0:00 You know, there's a lot of noise with 0:02 AI. Just like you said, there's people 0:04 on LinkedIn or Twitter or YouTube saying 0:07 it's crap or you've got doomsday, you 0:10 know, everyone's going to get wiped out. 0:12 What's the point? And so there's a lot 0:14 of this like there's fear, there's FOMO, 0:18 there's I don't know what to do. And 0:20 then there's also, wow, this looks very 0:23 complicated to really do the nerdy the 0:25 nerdy stuff. 0:26 >> How good you are with AI at this point, 0:28 and you kind of alluded to this before, 0:30 is really a function of how curious you 0:33 are and a little bit your willingness to 0:36 or your tolerance for pain at this point 0:39 because [music] 0:39 it is a very iterative process. It is 0:42 not going to work beautifully first time 0:44 out of the box. 0:45 >> All right, I've got the Godfather of 0:49 Google Ads himself with me today. 0:52 Um, Mike Rhodes. I, you know, I think 0:55 it's probably been 10 years, maybe even 0:57 longer. I was learning Google Ads and I 1:01 found your community. Um, 1:04 it's terrible. I forget the name now. It 1:06 was like Web Savvy or that was your 1:07 agency. 1:08 >> It was PPC Savvy way back then and then 1:11 it became Agency Savvy. 1:13 Um, yeah, it's a different one now, but 1:15 we'll get to that. 1:16 >> And Mike, I think 1:17 I think I was 1:20 uh I think I paid $30 a month. I was 1:22 like one of the 1:24 first people cuz I by the time you, you 1:26 know, had a lot of people in there. I 1:28 think it was a few hundred dollars per 1:30 month. 1:31 >> Ended up at 300 a month, yeah. 1:32 >> Yeah, so I was paying this grandfathered 1:34 rate of like $30 a month out of my own 1:36 money when I had no money back then. I 1:38 was just trying to learn. And so I think 1:40 I was one of the first, you know, at 1:41 least probably 100 people in there. And 1:45 uh it's cool. I met um I met Alex 1:48 Fedheim in that community. I've never 1:50 met him in person, which is wild. And 1:52 we've worked together now for for 10 1:54 years. And he introduced me to my 1:59 business partner in Grow My Ads. And and 2:02 so Luke, my business partner, Alex is 2:05 the one who introduced us through Skype 2:08 back then. You know, Skype's not even 2:10 around anymore. 2:11 And I started just kind of doing work on 2:15 the side for Luke. So I'd manage the 2:18 accounts that he would sell. 2:20 And this is I was working at another 2:23 agency that fast forward today I I now 2:26 own half of and uh 2:28 uh we ran that and built it to a nice 2:30 seven-figure thing. And then COVID 2:33 really rocked that agency. Our sales 2:35 engine completely changed and that's all 2:37 we had. And so that's when I made a 2:40 pivot and started the brand Grow My Ads 2:42 with Luke where we were just kind of 2:44 freelancing at the time and and now Grow 2:46 My Ads is, you know, 2:48 really exploded over the last few years. 2:50 So anyways, it all comes together now on 2:53 this call and it's all because of of 2:56 you. So there's probably no introduction 2:58 needed here, but I guess a little 3:00 background Mike and then we can kind of 3:01 get into what you're doing with AI. 3:04 You've got insane amounts of experience 3:06 and history with Google Ads, but not 3:08 only that, you were way ahead of the 3:09 curve on AI even back in the the day in 3:12 the community you would talk about AI, 3:14 you know, before it was cool. So 3:16 >> I I remember doing a talk at Traffic and 3:18 Conversion. I remember Justin Brooke 3:20 sitting in the front row. That was 2017, 3:23 2018. And the first time I talked about 3:26 AI sort of 3:27 proper. 3:28 Um machine learning and AlphaGo had just 3:32 happened. And I kind of concluded then 3:34 we're we're too early. We'd I think I 3:36 said this line of like one day 3:39 AI will be like LEGO bricks that we get 3:41 off the shelf and put together and we'll 3:42 be able to and it'll be that easy. But 3:44 right now you've got to learn Python and 3:46 build a neural network and 3:48 no one's going to do that. 3:50 Focus on automation. Come back later for 3:52 the AI. I did not predict 3:54 Chat GPT 3:56 Um but I do remember being on stage in 3:58 '22 talking about images and you know, 4:01 there was this website back then um this 4:04 person does not exist. It probably still 4:06 is out there. This person does not 4:07 exist.com and it kind of showed the 4:09 progression of how AI could create 4:13 images of people and it was getting more 4:14 and more realistic from '20 4:17 13, 14. By 2018 it was pretty realistic. 4:20 By 2020 it was like you can barely tell 4:22 the difference. 4:24 And here we are now where AI is 4:28 running huge chunks of 4:30 lots of agencies. It's not replacing us, 4:32 but it's helping with 4:34 huge amounts of work. Um yeah, I'm I'm 4:39 fascinated by I get to do what I've 4:41 always loved doing, which is be the 4:44 scout and and run four or five hills out 4:46 into the future. 4:48 I kind of live 6 to 12 months in the 4:49 future I think and every now and again I 4:51 come back to the present. But I don't 4:53 like to come back and report on that. 4:55 Say here's what I think I saw and I like 4:57 to come back and make it really useful 4:59 and practical for people and and give 5:03 tools, build tools. Somebody on a 5:05 podcast yesterday said you you just 5:07 you're a builder, right? Um I don't I 5:08 caught myself smiling 5:10 cuz that's what I did when I was a kid. 5:12 I would build stuff out of LEGO and 5:15 bricks and build stuff in the backyard 5:17 and um 5:18 and that's what I'm doing now all day 5:20 every day. I just get to build stuff, 5:21 but I I have this need to be useful. So 5:24 I want to build stuff 5:25 and then 5:26 give it away or maybe sell it 5:28 occasionally. Like the display grid way 5:30 back in the day when we first were 5:31 talking. Um 5:33 um 5:35 the profit curve tool and now it's AI 5:38 stuff. So yeah, I'm sure we'll talk 5:40 about the brain at some point. Um but 5:42 it's 5:43 tools tools to to help agencies, to help 5:46 businesses, to help freelancers, and and 5:50 to help people with their careers, too. 5:52 I'm really, really 5:54 keen on this at at at the moment, 5:56 because there's far too much crap out 5:58 there. There's two ends of the spectrum 5:59 out there. There's either like, "Oh, 6:01 it's going to kill us all and and what's 6:04 the point?" Or it's 6:06 "AI is crap. I tried it once. I did a 6:08 Dr. Seuss poem, and it it can't do what 6:11 I do, and it's crap." And there's these 6:12 two schools of thought, and it's like, 6:13 "No, 6:14 there is a middle ground here. Let's be 6:15 sensible about this. It can't do 6:17 everything. It's spiky. It's jagged." 6:21 The only way to figure out what it's 6:22 great at 6:24 and what it's terrible at, and sometimes 6:25 those two things are really close 6:26 together, which doesn't make any sense, 6:28 right? You think, "Oh, it's great at 6:29 this. It'll be great at the next thing 6:30 and the next thing." But it's terrible 6:32 at that. But then it's great at the next 6:34 thing. And it's 6:35 it doesn't make any sense, but the only 6:36 way to find that out is to play and 6:38 experiment. And and yeah, 6:41 be in a community of people sharing 6:42 ideas. Sorry. 6:43 >> Yeah, you've been doing it for a very 6:44 long time, too. You had So, you had PPC 6:48 Savvy for How many years did you run 6:50 that agency? 6:52 >> Oh, so Web Savvy was the agency. I ran 6:54 the agency I ran the agency for about 17 6:58 years. So, yeah, sold that about 2 and 7:00 1/2 years ago. 7:01 Uh I had two kids that were still just 7:04 in primary school at that point. Still 7:05 have the two kids. Uh but one of them 7:07 isn't in primary school anymore. 7:09 Um 7:09 and I was just very, very aware. I had 7:12 this thing with my elder daughter, 7:13 Kiara, where we were counting the Monday 7:15 mornings, cuz we live about 3K's, what's 7:18 that? 2 mi 7:19 from the school. And so, I would manage 7:21 to persuade her to walk once a week. And 7:23 the rest of the time, Mom would take her 7:24 in the car. 7:26 And we started this at about 100 weeks 7:29 out. So, halfway through grade four, we 7:31 had 100 Mondays left together. And we 7:34 we're counting these down. I think we 7:35 had a little chart on her bedroom wall. 7:37 Um 7:39 and then yeah, fast forward to the point 7:41 where I was sort of 7:44 starting to think about either either 7:45 growing and buying another agency or 7:48 selling. 7:49 And with my younger daughter, we worked 7:52 out I had 134 Mondays left. 7:56 Uh we're down to about 37 at the moment. 7:59 Um so that's how long ago it was that I 8:01 sold, but it was just this realization 8:03 of once I'd started the thing with Kiara 8:06 and when I sold, I think we were we 8:08 literally had 8:09 What was it? Like 16, 17? 8:12 >> Yeah. 8:13 >> Left. And it was just I am going to milk 8:15 this moment because this ain't going to 8:18 happen again. Like once these 16 are 8:20 done, she is never going to primary 8:22 school ever again. I don't get this walk 8:24 with her because she doesn't want me 8:25 anywhere near her when she's going to 8:26 high school, right? You don't walk your 8:27 kids to high school. They're like, "Get 8:29 Get away from me." 8:30 >> No, yeah. 8:31 >> And once you see that, you can't unsee 8:32 it. And I was just 8:34 very very aware that 8:36 the the the this little period of time 8:39 isn't going to happen again. So, 8:41 uh we'll just 8:44 get out of all of that and focus on this 8:47 for a bit. 8:48 And 8:50 then wouldn't you know, AI comes along 8:52 in a completely different form 8:54 a little bit later and just 8:56 and [snorts] 8:57 this is the stuff that I would do 9:00 anyway, right? This is the stuff that I 9:01 would do if I wasn't being paid to do 9:03 it. I love this stuff. I love building. 9:06 I love running 6 to 12 months into the 9:07 future and and figuring this stuff out. 9:11 I loved back in the day climbing inside 9:13 the Google engine and working out how it 9:16 worked 9:17 and teaching that. Along the way, 9:18 discovered I was a bit of a teacher at 9:20 heart. 9:21 So yeah, it's it's it's I'm having so 9:23 much bloody fun. 9:25 >> That's amazing. So you sold and then you 9:27 started I mean what is now I I don't 9:29 think it had the name that it has today, 9:31 but Ads to AI, correct? 9:33 >> Correct. Yeah, sorry, just to finish the 9:34 timeline. So yeah, I I had released this 9:37 free PMax script that something like 9:40 15,000 agencies have since 9:42 >> blew up, exploded. Yeah, by the way, the 9:44 greatest script my typical LinkedIn post 9:47 >> My typical LinkedIn post would have 9:48 about seven impressions. Not not not 9:51 comments or anything, impressions. 9:53 >> Wow. 9:53 >> And then that PMax post, I think between 9:57 the three of them I ended up having had 9:59 a million impressions. It was 10:01 ridiculous. 10:02 >> Amazing. 10:02 >> And so, yeah, right after I sold it, it 10:04 was my IP. 10:05 Um I talked about it with the new 10:07 owners, "No, no, that's your stuff, you 10:08 take it." Okay, then. 10:10 And I was speaking at Gianpaolo's event 10:12 in Bologna, so I figured, "Well, that's 10:14 a good time to launch it." 10:16 Um started selling the free version of 10:19 the script and then I'm sorry, so 10:20 started selling the paid version. 10:22 Where are we now? October 2023. 10:24 And it just went bananas. And it just 10:26 kept going bananas because Google didn't 10:28 come up with their version of that until 10:30 what? Almost 2 years later. 10:33 And then off the back of that, lots of 10:34 people were like, 10:36 "Scripts, I need to learn scripts. Can 10:37 you teach me about scripts?" So, then I 10:39 had workshops and stuff teaching scripts 10:41 and Google Sheets. 10:43 That led to me starting to build more 10:45 and more stuff with AI in 24. 10:48 So, then I built it so I would learn how 10:50 to do it cuz I've never taught 10:53 textbook stuff. I want to know how to do 10:55 the thing if I'm going to teach it. 10:57 So, I started building 10:58 agents. Um 11:00 not what we would call agents today. 11:01 It's changed a bit since then, but 11:03 started building a whole bunch of stuff 11:05 with AI. 11:06 And then started teaching that and 11:08 bringing these build the agent courses. 11:11 This is a little over a year ago. And it 11:13 was off the back of that that everybody 11:14 went, "That was awesome. That was a 11:16 great 8 weeks, but now what? What have 11:18 you got?" So, that's when I started Ads 11:21 to AI, beginning of last year, 11:23 as the community of scouts and builders 11:26 and people that just want to get their 11:28 hands dirty and play with this stuff, 11:30 but don't want to do it by themselves 11:32 and want a head start. I don't want to 11:34 figure everything out from scratch. 