all do. the better you are at prompting the fewer tokens you're gonna use to get the output you're looking for, that's probably the most ho honestly, like the most immediate action most people can take.
[00:15:38] Other than that. I don't know. I'm opt I believe in Demis Hassabis and Google DeepMind in particular. And I know that they are focused on energy. It's one of the key things they're thinking about, and I think if anyone's gonna solve it, I think Demis has a decent chance to do that, you know, in the next five to 10 years.
[00:15:54] so I don't know. I choose to be optimistic. I just, I don't know exactly how it happens.
[00:15:59] Mike Kaput: Yeah, I [00:16:00] would just add there too, related to the previous question, like, my gosh, you will start to see how many tokens at the moment agents start using when you start running these in Claude code or something. I'm like, oh my gosh, this is highly compute intensive.
[00:16:16] Paul Roetzer: Yeah, we talked about, I think it was the Jensen Wong quote.
[00:16:19] Mike Kaput: Yes,
[00:16:19] Paul Roetzer: on Tuesday's episode. About OpenClaw being like the most important software. What did you, what was the, do you remember
[00:16:25] Mike Kaput: what was, it basically was like the most important software I'd ever been
[00:16:28] Paul Roetzer: released, basically. And then the open source diversion of something like that yesterday.
[00:16:31] I think we'll talk about it on the show next week. But yeah, that's the idea is like agents, so something that's taking all these actions you, I think requires way more computing, power inference, which is when you and I would use it than like a standard text. Just like video requires more tokens than, you know, text chat.
[00:16:50] images, things like that. Yeah, agents are gonna be a massive drain. It's why they're building all these data centers and getting prepared for a world [00:17:00] where agents are in everything, in essence.
[00:17:04] Mike Kaput: Question four, it seems like there's an assumption that AI will enable us to produce better work, but a lot of us feel like we're expected to take that on faith.
[00:17:15] What do you say to people who aren't convinced that AI will lead to better performance or will make their work, more meaningful or, you know, more valuable? Or have a bigger impact? Basically,
[00:17:26] Paul Roetzer: I think about this just on, we have, we talk a lot about personalized use cases.
[00:17:32] Mike Kaput: Yeah.
[00:17:32] Paul Roetzer: I would just say, okay, like if there's something where you are feeling that way, what is, what is the use case exactly.
[00:17:38] That you're not sure. It's on par with like a really qualified human to do that job. And I would develop some, you know, evals as like the in term just means evaluations or benchmarks. That's like, okay, I write this research report every month, or I do this performance report every Sunday night, or I'm in charge of these blog posts or these emails, or [00:18:00] this proposal, or this talent review, or this meeting summer, or whatever it is that like, you know, just take a basic use case.
[00:18:07] I would make sure you're using the best model available. Oftentimes people who have this concern are using the free versions of these tools, and they haven't used the advanced reasoning models like a 4.5 0.4 thinking from ChatGPT versus whatever the baseline model is today. so I would say take personalized use cases and then solve for it.
[00:18:28] Now, I think more and more, you know, we talk about this move 37 moment. we touched on it again on the, on episode 2 0 1. I think more and more it's becoming very hard to, to say that a human just is way better at a knowledge term than a than machine. We'll talk about, on episode 2 0 3, New York Times just ran an experiment with AI writing, and it was a blind taste test in essence, like, here's a paragraph from an ai, here's a paragraph from a human writer, which you prefer.
[00:18:58] And the AI wins. [00:19:00] And it was like across like 86,000 votes, I think. Like it was not a small sample size.
[00:19:05]
[00:19:05] Paul Roetzer: And that that's, you know, writing is just one use case. And, but I think more and more over the next, you know, one to two years, there's just gonna be very few tasks left. Where if you did a blind taste test that the AI isn't gonna be at least on par with a highly qualified expert in that field.
[00:19:29] I , I t's a very hard reality for all of us to Yeah. Come to, but I feel pretty deep conviction about that one.
[00:19:38] Mike Kaput: And, you know, I really appreciate this question, so I'm not knocking it at all. But when you ask like. Are we, a lot of us feel like we're expected to take that on faith. The solution to that is just go use the tools and like kick the tires.
[00:19:50] Right? You don't have to take anything on faith. You can go try for yourself.
[00:19:53] Paul Roetzer: Yep.
[00:19:54] Mike Kaput: Alright.
[00:19:55] Mike Kaput: Question five. When using AI as a thought partner, is chat GBTs tend or another models tendency? To be agreeable. An issue here, like how can it give you valuable feedback if it's configured to essentially agree with you by default?
[00:20:10] Paul Roetzer: So the term here we've touched on on the show is s of Fancy. Yeah. Where it just like, Hey, that's a great question, or that's a great insight. Like let's build on that. And it never tells you like you're an idiot. That's actually a terrible idea. I would say the simplest thing is one, they're aware of that issue and they are, they have adjusted the system prompts behind the models so that they supposedly aren't that way.
[00:20:30] The other is just say it in your prompt. Like, I would like you to challenge my ideas. I want you to function as a critic. I want you to challenge this as though you are this person. Like, just tell it that. So even if it's system prompt enables it or, or it's default is to tell you you're brilliant. And that like every idea you have is great.
[00:20:51] Just tell it to take the opposite point, steelman this, you know. Position on this idea or, you know, assess this. Writing as though you're [00:21:00] a n editor at the New York Times. Like, just tell it to be critical and it will function in that way. So that is the fastest solution. Do you have any tips, Mike, that you've used with yours?
