1 00:00:00,583 --> 00:00:04,543 Can AI powered filmmaking teach us to rethink business processes? 2 00:00:05,323 --> 00:00:09,293 Today's guest takes us on his journey from traditional video production 3 00:00:09,583 --> 00:00:11,703 to crafting AI powered films. 4 00:00:12,423 --> 00:00:16,113 Along the way, he shares how tools like custom GPTs didn't just 5 00:00:16,113 --> 00:00:20,483 streamline his creative workflows, they reshaped the way he tells stories. 6 00:00:21,093 --> 00:00:24,603 If you're looking for inspiration to automate your own processes and 7 00:00:24,603 --> 00:00:28,643 unlock new possibilities with AI while staying grounded in what works, 8 00:00:28,973 --> 00:00:32,733 this conversation is packed with practical insights you can put to use. 9 00:00:33,293 --> 00:00:37,373 My name is Alexandre Nevski and this is Innovation Tales. 10 00:00:52,426 --> 00:00:56,286 From creating Emmy winning videos to pioneering generative AI tools, 11 00:00:56,556 --> 00:00:59,976 Erich Archer has spent nearly two decades at the intersection 12 00:00:59,976 --> 00:01:01,626 of storytelling and technology. 13 00:01:02,416 --> 00:01:07,216 But a year ago, he took a bold leap, creating his first AI powered short film. 14 00:01:07,746 --> 00:01:08,186 Why? 15 00:01:08,736 --> 00:01:13,716 To explore how tools like custom GPTs, mid journey and runway could revolutionize 16 00:01:13,726 --> 00:01:18,966 creative workflows, making high quality production more accessible than ever. 17 00:01:19,856 --> 00:01:23,816 Today, Erich joins us to share not only how these tools are reshaping 18 00:01:23,816 --> 00:01:27,536 the media industry, but also how they can be applied to transform 19 00:01:27,536 --> 00:01:30,126 business processes across industries. 20 00:01:30,986 --> 00:01:35,851 If you've ever wondered how to turn AI hype into real actionable outcomes, 21 00:01:36,181 --> 00:01:40,171 whether it's automating repetitive tasks, improving creative consistency, 22 00:01:40,401 --> 00:01:45,081 or just scaling innovation without scaling your team, this episode may 23 00:01:45,091 --> 00:01:48,851 challenge the way you think about integrating technology into your work. 24 00:01:49,851 --> 00:01:54,791 Erich's hands on experience proves that AI isn't about replacing jobs. 25 00:01:55,121 --> 00:01:59,261 It's about unlocking new ways to think, create, and connect. 26 00:01:59,791 --> 00:02:03,071 Without further ado, here's my conversation with Erich Archer. 27 00:02:03,399 --> 00:02:04,649 Erich, welcome to the show. 28 00:02:04,818 --> 00:02:05,508 Thanks for having me. 29 00:02:05,518 --> 00:02:06,228 Happy to be here. 30 00:02:06,539 --> 00:02:09,489 Well, I'm super excited about our conversation today. 31 00:02:09,969 --> 00:02:15,059 I expect it to be a little different because you come from the film and TV 32 00:02:15,059 --> 00:02:19,759 industry, but a year ago you made your first film using artificial intelligence. 33 00:02:20,249 --> 00:02:23,919 So what inspired you to explore AI tools for video production? 34 00:02:24,401 --> 00:02:29,591 Well, uh, by day I run a nonprofit TV station and so I was very excited 35 00:02:29,591 --> 00:02:34,231 about ChatGPT for the resource gain that I felt it presented, you know, as 36 00:02:34,231 --> 00:02:38,411 a nonprofit executive director, we're always a little starved for resources. 37 00:02:38,411 --> 00:02:42,071 So it's like, Oh wow, this can help me work better, faster. 38 00:02:42,306 --> 00:02:44,196 You know, uh, output more. 39 00:02:44,196 --> 00:02:45,836 And so that was attractive. 40 00:02:45,836 --> 00:02:49,036 And then mid journey came out the image generation platform. 41 00:02:49,606 --> 00:02:53,446 And because I also work in a visual medium, that was very attractive too. 42 00:02:53,586 --> 00:02:58,186 I mean, you, you're telling me I can generate any image I can think of like, 43 00:02:58,186 --> 00:03:03,346 that's a really big deal when those are assets that I work with day-to-day. 44 00:03:03,836 --> 00:03:09,756 So it was very clear early on that those tools were going to change my industry. 45 00:03:10,151 --> 00:03:10,851 In a big way. 46 00:03:10,908 --> 00:03:11,168 Yeah. 47 00:03:11,188 --> 00:03:13,768 Making projects more accessible, I guess. 48 00:03:13,888 --> 00:03:14,328 Very much. 49 00:03:14,328 --> 00:03:15,688 So, yeah, very much so. 50 00:03:15,688 --> 00:03:19,398 And production workflows are starting to change really quickly. 51 00:03:19,498 --> 00:03:23,258 Just leveraging these tools for little efficiencies and a 52 00:03:23,268 --> 00:03:26,848 little bit of, you know, extra research or rigor here and there. 53 00:03:26,848 --> 00:03:28,428 And then, oh, maybe these. 