11:36 Give me a starting point. 11:38 Filter out the 11:40 the noise. Give me the signal. So, I do 11:42 like a weekly video each week, and we 11:43 have a monthly Q&A, and then I'm 11:45 building stuff all the time, skills and 11:47 tools and all these different things, 11:49 and it's just so much bloody fun. It's 11:51 great. 11:53 >> We We were chatting before I, you know, 11:56 >> [clears throat] 11:56 >> we started recording this, but uh I was 11:59 saying that, you know, 12:02 uh you know, like I here at Grow My Ads 12:03 have a team of 20 employees. I'm very 12:06 much in the weeds of the agency still. 12:09 Um 12:09 >> Right. 12:10 >> So busy, and [clears throat] 12:12 >> [laughter] 12:13 >> you know, you you've been through it, 12:15 right? Uh 12:16 and 12:18 you know, there's a lot of noise with 12:20 AI. Just like you said, there's people 12:22 on LinkedIn or Twitter or YouTube saying 12:25 it's crap, or you've got doomsday, you 12:28 know, everyone's going to get wiped out, 12:30 what's the point? And so, there's a lot 12:32 of this like there's fear, 12:35 there's FOMO, there's I don't know what 12:37 to do, and then there's also, wow, this 12:40 looks very complicated to really do the 12:42 nerdy the nerdy stuff, right? Like, 12:44 sure, I have a 12:47 I had a ChatGPT account, Claude account, 12:49 obviously Jim and I, and I 12:51 I would kind of use them to just search 12:55 like most people, and then and then 12:57 Claude I started gravitating naturally 12:59 more to Claude. I just enjoyed it 13:01 better. 13:01 >> Mhm. 13:02 >> And then yeah, sure, I started adding in 13:04 spreadsheets and doing things through 13:06 through chat. But, when I see the 13:09 terminal, which Luke, my business 13:11 partner here at Grow My Ads, he's for 13:13 the last couple years been really 13:14 nerding out on AI. 13:17 >> Of Luke. 13:17 >> And 13:18 he's in you know, he was showing me the 13:20 terminal. He's got multiple, you know, 13:23 screens running, you know, he's got 13:24 codex running here and Claude code here, 13:27 and he's going back and forth, and it 13:29 looks 13:31 chaotic and complicated, and I'm like, I 13:33 don't have the time, right? It It's It's 13:35 just like I'm not going to spend the 13:37 time to learn this. Then, 13:40 Alex Fedheim, who is a part of Ads to 13:42 AI, started talking to me about what 13:45 he's doing, and he He's like, "Hey, 13:46 listen, Mike has actually made this very 13:49 easy to learn. Go spend like a few hours 13:53 and just go through his tutorial videos 13:55 in the group, and just play around with 13:58 it." So, I decided, "You know what? I 14:01 I'm going to spend a weekend. This is my 14:03 weekend. I'm just going to learn what 14:04 Mike has. I'm going to go step-by-step 14:06 through your videos, and I'm just going 14:08 to see what it's about." And you made 14:11 that 14:12 unbelievably easy. I think I told you I 14:14 still don't really know what I did, 14:16 because, you know, you you you get the 14:18 terminal up, and you've got to download 14:19 these things, and it's doing all this, 14:21 and I'm like, "It's working. Whatever's 14:23 happening is working." 14:25 And 14:27 uh lo and behold, I I got hooked, right? 14:29 And then I started playing around, but 14:31 the brain made it very simple. It was 14:33 like it Like you said, it's like an 14:34 agency OS in the box, and 14:38 um it's because then of how easy you 14:40 made it that I you know, Claude code to 14:42 me is like I I'm in that every single 14:45 day. It's just a part of my workflow 14:48 now, and I'm barely barely even 14:50 scratching the surface now. And we're 14:53 now at you know, I'm at a point now 14:55 within the organization of like 14:57 we can get into this a bit later is 14:59 like, how do I deploy this with a team 15:02 of 20? You know, it gets 15:03 >> [laughter] 15:04 >> It gets a bit uh 15:05 complicated in that, but anyone who's 15:08 watching this, and you're just using 15:10 chat, right? You're just using going to 15:12 Claude on your web browser and chatting 15:15 or or chat GPT, 15:17 um it's it's really not that difficult, 15:21 and it's it's once you get it set up, 15:23 especially your video training on how to 15:25 do it, it it's very simple. In fact, I 15:28 love it now. I I like it to me it's it I 15:31 would rather spend my time in Claude 15:33 code than a Google Ads interface today. 15:35 I would rather just work within Claude 15:37 code itself. 15:38 >> Yeah. Um and so coming from a guy 15:40 >> started. 15:41 >> Yeah. 15:41 >> That's how it started. I I would be 15:43 doing work and every time I would go to 15:46 a different window outside of this 15:48 horrible thing called terminal. And 15:49 don't worry, um dear listener, if you're 15:51 listening to this and going, "Oh, I 15:52 don't want to do this thing in this 15:54 thing that looks like it's from 1987, 15:57 this black screen." I have a fix for 15:58 that. But whenever I would go outside of 16:01 that, and I got used to it, and you can 16:03 turn it from black to white or purple if 16:05 you want. 16:06 I would be asking myself like, "Why am I 16:08 going here? What am I doing here? Why am 16:11 I logging into Gmail? Hey brain, can you 16:13 do that?" Yeah. "Okay, we can just pull 16:15 in all of the emails and you can alert 16:17 me when there's one that looks 16:19 important, not that Amazon just 16:21 delivered your thing, or this thing just 16:23 billed you more, but you need to reply 16:25 to this in the next 2 hours. Can you let 16:26 me know so I don't have to keep going 16:28 over to Gmail just to check just in 16:30 case? And all those emails that come in 16:32 overnight, can you write draft replies 16:34 to those? Don't send anything, you're 16:35 not allowed to send anything, but could 16:36 you write drafts 16:38 so that I wake up to 13 drafts ready to 16:41 go rather than 16:43 an inbox full of noise that I've got to 16:44 spend the first 2 hours of every day on 16:46 and get knocked out of flow? Yeah, I can 16:48 do that. 16:49 Great. And it's just like you have this 16:51 conversation with this 16:53 occasionally incredibly stupid, but most 16:55 of the time incredibly brilliant 16:58 Yeah, there's there's lots of ways to 17:00 think about it. I think AI should stand 17:02 for alien intelligence because it's it's 17:04 it's human-like, 17:06 but it's also really weird, and it will 17:08 do the strangest, dumbest things. We 17:11 tend to What's the word I can never 17:13 pronounce? Anthropomorphize 17:15 a lot of stuff, right? We see faces in 17:17 the clouds and 17:19 that's just a million years of evolution 17:21 so that we can spot the tiger or that is 17:23 a snake or a stick. I need to know 17:25 really quickly. Yeah, that's just the 17:27 way our brain works and so we tend to 17:29 think these things 17:31 are human, but they're not. They're just 17:33 they're mimicking us. They're just 17:35 holding up a a mirror. So when it does 17:38 something wrong, it goes, "Oh, I'm yes, 17:39 I'm sorry. I shouldn't have done that. 17:41 You're absolutely right. I messed up." 17:43 Because it thinks that's what we want to 17:45 hear and that's what it's seen in all of 17:47 the training when people mess up, they 17:48 say stuff like this. 17:50 It's not sorry. It doesn't have 17:51 feelings. It's just mimicking and it's 17:54 really bloody annoying when it deletes a 17:56 bunch of your code and it says, "Oh, 17:58 you're absolutely right. I shouldn't 18:00 have done that." 18:02 Okay, what are you going to do now? It 18:04 just like cost me 500 bucks. I'm like, 18:06 "Yeah, that's terrible, isn't it?" Yeah, 18:08 you know, it's it's frustrating at 18:10 times, but 18:11 >> Yeah. 18:11 >> it is absolutely incredible technology. 18:15 I I promised to an interim step. Can we 18:18 go there? 18:19 >> Yeah, let's do it. 18:21 >> Um 18:23 What's the best framing for this? 18:24 Skills, I think. 18:26 I'm jumping ahead, I think, by doing 18:28 this, but I think it's worth mentioning. 18:29 I'll just mention I'll throw it out 18:30 there now. 8020skill.com 18:33 is a site I built 18:35 to show people to to make you play with 18:39 it in a very easy and 18:42 fun way because there's some games 18:44 hidden in there if you go through the 18:46 steps 18:47 to 18:49 viscerally show you 18:51 how good AI can be. I've made it as drop 18:53 dead simple as I can. There's three 18:56 little demos. You don't have to do 18:57 anything. You just watch. Then you get 18:59 to level two 19:00 and for each of the things on level two, 19:03 you download a skill. 19:05 You download some sample data. It's like 19:07 literally click and it downloads. Click, 19:09 it downloads. And there's a prompt. and 19:12 the prompt is on purpose really, really 19:15 short cuz there's no like all magic 19:17 stuff. A lot of people when you see 19:18 these like 14-page prompts, they're 19:20 like, "Well, I could never write that." 19:21 Exactly, but you can write a one-line 19:23 that says, "I want to do this with that. 19:24 Can you do it?" Great. 19:27 And then you use a tool 19:29 which you mentioned before. Claude has 19:31 this desktop app. So, you go download 19:34 that. Takes you 10 seconds to set up. 19:37 And we've got chat, which like you said, 19:40 chat GPT. Claude has their own version 19:42 of that, Claude.ai. That's at this end 19:44 of the scale. 19:45 At the other end of the scale, you've 19:47 got Claude Code running in a terminal 19:49 window like Luke was doing, like you're 19:51 now doing, like I'm doing every day. 19:53 That's the uber geeky version that looks 19:55 like it's from 1987. 19:57 Code work 19:58 is about here. It's slightly less 20:01 powerful, but it's got the same 20:03 interface as chat. So, everybody knows 20:05 how to use it. It's just a text window, 20:07 but there's a ton of power hiding under 20:09 the hood. 20:11 And all I want listeners to do, if 20:13 they're still scared of AI, is make make 20:16 it to level two of 80/20 skill. 20:19 Click, click, click, download these 20:21 strings, and just drag them into Claude 20:24 Code, and watch it go to work using a 20:27 skill. 20:29 So, you pick who you are. Are you an 20:31 agency owner? Are you a freelancer? Are 20:33 you a business owner? So, that the use 20:35 cases are more tailored to you because 20:37 I've discovered that the closer 20:40 a use case you can show someone that's 20:42 that's as close as possible to their 20:43 world, 20:45 it's this moment of like 20:47 "Oh, AI can do that." But if you show 20:49 them a use case that's got nothing to do 20:51 with them, it's like, "AI is AI is 20:53 shit." This is This is my world here. 20:55 This is my Venn diagram with hands. This 20:57 is my world or your world, and this is 20:59 AI me showing you AI. And it's like, if 21:01 I show you AI that's not your world, AI 21:04 is rubbish. AI is rubbish. Even if it 21:06 overlaps a little bit on that Venn 21:07 diagram, AI is still rubbish. Unless 21:09 it's almost exactly overlapping 21:12 and then there's this moment of 21:14 Oh my god. 21:16 And once you see it, you can't unsee it. 21:19 And then, like you did, you get really 21:21 curious. 21:23 You spend a few hours, yes, mucking 21:25 around. You don't know what you're 21:26 doing, but you hit this moment of like 21:29 Wow. Okay, and now I want to lean in. 21:32 Now I want to spend a couple of hours a 21:35 week learning how this stuff because I 21:36 can see how much this is going to help. 21:39 And once you reach that point, you 21:40 really almost can't imagine going 21:42 backwards. You can't go back to a world 21:45 where I have to do everything by hand. 21:47 That's what this website is designed to 21:49 do. There's There's a few levels. 21:51 There's a few different 21:53 demonstrations, examples in there 21:55 with sample 21:57 um meeting notes or a sample 21:59 spreadsheet. So, you don't have to 22:00 wander around and oh god, now I've got 22:02 to download some data in order to test 22:03 this thing out. 22:05 But please just play with it. Please 22:07 make it to level two. Have a look for 22:09 the Super Mario option and um and have a 22:11 play with that. I've made it fun. That's 22:14 kind of the the free stepping stone so 22:16 that then you 22:17 experience it and then if you gain then 22:20 yeah, go play with Cloud Code, which is 22:22 just another button in that app. You can 22:25 start using Cloud Code in there on a 22:27 wide screen instead of the scary-looking 22:28 terminal. But that I think is the It's 22:31 that moment. You got to go from zero to 22:32 one. You got to see what's possible and 22:35 and feel it rather than just be told it. 22:38 It's not a course. It's not It's like 22:40 riding a bike, right? You can't learn to 22:41 ride a bike by reading a book. You got 22:43 to get on the bloody bike, fall off a 22:44 few times, and then 22:47 there's this moment where you're like, 22:48 "Oh, I'm riding, Dad. Oh, this is 22:50 awesome." 