[00:21:09] Mike Kaput: Yeah, no, I think that you hit the nail on the head, giving it. Explicit instructions to be that kind of skeptical critic, to argue with you whatever level of this you feel comfortable with. Trust me, you will have the opposite problem if you do this right, where you're like, no, I actually wouldn't like you
[00:21:26] Paul Roetzer: go back to telling me I'm brilliant.
[00:21:28] Mike Kaput: Yeah. You come away thinking like, I'm not good at anything, you know?
[00:21:32] Paul Roetzer: I did this when I was creating courses for AI Academy. Yeah. Last summer and fall. I built a AI learning assistant. I've talked about this on the podcast, but in its system instructions, it was very specifically told, I want you to challenge everything I give you.
[00:21:46] When I give you a course that I've created and you're reviewing the deck for me, I want you to challenge my ideas. I want you to ask difficult questions. And it works. And it does. Like, at some point you're like, all right, dude, that's a, that's enough back off. Like I,
[00:21:59] Mike Kaput: right.
[00:21:59] Paul Roetzer: I get [00:22:00] it. But yeah, you gotta, you kind of turn the temperature down a little bit.
[00:22:05] Mike Kaput: All right.
[00:22:28] Mike Kaput: Question six, what has been any of the efficiency gains reported by the use of generative AI in the marketing world? This person said, we are jumping into AI use completely, but in the interim, our staff seem to think we have to get this job done the old fashioned way. So I think they're trying to kind of look for some proof here about what are we hearing when it comes to efficiency gains in marketing, thanks to ai.
[00:22:30] Paul Roetzer: I think this goes back to the question we answered earlier, Mike, where just pick your own benchmark. Like you can go find all these different reports, like Mike and I could give you stories of things we're doing in two hours that used to take 20 or 50 or a hundred. And like you can have all those stories you want.
[00:22:47] Just pick something internally, do a pilot of it and say, okay, traditionally here's something we do every month. It takes us 17 hours on average. We looked at the, you know, previous data or we went through on a [00:23:00] task by test level and estimated best knowledge we have of how long it would take in a normal environment.
[00:23:05] And then we're gonna do this with AI and we're gonna make sure that people are trained to actually use the AI properly, you know, best, you know, how, how they handle prompts, things like that. And so, and then run your own pilot and say, wow, we, we did it in two hours instead of 17. So like, create a few of those and now you've got the business cases.
[00:23:25] Now you've got the proof internally, and there's just, there's nothing better than your own proof. You can recite all the reports you want. Just, and to your point earlier, Mike, it's 20 bucks a month. Like just spend the 20 bucks a month on the paid version. If you don't have it, get approval. If you're in a bigger enterprise and you're trying to prove out the reason for buying like a Jasper or a writer.
[00:23:48] Or you know, a ChatGPT enterprise or Google, Google Gemini for the team, whatever it is. pick a business case that means something internally. Yeah. And show them that, and say, [00:24:00] we can stack these and then we can get to 10% efficiency gains, 20%. You can get to like 90, a hundred pretty easily. but no one's gonna believe that.
[00:24:09] So start with believable numbers. Sure, yeah. Or just show like, you know, the reality.
[00:24:14] Mike Kaput: Yeah. I don't know about you Paul, but I wanna kind of communicate to our listeners. Like I literally. Speak to leaders and audiences as part of my job. And I can tell you there's no study or stat that's gonna play from McKinsey or whoever.
[00:24:28] It's helpful information. Yeah. But it's not gonna make people wake up more than you saying like, Hey, this thing we all are familiar with, we used to do it this way, now we do it this way. And look at the difference.
[00:24:38] Paul Roetzer: Well, and to one of the earlier questions, like if people aren't. Don't have a high degree of AI literacy yet.
[00:24:46] Mike Kaput: Yeah,
[00:24:46] Paul Roetzer: they are. They're inherently going to be like, it can't do what I do. Mm. And we hear it all the time. And it can be something as basic as like writing a newsletter or something like, oh, I can't write the newsletter the way I do. It's like, yeah, I can, I'm sorry. Like [00:25:00] it can,
[00:25:00] Mike Kaput: yeah.
[00:25:02] Paul Roetzer: so that, that's something you have to deal with as well.
[00:25:04] But I would say again, like focus on individual use cases, business cases, prove that out through your own pilot, your own data. Do a few of 'em if you need to. And then that's actually one of the best ways to drive adoption across other departments. Like if marketing's leading the way and you want to get sales on board, or the success team or the finance team just show 'em a business case of something they do.
[00:25:25] What, what? What about your job don't you enjoy? Like, what's the thing you would love for AI to help you with that? You'd rather let go. Don't take the, you know, the thing that they love and care about that gives them fulfillment. Take the thing that they hate and then show them how to do that.
[00:25:42] Mike Kaput: All right, so Paul,
[00:25:42] Mike Kaput: question seven, I kinda selected because it is very related to this.
[00:25:45] You touched on this a bit and I kind of just want to close the loop here. How are people tracking and counting the TAM time saved by ai? Is it really just that benchmarking you're talking about? Just fire up a spreadsheet and basically start writing this stuff down? [00:26:00]
[00:26:00] Paul Roetzer: For most organizations, probably, I mean, you and I might come from the agency world where you tracked everything
[00:26:05] Mike Kaput: anyway.