54 00:03:28,788 --> 00:03:31,998 Image generators can help with storyboarding and things like that 55 00:03:31,998 --> 00:03:36,048 and processing different creative briefs or whatever the case may be. 56 00:03:36,048 --> 00:03:41,728 And, slowly by slowly, but surely sort of starting to understand some of the 57 00:03:41,728 --> 00:03:46,068 different ways these different tools could help me and it started to become 58 00:03:46,068 --> 00:03:51,948 very clear very quickly that generative production was a growing space that 59 00:03:52,313 --> 00:03:56,333 know, these images were just the tip of the iceberg and that it was going 60 00:03:56,333 --> 00:04:01,943 to be moving toward generative video production and just more sophisticated 61 00:04:02,123 --> 00:04:03,583 versions of everything we were seeing. 62 00:04:04,023 --> 00:04:08,023 So I wanted to get in early and start to experiment and learn 63 00:04:08,023 --> 00:04:09,603 it and, and get a feel for it. 64 00:04:10,153 --> 00:04:14,993 And so the first short film you made is about your great, great, great 65 00:04:14,993 --> 00:04:16,753 grandfather, if I'm not mistaken. 66 00:04:17,253 --> 00:04:18,853 And so what's the story behind that? 67 00:04:19,300 --> 00:04:24,000 I had the opportunity to enroll in a month long fellowship program at an 68 00:04:24,210 --> 00:04:29,060 organization called the AI exchange and what was great about that was they said 69 00:04:29,520 --> 00:04:31,870 you have to make a personal AI project. 70 00:04:32,697 --> 00:04:38,157 That gave me the push to put the time and effort into really try to 71 00:04:38,157 --> 00:04:43,677 make something start to finish, that was good, that I may not have taken 72 00:04:43,727 --> 00:04:45,797 otherwise had I not been in that program. 73 00:04:46,027 --> 00:04:52,107 My goal was really, come up with a project that could demonstrate the, 74 00:04:52,207 --> 00:04:56,727 the level of storytelling that the technology was enabling at that time. 75 00:04:57,577 --> 00:05:02,677 So it was sort of a proof of concept more than a short film. 76 00:05:02,916 --> 00:05:07,526 But I had encountered some of my dad's ancestry research about our family 77 00:05:07,766 --> 00:05:12,506 and he had a stack of notes and dates and names and some photos, and one of 78 00:05:12,506 --> 00:05:14,696 like the family trees that you see. 79 00:05:15,296 --> 00:05:20,926 So I started to kind of use that as the creative briefs and, uh, I told the story 80 00:05:20,926 --> 00:05:26,036 about an ancestor that I had the most information about: this guy named Rufus, 81 00:05:26,036 --> 00:05:28,226 who is my great, great, great grandfather. 82 00:05:28,226 --> 00:05:32,866 And I had two pictures of him: one in his volunteer firefighter uniform. 83 00:05:32,970 --> 00:05:34,895 Another of his, like a head shot. 84 00:05:34,915 --> 00:05:39,465 And then I had a picture of his gravestone and I knew what he did for a living. 85 00:05:39,475 --> 00:05:43,735 He made barrels and he had four kids and I knew where he lived and when. 86 00:05:43,735 --> 00:05:50,075 And so I was able to build a timeline of his life and understand this area 87 00:05:50,095 --> 00:05:55,015 and kind of infer, with ChatGPT and Perplexity's help, what his life was like. 88 00:05:55,615 --> 00:05:59,485 And I made a custom GPT to actually be able to interview him and use 89 00:05:59,705 --> 00:06:02,735 the material from that exchange and the narration and stuff. 90 00:06:02,735 --> 00:06:08,705 And I would use the voices from Elevenlabs (the AI audio tool). 91 00:06:08,705 --> 00:06:10,405 And that gave me all the sound effects. 92 00:06:10,405 --> 00:06:13,405 And so I was able to just. 93 00:06:13,745 --> 00:06:18,105 Use a bunch of different tools and generate images that looked a lot like 94 00:06:18,105 --> 00:06:23,435 him, a lot like Salem, in sort of a Ken Burns, very simple documentary style. 95 00:06:23,885 --> 00:06:27,055 That was an example of what I felt at the time was where, where the 96 00:06:27,055 --> 00:06:29,105 tools would allow you to work. 97 00:06:29,612 --> 00:06:30,392 It looks amazing. 98 00:06:30,392 --> 00:06:32,772 And of course we'll provide a link in the description. 99 00:06:32,991 --> 00:06:37,381 There's a number of other videos that are in a different style. 100 00:06:37,391 --> 00:06:39,031 I think some of them are commissions. 101 00:06:39,131 --> 00:06:40,231 Can you talk a bit about them? 102 00:06:40,676 --> 00:06:41,036 Yeah. 103 00:06:41,036 --> 00:06:45,886 So I started to get a little bit more attention for the, that work 104 00:06:45,976 --> 00:06:51,256 and got in touch with an agency that had some clients interested in 105 00:06:51,476 --> 00:06:56,096 early AI pioneering work, you know, let's, let's try to make some stuff. 