22:51 It's the same thing with this. 22:53 >> It is that um 22:55 it's I had that moment when I actually 22:57 went through 22:58 uh your course and 23:01 uh you get hooked. and now and then 23:04 you're energetic. It's probably it's 23:06 probably the same feeling of like your 23:08 first time setting up a Google ads 23:10 campaign that works and you get like 23:12 conversions and you're like 23:13 really straight and just like 23:15 ball straight. You yeah, like it is that 23:18 moment and then you can't sleep because 23:20 your your your brain's racing. You've 23:22 got a million different ideas and and it 23:26 just takes I think like you said 23:30 in your world seen what it can actually 23:33 do. Now anyone listening to this video 23:34 is obviously going to be a part of like 23:36 running Google ads accounts. So your 23:39 what's the website again? 23:40 8020skills.com? 23:42 >> 8020skill.com. 23:43 I should probably buy the one with the S 23:45 as well. I've got a million domains. I 23:46 should probably buy the other one. 23:47 >> I'll have skill 8020skill.com. I'll have 23:50 the link in the description below on 23:52 this video so people go check out all of 23:54 them. 23:55 Yeah, I'll link to ads to AI your group 23:57 as well. I think everyone who is trying 24:00 to self learn this right now just go 24:02 take Mike's The course is free too. 24:04 >> That course that you went through I've 24:05 since put it up on YouTube so I'll send 24:07 you the link to that too. It's like 50 24:09 videos. That whole thing is free to 24:11 learn. 24:12 >> Yeah, well that's what what that's 24:13 amazing. 24:14 Why would you do that? That that was 24:16 that right there that that was such a 24:17 like that course itself was worth I 24:19 forget what I paid but it's like worth 24:20 10x that now so 24:22 >> Oh so yeah so ads to AI is is really 24:24 it's the community and it's all the code 24:26 for the brain to get you started. I'm 24:27 not giving the brain 24:29 >> Okay. Okay. Yeah, yeah. Yeah, yeah. 24:30 >> the 50 videos that explain what it is 24:32 and how to use it and how to build on it 24:35 and how to start doing Google sheets and 24:36 Google ads and 24:39 anything you [clears throat] want my 24:40 Gmail, calendar, meeting prep, email 24:44 triage. Like these are all things that 24:45 you you mentioned the good morning 24:47 command just before we started 24:48 recording. 24:50 There's heaps of us in the community 24:51 that start our day by walking up to an 24:53 AI and going /gm and saying good morning 24:56 to a bot. Not on Twitter or somewhere 24:59 else, but saying good morning to a bot 25:01 and it goes 25:03 Yeah, first thing I do, too. And it 25:04 tells me 25:05 >> edit it. I I'm editing every day. I'm 25:07 like, "Well, actually, let's take this 25:08 cuz I you know, I have it scrape YouTube 25:10 channels I like and give me summaries of 25:12 the new videos and like emails." It's 25:14 So, I'm like every day like it's been 25:16 now well over a month and so it's 25:19 evolved into this thing. 25:21 >> Great. That's exactly what it should be. 25:23 >> to me. Yeah, it's it's incredible. 25:26 >> You imagine walking into an office and 25:28 you've got the world's perfect assistant 25:30 who just knows you and they know exactly 25:33 what you want. Do you want to know about 25:34 the day ahead? Do you want to know what 25:36 your competitors are doing? Do you want 25:37 to be given ideas 25:40 um or better yet, know what your 25:42 competitors are doing, which one of 25:43 those performed the best and here are 25:45 some ideas for your next video that will 25:47 actually blow up on LinkedIn or Twitter 25:50 or YouTube. And like, what is it that 25:53 you want? Do you want meeting prep notes 25:56 for that meeting that you have at 10:00 25:57 a.m.? When do you want those delivered? 25:59 Do you want that now or do you want it 26:01 at 10:00 to 10:00? Do you want me to 26:02 text it to you if you're in the car? Do 26:04 you want me to read it to you as a 26:05 podcast on your way to the meeting? 26:08 Like, 26:09 that's that is the 26:10 that's the hardest bit. 26:12 Is just knowing what you want. 26:13 Imagination is the only constraint now. 26:17 It's 26:17 >> Yeah. 26:18 >> I describe it, we mentioned Lego before. 26:20 The way I describe it is we now have the 26:21 world's biggest Lego kit. 4,000 pieces 26:23 in this thing, but you only need to 26:25 learn about 20 of them to get tons and 26:28 tons of value. 26:30 >> I think everyone who's watching this 26:33 start like they shouldn't even go try to 26:36 self learn. I'm telling you, your brain 26:39 that you've already developed because 26:40 you can do what I'm doing. I'm I'm 26:43 I'm starting with the brain, but then 26:45 I'm building on top of it, right? For my 26:48 own needs. Like, I know what you have 26:50 preloaded skills in there for Google Ads 26:52 stuff. 26:52 >> Just to get people going, yeah. 26:53 >> Right. So, I think like one of them was 26:55 a negative keyword thing, but I have I 26:57 already was had a negative keyword sort 27:00 of skill I was trying to create. And so, 27:03 I've I actually cross-referenced what's 27:06 Mike's do that mine doesn't. And then I, 27:09 you know, it took 2 weeks of reiteration 27:11 cuz I And that is a part of the problem 27:13 with all of this is you end up becoming 27:15 a like it's never going to be perfect. 27:17 [laughter] 27:18 >> And the hardest thing is knowing when to 27:19 stop, right? 27:21 You can go from idea to implementation 27:23 in literally 5-10 minutes and build 27:25 version one. But once you see version 27:27 one, you look at it and you go 27:29 Oh, wow. Um that was cool. That was 27:30 easy. Um oh, you know what would be 27:31 really cool if we had a drop-down there. 27:33 I I want to be able to choose between 27:35 all campaigns and choose an individual 27:37 campaign. 27:38 Oh, done. Um I actually want to be able 27:40 to color code those. Oh, I need filters. 27:42 I want a slider so I can say like only 27:44 show me the search terms with more than 27:46 100, 500,000 impressions. Can you Oh, 27:49 it's already built it. Right. Um and 27:51 it's just like 27:52 because it's so fast to get the idea out 27:54 of your head and build the thing, 27:56 the ideas keep coming. You get in flow. 27:58 And the hardest thing is walking away 28:01 from the screen at the end of the day. 28:03 It's like Yeah, I'm I'm coming I know 28:05 dinner's ready, but I 28:06 I just like quickly want to give the 28:08 machine more stuff to do while I'm not 28:11 here. Can you build this? Can you build 28:12 that? Like 28:13 >> It is incredible like I'm going to share 28:14 my screen real quick. This is coming in 28:16 from a guy and that you know, you won't 28:18 even probably be uh too impressed by 28:20 this, but um this is coming from a guy 28:23 >> I keep top of it. That's exactly what I 28:25 built it for. Is I don't want to be 28:27 prescriptive, right? It's like here's 28:29 the Lego kit. Let me explain the most 28:31 important pieces. Let me build the first 28:32 couple of models for you to show you and 28:34 then go build. 28:36 >> I want to show people like I'm a guy who 28:40 uh was afraid to use cloud code, right? 28:43 And get in the terminal. Now, I this is 28:47 This is like our audit process. I'll do 28:49 our large audits that come through. So, 28:51 these will be like big spenders, 50K 28:53 plus per month usually is where like 28:55 it'll get sent to me. 28:56 And um you know, I honestly used to just 28:59 do only a Loom video and then if I had 29:03 if we had the resources and time, I 29:05 would have like a slide deck created 29:07 from that. All manual work right there. 29:09 >> Yep. 29:09 >> I now have by giving it my brand 29:13 guidelines, so my my brain now knows 29:16 Grow My Ads brand very well. Anything I 29:18 create, spreadsheets, whatever, it it 29:20 knows what to go get for the brand 29:22 guidelines. 29:23 >> Yep. 29:24 >> And I now was I've been able to take 29:27 This was built from a Loom video and 29:28 then working with some previous um you 29:31 know, reiterations of this. This is not 29:34 the first one. This audit was created 29:37 essentially in 5 to 10 minutes. There's 29:40 some reiterations here. This 29:42 >> is a This was done I mean, for me to 29:44 have my team manually build something 29:46 like this in the past would have taken 29:48 so much time and resources. This was 29:50 done in 10 Now, the audit itself still 29:53 took time um 29:55 and I'm not I With audits, I I actually 29:58 don't use Claude code cuz like I do I I 30:00 don't even download data from their 30:02 account cuz I I just I feel like they 30:04 gave me access. I'm not even going to 30:05 download a CSV of their data here. So, I 30:07 do manually review their accounts cuz 30:09 it's not hooked up to our MCC. So, I 30:12 don't have the API integration there. 30:14 >> But, I can take a Loom video of my audit 30:18 and the sales All of our sales calls are 30:20 recorded. It take I take that, I take 30:23 transcripts, feed it in, and then I go, 30:25 "Go build me an audit deck." It now 30:27 knows what that means, right? Uh based 30:30 off of this audit. And then sure, 30:32 there's a few things I got to go 30:33 reiterate. It nails it almost every 30:35 time. And they I'm talking I've closed 30:37 We've closed like 10K plus per month 30:39 deals with this exact process I'm 30:41 talking about. And this is coming from a 30:43 guy 1 month ago who wouldn't even 30:46 think of using Claude code. And now I'm 30:48 closing 10K per month deals with the 30:50 power of it. Very simple. 30:52 >> use that quote. 30:53 >> There [laughter] we go. 30:55 You can use whatever quote you want. 30:56 >> You but you had this really cool line in 30:57 there. If I had the time and resources 31:00 to do it, that is a great way to attack 31:02 AI, right? It's like, what are the 31:05 things you do? Not just let's automate 31:07 the stuff we're already doing. That's 31:09 That's an okay place to start. Start 31:11 with that. Free up an hour or two a 31:13 week, but then use that hour to start 31:17 thinking a little bit broader. What are 31:19 those things that you occasionally do 31:21 that you would love to do more of, that 31:22 you would like to deliver to a client if 31:25 only you could spend 20 hours on it. But 31:28 what do you know? We don't have a spare 31:29 20 hours lying around, and I don't want 31:32 to 31:33 pay for 20 30 hours of the team time to 31:37 put this thing together. But my god, 31:39 that client would love it if we did, but 31:41 they're just not paying us enough. But 31:43 now that is possible. We did this 31:45 challenge back in February in the 31:48 community. 31:49 Um 31:50 We built a couple of different things, a 31:52 landing page audit tool, and a 31:53 competitor intelligence report that went 31:56 off and you had told it, "This is me and 31:58 here are five of my competitors." And it 32:00 went off and looked at all of their 32:01 sites, read all of their copy, took 32:03 screenshots of the site, and bundled 32:05 that all up into a beautiful 18-page 32:08 PDF, as you say, fully on brand. 32:12 You don't send it without checking. You 32:14 check everything. 32:16 But it solves the blank page problem. It 32:17 gets you 80 90 maybe 98% of the way 32:21 there, 32:22 and you need to change it a bit, tweak 32:24 the voice, but the more you do it, the 32:26 more it learns who you are, how you 32:27 speak. You give it rules and say, "Never 32:29 use the word genuinely again, please. 32:31 Just Just get rid of that. 32:33 Get rid of the M dashes, all those silly 32:35 things." 