[00:26:05] Paul Roetzer: Yeah. And so if you're in a professional services firm, you know, like a law firm or a consulting firm or an agency, you're probably used to it and you have benchmarks to look at. Like when I was running the agency, we had 16 years of time data so we could go back and look at anything and be like, oh yeah, the strategy's taken on average 44 hours, and the blog post taken average 3.2 hours.
[00:26:24] And like we knew that if you don't track time, which in most enterprises you aren't, then pick very, distinct use cases and develop a benchmark. Or like I said. Go through and break any workflow into tasks, any project, into a series of tasks. And then at least best case estimate, like, all right, like if I was tracking my time, this planning part would take me about three hours.
[00:26:51] and then go from there. Now, I don't know, I should probably get into the Fibonacci sequence here. It's probably like overboard, but we used to use like Fibonacci. 'cause what I learned in 16 years of running an [00:27:00] agency is humans suck at estimating how much time it takes to do something. It's almost always wrong.
[00:27:04] And so we used to use Fibonacci, which is, the two previous numbers total, the next number, so like 1, 2, 3, 5, seven, you know, I'm down the line. Yeah, I forget 'em now. 11, like 13, 21, 34, 55, 87. I don't know. Yeah, I used to know 'em by heart. 1 21. But we would do that. So I would be like, okay, this is a one hour, this is a two hour, this is a three hour, this is a five hour.
[00:27:29] And you just sort of like ballpark based on it because it goes up by the same percentage in essence each time. So it removes the human error from like, this takes one and a half, or yeah, two points that like nobody knows and we get distracted all the time. And so it's like that those, those estimates are never accurate.
[00:27:47] Mike Kaput: Question number eight, how do you manage the information on each AI platform? Do you keep prompts separately? Each AI platform is getting better and they constantly move positions in [00:28:00] terms of who is the leader. So like are you porting information between models? Like how have you managed that?
[00:28:05] Paul Roetzer: I would be interested to get your take on this one.
[00:28:07] Yeah. As well, Mike, because this I, this is a daily issue for me 'cause I am now actively working in three models every day. Yeah. So I do use Claude all the time. I use ChatGPT all the time and I use Gemini all the time. Gemini is baked in Google Workspace. We are a Google Cloud Google Workspace customer, so that is native right in the productivity tools we use all the time.
[00:28:27] But I then use the Gemini app separately. chat. GPT came to market first with the best model, and so I have a history of three plus years now with chat. GPTI have a bunch of custom GPTs built in there, and so a lot of times I gravitate back to there where I just have a history and a memory and like I know it's gonna do it the right way if I'm working on something new, especially if it's a high level cognitive task, like a strategy.
[00:28:54] I will often use all three models or sometimes multiple models, even [00:29:00] within the platforms. So I'll use like a Claude 4.6 Opus and a Claude 4.6 sonnet, and I'll like compare the differences. So I, this is very messy for me right now. Yeah. I don't have an answer. I don't know. Mike, have you come up with any ways you're handling this differently?
[00:29:15] Mike Kaput: It's getting better, but it's still super messy. I think the first thing I started doing for the last couple years is just documenting all of my workflow. So it's not just about which AI tools, it's like, oh, at this step we go into this GPT to do the thing. So it's like, even if worse came to worse and I had to switch, I'm like, okay, I just need, can grab the instructions of that GPT and spin up a project or a gem or whatever.
[00:29:37] So that's been really helpful. What's really interesting, and I'll see how far this goes, is. As I use Claude Code much more. Claude Code will spin up different skills like these markdown files that tell you how to do something. It's essentially, you can think of it as a prompt, honestly, your instructions for a GPT.
[00:29:53] What's cool about this, I have all those skills being created, updated, and logged in a personal, shared [00:30:00] Google Drive, like the personal uses of Claude Code. So it's like, okay, even if Claude Code got nationalized by US government tomorrow or something, right? Is like, I can then take those skills and give them to ChatGPT or Gemini.
[00:30:15] It's like my internal knowledge and skill architecture that could port over. It's not gonna be exact, but I've tried it before and it works. So that's been nice. But also that takes a lot of work to build as well. Yeah.
[00:30:28] Paul Roetzer: I don't remember if we talked about this on the Weekly, but. Anthropic last week announced an import function.
[00:30:35] Yeah. Where they give you a prompt that you can put into one of the other models and it'll basically summarize everything in its memory and then like you give it to Claude and then it in theory remembers everything. the other thing, you kind of hit on this, but the way I do this internally is, so again, I'm the CEO of the company.
[00:30:52] I dabble in all the departments, and Mike can attest to this, like, sometimes something will bubble up and become a [00:31:00] priority for me that isn't part of my daily job.
[00:31:02]
[00:31:02] Paul Roetzer: And so I'll have like two days to grind on something while I'm traveling. I've, I've shared a couple of stories like this, like our success score for customer success as an example of this.
[00:31:12] And so I'll spend like 48 hours and I will get it to a certain point. It's like, okay, that's it. Like, I'm tapped out, I gotta move on to like my other CEO priorities. And I have to hand it off to the team.
[00:31:24]
[00:31:24] Paul Roetzer: Well, what I'll do is exactly what Mike explained. I'll create a Google doc. I will create tabs within that Google Doc.