106 00:06:56,096 --> 00:06:58,776 It was a lot of um, cryptocurrency companies. 107 00:06:58,776 --> 00:07:01,853 They seem to be, interested in pushing the envelope. 108 00:07:01,853 --> 00:07:03,763 They don't have a lot of visuals to begin with. 109 00:07:04,393 --> 00:07:06,813 They're, I don't know, adventurous, I guess. 110 00:07:07,173 --> 00:07:09,563 So, there were a few of those. 111 00:07:09,593 --> 00:07:13,733 And what's so interesting about the tools is they make you so versatile 112 00:07:14,026 --> 00:07:22,076 as a creator, because you can pull in any inspiration directly into the 113 00:07:22,076 --> 00:07:29,466 process, whether it's black and white photographs from the 1800s or graffiti 114 00:07:30,106 --> 00:07:37,846 skateboarding images from today, or in another piece I made, 35, 000 115 00:07:37,846 --> 00:07:40,886 years in the past, it doesn't matter. 116 00:07:41,006 --> 00:07:43,966 You can now generate these assets and work with them. 117 00:07:44,806 --> 00:07:48,656 So I've made, I don't know, five or six AI videos now, and they're all 118 00:07:48,696 --> 00:07:51,998 very different aesthetically, which is exciting, you know, it really makes 119 00:07:51,998 --> 00:07:53,118 you feel like anything's possible. 120 00:07:53,647 --> 00:07:58,667 How did you initially start integrating these various tools into the 121 00:07:58,667 --> 00:08:00,247 workflows that you're familiar with? 122 00:08:00,507 --> 00:08:02,787 Was that particularly easy? 123 00:08:02,797 --> 00:08:03,757 Straightforward? 124 00:08:03,797 --> 00:08:04,457 Not so much? 125 00:08:04,665 --> 00:08:08,155 One of the earliest things that I discovered was that I could take a 126 00:08:08,155 --> 00:08:13,335 transcription from an interview that I would do like this, but with video, 127 00:08:13,425 --> 00:08:17,105 video cameras going out and shooting interviews, coming back with a day's full 128 00:08:17,105 --> 00:08:22,735 of transcriptions, being able to run those through Claude and say something like 129 00:08:23,525 --> 00:08:28,945 "I need a soundbite that is emotionally engaging, and then I need the who, what, 130 00:08:28,945 --> 00:08:33,615 when, where, why information soundbites, and then I need an emotional closer," and 131 00:08:33,615 --> 00:08:40,025 it'll give you that paper edit with time codes, and that, that process of going 132 00:08:40,025 --> 00:08:43,889 from the raw material of an interview or a transcription down to what you 133 00:08:43,889 --> 00:08:46,469 want to use can be a very long process. 134 00:08:46,479 --> 00:08:50,679 I mean, we'll come back with hours and hours of interviews. 135 00:08:51,369 --> 00:08:53,079 So something like that. 136 00:08:53,569 --> 00:08:57,926 All of a sudden, you're you're going from something that might have taken 3, 137 00:08:57,926 --> 00:09:04,615 4 rounds of editing down, you do almost instantly and it might not be 100 percent 138 00:09:04,935 --> 00:09:07,825 what you want, but it's way closer. 139 00:09:08,155 --> 00:09:14,855 Or if you need a B-roll, now you can generate it instead of buy it or shoot it. 140 00:09:15,635 --> 00:09:20,165 It's getting harder and harder to kind of figure out what we would need to shoot 141 00:09:20,195 --> 00:09:22,405 with cameras, as opposed to generate. 142 00:09:22,497 --> 00:09:26,647 There's some obvious ones, like if you're going to go shoot something, a live event 143 00:09:26,647 --> 00:09:28,427 or a sports game or something like that. 144 00:09:28,427 --> 00:09:32,877 But I can make a spice commercial using spice B-roll from all around 145 00:09:32,877 --> 00:09:37,007 the world without flying any camera operators all around the world. 146 00:09:37,110 --> 00:09:38,150 It's just amazing. 147 00:09:39,195 --> 00:09:45,035 Yeah, there is an immediacy where you can turn an idea into something 148 00:09:45,095 --> 00:09:54,310 usable with much less process, let's say, you don't need as big of a team 149 00:09:54,370 --> 00:10:00,580 in some cases, that's my experience, which enables you to also run with 150 00:10:00,580 --> 00:10:02,910 ideas a bit further on your own. 151 00:10:03,090 --> 00:10:09,765 But I know that you've also had to use image-to-video rather 152 00:10:09,765 --> 00:10:11,465 than directly text-to-video. 153 00:10:12,015 --> 00:10:17,337 But with the announcement of Sora, or others, everyone's talking about it. 154 00:10:17,410 --> 00:10:21,270 An example is just like I get a creative brief from a brand and the 155 00:10:21,270 --> 00:10:29,687 brand says "we we have an aesthetic we need you to create that is, ancient 156 00:10:29,687 --> 00:10:32,277 Greece, 300 years in the future. 