32:37 But people were then able to go give 32:40 that to a favorite client, build 32:42 reciprocity, build the relationship, 32:45 and position it as we can do more of 32:47 these. Next one's not going to be free. 32:49 First one's free, but do you want us to 32:52 do one of these once a month, once a 32:53 quarter? Do you want to do the same five 32:55 competitors every time? Do you want to 32:57 mix it up? What would you like? Because 32:59 we now have the resources to do that. 33:01 And there is this gap where that is 33:03 going to work, right? There comes a 33:05 point, I don't know when, 33:07 couple of years from now, 33:09 where 33:10 most small businesses are going to say, 33:12 "Oh, that's just AI." 33:14 Um yeah, that was just a thing that you 33:16 did, only took 5 minutes. Why would I 33:18 pay 500 bucks for that? 33:20 But right now, the fact that you know 33:22 how to use these tools and you can wield 33:24 them in that way and AI is just one tool 33:27 that you're using as part of the mix. 33:29 It's not like anybody can just walk up 33:31 and go, "Oh, competitor intelligence 33:32 report, please." and get this 18-page 33:35 gorgeous report that looks and sounds 33:37 like you and is what you would create 33:39 and put your name to. 33:40 Most people 33:42 can't do that. 33:43 Yeah, how 33:45 how good you are with AI at this point, 33:47 and you kind of alluded to this before, 33:49 is really a function of how curious you 33:52 are and a little bit your willingness or 33:55 or your tolerance for pain at this 33:57 point. Because it is a very iterative 34:00 process. It is not going to work 34:02 beautifully first time out of the box. 34:05 But you can build a really quick version 34:06 one or version 0.1 really quickly. And 34:10 if you're willing to just spend a little 34:11 bit of time, as you said, to learn, 34:13 spend a weekend, to understand the LEGO 34:15 bricks, to learn how the moving parts 34:18 work. And yet, 34:20 the thing I've built is designed to get 34:22 you from zero to one very quickly, but 34:24 not from one to 100. That's That's on 34:27 you, right? Then you go iterate. You do 34:29 what you've done with this thing. You 34:30 add your own branding. You add your own 34:32 sections to the report. I don't like 34:34 that thing that Mike built, throw that 34:36 out, but I want this new thing. 34:37 >> Yep. 34:38 >> And it just But it But then 34:40 it gets it perfectly right. It's like 34:42 that dream employee you wish you could 34:44 have hired. Of once you say, "Oh, 34:46 remember we always do this as section 34:48 five." it never forgets that. And people 34:51 will be yelling at the screen going, 34:53 "Yeah, but it hallucinates and it gets 34:54 stuff wrong all the time." It depends 34:56 how you build it. 34:57 >> Right. Right. You know, um 35:00 I I have a comment on cuz you said 35:02 businesses or clients, you know, will 35:05 say, "Oh, you just did that with AI, 35:07 right?" 35:07 >> Mhm. 35:07 >> And we've already had that um with some 35:10 clients. Like, "Oh, that was just 35:11 generated with AI." But uh to me, I like 35:16 to attack that. So, first, like I if I 35:19 send an email and I recap it, like cuz I 35:22 do a lot of Loom videos, right? And so, 35:23 I will I'm just up front in a lot of 35:26 times. AI summary. 35:27 >> Yeah. 35:28 >> So, I'll just I I'm not going to fake 35:29 it. This is a summary from AI based off 35:31 of this video. If you want the nice 35:33 >> useful. 35:35 >> Right. Now, that audit, if they said, 35:38 "Well, AI did that." Yeah, you're 35:41 absolutely right. We used a tool to help 35:44 build our knowledge that we just gave 35:47 you, right? 35:49 >> Yeah. 35:49 >> Who cares if a human wrote that PDF doc 35:52 >> 10 years of wisdom. 35:54 >> AI? You're not paying You're paying for 35:56 an outcome that we're able to achieve. I 35:58 mean, an agency, I don't care what 35:59 anyone says, an agency is nothing more 36:00 than a brain trust of human beings, 36:02 right? So, you're you're going to this 36:04 group collective group of humans that 36:07 can do something, and in this case ads, 36:09 right? Better than what you can 36:11 internally. And so, it's more valuable 36:13 for you to just go pay them to go do 36:15 that to get the outcome you're looking 36:16 for. And so, I'm I've you know, I've 36:19 told my team, like, just Yeah, attack 36:22 that. Tell them you cannot replicate 36:24 what I just did. So, please go go to 36:26 chat GPT, Claude, whatever, and see if 36:29 you can replicate that. You can't 36:30 because we have our own brain, right? 36:34 So, the brain trust of Grow My Ads, the 36:36 Thank God we've recorded 36:39 uh for the last 5 years almost 36:41 everything with Loom. So, we have 36:44 thousands of thousands of videos and we 36:47 have already transcribed all of them and 36:50 fed it into now what is turning into the 36:53 Grow My Ads brain. Right? Yeah. So, 36:56 yeah, you can't You cannot go just do 36:59 what we're doing. Yes, we're using AI to 37:01 help produce a final asset, but the 37:05 data, the knowledge, the expertise, all 37:08 of that is us and this is the outcome of 37:11 that. So, I I just kind of attacked 37:13 that. I'm not afraid of someone saying, 37:15 "Well, just AI did that." That's not 37:16 that's not the case. 37:18 >> Yeah, Kirk Williams had a lovely post 37:19 the other day. I think he's He put a 37:21 little registered trademark after it. I 37:23 think 37:23 "a Claude it." Cuz people are doing 37:25 >> Oh, yeah. I did see that. Yeah. 37:27 >> dumping data into Claude and going, 37:29 "Audit this." And 37:31 the the way I like to explain this is 37:34 AI the output from AI is absolutely just 37:38 a function of what you put in, right? 37:40 And if you don't put in much effort or 37:42 much it 37:44 I like to picture it as this universe 37:45 full of millions and millions of 37:46 planets. There's this huge big planet in 37:49 the middle that I call planet average. 37:51 That's what you're going to get back. 37:52 You're going to get an answer from 37:53 planet average every time unless you 37:56 take the time to push the AI out to the 38:00 edges of the universe. There's a Think 38:02 of it this way, there's an answer on 38:03 every single one of those planets. 38:06 But if you don't try very hard, you get 38:07 the answer from planet average. And 38:08 guess what? AI has been trained on the 38:10 entire internet, all of Reddit, all of 38:12 Twitter. That wasn't a great idea. And 38:14 it's going to give you a meh, very 38:16 average answer. But if you add all of 38:20 the context about that client, if you 38:22 add all of your 10 years of wisdom, all 38:25 of these transcripts, all of the ah, but 38:27 we do it this way because of this and 38:28 the you add all you start to push it out 38:31 to the edge and you get a very, very 38:33 different, unique answer that no one 38:36 else is going to be able to get or 38:38 replicate because that's your wisdom. 38:40 It's your way of doing things. You've 38:42 just codified that. We've been codifying 38:45 stuff for years. Nobody looks at the 38:48 financial reporting, oh, you used Excel 38:50 to do that. 38:52 >> No. 38:52 yes, I did. 38:53 >> Because 38:54 >> [laughter] 38:55 >> Because it would take me about 14 weeks 38:56 longer if I didn't. So, yes, of course, 38:59 as you say, of course I used AI to help 39:02 with that, but AI didn't do 39:05 all of that. It's just one of the tools 39:07 that we used. 39:08 >> I like the way you frame it. 39:09 What have you seen? So, all right, the 39:12 someone who's like 39:14 totally like, I've only used like chat 39:17 or cloud AI and just the the chat 39:19 interface. 39:20 So, you said, hey, go to 8020skill.com, 39:23 get to level two, right? And then start 39:25 start learning at least and and and 39:27 seeing what the power is there. I I'm 39:30 saying everyone should just go buy the 39:32 brain and your course. Now, 39:35 so, if you haven't played around with 39:37 any of this, just go do those things. 39:38 But, those people that are dabbling, um 39:42 are already learning, right? What are 39:44 you seeing in the group right now 39:47 uh in regards like case studies or 39:49 outcomes where they are now automating a 39:53 large part of what they might do with 39:55 managing Google Ads accounts or they're 39:58 an agency and they're automating a large 40:00 portion of, you know, how they run their 40:02 agency. 40:04 >> Um oh, there's so many. 40:06 Um 40:07 Benny comes to mind immediately. Benny 40:09 started playing with the brain at 40:11 Christmas 40:12 and within 3 days replaced Motion. 40:15 So, 40:17 most of us use a software tool, be it 40:19 ClickUp, Asana, Notion, HubSpot, 40:21 something. 40:22 But we have this nagging feeling that 40:24 I'm only scratching the surface. I'm 40:25 only using 10% of it. I remember 40:27 thinking that way about Infusionsoft. I 40:29 remember going to a conference thinking, 40:30 I only use 10% of this Infusionsoft 40:32 thing. And then the guy on stage, who's 40:36 a genius at it, says, "I'm only using 40:38 10% of this thing." I'm like, "Oh, 40:40 I'm only using 1% of it if he's using 40:41 10." Like, but I'm paying 100%. 40:44 Right. All using tools like that. And 40:46 now, this isn't where you start, 40:48 absolutely, but if you're really curious 40:50 and you have a play, you can do that. 40:52 You can replace Supermetrics. You can 40:54 replace Asana. You can build it to be 40:57 just the exact way that you want it to 41:00 be, that just does the thing that you 41:01 want. 41:03 And it works for you. 41:06 Um 41:07 audits and creating, you know, audits 41:09 and reports and all that stuff. 41:10 Once you wire up the Google Ads API, and 41:13 yes, you've got to put the guardrails in 41:15 place, and 41:16 it it comes down to trust. So, I'm I'm 41:18 making a lot of assumptions here that 41:20 we've built trust over time, right? 41:22 Which you do by observing the AI, by 41:25 looking very carefully at what comes 41:26 out. But it's very similar here to 41:29 working with a junior employee. We used 41:31 to have this thing when we would uh 41:33 hire maybe a VA for someone in the 41:35 Philippines. Of me, we, you. Right? This 41:38 is me doing it. Loom comes in here a 41:40 lot. This is me doing it. I've recorded 41:42 it. You've got the transcripts. You can 41:43 go back and watch those. And now we're 41:45 going to do it we Well, we'll do it 41:47 together, and we'll see if you hit 41:49 roadblocks. And once you do it perfectly 41:51 five times in a row, now we trust you. 41:53 Now you can do it. You need to follow 41:55 the checklist every time, please, which 41:56 of course nobody does. But now you're 41:58 doing it because we've built the trust, 42:00 and we have some sort of 42:02 observability. We've got some dashboard 42:04 somewhere. We can see that people are 42:07 following the systems, doing things 42:08 right. 42:09 And then it becomes what Gerber used to 42:11 call management by exception. Like I if 42:14 everything's a green light, I don't need 42:16 to know about it as the manager. I just 42:18 need to know when things go wrong or are 42:20 going wrong. 42:21 >> Right. 42:21 >> Let's bring that to my attention and 42:22 then we can problem-solve that. 42:25 So you've got to have this sort of 42:26 observability. So assuming that's there, 42:28 you observe it, you build trust, then 42:30 you kind of jump to the next layer of 42:33 the stack. You start to abstract away 42:35 all of the the crap. You don't download 42:37 CSVs anymore and pull them into Google 42:40 Sheets and delete the top two rows and 42:42 change the formatting. All of that stuff 42:45 just gets done automatically. And once 42:46 you're doing that, 42:47 we move up a level. I've 42:49 Maybe this is a good way to explain it. 42:50 I've had this sort of 42:52 chart in my head for 20-odd years. 42:55 And it goes It's very easy. Five words. 42:56 But data, information, insight, action. 43:01 I'm going to cheat. Expected business 43:02 outcome. It's not one word. But let's 43:04 just call it outcome. Right? And you you 43:06 move You want to move up that stack. 43:08 >> Yeah. 43:08 >> Scripts were a wonderful way of going 43:10 from data to information. Google Ads 43:12 scripts. That was the automation piece. 43:15 We take this data and we 43:17 massage it in these ways and we format 43:19 it like this and we output it over 43:20 That's automation, not to be confused 43:23 with AI. Just that AI's really good at 43:25 writing the automation, at writing these 43:28 little things to go talk to APIs. 43:32 Then you go from information to insight. 43:35 And that's where a lot of people would 43:36 start with AI. 43:38 Here's a bunch of data and you ask 43:40 questions of it and it pops back with 43:42 insights. 43:43 To begin with, most of them are from 43:44 planet average. They're a bit meh. 43:47 But the more Again, the more context you 43:50 add, the more information you add about 43:51 that client, about their budgets, about 43:53 what's happening on the website, about 43:55 when tracking broke, you bring all of 43:57 that to the table and now the answer 43:59 gets better and better. 