[00:31:32] I'll put the prompts I used across the different models I've been experimenting with. I will have a tab for each of those outputs. Then the team can actually go in and see the different flows. And then if I have the time, I will do a curation of all those outputs into a single. Here is my CEO stamp of approval.
[00:31:52] Best summary of all, what all the different models are saying and which ones I went and why I did it.
[00:31:56] Mike Kaput: Yeah.
[00:31:57] Paul Roetzer: Part of that I do, just for clarity for the team so they can [00:32:00] see the thinking. Part of it is to, to model behavior. It's to say like, listen, I don't want you to just give me an output. Like if I ask you a coworker of mine to do something for me, I don't just wanna see a copy and paste outta Gemini.
[00:32:13] Like, I wanna see what you thought about how you prompted it. And so in essence, it becomes like an audit trail. And I found those to be extremely valuable. And I do that now for almost every strategy doc. I'm working on all these like innovations I'm working on for the company. I start with a Google doc and then I just, I journal everything.
[00:32:32] So I don't like lose track of which model did I do this in and that in, 'cause that happens all the time too. It's like. Where did I work on this at? Which, which model was I playing around with Success score in? And I'll, I'll lose track of things. So I also keep a business journal that's just for me. And I'll note in there like, Hey, I worked on success score these two days.
[00:32:50] Here's a link to that doc. And here, like I, because I'll forget, I had this happened yesterday. I like found this doc. I was like, oh shoot, I don't even remember doing it. This is really good. Like I [00:33:00] guess I started this like three weeks ago and I forgot about it. 'cause so many times you can just start all these projects and the agent stuff's gonna make this worse.
[00:33:06] You can just start things. And then like two weeks later you're like, I did, I do that? I feel like I started that project somewhere.
[00:33:13] Mike Kaput: You know, it's interesting, I preach quote increasingly to our teams internally, like this really unsexy but important skill is document discipline. If you can manage it because like having this, a system of effort that works for you to keep all this stuff, your notes, your context, your knowledge.
[00:33:33] Your working docs all in like a consistently organized place. That is how you get value out of these tools because you kind of solve that problem after a while where, you know, as we increasingly have agents or coworker or whatever, just pointing at folders or Google Drive or whatever, as long as you have this stuff organized well, you're not gonna have to worry about like necessarily where your prompts are or things like that.
[00:33:56] Paul Roetzer: Yeah.
[00:33:57] Mike Kaput: Yeah. Okay.
[00:33:59] Mike Kaput: Question number nine. How can we balance using AI and putting a lot of company content into AI with balancing that with privacy? How can we leverage AI in industries that require a lot of security? That gave the example of finance.
[00:34:14] Paul Roetzer: Yeah. I don't know a way around this other than working through it and legal, like they, they have to.
[00:34:20] You know, be not only be in the loop, but probably be in the driver's seat from a governance standpoint when it comes to sensitive information and confidential information, personally identifiable information, what, you know, whatever. What I often encourage people is like, find all the use cases that don't have to touch any of that though.
[00:34:37] Like let it and legal do their thing and let them protect the organization and the users and put degenerative policies in place that provide the guardrails for responsible use. But don't let that slow you down from all the use cases that don't require that data. And there's, I mean, in marketing and sales and customer success, there's literally thousands of things you can be doing every day, even in a [00:35:00] bank or a hospital system, a law firm.
[00:35:02] Like there's all these uses that don't have to touch any of that data. So definitely something you have to think about. Definitely something you have to collaborate with the right people internally on. But you need to own finding all the safe use cases that can let you race forward while they're figuring all this other stuff out.
[00:35:22] I just, I've talked to way too many companies, even in the last couple months that are just sitting on the sidelines because it and legal still have to approve like every use case or tool. And that is, that is not a sustainable model in with the rate of accelerated change we are going through.
[00:35:39] Mike Kaput: And I certainly sympathize with how hard that can be to figure out, but I honestly think some individuals or companies are using this like data thing as an excuse, like to not take action.
[00:35:49] It's like you can be doing so much with just the knowledge in your head.
[00:35:54] Paul Roetzer: Yeah, and if you hear us saying that and you're like, but I don't know what that is, just come to the free intro class, honestly, right? Like, [00:36:00] if you just attend the Intro to AI class, you will have the frameworks to go figure this out and like, move the ball forward.
[00:36:09] Just do not let waiting for it and legal stop you from making progress.
[00:36:16] Mike Kaput: Question
[00:36:17] Mike Kaput: number 10, are there certain roles or role types within the marketing function that you envision being rapidly undercut or impacted as AI evolves?
[00:36:29] Paul Roetzer: So of them, I really struggle with how to answer this. Sometimes all of them is like my, if I'm giving the multiple choice and I get an all the above, I would choose all the above.
[00:36:37] Mike Kaput: Yeah,
[00:36:38] Paul Roetzer: I think the ones that actually affect. Job security and job opportunity is any entry level role that completed tasks.
[00:36:48]
[00:36:48] Paul Roetzer: and like I don't know how to, I'm trying to think like how to frame that. like if, if you, if someone was giving you a campaign and you were just executing and all you [00:37:00] do for your job is like build landing pages and write email copy.
[00:37:04] Write ad copy. You know, if you, if you only do one of those narrow things and you just do it a bunch of, like, all day long
[00:37:09] Yeah.