157 00:10:33,297 --> 00:10:37,537 And our brand is purple and we need these magical orbs. 158 00:10:38,107 --> 00:10:43,791 How am I going to get that once, let alone in every single shot in a sequence 159 00:10:43,851 --> 00:10:46,971 that's maybe 25 different shots. 160 00:10:48,626 --> 00:10:51,106 So that's that's the challenge you start with. 161 00:10:51,176 --> 00:10:55,676 Okay, I've got to make this look and it's got to be brand exact. 162 00:10:55,956 --> 00:10:56,976 We're making a commercial. 163 00:10:56,976 --> 00:10:58,146 This can't be close. 164 00:10:58,146 --> 00:10:59,196 It has to be exact. 165 00:10:59,269 --> 00:11:01,429 It has to really be consistent. 166 00:11:01,429 --> 00:11:02,919 So that's the bar. 167 00:11:03,419 --> 00:11:04,179 You know what I mean? 168 00:11:04,429 --> 00:11:09,544 You can't have anyone watch it and go what the, what, you know, what was that? 169 00:11:10,044 --> 00:11:13,614 There's a couple of ways you can approach it basically, which is you can 170 00:11:13,734 --> 00:11:20,804 go to a platform like Sora and you can just give it a text prompt, you know, 171 00:11:20,934 --> 00:11:26,699 I'm looking for ancient Greece with a futuristic vibe and a magical purple 172 00:11:26,739 --> 00:11:34,639 orb and try to prompt out a video of that idea and you'll see what you get. 173 00:11:34,752 --> 00:11:36,292 So that's, that's the difference. 174 00:11:36,292 --> 00:11:43,582 It's like, in Midjourney, you can upload a picture, or multiple pictures, and 175 00:11:43,582 --> 00:11:52,107 try to craft the image, out of pictures, which is more typically how I will work. 176 00:11:52,204 --> 00:12:00,774 I'm, probably for every 1000 images I generate, using 50 of them. 177 00:12:02,889 --> 00:12:06,789 Yeah, there is a lot of trial and error. 178 00:12:07,067 --> 00:12:10,917 How much difference is there between what's on the package, 179 00:12:10,927 --> 00:12:12,487 what's promised and reality? 180 00:12:12,552 --> 00:12:16,262 I think that that's really one of the biggest takeaways that I could share 181 00:12:16,262 --> 00:12:21,172 is that it's a very big difference between the highly cherry picked 182 00:12:21,242 --> 00:12:28,472 outputs you see in the promotional pieces and the real outputs you get 183 00:12:29,017 --> 00:12:32,097 in platform when you go to use it. 184 00:12:32,997 --> 00:12:35,437 They're all pretty wonky still. 185 00:12:35,817 --> 00:12:39,997 They just give you all kinds of weird stuff and errors. 186 00:12:40,037 --> 00:12:42,097 And sometimes it's great. 187 00:12:42,317 --> 00:12:46,627 I mean, I've used a lot of unexpected outputs in my work. 188 00:12:46,897 --> 00:12:52,722 So there's a real coolness in the unexpectedness of it. 189 00:12:52,732 --> 00:12:57,852 Sometimes that can be useful, but most of the time it's just like, Oh my God, it, 190 00:12:57,912 --> 00:13:03,992 you know, just this absolute battle with the platform to try to give you something. 191 00:13:03,992 --> 00:13:06,222 And sometimes you'll get tantalizingly close. 192 00:13:06,262 --> 00:13:09,832 Like it's exactly the image you want, but it's got six fingers or something. 193 00:13:09,832 --> 00:13:13,602 And you're like, there is no way to undo that one little piece. 194 00:13:13,602 --> 00:13:18,032 Or sometimes there is, with painting and stuff, but like, it all depends on what 195 00:13:18,032 --> 00:13:19,702 platform you're using and everything. 196 00:13:19,702 --> 00:13:28,072 So I think, somebody really smart told me early on not to let all of that hype 197 00:13:28,082 --> 00:13:34,592 stuff that's constantly coming out stress me out because all of those companies 198 00:13:34,872 --> 00:13:38,732 are, uh, cherry picking in a major way. 199 00:13:39,262 --> 00:13:44,672 That's been my experience also that, you know, I go and I, I try all these tools. 200 00:13:44,782 --> 00:13:45,582 Oh, this is great. 201 00:13:45,582 --> 00:13:47,222 New things come out and look what it can do. 202 00:13:47,222 --> 00:13:53,637 And I run over and try it and I'm like, oh, okay, it's, it's, it's still just 203 00:13:53,647 --> 00:13:56,247 as rough to work with as the rest of it. 204 00:13:56,927 --> 00:14:00,187 Um, and hopefully that's freeing for people to hear. 205 00:14:00,187 --> 00:14:01,027 It was for me. 206 00:14:01,057 --> 00:14:04,597 It's like, okay, I'm not doing anything wrong. 