44:00 But we're not ready to take action yet 44:04 because we don't trust it. But once it 44:05 starts giving us really, really, really 44:07 good insights 44:09 and we're saving a bunch of time, now we 44:11 can move up the stack and we can 44:13 actually have it take action. And I'm 44:15 sure you've experienced this. Once 44:16 you've got the Google Ads API working in 44:18 the brain, 44:19 you can chat to the brain in plain 44:21 English or whatever language it is that 44:23 you speak 44:24 and say, "Oh, find me the best ad from 44:27 the best ad group in that campaign and 44:29 make a copy of it and put it in that ad 44:30 group." And it will just do it. 44:33 And the first few times you do that, you 44:35 go open up the interface and you're 44:36 like, 44:37 "Holy it actually 44:38 it actually did. Um 44:40 I Okay, try this one, brain. I've got 44:42 this list over here in this spreadsheet. 44:44 Here you go, here's the link to the 44:45 spreadsheet. Um can you add all of those 44:47 as exact match negs to that list over 44:50 Oh, it's already done it. 44:52 Wow." And you just 44:54 the more you see it do, the more you 44:55 trust it, the more it doesn't go wrong, 44:57 and you just keep 44:59 doing more and more. And then, yeah, you 45:01 can do things like replace software or 45:04 build entire new services. I think 45:07 there's going to be a huge Well, 45:10 not that going to be, there is this huge 45:11 opportunity right now 45:14 to help your 45:17 clients, those small businesses that are 45:20 feeling really overwhelmed. And And 45:23 Yeah, 45:24 we're digital agencies, right? I'm I'm 45:26 not anymore, but um 45:28 Listen to this. We understand this world 45:31 and 45:32 small businesses look to you and say, 45:35 "Well, you're the digital guy. You're 45:36 the computer person. Like, 45:38 you must understand this cuz we sure as 45:40 hell don't in the business. We're like, 45:42 yeah, we've played with ChatGPT, but no, 45:45 we are behind. We don't know where to 45:46 start." There's this huge opportunity 45:49 there for you as the agency to help them 45:53 understand it, to help them use it. 45:56 And there may well be a an ongoing 45:58 retainer thing there where you build 46:01 something like the brain for them 46:04 >> [clears throat] 46:04 >> and then help them get set up and 46:06 configure it all and do all the plumbing 46:08 and hook it up to their 46:10 email and their calendar and and then 46:12 build skills for them. 46:14 And say, "What is the stuff that you 46:16 guys do 46:17 all the time? Where does all of the time 46:19 go? What are those things that you would 46:21 love to do if you had the resources and 46:24 the time, like you said before, but you 46:27 just can't do now? Let me help you do 46:29 that. Let me help you use it in the 46:30 right way. And I think the tool for 46:32 that, and this is why I built 80/20 46:34 Skill, 46:35 is going to be Co-Work. 46:37 Because it's so much easier to use 46:40 with almost all the power, because no 46:42 small business is going to get all of 46:44 their team to use Terminal. That's just 46:45 never going to happen, right? It's hard 46:47 enough to get all of your team inside 46:50 You mentioned a team of 20, right? 46:52 There's going to be people in that team 46:53 of 20 that are just like, "There is no 46:54 freaking way I'm ever opening that. 46:55 [clears throat] I'm not going to work in 46:56 there. Nah-uh." 46:58 But, 46:59 this huge benefits if you do. Co-Work is 47:01 this beautiful sort of middle ground. 47:04 But, to get the most out of it, you need 47:06 skills. Hence, 8020skill.com. So, that 47:09 you see it and you go, "Oh, wow." Now, I 47:11 don't expect most people, or even most 47:13 people in your agency, to go build their 47:15 own skills. 47:17 But, if you have somebody like you, like 47:19 Luke, and you'll probably have two or 47:21 three others that lean into this and go, 47:23 "I just want to spend all my time doing 47:25 this stuff." 47:26 that build skills for everybody else. 47:30 Now, if you can build that shared 47:32 knowledge base, that shared brain, which 47:34 is what I'm building at the moment. I've 47:36 got a bunch of agencies beta testing 47:38 this at the moment. It's due to roll out 47:40 in the next week or two. And that'll be 47:42 free for everybody inside Ad Astra AI. 47:44 I'm not charging more for it. 47:46 But, the team brain, where you have that 47:48 shared context, cuz if one person's 47:50 going to write up all of the notes about 47:51 a client once, you don't want everybody 47:54 going off and doing that. Or, you don't 47:55 want those notes sitting on their 47:56 computer and when they're on holiday, 47:58 nobody can access it. And when somebody 48:00 improves that skill, that audit skill 48:02 that you've got, 48:03 and somebody adds a an AI max section to 48:06 the end of that, you want that to be 48:09 there 48:10 for all audits, if required. 48:13 Every more, you don't want that new 48:14 skill to be over there and version three 48:16 to be over there and you've got version 48:17 seven and that's just a mess, right? But 48:19 if all of that's there, so now imagine 48:21 taking that 48:23 and building that for clients. Maybe 48:26 they've got that scout internally that 48:28 can build skills over there internally 48:30 for their team, but maybe you, the 48:32 agency, become 48:35 the skill creator for them. You 48:37 interview them about their processes, 48:39 you get them to record looms and send 48:41 you the loom and the transcript. You 48:43 have a process for dropping the 48:44 transcript and the loom and everything 48:46 into a system and building a skill and 48:48 delivering that skill back. That's 48:50 hugely valuable right now. 48:53 Most agencies could charge thousands and 48:55 thousands per month doing that. 48:56 >> Well, I've already seen some that are, 48:58 you know, their retainers are pretty 49:00 high, like 10 $10,000 plus retainers on 49:03 some of the 49:04 >> will really value it and others won't. 49:05 Others will give that away for a couple 49:06 hundred bucks and some people will 49:07 charge five figures. It's just how much 49:10 you value the work that you do, but the 49:13 it it reminds me so much of when I got 49:15 into Google Ads, which we we won't we 49:17 won't mention the year, but it was a 49:18 long time ago. Um, 49:20 they want it and they need it. That was 49:23 when I started the agency because all 49:25 these people were coming to me for a 49:26 year going, "Mike, we need you. Stop 49:28 telling me like like you're just here to 49:30 help and like like I want to give you 49:31 money. Please, can you do this for me?" 49:33 Because I was trying to teach everybody. 49:35 And nine out of 10 people would say, 49:37 "Mate, I don't give a how it works. Can 49:39 you just do it for me, please?" 49:40 >> please. Do it for me, yeah. 49:41 >> Yeah. [laughter] 49:42 >> Yeah. Yeah. And they it's going to be 49:44 very similar with this. Like I don't I 49:45 don't want to learn how to build a skill 49:47 and how to connect this and API what? 49:51 API what does that I don't care it 49:53 stands for. Just set it up for me. I 49:55 want to come in 49:57 to 13 drafted replies, not a messy 50:00 inbox. Also, I want 50:03 >> Uh sorry to cut you off. I'm I'm already 50:05 testing, hiring. Uh 50:07 he's on a bit of a you know, 50:10 probationary test period. Uh 50:13 it's essentially an AI integrator within 50:15 Grommets because from an organization 50:18 and he knows Google Ads, too. So, it's 50:20 like I found, you know, some that cuz 50:22 like it's one thing if you know AI, but 50:24 then if you don't know Google Ads, 50:25 there's like a lot of training there. 50:27 So, uh he knows Google Ads. He's gone 50:29 through my training on Google Ads. And 50:32 so, it it's it's worked out so far. But, 50:35 um 50:37 it's what like from an 50:38 it's 50:40 from an organization for Grommets. Like 50:42 for me right now, and so for us here at 50:44 Grommets, it's very siloed still. Luke's 50:46 off doing his thing. I'm kind of doing 50:47 my thing. I've shared some skills like 50:50 the negative keyword thing I built, like 50:51 I shared that with my school group. And 50:53 I showed like, "Hey guys, if you don't 50:55 have cloud code and it's too nerdy, 50:57 here's how you just dump a CSV, use this 50:59 skill inside of a project inside of 51:02 cloud, right? Through chat." So, just go 51:03 play with that, right? Now, if you're 51:05 nerdy and you've got this integrated 51:08 with the MCP or an AI integration, then 51:11 or API integration, then here's how you 51:13 would just do this inside cloud code and 51:15 and automate it. So, I I kind of showed 51:17 and I just shared here's the the full 51:19 skill, 51:20 >> Right. 51:20 >> Yeah. 51:20 >> right? From our standpoint, what we have 51:24 found right now, and what we're dealing 51:25 with is A, how do we deploy 51:28 all of this to the team cuz there's like 51:29 a huge training piece there. 51:31 >> Mhm. 51:32 >> It's would require a lot of time for me 51:34 or Luke. And I don't know if that makes 51:37 complete sense and that I'll I'll I'll 51:39 lead into a question here then, too. Um 51:42 there's probably a million questions 51:43 from all of this. But, the other thing 51:45 what we have found is, you know, Claw 51:48 and it sounds like you're working on 51:49 this with a shared brain concept, but I 51:52 cannot give the API integration into 51:56 Claw code to my entire team. That 51:58 wouldn't make sense. I like from a 51:59 security standpoint, I'm not going to 52:01 give them access to the entire MCC. So, 52:03 our strategists all work within sub 52:05 MCCs. Now, the problem with 52:09 the MCP is it does use a ton of tokens. 52:14 >> Mhm. 52:15 >> And so, we've ran into that. Now, Luke's 52:17 working on a way of figuring that all 52:20 out so it doesn't just burn through 52:22 usage just when you're trying to pull 52:23 like, you know, some data 52:25 >> common. 52:25 >> from one account. And so, I'm we're 52:28 still at the stage like I'm playing with 52:30 it. I've got skills. I can easily spin 52:33 out audits and then just sharing that 52:35 with the team, but it's still it's this 52:37 siloed thing and it's not totally 52:39 organized. And then I saw I'll give him 52:42 a shout. I forget his last name, Daniel, 52:44 who owns a Brain Labs Digital. 52:47 >> Daniel Gilbert, yeah. 52:48 >> Yeah, incre- I don't I don't know him. 52:50 I've never talked to him. Incredible 52:51 what he's built. Incredible guy. I 52:54 Again, I don't know him, but I've seen 52:55 him on his I follow him on LinkedIn. And 52:58 he came out with this really cool post. 52:59 I'll probably put that in the 53:00 description of this video below, too. 53:01 You guys should all check out. But how 53:04 he has built I mean, this is a thousand 53:07 plus person agency. So, the you know, 53:09 this is not a small thing that they just 53:11 accomplished, but they're using 53:13 essentially Notion as the data layer, it 53:15 sounds like, for their their entire 53:17 organization in regards to Now, how how 53:20 they hook it all up with like a thousand 53:22 employees and all that, I'm not entirely 53:24 sure. It's He doesn't give you every 53:25 single step, but the infrastructure that 53:27 he's developed within a that big of an 53:29 organization was very cool to see like a 53:32 true case study of it. Cuz you hear you 53:34 hear all the hype 53:35 >> Yeah. 53:36 >> like, I automated my agency and fired, 53:38 you know, half my employees and all you 53:40 know, I'm not buying most of that. I 53:42 think a lot of that is just like people 53:44 flexing for for clicks and impressions. 53:47 Um and and uh and I like his standpoint, 53:50 too. He's like, "If I can get a 20%, you 53:53 know, more output per employee, I'm not 53:55 looking to decrease my workforce by 53:57 20%." 53:58 >> Not at all. 53:59 >> Not at all, right? I would rather hire 54:01 someone and get 20% more out of them as 54:03 soon as I hire them. 54:05 >> and then we'll get 30% next year, and 54:06 then 50% the next year. And this what a 54:09 lot of businesses will do, they're 54:10 freezing hiring. Not great for juniors 54:13 coming in, that's a whole different 54:14 podcast. Um 54:17 but yeah, if you can but I think 20% 54:19 will be the table stakes efficiency gain 54:22 because all of your competitors or some 54:25 of your competitors 54:26 will be increasing 20%, 30%, 40%. 54:29 A whole bunch won't. A whole bunch will 54:30 stick their head in the sand and go, 54:32 "This whole AI thing, it's rubbish. It 54:34 just tells dad jokes badly, and I'm just 54:37 going to ignore it." And they may not be 54:39 in business in two or three years' time. 54:41 I think 20, 30% is table stakes to stay 54:44 in the game. 54:45 But yeah, Dan's awesome. Um every time I 54:47 used to go through London to visit my 54:49 folks, um I'd pop in and see him. When I 54:52 first saw him, he had a 60-person team. 54:55 Uh by the time I went back the next 54:57 time, it was 200, 250. Like he is a 55:00 unicorn. Very few people can grow that 55:03 fast. 55:04 >> Incredible. I would love I would love to 55:05 talk to him. I I I might hit him up 55:07 after when I when this is out. Say, 55:08 "Hey, we gave you a shout-out." But I 55:10 would love to talk to him because it's 55:11 what he's built is just truly 55:12 incredible. And you know, I I run a 55:14 20-person team, and it's hard, right? 