[00:37:10] Paul Roetzer: you're cooked like that is, that is not a job one to two years out. so I think anything that has very narrowly defined role that is this, like these are the 10 to 15 tasks and they're all like AI's really good at all of them right now.
[00:37:31] and so I think a lot of times that is the entry level. That's why we're starting to see some early data that entry level jobs are very difficult right now because firms like ours, I've said this on the podcast, I would love to hire a ton of entry level people. Like I want to create job opportunities for students straight outta college.
[00:37:46] I don't know what they are right now, because when I have an idea to build something, say I wanna build a new app, or I wanna build a new, you know, score, you know, success score, whatever. When I do that, I will just then [00:38:00] go in and say, okay, great, we finished it. Now write the landing page, write me the emails, do all these things, and I'm gonna hand that to the marketing team.
[00:38:06] And so I hand the marketing team, a almost fully baked campaign that they just need to edit and execute.
[00:38:14] Mike Kaput: Yeah.
[00:38:14] Paul Roetzer: So all the work that used to get pushed down as the CEO, I do it in like seven minutes. That would've taken seven weeks for the team to do. And so once you have. You know, managers, directors, VPs, who realize they can just click a button and do most of the work the entry level people did, that's going to rapidly disrupt the job market.
[00:38:38] So I don't, I don't know, like, you know, copywriters obviously a role that's been under attack for a while here, and I think that's gonna continue to be, you're just gonna need fewer copywriters. You're gonna need AI forward copywriters. If you have two of those, that's the equivalent of like 20
[00:38:54] Mike Kaput: for sure.
[00:38:54] Paul Roetzer: Traditional copywriters. So I think that's what's gonna happen is you're just gonna have AI infused [00:39:00] into a lot of roles. They will evolve, may not have AI in the title, they're just gonna become AI forward versions of whatever that role was, and they're gonna be able to 10 x and I honestly don't even think 10 x is an exaggeration
[00:39:13] Mike Kaput: at all.
[00:39:14] No, it's not.
[00:39:14] Paul Roetzer: Based on what Mike and I are seeing every day in our own company.
[00:39:18] Mike Kaput: Yeah, and I would, I would just add to that something increasingly goes through my mind, and you know, this may be uncomfortable to say, but not in a negative way. But it's like with the entry level thing, with the kind of like tasks thing.
[00:39:29] If I'm the one who has to sit here and give you the workflow or the series of steps, it's like increasingly irresponsible of me to do that for a human. That's got to be so something I should be giving to an agent that can then scale, not to replace a human, but like if that is your job, if you're like, well, okay, I take the steps my manager gives me and go and do those, that's a really dangerous place to be in.
[00:39:55] Right?
[00:39:55] Paul Roetzer: Yeah. And I think, you know, we, we talked about OpenClaw a few times on the podcast. It's, it's been a little bit [00:40:00] more of a technical topic, so we haven't gone super deep on it, but we, maybe we should. Connect the dots a little bit better on an upcoming episode.
[00:40:08]
[00:40:08] Paul Roetzer: The way I think about this is whatever they're doing right now with that, you know, building these swarms of agents that you can just put a task, like direct it to a knowledge base and it can just go do the thing.
[00:40:18] It is only a matter of months until like you can see Claude Cowork would be an early example of this. Yes, it is just a matter of months until these SaaS companies are selling marketing agent swarms. Here's your marketing team out of the box. It's got an a media buying agent and a copywriting agent and like all these things that used to live in some of these SaaS companies as templates.
[00:40:42] Like, hey, go in and or GPT. Basically. Just imagine like you're just paying for a marketing team and like maybe they sell it for 250,000 a year or whatever, but it's everything you need. Just plug in. And these are. Agents that have harnesses attached to them that like give them, what tools do they get access to?
[00:40:59] What are [00:41:00] the system prompts? And then you're, you're literally just buying teams. And I 'm honest to God, I'm not at exaggerating. Like I think by the end of this year I could absolutely see companies starting to sell their software in that way where they just pre-bake agent swarms to do specific things.
[00:41:18] And that is gonna be very disruptive, but it is a hundred percent coming. Like that is. Again, like when I think about things I'll say on the podcast, I generally only say things like, I have a high level of conviction of, yeah, yeah. That one is like. On a one to 10 scale, I'm at like an eight or a nine that by the end of this year, you won't be able to do that.
[00:41:35] I'm, you could do it right now. I mean, Mike, you, if you and I had a week, right, we could turn Claude Cowork into that. Like, this is not hard.
[00:41:41] Mike Kaput: This is functionally what we are doing piece by piece. It's stuff like Claude Co or Claude Cowork when you're building these skills to train it, to do these things increasingly autonomously.
[00:41:50] It's just harder to do because it's not integrated right into the systems and the software you pay for every day.
[00:41:56] Paul Roetzer: Coding is the canary in the coal mine. We've said this many times, like it's, everything is [00:42:00] happening in coding first, and then it very quickly comes to the rest of knowledge work.
[00:42:04] Mike Kaput: And I think this could happen even faster because half the battle here is these software companies just trying to figure out the pricing of it.
[00:42:12] It's like once they crack the code on that, we're gonna start seeing this happen.
[00:42:15] Paul Roetzer: and then honestly, how do you position that? So like, like let's say you're a CRM company that sells software and you're, you sell licenses to marketing teams. And you find a way to use your CRM as the source of truth, and you realize you can build agents on top of Claude or openAI's or Gemini or whatever you're building the agents on top of, and you can create these agents swarms that function as a marketing team.