207 00:14:04,887 --> 00:14:05,977 Like this is hard. 208 00:14:06,867 --> 00:14:07,567 It's hard. 209 00:14:07,677 --> 00:14:11,277 And I'm, I've been doing this work a long time. 210 00:14:11,277 --> 00:14:13,437 So there's parts of it that I'm very fast at. 211 00:14:14,102 --> 00:14:18,182 Like editing and I can, I know what elements I need for any given video pretty 212 00:14:18,182 --> 00:14:19,672 quickly and certain things like that. 213 00:14:19,882 --> 00:14:22,342 So I can still go fast and it's still hard. 214 00:14:23,242 --> 00:14:26,592 You know, it's, is my core competency and it's still hard. 215 00:14:27,532 --> 00:14:33,482 So, you know, it can look like this very 1 click and oh, my God, it's 216 00:14:33,482 --> 00:14:37,282 all done for you and beautiful, but it's not like that at all. 217 00:14:38,492 --> 00:14:38,752 You know? 218 00:14:40,275 --> 00:14:42,595 The human in the loop is definitely a must have. 219 00:14:42,615 --> 00:14:48,678 But I know that you've also managed to use some of the tools we've already 220 00:14:48,678 --> 00:14:51,738 mentioned in a more reliable way. 221 00:14:51,748 --> 00:14:53,858 Would you mind talking about this for a bit? 222 00:14:54,158 --> 00:15:00,438 So pretty early on, I found a use case for custom GPTs, that was pretty 223 00:15:00,438 --> 00:15:08,008 powerful, where I was getting these creative brief documents and able to use 224 00:15:08,008 --> 00:15:11,288 ChatGPT to process the language in them. 225 00:15:11,578 --> 00:15:15,088 Basically, these were documents full of insider knowledge 226 00:15:15,088 --> 00:15:16,998 that I wanted to understand. 227 00:15:17,818 --> 00:15:21,088 But it was vague language and sort of difficult to figure out 228 00:15:21,088 --> 00:15:22,418 what they meant all the time. 229 00:15:22,668 --> 00:15:25,488 And I'm like, this is probably something ChatGPT can help me with, you know. 230 00:15:25,498 --> 00:15:30,788 Put this PDF in there and be like, you know, clean up this language and add some 231 00:15:30,798 --> 00:15:33,368 brand guidelines and do some research. 232 00:15:33,368 --> 00:15:36,758 And I had all these ideas of how I could like process this document 233 00:15:36,768 --> 00:15:38,683 really quickly using ChatGPT. 234 00:15:39,138 --> 00:15:40,518 And a lot of it worked. 235 00:15:41,128 --> 00:15:44,758 And then I tried to build this like super custom GPT that 236 00:15:44,758 --> 00:15:46,823 would just do it all the time. 237 00:15:46,823 --> 00:15:47,993 And that didn't work. 238 00:15:48,943 --> 00:15:52,778 And in the process of trying to figure out why it didn't work, I just wound 239 00:15:52,778 --> 00:15:58,788 up making like 10 GPTs out of the one, because what I realized is I was 240 00:15:58,788 --> 00:16:00,918 asking it to do 10 different things. 241 00:16:01,378 --> 00:16:06,608 And if I just asked 10 different custom GPTs to do one thing each, I'd 242 00:16:06,608 --> 00:16:09,198 get a more reliable output each time. 243 00:16:09,198 --> 00:16:14,968 And then I, I didn't mind manually going from one to the next to like clean up 244 00:16:14,968 --> 00:16:19,648 the document and then add some research and then add some brand guidelines 245 00:16:19,648 --> 00:16:24,028 and, you know, make some thumbnails and generate ideas and, you know, 246 00:16:24,388 --> 00:16:26,258 all the stuff that I wanted it to do. 247 00:16:26,258 --> 00:16:27,948 I just had to do one at a time. 248 00:16:28,358 --> 00:16:33,178 And that was effective, but also enlightening. 249 00:16:33,178 --> 00:16:35,948 I was like, Oh, that would work in other areas too. 250 00:16:36,488 --> 00:16:40,838 And so I started to apply that idea of just making these little precision 251 00:16:40,858 --> 00:16:44,928 custom GPTs that did one thing well. 252 00:16:46,208 --> 00:16:51,028 And I've now made a couple hundred of those and kind of map them all out on this 253 00:16:51,438 --> 00:16:54,508 mind map platform online that I found. 254 00:16:54,558 --> 00:16:58,848 So that's become a fun process and something that I just sort of treat, like, 255 00:16:58,848 --> 00:17:03,318 as a hobby, but people are using them now. 256 00:17:03,318 --> 00:17:06,188 And, like, they're available on Gumroad for free. 257 00:17:06,188 --> 00:17:09,018 If anyone would like to try them out, but they're for all 258 00:17:09,588 --> 00:17:12,338 different business frameworks and models and things like that. 259 00:17:12,348 --> 00:17:17,453 That when I took over this business, my day job as an executive director, 260 00:17:17,453 --> 00:17:19,103 I had never been in management before. 261 00:17:19,103 --> 00:17:22,973 So I didn't know a lot of these frameworks and things existed or how to do them. 262 00:17:23,523 --> 00:17:26,923 This is designed to make it very easy to do those things and not have to pay a 263 00:17:26,953 --> 00:17:28,953 consultant to teach you how to do them. 264 00:17:29,493 --> 00:17:33,713 I also, I was a startup founder for a few years, too, and I had 265 00:17:34,483 --> 00:17:36,603 friends and family money invested. 