55:16 You know, so I can't imagine getting to 55:17 a thousand people. 55:19 >> So, a couple of things on that. Um yeah, 55:21 so they chose Notion because that was 55:23 where the work already lived. Because a 55:25 huge, huge part of this, which isn't on 55:28 the LinkedIn posts, is behavior change. 55:31 And trying to change the behavior of a 55:33 thousand people 55:35 is really bloody hard. I've never tried 55:37 to do it, but I hear it's really, really 55:38 hard. It's hard enough to change the 55:39 behavior of 20 people. So, the less 55:42 behavior change, the better. Meet them 55:44 where they are, 55:46 make things better, build stuff for 55:48 them, and they'll come along. A few 55:50 won't, and okay, that's okay. We may 55:53 need to have a conversation, 55:54 but most will. The model I use, cuz as 55:57 you can see, I like mountains, Whistler. 55:59 Um 56:01 is summit scout trekkers. This is the 56:03 framework I've sort of the mental model 56:05 I've created for this. So, 56:07 summit, that's your job. Leadership's 56:09 job is to to what is the view going to 56:11 be from the top of the mountain? Let's 56:13 go out a year or two. There's no point 56:15 going any further than that. 56:17 Why? Why are we doing this? Why is AI so 56:20 important to the business? Why is it an 56:22 existential risk? What is this going to 56:24 be like? 56:25 We are not doing this so we can get rid 56:27 of half of you. Like, people need to 56:29 hear that because 56:31 people are worried. They're worried 56:32 they're not using AI correctly. They're 56:34 worried they're breaking some sort of 56:35 policy that they don't know about. They 56:37 don't even know if they're supposed to 56:38 paste client data in there or not 56:40 without redacting it first. Like, 56:43 why are we doing this, boss? Are you 56:44 doing this so you can get rid of half 56:45 the people so you can pay me the same 56:47 and I do twice the work? That's what's 56:49 going through people's heads. So, it's 56:50 got to start with that vivid vision of 56:53 this is why we're doing it and this is 56:55 why we're resourcing it properly. This 56:57 is why we're going to give you each 56:59 Well, we're not going to count tokens, 57:01 but why we're going to give you access 57:03 to all of these tools and the proper 57:05 training that goes with it. 57:07 Then, scouts. You figure out who the 57:10 scouts are in the business. Sounds like 57:12 for you, with 20 people, it's probably 57:14 you, Luke, and two or three others. It's 57:16 generally somewhere between 10 and 30% 57:18 of the business. 57:19 Those are the people that are like, I 57:21 would love to play with that. It's the 57:23 super curious. It's the high agency. I 57:25 wish there was a hiring test for high 57:28 agency. I would pay a a for that test 57:31 and only hire those people. If I think 57:33 back, I had the very, very good fortune 57:37 of hiring some incredible people. The 57:38 vast majority were incredible. 57:41 The ones that were 57:43 those incredible people were always the 57:44 ones high agency, not "Boss, I've got a 57:47 problem. Can you fix it for me?" 57:49 "Boss, I had a problem. I tried these 57:51 three things. This was the best way. I 57:52 think we should do this from now on." 57:54 >> Yes. 57:54 >> Awesome. Thank you very much. Right? 57:57 >> So, you figure out who the scouts are. 57:59 And then the scouts and someone from 58:01 leadership has to be a scout. You can't 58:04 delegate this. You can't abdicate this 58:06 and just say, "Right, you guys are the 58:08 scouts. Go figure out how to completely 58:10 change our business." Dan Gilbert is in 58:12 the weeds doing this stuff with those 58:15 scouts on his team. 58:17 And then the scouts figure out how to go 58:19 from base camp to camp one. 58:21 Once we get to camp one, once we've 58:22 paved that path, it's now on you, 58:25 leadership, to go back and get the third 58:27 group of people who are the trekkers. 58:30 And that's the 80% of your business who 58:32 aren't going to open a terminal window, 58:34 who aren't going to build a skill from 58:36 scratch, but are quite happy if the 58:38 paved road is in front of them. They're 58:40 quite happy to go, "This way? With Okay. 58:43 I 58:44 Let's go." Meanwhile, the scouts are off 58:47 figuring out how do we get from camp one 58:49 to camp two and building that path. And 58:52 then rinse and repeat. So, that's the 58:54 mental model that that works. You 58:56 mentioned MCPs. I know we're getting 58:58 into geeky territory, but you also said 59:00 the word nerdy, and I want to come back 59:01 to that word cuz I think it's not if you 59:03 are nerdy. I think there's too much of 59:05 this that's 59:06 if you're technical, you can do blah 59:08 blah blah. Technical Being technical has 59:10 got nothing to do with I don't think of 59:12 myself as technical. 59:14 I I realize the average person in an 59:16 agency probably looks at me and goes, 59:17 "Yeah, but you write scripts and you 59:18 play in terminal. You're a geek." 59:21 Probably, yes, but I don't consider 59:22 myself technical. A proper technical 59:24 person, a proper software developer, 59:26 would look at what I'm doing all day 59:27 every day and go, 59:28 "You're a numpty. You haven't got a clue 59:30 what you're doing." 59:31 But it's creating a ton of value. So, 59:33 there's that. So, I don't think you have 59:34 to be technical. I think you have to be 59:35 curious, and you have to be high agency. 59:38 And then, just to cover the thing on 59:39 MCPs, yes, I kind of use So, an MCP for 59:44 if you're still playing at home and 59:45 you're still listening to me an hour in, 59:46 good on you. 59:47 Um 59:48 model context protocol, 59:50 something that Anthropic created over a 59:53 year ago. 59:54 I think there was a very early version 59:58 that is kind of morphed into skills. 1:00:00 Custom GPTs, 1:00:02 vaguely similar, but they're kind of 1:00:03 going away. That was an early early 1:00:05 version. 1:00:07 An MCP is essentially a way to connect 1:00:10 to something else, usually some form of 1:00:13 data. So, Google Ads, thankfully, have 1:00:16 finally built their own MCP, and it's an 1:00:19 easier way to connect to something 1:00:22 instead of writing all of this API code, 1:00:25 often. 1:00:26 It's a way for AIs to talk to other 1:00:29 machines. However, as you said, 1:00:32 tokens. 1:00:33 Uh and I don't want to go too deep into 1:00:35 the weeds on this and talk about context 1:00:37 windows and stuff, but most Google Ads 1:00:39 reports 1:00:41 have a lot of data. Well, if you're 1:00:43 trying to pass that data through a 1:00:46 conversation, 1:00:48 that is not a good way to do it for two 1:00:50 reasons. One, the amount of 1:00:53 words, numbers, data that you can have 1:00:56 in a conversation is limited, but two, 1:00:59 it's called an LLM. It's a large 1:01:01 language model. We don't want a language 1:01:04 model to be doing maths. It's not what 1:01:06 they're good at. They're good at 1:01:07 predicting the next word. They're not 1:01:09 good at doing maths. Much better than 1:01:11 they were, 1:01:12 but they're not good at maths. They're 1:01:13 good at writing code that can do maths. 1:01:17 So, they can write a little Python 1:01:18 script, whatever the hell that is, that 1:01:21 can do maths. It's the way geeks used to 1:01:24 do or still do do data analysis, right? 1:01:27 They do stuff in R and Tableau and they 1:01:30 build code to analyze data. Us normal 1:01:34 humans, we don't do that. We open up a 1:01:35 spreadsheet and we go, "What's going on 1:01:37 here?" 1:01:38 So, MCPs not a good way to deal with 1:01:40 most Google Ads data. If it's a quick 1:01:43 question, which campaign spoke spent the 1:01:44 most yesterday, 1:01:46 great, use an MCP. But, if you're saying 1:01:48 go grab me the search term report the 1:01:50 last 90 days, 1:01:51 it's not good because that's a lot of 1:01:53 tokens. It's a lot of numbers. So, the 1:01:55 way to do it there is to use the API 1:01:58 instead. So, you can go grab data. So, I 1:02:02 I want to make it like 1:02:03 really relevant to you cuz that will 1:02:05 make it more real and concrete for 1:02:06 listeners. 1:02:07 >> Yeah. 1:02:08 >> You've got a bunch of pods, a bunch of 1:02:09 teams in different sub-agencies, right? 1:02:11 So, you might create up the top here 1:02:14 some sort of automation that 1:02:16 every morning at 4:00 a.m. 1:02:19 goes and pulls the data using the API 1:02:22 and saves that somewhere, probably as a 1:02:25 CSV, so that it's there and ready for 1:02:28 the team to use and they can only see 1:02:30 the data that they are allowed to see. 1:02:33 >> Right. 1:02:33 >> Whether you use We probably wouldn't use 1:02:34 Notion for that, but whether you use 1:02:35 Google Drive, whether you use The Brain, 1:02:38 or whether you use BigQuery, who knows, 1:02:41 right? But, data is stored somewhere. 1:02:43 Here's the data that you're allowed to 1:02:44 see. Now, you can go do stuff with that 1:02:46 data. You can pull that data into a 1:02:49 skill that we built and gave you. You 1:02:51 can pull it into an audit. You can pull 1:02:53 it into the 1:02:55 negative keyword tool that we built and 1:02:57 gave you and do the work on it. 1:03:00 And because we're still building trust, 1:03:02 the tool is going to give you some 1:03:05 output. Maybe it's a file formatted in 1:03:08 Ads Editor format. You can just drop 1:03:10 that into Ads Editor. 1:03:12 Or go to a bulk upload. Or maybe it's 1:03:14 going to give you a Google Sheet with 1:03:16 some ad copy. And we're not quite ready 1:03:19 to have this automatically change ads, 1:03:20 so we're going to copy and paste the ad 1:03:23 copy from the sheet into the ad account 1:03:26 cuz that's what we're comfortable with 1:03:27 yet. 1:03:28 But once you build the trust, you go up 1:03:30 a level, and then we get to the point of 1:03:33 yeah, just make all those changes for 1:03:34 me. Just take that and do that there. As 1:03:36 long as we've got some guardrails in 1:03:38 place 1:03:39 and some observability, we can check 1:03:41 that it has done the thing that we asked 1:03:43 it to, and we build the trust. We keep 1:03:46 moving up a level, and then the team are 1:03:48 going, "This is great. This is more fun. 1:03:51 I'm not copying and pasting crap." 1:03:53 That's, by the way, a great place to 1:03:54 look for anyone listening to this that 1:03:56 hasn't really 1:03:57 done much with AI at this point. Where 1:03:59 are you copying and pasting in your 1:04:01 business? That's a great place to start 1:04:04 because there's 1:04:05 a huge time saving there. 1:04:07 It's just integration. It's just 1:04:09 computers that need to talk to other 1:04:10 computers. You're copying stuff from 1:04:12 Excel and putting it in Xero. You're 1:04:14 taking stuff out of HubSpot and Google 1:04:16 Analytics and putting it into a sheet. 1:04:18 All of that stuff, start with 1:04:20 automation. We don't need AI for that. 1:04:23 AI can help create the automation. It 1:04:25 can even help suggest the automation. 1:04:28 Record that thing. I love that you use 1:04:30 Loom and have been for so long and 1:04:32 record everything. 1:04:35 There was this wonderful YouTube. I'll 1:04:36 try and find the link for you. It was 1:04:38 between Tim Ferriss and the CEO of a 1:04:41 company called Levels. 1:04:43 And his name began with a C, but I've 1:04:44 gone completely blank on that. 1:04:46 It was about 3 hours long. This guy has 1:04:48 four EAs just for himself. 1:04:51 >> Wow. 1:04:53 >> But now think of that as agents 1:04:55 >> Yeah. 1:04:55 >> or a brain or an AI. Like, what am I 1:04:59 doing that I don't need to do anymore? 1:05:02 And the question I'm asking myself more 1:05:04 and more these days is why am I doing 1:05:06 it? Like, what's the trigger? 1:05:07 When I get this email from Stripe, I go 1:05:10 and tell the brain, "Hey, can you do 1:05:12 this and add that person there and 1:05:14 deliver this?" 1:05:15 Why am I doing that? It's just Surely 1:05:17 you Hey brain, can you see this? Can you 1:05:19 Yeah, I've got access to Stripe. I've 1:05:21 got access to Gmail. Like, great. When 1:05:23 this happens, do this. 1:05:25 Great. Add it to my dashboard so I can 1:05:27 see that you're doing it and I can see 1:05:28 that you've done it right and you 1:05:29 haven't messed up. I've got 1:05:31 observability. I build trust and we move 1:05:33 on to the next thing. 