[00:42:38]
[00:42:38] Paul Roetzer: How do you possibly go to market with that message? Because it is a pure replacement play. Yep. We are for X dollars, instead of buying a hundred seats, we're gonna charge you, you know, a hundred thousand dollars a month or whatever that number is. And you're gonna have an in-house marketing team and all you really need is people to be the human in the loop and oversee and guide those agents and keep 'em [00:43:00] on track.
[00:43:00] And that's why I think like it's that that is where software companies, the legacy companies could get disrupted real fast is because their inability to say what has to be said that the AI native companies Yeah. Will have no problem saying that. Like a Silicon Valley startup totally could come in and say unapologetic about it.
[00:43:20] Yeah.
[00:43:20] They're gonna raise $50 million 'cause they're gonna go after. You know, let's say a $90 billion labor market of market. I don't know what the actual labor market for marketers is, but let's just make a number up and say it's 90 billion a year in payroll.
[00:43:31]
[00:43:31] Paul Roetzer: You just go after it and you get, you know, 90 billion, you get what?
[00:43:37] 10%? One 1% of that's 900 million, I think. Not bad. That's an E. That's an easy raise if you're trying to get money from VCs and they're, trust me, it's already happening like that. Yep. That's how they're doing it.
[00:43:50] Mike Kaput: All right, this next question,
[00:43:51] Mike Kaput: question 11 is a bit sales specific coming from the AI for Sales webinar.
[00:43:55] So they, this person asked when AI can start making sales calls, won't people associate [00:44:00] that with spam robocalls? I can't see how this would make a cold prospect trust my business because I wouldn't trust any company that had an AI call me. They can't be bothered to pick up the phone, so I can't be bothered to buy from them.
[00:44:13] So certainly one perspective. I am curious, Paul here, like how. Do you look at this because we are seeing a lot of companies start to experiment with this type of thing.
[00:44:20] Paul Roetzer: Yeah, I mean, I think it's just a pure numbers game. Cold calling. Most of us hate it. It does work at some small percentage and, you know, it's like that I 'm not pure sales person, but it's that, that the, how many calls do I make?
[00:44:34] And like we know my numbers. If I make a hundred calls, I'm gonna get three people to actually talk to me, or 10, whatever that number is. And now imagine it's a agent swarm that's doing this and they can make 10,000 calls in a day. And like now, now maybe you can just, so you're just playing percentage games.
[00:44:52] Yep. And yes, most people hate it and wouldn't trust that company, but if it's a company that isn't built on trust, which they're unfortunately aren't many of them [00:45:00] that are just in it to, you know, make money and drive those sales, then they're totally going to use. This method and just flood the market and just play the numbers.
[00:45:10] Game cost them way less to, you know, take shots on goal with a machine rather than a human. So I 'm sure this is already happening. Again, I'm, we don't live in this world, but I can almost guarantee you there are, I don't even wanna tell you unethical, like that's, that's probably not fair. There are companies that play this game, just the cold call, outbound spam game, and they play it well, and there's no way they're not playing it with AI right now.
[00:45:37] Mike Kaput: Yeah, for sure. And you know, we've talked about this on past episodes, it's like, okay, are we really talking about the issue here being AI or the fact you don't like being cold called, like you just mentioned. It's like, think about in a different context via chatbots. People don't care that chatbots exist.
[00:45:53] They care that they're bad. If they're good, if these calls are good and relevant and personal and less annoying, then [00:46:00] a human that knows nothing about you anyway, who knows.
[00:46:04] Paul Roetzer: Yeah. If, if they can target them based on needs, if they can get the right data set.
[00:46:09] Mike Kaput: Yeah.
[00:46:09] Paul Roetzer: And they can predict that you're someone who is, you know, a captive buyer of what they offer.
[00:46:14] Mike Kaput: Yeah.
[00:46:15] Paul Roetzer: And you get 'em at the right moment, then yeah, it's just like advertising. It's like a lot of times you turn off ads, but if it's something relevant to you, stop and listen. So,
[00:46:22] Mike Kaput: yeah. If
[00:46:22] Paul Roetzer: they're in the market at the right time and they can use predictive modeling to figure that out, then yeah, who's to say they're not gonna be successful doing it?
[00:46:29] Mike Kaput: Question number 12. If someone suddenly finds themselves with so many extra hours each week because AI is doing things like handling admin and support work, what strategies would you recommend to make sure that time is reinvested into real growth and competitive advantage rather than just more busy work?
[00:46:48] Paul Roetzer: We often guide to have a sandbox. So as you're working through your AI adoption plan, especially like when you think about scaling AI within an organization, the need to coach people [00:47:00] like one is you can give 'em some time back, like maybe Yep. They don't have to work as many hours. but the other is to, and you might have to do this in like a workshop model where you're helping people idea.
[00:47:10] But it's like, what other value can I be creating? What are the other projects that I can be working on? What are new ideas? Like maybe I can take 20% of my time savings, I can put it into innovation. Like that would be an amazing thing if we had a, you know, a 20% innovation budget for people's time, and you have innovation workshops every quarter and everybody comes up with innovation ideas.