266 00:17:36,703 --> 00:17:38,063 I built an app. 267 00:17:38,063 --> 00:17:40,973 I was trying really hard, but like, there was so much I didn't know. 268 00:17:40,973 --> 00:17:43,803 And I knew I didn't know it because I was the 1st time founder. 269 00:17:43,803 --> 00:17:46,633 And I was like, man, I hope I'm just not, I hope I'm not making 270 00:17:46,633 --> 00:17:47,933 mistakes because I don't know. 271 00:17:48,598 --> 00:17:50,148 The right way to look at this or whatever. 272 00:17:50,148 --> 00:17:56,758 So I just have had that experience where like, I just want to know how to think 273 00:17:56,758 --> 00:17:58,168 about things, you know what I mean? 274 00:17:58,168 --> 00:18:00,178 So I made all these tools to just help with that. 275 00:18:00,588 --> 00:18:03,088 Like, Oh, this is a SWOT analysis. 276 00:18:03,298 --> 00:18:04,488 I just have to click this button. 277 00:18:04,508 --> 00:18:05,348 It'll walk me through it. 278 00:18:05,398 --> 00:18:06,258 It'll do it for me. 279 00:18:07,030 --> 00:18:07,930 Yeah, that's amazing. 280 00:18:08,056 --> 00:18:11,356 Yeah, you don't ask it to do the work for you. 281 00:18:11,356 --> 00:18:14,836 You, you, You really have to be very specific. 282 00:18:15,056 --> 00:18:21,229 But I really am curious, cause I'm a techie, so I know what's behind them, but 283 00:18:21,479 --> 00:18:29,154 when you talk about those custom GPTs and the processes that you are able to 284 00:18:29,194 --> 00:18:33,814 automate partially with them, how do you explain what the custom GPT is, I wonder? 285 00:18:34,350 --> 00:18:39,970 Well, so ChatGPT is a general purpose tool and you can ask it anything 286 00:18:40,620 --> 00:18:42,760 and get an answer, which is amazing. 287 00:18:43,440 --> 00:18:49,720 But custom GPTs offer you a way to make a much smaller version of it. 288 00:18:50,830 --> 00:18:56,380 You basically can control all of the parameters of how you want it to 289 00:18:56,390 --> 00:18:57,720 behave and what you want it to do. 290 00:18:58,240 --> 00:19:01,320 Just through conversation, which is pretty wild. 291 00:19:01,410 --> 00:19:03,130 I don't know how to code, at all. 292 00:19:03,580 --> 00:19:11,020 You can just be like, Hey, I want to make a custom GPT that 293 00:19:11,950 --> 00:19:13,820 helps me do a SWOT analysis. 294 00:19:15,533 --> 00:19:15,990 Okay. 295 00:19:15,990 --> 00:19:21,860 Well, what would a tool need to do in order to do that? 296 00:19:22,020 --> 00:19:24,310 Like it would have to ask you about your business. 297 00:19:24,850 --> 00:19:26,030 What information would it need? 298 00:19:26,030 --> 00:19:26,250 Right? 299 00:19:26,270 --> 00:19:30,170 Like, if it had all of this information, if you could provide all this raw 300 00:19:30,170 --> 00:19:36,930 information, it would be able to tell you with some web searching, also, you'd 301 00:19:36,960 --> 00:19:38,909 be able to help you do a SWOT analysis. 302 00:19:39,050 --> 00:19:39,580 You know what I mean? 303 00:19:39,680 --> 00:19:47,740 It's basically just putting whatever parameters in place that you want, using 304 00:19:47,750 --> 00:19:52,965 the power of this general purpose tool to just look through whatever scope 305 00:19:52,965 --> 00:19:55,815 you want to sort of create for it. 306 00:19:56,677 --> 00:20:01,587 And I love what you were saying about using tools like this when you're like, 307 00:20:01,597 --> 00:20:07,387 in your case, a first time founder and needing to learn to do a whole set of 308 00:20:07,397 --> 00:20:13,317 things that this can be a complement, let's say, where maybe before you'd 309 00:20:13,367 --> 00:20:19,872 have to go and read a book, maybe get a consultant, do a training or something. 310 00:20:20,210 --> 00:20:26,030 We're seeing now the rise of these AI-powered tools as kind 311 00:20:26,030 --> 00:20:29,010 of a learning upskilling tool. 312 00:20:29,420 --> 00:20:35,101 I wonder from your experience, like if you were talking to other business 313 00:20:35,101 --> 00:20:37,604 owners, how would you advise them? 314 00:20:37,604 --> 00:20:39,689 What would you tell them? 315 00:20:39,689 --> 00:20:43,443 Like, how should they be exploring these custom GPTs? 316 00:20:43,493 --> 00:20:48,163 Well, they're certainly welcome to jump in to, uh, to my 317 00:20:48,433 --> 00:20:50,613 ecosystem and try these things. 318 00:20:50,613 --> 00:20:54,963 There's also a number of custom GPTs available within ChatGPT. 319 00:20:54,983 --> 00:20:59,423 You can go in there and search by category and they'll have probably, you know, 320 00:20:59,423 --> 00:21:02,583 if you're in real estate, they'll have a whole bunch of real estate, custom 321 00:21:02,583 --> 00:21:04,513 GPTs, and you can check those out. 322 00:21:05,331 --> 00:21:12,501 I highly recommend just starting with a conversation in ChatGPT. 