1:05:34 >> Do you 1:05:36 So, going back to like founders 1:05:38 [clears throat] that need to be involved 1:05:39 like you said, "Hey, you have a 1:05:41 20-man team 1:05:43 and you you and Luke probably should be 1:05:46 a part of that, right?" And 1:05:48 you you should be leading this along 1:05:49 with probably a a couple of others. Now, 1:05:52 one thing that I have started to see and 1:05:57 I've you know, I I I can see where 1:06:01 some people will catch 1:06:04 the bug that this will give you, this 1:06:06 like lightening of inspiration and you 1:06:09 get hooked on this, right? But, I'm also 1:06:13 observing and seeing what I would say 1:06:15 are a lot of people playing with a new 1:06:17 shiny object and they're spending now, 1:06:22 you know, their full week or more than 1:06:24 half of their week on it. So, there's 1:06:27 like In in In the way I'm looking at it 1:06:29 is 1:06:31 Okay, I know this is important, right? 1:06:34 So, we have to work on it. But, what 1:06:36 time allocation 1:06:39 makes sense for me personally or 1:06:43 my Luke, the the the you know, my 1:06:46 business partner to be involved because 1:06:48 I see a lot of small smaller agencies or 1:06:50 smaller than us, um probably even, you 1:06:52 know, those that are bigger than us, 1:06:54 where it's like, if I looked at their 1:06:56 P&L sheet and I looked at their sales 1:06:58 volume and retention, I would say, you 1:07:00 know, it's probably not that smart to 1:07:03 only focus right now on AI when you 1:07:07 would need more sales volume and you're 1:07:09 doing nothing to generate that and your 1:07:11 retention's not really good and you're 1:07:13 doing nothing to fix that right now. 1:07:14 Now, I know learning AI is going to 1:07:18 absolutely help you, right, in the 1:07:20 future, but you've got to exist to the 1:07:23 future. 1:07:25 >> Uh, 1:07:26 that's a great point and a really good 1:07:28 question and I've got so many different 1:07:30 answers bouncing around my head. 1:07:32 Um, 1:07:32 there's a line from Dan Gilbert I want 1:07:34 to quote, theory of constraints and 1:07:38 useful. 1:07:39 Theory of constraints, right? Your 1:07:40 business is a pipe 1:07:42 and 1:07:43 some [clears throat] bits of the pipe 1:07:44 are wider and some bits are thinner. 1:07:46 There is no point spending any time on 1:07:48 the wide part of the pipe making that 1:07:49 wider when you've got this thin part of 1:07:51 the pipe because you make this part of 1:07:52 the pipe wider, doesn't matter that more 1:07:54 stuff can flow through there, you've got 1:07:56 a thin bit of the pipe, that's the 1:07:57 bottleneck, that's the constraint. 1:08:00 The thing to work on in your business 1:08:01 right now is the constraint. For so many 1:08:04 businesses, 1:08:06 for businesses out there, that 1:08:07 constraint is we just need more. We need 1:08:10 more customers, we need more sales, we 1:08:11 need more money in the bank and that's 1:08:12 why Google Ads has exploded over the 1:08:14 last 20 years, that's why our agencies 1:08:16 have been successful because 1:08:18 it works so much better than Yellow 1:08:19 Pages and letterbox drops and radio 1:08:21 advertising. Like that is the best way 1:08:23 to get more. 1:08:25 But 1:08:26 it's funny, when I was an E-Myth coach a 1:08:28 long time ago and trained with Gerber 1:08:30 and his team, 1:08:32 everybody thought the first module would 1:08:35 be let's go get more. It would be lead 1:08:37 generation, but it wasn't, that was 1:08:38 module five. 1:08:40 You started off with mindset and systems 1:08:44 and all of this stuff that prepared you 1:08:47 for the point 1:08:49 where leads would explode, but 1:08:52 if you explode leads in most business, 1:08:53 I'm sure you've done this for businesses 1:08:55 in the past, right? You double or triple 1:08:58 the amount of business coming in, 1:09:00 what breaks? Cuz something breaks really 1:09:02 quickly when that happens and you get 1:09:04 the phone call going, "I don't know what 1:09:05 you guys are doing, but can you turn it 1:09:06 off?" Because the guys on the factory 1:09:08 floor can't keep up. That's a wonderful 1:09:09 phone call to get, but you just lost a 1:09:11 client. But where's the constraint in 1:09:13 business? 1:09:15 Dan had a line in that article that 1:09:16 you're going to put in the notes, the AI 1:09:19 native thing, of like, "This is the most 1:09:22 important thing in our business right 1:09:24 now, and any time that I'm spending 1:09:26 doing anything else, I feel it's 1:09:28 wasted." But he's got a thousand-person 1:09:31 organization. I met Ken once, his head 1:09:34 of 1:09:35 uh acquisitions. I won't say why. 1:09:38 Um 1:09:40 amazing guy, but he's got this 1:09:41 incredible team around him that are 1:09:44 helping run the day-to-day business. His 1:09:46 job is just to sit on top of that 1:09:49 mountain and figure out the vision. 1:09:51 That's basically it. Yeah. He probably 1:09:54 hasn't touched an account in years. He 1:09:56 probably does the odd sales meeting and 1:09:58 brings in some whales, but that is his 1:10:01 job. For a smaller, typical 1:10:04 10-to-30-person agency, 1:10:08 the reason I said you need to be a scout 1:10:10 is certainly 1:10:12 one of you two. So, if you've got two 1:10:14 business partners, if you've got a 1:10:15 management team, at least somebody from 1:10:17 that top level needs to be one of the 1:10:20 scouts, because you can't just abdicate 1:10:22 it and say, "Go figure it out." 1:10:24 >> Sure. 1:10:24 >> Because you have to be aware of the 1:10:27 limitations of how it's going to be 1:10:28 used, of what it can do now. Just to to 1:10:31 build that intuition, to build that 1:10:34 feeling of "I think this is the right 1:10:36 path to go down." 1:10:38 Roughly where are things going to be in 1:10:40 six months' time. I think it's really 1:10:42 important to build that intuition. 1:10:45 None of us are going to be right 100% of 1:10:47 the time with that, but I think it's so 1:10:49 important to the business that you owe 1:10:51 it to yourself and the business and the 1:10:53 people that are depending on you to 1:10:55 build that intuition for this stuff, to 1:10:58 have a feel for it, not just read a 1:11:01 bunch of LinkedIn posts and go, "We're 1:11:03 going this way." Like, my team used to 1:11:05 suffer from what I used to call um 1:11:07 entrepreneurial whiplash. 1:11:09 Which is like, imagine you're the head 1:11:11 of the snake and like, "We're going this 1:11:13 way." And the whole business has to 1:11:14 orientate behind you to go this way. And 1:11:17 then you come in on a Monday morning and 1:11:18 go, "No, no, we're going this way." And 1:11:19 the whole team's like, "Whoa, 1:11:21 okay, now we're going over here." And 1:11:23 there's this like 1:11:25 And then 1:11:26 oh, actually no, I don't like that way. 1:11:27 No, we're going this way again. Ah, and 1:11:30 it's just like, yeah, the people at the 1:11:32 end of that long line are just getting 1:11:33 swept out into the distance every time 1:11:35 you It feels at the front of that line 1:11:38 like that's easy, I just change from 1:11:39 here to here. But you're you're turning 1:11:41 an oil tanker, you're not turning a 1:11:42 speedboat. It takes time for that 1:11:45 to move around. So, I think you have to 1:11:47 be 1:11:49 experiencing that and resource it 1:11:51 properly. Like, I am so sick of hearing 1:11:54 of agencies like, "Oh, yeah, we've got a 1:11:56 we've got a chat GPT account. Everybody 1:11:57 logs into the same thing. They they 1:11:59 share all that. Yeah, we're spending 20 1:12:00 bucks a month. What what more do you 1:12:02 want us to do?" 1:12:04 I want you to buy a proper subscription 1:12:05 for everybody on the team. 1:12:08 So, that they can play and experiment, 1:12:10 and they can take it home with them. One 1:12:12 of the best ways to get your team really 1:12:13 into AI is they take the subscription 1:12:16 and go build something little 1:12:20 that is meaningful to you. Something 1:12:22 from home life, something with your 1:12:23 kids. Maybe it's an illustrated bedtime 1:12:25 story thing, or it's something to do 1:12:28 with the Dungeons and Dragons game that 1:12:29 you like to play on Tuesday night, or 1:12:32 whatever. But, build that thing 1:12:36 and then you're going to go cuz you'll 1:12:37 you'll tip more time into that. You'll 1:12:39 you'll spend a couple of hours in the 1:12:40 evening building the thing that's 1:12:42 meaningful to you, not building an 1:12:44 automation for the boss. 1:12:46 That's a wonderful way for the team then 1:12:48 to experience this and say like, "You 1:12:50 can use this. We've bought you a team 1:12:52 license. Knock yourselves out. Use it as 1:12:54 much as you can. Use it in the evenings. 1:12:56 Please, go discover what this can do. Go 1:12:58 play with some LEGO bricks. 1:13:00 Because we have this amazing LEGO kit 1:13:02 that we keep talking about. The good 1:13:04 news is 1:13:06 you can build anything that you imagine. 1:13:08 The bad news is 1:13:10 there's no manual that says put this 1:13:12 brick here and this brick here. The only 1:13:14 way you learn this stuff 1:13:16 is to play with it and to experiment and 1:13:19 to figure out what it's good at and what 1:13:20 it's bad at. 1:13:21 >> Do you see like in the future cuz there 1:13:24 is a part of me, you know, for example, 1:13:26 I spent 2 weeks building negative 1:13:28 keyword skill, right? Now, would it take 1:13:30 me 2 weeks now? Probably not cuz part of 1:13:32 it was learning how to build skills and 1:13:34 whatnot. But 1:13:36 do you see and it's built off of, you 1:13:38 know, my trainings and the way I think 1:13:41 in how to do it. So, I do like how it 1:13:43 works, right? But, you know, the reality 1:13:46 is 1:13:47 there's 1:13:48 probably a thousand other Google Ads 1:13:51 people out there that could build just 1:13:52 as good of a negative keyword sort of 1:13:54 skill, right? So, there's a part of me 1:13:56 that thinks 1:13:57 okay, 1:13:59 you know, 1:14:00 and it kind of goes back to where like 1:14:02 the audit report that I 1:14:04 that that's probably one of the most 1:14:06 like time-saving useful things actually 1:14:08 that I've [clears throat] done recently. 1:14:09 And I'm like, "Wow, that's really cool." 1:14:10 When When I've been looking at it more 1:14:12 when I first started, it was like, "All 1:14:14 right, here's all the, you know, 1:14:17 work 1:14:18 workflows and optimization checklist 1:14:20 that we have. One by one, we got to go 1:14:22 build skills for all of that." And And 1:14:24 we are. We'll play around and do that. 1:14:26 But part of me is like 1:14:28 man, do most people need to go do that 1:14:30 cuz like a year from now I could see, 1:14:33 you know, Fred from Optimizer has like 1:14:35 the perfect AI tool for all of those 1:14:38 that's, you know, 80% there, good enough 1:14:40 for most people who are like working 1:14:42 within Google Ads accounts. It doesn't 1:14:44 need to be this entire sort of 1:14:46 ingredient playbook that your own brain 1:14:49 has built. And so, I'm looking at now 1:14:52 more of like like the audit thing that I 1:14:55 did. Like to me, it's more like where 1:14:56 exactly what can I do where I don't have 1:14:59 the resources for with like my team? 1:15:02 What can I just do now that I would love 1:15:03 to be able to do? And I'm starting to 1:15:04 create more there versus like individual 1:15:08 Google Ads sort of like uh 1:15:11 optimizations. 1:15:13 >> We had this exact discussion in my 1:15:16 mastermind group. So, I've got a a 1:15:17 mastermind that's limited to 50 people. 1:15:20 You'd be a great fit, actually. Um 1:15:22 where we talked about this. And 1:15:26 the upshot of that was kind of take the 1:15:28 pressure off yourself a bit. Um 1:15:31 not everything you build will get used 1:15:33 by the whole team. Not everything you 1:15:34 build will be valuable. But, the process 1:15:36 of going through that And it wasn't 2 1:15:38 weeks, right? It was on and off during 1:15:39 the 2 weeks. People shouldn't hear it as 1:15:41 like I didn't spend 100 hours on this 1:15:42 thing. But, 1:15:45 it took 2 weeks to get to the point 1:15:46 where it was useful. 1:15:48 But, all the skills you built after that 1:15:50 would have been quicker. 1:15:52 We have a choice. We could wait around 1:15:54 for that year. And there will be 1:15:57 incredible tools out there in a year's 1:15:59 time. 1:16:00 But, we'll be so far behind the curve. 1:16:03 We'll be just 1:16:04 grabbing things from LinkedIn and 1:16:06 commenting on and downloading it 1:16:07 and and stuck in the whole shiny object 1:16:10 merry-go-round. 