[00:47:31] And as they save more time, they've got their wishlist that that management has signed off and prioritized of like, yeah, if you get time, like these are the innovations we're really excited about. Let's do these. That that would be a great use of it and be, and again, because I've said this many times, but to me the only way we slow down the job disruption is through innovation and growth.
[00:47:51] Mike Kaput: Yeah.
[00:47:51] Paul Roetzer: And so an innovation growth mindset. is essential. And so the idea of having an innovation sandbox, which now I'm saying this a lot, I've never [00:48:00] actually kind of verbalized it quite in this way. Yeah. And it makes a ton of sense that I'm saying it. We need these internally. having an innovation and growth sandbox of ideas that's like, Hey, I, the thing I thought was gonna take me all week, I actually just did, and it's, it's 11:00 AM on Monday.
[00:48:14] What do I do this week now? Innovation sandbox, like that would be, that would be a great use of it.
[00:48:19] Mike Kaput: Yeah. And there was a recent post actually from a software entrepreneur who basically was like, Hey, like months are now weeks and days, like with the ability of what AI enables. And he is kind of like, I'm doing our company planning and all this quarterly stuff, like we just did it in an afternoon.
[00:48:34] Yeah. so that innovation like. Getting the wheels turning in that way. You have to ask some kind of insane questions sometimes of like, Hey, this is my year goal. Could it be done in a day? You know, something like that.
[00:48:46] Paul Roetzer: Yeah. And this is, this is a hot topic for us 'cause we're having an annual meeting next week.
[00:48:49] Yeah. And we're doing an innovation workshop. Like I'm leading an AI innovation workshop. Mike's doing one on productivity. We're talking about rocks for the upcoming quarter. And it is like you have to, [00:49:00] things that seem crazy aren't anymore. Yeah. Whether they're the goals for the company. What can be achieved in a quarter.
[00:49:07] So what used to be like, all right, I'm gonna have these five rocks and this will get me through these next three months, and you know, if I accomplish this, that'll be great. And as a manager, as a leader, I'd be like, yeah, that, that would be great. But what if you did those in the first three weeks of the quarter, right?
[00:49:20] Instead,
[00:49:20] Mike Kaput: right.
[00:49:21] Paul Roetzer: Because I'm looking at 'em thinking those aren't three month projects anymore. And so I do think it again, is a mindset shift, but it's challenging people to think much bigger about what they can do.
[00:49:33] Mike Kaput: Question number 13, there is this larger conversation around the SaaS apocalypse, which we talked about on previous podcast episode.
[00:49:41] When do you personally think it makes sense to purchase software versus try to do things yourself or build things yourself with tools like Gemini, et cetera, for example, like if I want to analyze calls for call client sentiment, right? Like that's something AI might be able to do outta the box with the right prompt, or I may need.
[00:49:58] To consider buying [00:50:00] software for it. How do you look at that?
[00:50:02] Paul Roetzer: Yeah, I think generally you're still just gonna be buying software in a lot of these cases. I was trying to find, I'm looking for a tweet right now. There was a, there was something I saw, I think it was this morning. I couldn't sleep last night.
[00:50:15] I was up at like two in the morning when I was just looking at stuff. yeah, here it is. So this is, I think I put this on the list for us to talk about next week, Mike, but I'll just use this now. So, okay, so at a high level, you're, you're likely still just using the traditional tools, like there's a reason they're good.
[00:50:35] What the more likely scenario is how you use those tools and your, your pricing plan may evolve to the point where like, you're gonna have agents that are just gonna be logging in and extracting information, and you're not gonna need as many seat licenses. And so your relationship with that software company may change.
[00:50:52] But the reality is most companies aren't going and building a CRM. now, if they're distinct [00:51:00] narrow use cases where you can, you know, vibe, code something as someone with no coding ability, like, like me and Mike, then maybe you are like, I shared the example of like an org chart builder, right? Like I just couldn't find one, so I just built one myself and lovable.
[00:51:15] That's a more practical example, like these point solutions. Maybe they're for internal use only, stuff like that. Um hmm. So here is a tweet. This guy's, Todd Saunders, on, on X Broadloom, CEO of Broadloom. Okay. previously at Google. So he tweeted, we all knew this was coming, but today I heard about it actually happening.
[00:51:38] A seed stage company backed by a well-known vc, openly admitted in a board deck that their strategy is to get access to a large incumbent software from a customer clone. The entire coone, the entire thing, using Claude code and offer it for 90% less, not build something better, just copy it and offer it for less.[00:52:00]
[00:52:00] The VC endorsed this as the go to market strategy, huh? And even wrote back in writing that it was a good idea using a customer's licensed access to reverse engineer a product and clone it is ethical, ethically bankrupt. I don't know how else to put it. It likely violates terms of service. It may violate trade secret law as well, but I'm certainly not a lawyer and a reputable vc.
[00:52:23] Putting this in writing in a board deck is genuinely insane, but it's going to happen anyway everywhere all the time. I don't know where this ends, but we all knew this was coming and now it's here. So I don't know. Like I say, no one's gonna like. Code a c rm, but this is how the models work. It's how, right.
[00:52:41] You know, the rumor is that this is how the Chinese are currently, they're distilling,
[00:52:45] Mike Kaput: distilling,
[00:52:46] Paul Roetzer: yeah. Models from Gemini and Claude and Chet by just prompting hundreds of thousands of times and in essence, reproducing the weights and the code, how to build the model. So like you could certainly do, and so maybe in the US it's gonna be [00:53:00] illegal, but that doesn't mean other countries aren't gonna basically do that.