323 00:21:13,431 --> 00:21:18,341 What is something that you think about on a regular basis 324 00:21:18,371 --> 00:21:20,221 or process on a regular basis? 325 00:21:20,231 --> 00:21:24,431 So we're looking for, like, repetitive tasks and things like that. 326 00:21:24,921 --> 00:21:27,821 Or thought patterns that you nor your things you normally 327 00:21:27,821 --> 00:21:28,631 thinking about, whatever. 328 00:21:29,216 --> 00:21:33,358 And then, what would go into, like, building a machine 329 00:21:33,358 --> 00:21:34,768 to just do that for you? 330 00:21:35,333 --> 00:21:38,723 You know, um, it's it's often not very much. 331 00:21:38,723 --> 00:21:40,931 I'll give you an example of like a really simple one. 332 00:21:41,788 --> 00:21:48,659 For my day job here, all I did, was take our organizational documents, 333 00:21:49,089 --> 00:21:55,249 our handbook, our policies and procedures, our brand guidelines, 334 00:21:55,789 --> 00:21:56,749 you know, things like that. 335 00:21:57,129 --> 00:22:02,639 And I just put them as PDFs in a custom GPT, which is as simple as uploading. 336 00:22:03,449 --> 00:22:04,129 Here you go. 337 00:22:05,519 --> 00:22:12,729 And in the instructions, all I said was something like, you know, you are a 338 00:22:12,729 --> 00:22:18,539 custom GPT designed to answer questions for the staff of 1623 Studios using the 339 00:22:18,539 --> 00:22:21,319 documents that I've uploaded or whatever. 340 00:22:22,059 --> 00:22:27,489 And so the staff, with that little bit of instruction, and those documents can just 341 00:22:27,489 --> 00:22:31,699 say anything from what's the dress code to what's our strategic plan this year? 342 00:22:32,109 --> 00:22:33,249 And it can give you the answer. 343 00:22:34,369 --> 00:22:38,679 I mean, it took zero minutes to make that powerful tool for people 344 00:22:38,679 --> 00:22:42,569 that no longer have to come to me with those questions that are FAQs. 345 00:22:43,044 --> 00:22:43,899 They have answers. 346 00:22:44,559 --> 00:22:47,629 Those answers are now in the GPT and they don't have to ask me. 347 00:22:47,629 --> 00:22:51,359 So that I mean, there's 8 zillion examples like that. 348 00:22:51,849 --> 00:22:53,699 That could not be easier. 349 00:22:53,799 --> 00:22:54,879 You could do it yourself. 350 00:22:54,919 --> 00:23:00,099 And once you see that and experience that, you'd be like, okay, now I kind of get it. 351 00:23:00,129 --> 00:23:01,489 And the more you get it, the more you get it. 352 00:23:01,891 --> 00:23:02,131 Absolutely. 353 00:23:02,851 --> 00:23:04,911 Now, that's a terrific example. 354 00:23:05,141 --> 00:23:08,341 I think there's a lot of evidence that this is one of 355 00:23:08,341 --> 00:23:10,131 the first use cases to deploy. 356 00:23:10,241 --> 00:23:16,681 And if you are not a large enterprise with millions to spend that is a very 357 00:23:16,681 --> 00:23:19,166 effective way for leveraging these tools. 358 00:23:19,496 --> 00:23:24,346 And so what do you think about the role of human expertise in 359 00:23:24,346 --> 00:23:25,926 curating all these AI tools? 360 00:23:26,786 --> 00:23:27,646 Where do we stand? 361 00:23:27,726 --> 00:23:29,076 I think it's critical. 362 00:23:29,326 --> 00:23:34,636 I personally feel really lucky to have a couple decades of experience under 363 00:23:34,636 --> 00:23:36,856 my belt when this technology came out. 364 00:23:37,616 --> 00:23:40,776 Because I think subject matter expertise matters a great deal. 365 00:23:42,446 --> 00:23:53,636 Because so much of the known stuff is available to chat GPT now and 366 00:23:54,256 --> 00:23:59,076 therefore not really valuable to you. 367 00:23:59,846 --> 00:24:06,106 And so where subject matter experts have knowledge that is nuanced 368 00:24:06,236 --> 00:24:09,581 and proprietary and their's alone. 369 00:24:10,161 --> 00:24:13,871 And they connect dots that no one else connects because they've been doing it 370 00:24:13,871 --> 00:24:18,391 for 20-30 years and they have expertise and they get it and they can see the 371 00:24:18,391 --> 00:24:24,826 bigger picture and apply that knowledge or apply today's tools to that knowledge. 372 00:24:25,226 --> 00:24:26,906 That is where the magic is. 373 00:24:26,906 --> 00:24:30,446 That's why I've been having success because I'm applying AI 374 00:24:30,456 --> 00:24:33,086 to my subject matter expertise. 375 00:24:33,776 --> 00:24:40,686 And I'm able to just take everything I've learned over 20 years and, whatever I 376 00:24:40,686 --> 00:24:44,636 can learn with AI and try to smash them together and make something valuable. 