[00:53:04] They're just gonna get a license. I dunno. So
[00:53:08] Mike Kaput: yeah,
[00:53:09] Paul Roetzer: from an ethical perspective. Stick with the first answer. The second answer, or second con piece of context is to just give you a sense of what is actually happening in the world because it's kind of weird.
[00:53:19] Mike Kaput: And you know, one final point there we've talked about a bit in the past, I think is like, I also wonder, forget building or buying.
[00:53:27] I wonder also like how this will just change your expectations of your existing software. I'm already sometimes frustrated by stuff that's very valuable that we pay for. Because I'm like, I know that Claude or Gemini can give me a much better answer if I just had that layered over this data. It's like, why is your AI in this software we're paying for not as good as,
[00:53:47] Paul Roetzer: yeah.
[00:53:48] Mike Kaput: You
[00:53:48] know,
[00:53:48] Paul Roetzer: and I think that might be a good more narrow example to, to, to sort of stick on for a second, is, you know, imagine you have a CRM system and you wanna talk to your data.
[00:53:58] Paul Roetzer: And for whatever [00:54:00] reason they have yet to build the agent into it. That makes that super easy to do. So then you're like, screw it.
[00:54:05] I won't pay you on a token by token basis. Yeah. Or credits. However you wanna charge me to use your crappy agent. That doesn't gimme the information I want. I'm just gonna connect it to Gemini or Claude and I'll get the data myself and I'll just use that and I'm gonna pay them the tokens. So that's a more realistic thing is like these narrow plugins or use cases where you're just gonna like, I'm just gonna always track the data myself.
[00:54:28] I'm not gonna wait for this software company. Yeah. And if you do that enough times over, enough use cases, then maybe you don't need that software anymore.
[00:54:33] Mike Kaput: Yeah. All right.
[00:54:35] Mike Kaput: Question number 14. I'm nervous about others using AI agents irresponsibly. While it's connected to my personal information, is there anything we can do to protect ourselves from others who are basically implementing agents without guardrails?
[00:54:47] Paul Roetzer: Join the club on this one. I am, I'm also very worried about this. Yeah, there's, we'll share a story on Tuesday in the weekly. About like an all hands meeting at Amazon when their agents [00:55:00] apparently went haywire and racked a bunch of code at AWS. these things are not reliable and in many cases they are not safe and people are definitely racing ahead and using them regardless.
[00:55:13] And we, we all could be collateral damage in that grand experiment. I don't know how to protect ourselves other than the traditional ways we would monitor our personal information and our credit scores and things like that.
[00:55:29]
[00:55:29] Paul Roetzer: You know, having those fraud alerts set up personally, like it's probably a lot of the traditional stuff, it just becomes more important over time, I would imagine from a business perspective, you probably wanna be talking to your insurance agents about liabilities related to these things and how to protect yourself and your employees.
[00:55:45] If maybe your employees mistakenly use agents to do things like. It's a whole new world. and I think even asking this question is a good starting point for people.
[00:55:56] Mike Kaput: All right, our last question here, Paul.
[00:55:58] Mike Kaput: Question 15, when you are having conversations with leaders, what's your approach to communicate the fact that it shouldn't ideally be the one driving AI adoption in the organization?
[00:56:11] Paul Roetzer: It's not it's job it's job is not business strategy. It's not. Reskilling and upskilling people and dealing with change management. This, this isn't a technology thing. This is a complete business transformation that has to be fueled by AI literacy, you know, from the, from the leadership level down.
[00:56:33] And then you have to be able to personalize use cases. You have to be able to communicate with people who are afraid. Like there are so many layers to this that have nothing to do with it. Yeah. it is there to keep. People safe to use the technology responsibly to reduce risk, to make, make sure the data stays secure.
[00:56:53] Like they play a critical role. They are not the ones that should be telling marketing how to use Google [00:57:00] Gemini. Like that's not their job. So I just think you lay out what are the goals of our use of AI technology, what are the sample use cases, how are we gonna infuse it into workflows? And none of that is it's job.
[00:57:13] So I would just like lay out what needs to happen for true AI adoption and transformation. And it's very apparent at that point that it plays a critical role, but it is not to guide the strategy.
[00:57:26] Mike Kaput: All right, Paul. That's our 15 questions for this episode of AI Answers. again, if you need a reminder, go to SmarterX.ai/webinars and you can check out each of these awesome webinars.
[00:57:38] We did go to SmarterX.ai/blueprints You can go get an ungated copy of each of these blueprints for marketing, sales, and customer success in partnership with Google Cloud. We're so appreciative to them for making all this possible as well. So, Paul, thanks for my first, AI answers experience.
[00:57:55] Paul Roetzer: Cathy, the answers, but yeah, we, and just fyi like Kathy will be back, for the next AI [00:58:00] answers. But Mike and I did these webinars together. That's why Mike and I ended up doing this AI answers together. So Mike and I will be back with episode 203 of the podcast on Tuesday. So thanks for joining us for this special edition, and we will talk to you again next week.
[00:58:15] Thanks for listening to AI Answers to Keep Learning. Visit SmarterX.ai where you'll find on-demand courses, upcoming classes, and practical resources to guide your AI journey. And if you've got a question for a future episode, we'd love to hear it. That's it for now. Continue exploring and keep asking great questions about ai.