377 00:24:44,856 --> 00:24:48,066 I think that's a good position to be in as opposed to like a, you know, 378 00:24:48,066 --> 00:24:54,746 straight out of school and great with technology because good for you. 379 00:24:55,406 --> 00:24:59,806 You're great with technology, but, what expertise can you 380 00:25:00,486 --> 00:25:02,316 inject into it or get out of it? 381 00:25:02,508 --> 00:25:02,988 Absolutely. 382 00:25:03,178 --> 00:25:06,738 No, connecting the dots, I think is the title of this episode. 383 00:25:08,368 --> 00:25:12,048 And so as we are about to, uh, to wrap up, I tend to ask a 384 00:25:12,048 --> 00:25:13,488 couple of questions at the end. 385 00:25:13,853 --> 00:25:20,133 First, what's a book, tool or habit that has made a significant impact 386 00:25:20,133 --> 00:25:22,323 on you in the last 12 months and why? 387 00:25:22,653 --> 00:25:25,903 One of the things that I have found helpful is the simple idea 388 00:25:25,913 --> 00:25:33,273 of just daily reps. I am, I find it to be very helpful to address the 389 00:25:33,283 --> 00:25:39,883 overwhelm of the AI news and the sort of like the fire hose of stuff. 390 00:25:40,855 --> 00:25:44,419 It's just like finding where the daily reps need to come from. 391 00:25:44,429 --> 00:25:44,659 Like. 392 00:25:45,309 --> 00:25:46,839 I need to be learning every day. 393 00:25:47,609 --> 00:25:50,859 I just need to put in some learning reps and keeping a list of the 394 00:25:50,859 --> 00:25:51,999 things I need to learn about. 395 00:25:52,249 --> 00:25:55,959 And then making sure that every day I put in a couple of those reps. Or like, 396 00:25:56,719 --> 00:25:58,699 you know, doing the LinkedIn thing. 397 00:25:59,189 --> 00:26:00,939 Or doing the PodMatch thing. 398 00:26:01,019 --> 00:26:04,229 Or, you know, putting in time with ChatGPT. 399 00:26:04,384 --> 00:26:10,164 Or learning some new tools and just not getting bogged down in the hype and who's 400 00:26:10,164 --> 00:26:14,784 done incredible projects or how busy I am and can't get to it this week or whatever. 401 00:26:14,784 --> 00:26:20,214 Like trying to make a daily habit out of it with some daily reps is helping 402 00:26:20,214 --> 00:26:21,414 me address some of the overwhelm. 403 00:26:21,714 --> 00:26:24,684 I'm sure it's going to be very useful to our listeners. 404 00:26:24,724 --> 00:26:25,384 Absolutely. 405 00:26:25,384 --> 00:26:25,794 Yes. 406 00:26:27,674 --> 00:26:31,144 And then last, what is the one thing that you would expect to 407 00:26:31,144 --> 00:26:33,254 remain constant 10 years from now? 408 00:26:34,041 --> 00:26:38,981 I think there will be a role for people forever. 409 00:26:39,256 --> 00:26:43,706 You know, I think everyone's pretty freaked out that the humans 410 00:26:43,706 --> 00:26:45,446 are going to be less valuable. 411 00:26:45,446 --> 00:26:47,506 I think it's just going to be different. 412 00:26:48,346 --> 00:26:53,576 But in my experience, all the magic from these tools, and this 413 00:26:53,576 --> 00:26:56,266 isn't trying to toot my own horn, but it's coming from the people. 414 00:26:56,959 --> 00:27:02,390 They're just general purpose, dead tools, otherwise. 415 00:27:02,630 --> 00:27:05,740 It takes people to bring the life out of them. 416 00:27:06,260 --> 00:27:09,530 And I see these as people enhancers. 417 00:27:10,009 --> 00:27:12,599 And so with that, I'd like to thank you, Erich. 418 00:27:12,779 --> 00:27:13,589 Thanks so much for having me. 419 00:27:13,589 --> 00:27:14,199 It was a lot of fun. 420 00:27:14,637 --> 00:27:18,883 Erich's journey shows how AI tools become truly powerful when paired 421 00:27:18,883 --> 00:27:20,833 with human creativity and experience. 422 00:27:21,533 --> 00:27:25,813 From using ChatGPT to simplify workflows, to platforms like MidJourney and 423 00:27:25,813 --> 00:27:30,223 Runway for video production, Erich has demonstrated how experimentation 424 00:27:30,393 --> 00:27:32,083 can lead to meaningful results. 425 00:27:32,633 --> 00:27:36,493 His work proves that AI isn't here to replace us, but to help 426 00:27:36,493 --> 00:27:40,668 us create, innovate, and solve problems faster and better. 427 00:27:41,468 --> 00:27:42,988 So here's something to think about. 428 00:27:43,488 --> 00:27:48,688 How can you use AI in your own work to save time, unlock creativity 429 00:27:48,898 --> 00:27:50,278 and achieve bigger goals. 430 00:27:51,408 --> 00:27:54,838 To learn more about Erich's projects and insights, visit his 431 00:27:54,838 --> 00:27:57,728 portfolio at cgacreative.com. 432 00:27:58,518 --> 00:28:03,348 In the meantime, we'll have more exciting topics and guest appearances lined up, so 433 00:28:03,368 --> 00:28:07,758 stay tuned for more tales of innovation that inspire challenge and transform. 434 00:28:08,328 --> 00:28:10,248 Until next time, peace.