1:16:11 >> Yeah. 1:16:11 >> But, if you spend a little bit of extra 1:16:13 time now And it is extra time. It is 1:16:15 time on top of an already busy work 1:16:18 week. But, this is the most 1:16:20 transformative technology of our 1:16:21 generation, possibly of our lives. And I 1:16:25 think it is worth spending a little bit 1:16:28 of extra weekend and evening time to 1:16:30 understand it, to work out how to use it 1:16:32 because 1:16:34 it compounds. 1:16:36 You learn those first few LEGO bricks. 1:16:38 And then you start to understand how to 1:16:40 put those together. And then you start 1:16:41 to see that level up and like, "Oh, well 1:16:44 if I can do that, maybe it could do this 1:16:46 and this." And you learn how to ask the 1:16:49 machine, "What can you do? Could you do 1:16:51 this?" And you start to push it because 1:16:53 this was the weird thing for me. 1:16:54 I understood that my skills were going 1:16:56 to compound. And like for those only 1:16:59 listening, I'm sort of like drawing a an 1:17:01 exponential curve with my hand here. 1:17:03 I knew the skills were going to 1:17:04 compound, and they start to build really 1:17:06 quickly, right? Compound interest, 1:17:07 you've seen that exponential curve. It 1:17:09 gets really fun at the end of that 1:17:11 curve. 1:17:12 What I wasn't ready for is that my 1:17:14 ambition 1:17:15 would compound. I would start to push it 1:17:18 further and further. I would start to 1:17:20 ask for more and more. I would start to 1:17:22 ask for things that I thought would be 1:17:24 impossible. I built a CMS 1:17:28 in 2 hours the other day because a 1:17:30 member said, "I'm building all these 1:17:31 websites for clients now. We're 1:17:33 replacing WordPress with this thing 1:17:35 called Astro, which I'd never heard of 1:17:37 before, 1:17:38 because the brain is able to build all 1:17:40 of these pages in Astro. I've built six 1:17:43 websites in the past year. I'm not a web 1:17:45 designer, but I don't use WordPress or 1:17:47 anything. I just talk to the brain. I 1:17:49 need a new page. Can you build that? Run 1:17:51 it through the copywriting tool. Make it 1:17:53 better. Let me tweak it a little bit. Go 1:17:56 build an image for that, please. Put the 1:17:57 image in the right place. Now crop it. 1:17:59 Resize it. 1:18:00 Just with natural language. 1:18:02 And 1:18:03 this a few clients have been using 1:18:05 Astro. Sorry, a few members have been 1:18:07 using Astro to build sites for clients. 1:18:10 Albut, 1:18:11 now what do we do? We've got clients 1:18:14 saying, "Can we make this change?" And 1:18:15 like, "Okay. Well, let's let's ask the 1:18:17 brain." 1:18:19 Lots of people using Astro, could we 1:18:21 build a CMS so that clients could log 1:18:24 into this thing and be able to make text 1:18:26 changes to the blog post or whatever it 1:18:28 is. 1:18:30 And it built it. That was like four or 1:18:31 five prompts. It took less than 2 hours. 1:18:34 It was 1:18:35 big but but constantly 1:18:37 pushing it. There's a wonderful, 1:18:39 wonderful member, shout out to David, 1:18:41 member of my mastermind group. He won't 1:18:43 mind me saying that he's in his 70s and 1:18:46 he will sit in his car while his wife's 1:18:48 in the 1:18:49 shop on a Saturday. He'll sit in the car 1:18:51 park 1:18:52 on his phone because he's not in front 1:18:53 of his eight screens and he'll be 1:18:55 building stuff from the car on a 1:18:57 Saturday. 1:18:58 He has built stuff that is going to 1:19:01 change his industry. 1:19:03 He's started advertising on 1:19:07 Is it TV or radio? I think it was TV for 1:19:09 the first time ever. Been in business 40 1:19:11 years. Just started advertising on TV. 1:19:13 And the TV guys were saying stuff to him 1:19:15 about results and blah blah blah and and 1:19:17 he knows Google Ads really, really well. 1:19:19 So, he's like, "Yeah, but how do we 1:19:20 measure the performance of this?" to a 1:19:22 TV station. So, he goes home, fires up 1:19:24 the brain and says, "This is what they 1:19:26 said. This is what I want. This is what 1:19:28 I want to be able to track. Da da da da 1:19:29 da da." And the brain built him this 1:19:31 incredible dashboard. He shows it to the 1:19:33 TV guys and the TV station are like, 1:19:37 "Um could could you build something like 1:19:39 that for us that 1:19:42 that we could use?" He's done the same 1:19:43 with his accountant. He's built tools to 1:19:45 show his accountant stuff and they've 1:19:46 gone, "How did What did this cash flow 1:19:48 forecasting that you built? How? How and 1:19:50 how do we do that?" He's like, "I just I 1:19:52 just ask. 1:19:53 I just ask the brain. 1:19:55 Like, don't be 1:19:57 He said this to one of the other guys in 1:19:59 the mastermind the other day. He said, 1:19:59 "You're too smart for your own good. 1:20:02 Just get out of your own way. Just ask." 1:20:04 We tend to assume that like our way is 1:20:06 the right way. Just ask the brain if it 1:20:08 could do it. Ask it if it's the if 1:20:10 there's a better way. 1:20:11 Cuz once it has the context, it 1:20:13 understands you, understands your 1:20:15 business, understands your clients, 1:20:16 understands the way you work, it 1:20:18 understands what you built in the past. 1:20:20 And if you give it enough information, 1:20:21 it kind of understands what you need to 1:20:23 do next. My entire to-do list, all of my 1:20:26 projects are all in the brain. I can say 1:20:28 to the brain, "What are the three things 1:20:30 I should work on today?" 1:20:31 And it said, "Well, at 9:00 you got the 1:20:33 call with Austin and then da da da and 1:20:35 it tells me how to structure my day." 1:20:38 >> Yeah. 1:20:38 >> If I wanted to. 1:20:39 >> Yeah. It's uh you know, it's funny cuz 1:20:42 [clears throat] we've got I'll I'll 1:20:44 share a bit of a I can't share the whole 1:20:47 screenshot. I'll share my screen here in 1:20:48 a second but cuz it's funny we we were 1:20:51 on Notion for a time but uh 1:20:54 like you know, 1:20:55 >> It's over engineered, isn't it? It's 1:20:56 it's it's more than you need. I'm sure 1:20:58 if you're a thousand person company, 1:20:59 Dan, you need something like that. But 1:21:01 the typical 1:21:02 >> small business doesn't need a lot of 1:21:04 that. Same with ClickUp, same with 1:21:05 Asana. There's a lot of stuff in there 1:21:07 that 1:21:07 >> I'll have to I'll have my editor block 1:21:10 some of this but like we've built now 1:21:12 our and this is over the last year 1:21:15 um 1:21:16 so our own sort of operating system, 1:21:18 right? And it's funny cuz now we've got 1:21:21 uh 1:21:22 our like this we just built Notion now 1:21:24 and like we so we're getting rid of 1:21:25 Notion cuz we just built Notion. And so 1:21:28 now I do have I have developers. So but 1:21:31 well where I see the future for us is 1:21:33 like 1:21:35 I there will be a period of time where 1:21:37 the team is on 1:21:41 Claude, whether it's co-work or if they 1:21:43 want to run with code, that's fine. But 1:21:47 like just as an experiment for fun, I I 1:21:51 asked my one developer, "Can you build a 1:21:54 campaign builder tool, right? Like can 1:21:56 you just build a tool? We have an AI 1:21:59 workflow process based off of like our, 1:22:01 you know, research, keyword research, 1:22:03 training and all of that." And sure 1:22:05 enough, I did a YouTube video on it and 1:22:07 shared sort of a 1:22:08 a sneak preview of it to the group but 1:22:12 we we have a full tool. It's not 1:22:14 released yet cuz like I I do not want to 1:22:16 release that until I 100% confident 1:22:19 there's no bugs but 1:22:21 we were able to take essentially a 1:22:23 skill, an idea and now it's like a fully 1:22:26 fleshed built out tool that you can go 1:22:29 in and and use. And now that's turning 1:22:32 into hold on a second. 1:22:34 >> [laughter] 1:22:34 >> There's too many ideas. We got to go 1:22:36 priority here in regards to like what's 1:22:38 going to move the needle for the 1:22:39 organization. But I do see the future 1:22:42 where 1:22:43 my team for the most part will probably 1:22:45 be inside what we call metrics. And a 1:22:48 lot of the skills that are like working 1:22:51 really well for us, we will just have 1:22:53 the developers sort of build that out 1:22:55 within side of our operating system cuz 1:22:58 I could see I would rather have my team 1:23:00 live inside metrics versus Claude and 1:23:04 this and that. 1:23:05 >> Yeah. And and we'll we'll we'll just 1:23:07 abstract that layer away, move up the 1:23:09 stack, and we'll build interfaces that 1:23:12 sit over the top of all this stuff that 1:23:14 are exactly the way we want them, that 1:23:16 work differently. Now, I don't mean to 1:23:18 minimize that at all at all cuz you're 1:23:20 ahead of 99% of agencies. 1:23:22 >> Please do that. I mean right yeah, we're 1:23:24 >> But but but let me just take you three 1:23:25 hills into the future and tease 1:23:27 something that you will get next week in 1:23:30 the May update of Ads to AI with the May 1:23:32 update of brain. 1:23:34 What I'm playing with just this week and 1:23:36 I'm going to release next week 1:23:38 is imagine now that software 1:23:41 that can be changed by them. 1:23:43 So they can get metrics and that's 1:23:46 delivered as a fully functioning thing, 1:23:48 but they can right click anywhere on 1:23:50 that screen and say like we should have 1:23:53 this thing here or it'd be really useful 1:23:55 if we had these buttons here, not just 1:23:58 these. Make those suggestions 1:24:01 that 1:24:03 obviously to begin with we got to build 1:24:04 trust. There's going to be a human in 1:24:06 the loop there. Maybe that bounces to 1:24:08 your developer, but your developer isn't 1:24:10 going to be writing that code. Your 1:24:12 developer is going to come up a level 1:24:13 and be the orchestrator of that and he's 1:24:16 going to be basically looking at a whole 1:24:18 bunch of tickets and going that's a 1:24:20 really good idea. Drag that over across 1:24:22 to that Claude get on with this column. 1:24:25 Meanwhile, go back to the 1:24:26 That's a good ticket. Claude, get on 1:24:28 with this. Claude's going to build it. 1:24:30 Claude's going to test it all. It's 1:24:32 going to spin up the new UI. You or the 1:24:35 person that created the ticket or 1:24:37 somebody is going to glance at it and 1:24:39 go, 1:24:40 "That's really cool." Tick. And it's 1:24:42 going to go into production. Does it go 1:24:44 into production just for them or for 1:24:46 everybody? That's up to you. 1:24:48 >> Mike, I could talk to you, you know, for 1:24:50 many more hours, right? But I know 1:24:51 you're busy. So, 1:24:53 I will end it with one more question cuz 1:24:55 I see a lot of fear posting going on, it 1:24:58 feels like, in regards to security right 1:25:01 now. In fact, there was just that case 1:25:03 It was Was it today or yesterday where a 1:25:05 car rental software company it deleted 1:25:08 the Claude 1:25:09 went in and deleted the entire database 1:25:11 like, 1:25:12 you know, it sounds like they were able 1:25:14 to recover it all, 1:25:16 but it you know, became this huge thing. 1:25:18 Uh 1:25:19 I what like And I know that you're not, 1:25:22 you know, you're not a lawyer, this 1:25:23 isn't advice in regards to this, but 1:25:24 like in regards to the security aspect, 1:25:27 like here at Grow My Ads, we're not 1:25:29 Nothing's being auto like edited in 1:25:31 accounts or anything like that. And the 1:25:34 way I have interpreted Claude, it'd be 1:25:36 like if I added client, you know, stuff 1:25:39 into Google Drive, right? Um now, I'm 1:25:42 not, you know, if someone shared their 1:25:44 like customer list with us, I'm not 1:25:46 going to feed that into Claude, right? 1:25:49 Um obviously that's That gets hashed 1:25:50 into audience list inside Google or Meta 1:25:53 or however you're using it. But um when 1:25:56 it comes to the security and some of 1:25:57 this fear people are hearing are hearing 1:25:59 and and seeing, 1:26:01 uh what like what what do you know in 1:26:03 regards to like, "Hey, listen, if you're 1:26:04 just pulling in CSV data or even you 1:26:07 have the API integration and you're 1:26:08 working with an account, it's safe 1:26:11 fairly safe as long as you have 1:26:12 guardrails and you're not like telling 1:26:13 it to like go just automatically do as 1:26:16 it pleases and make edits inside of an 1:26:18 account. 1:26:19 >> I think 1:26:21 this is also a very very personal thing. 1:26:23 Some people 1:26:25 refuse to paste campaign names and and 1:26:28 like just campaign level data, just 1:26:30 campaign names and metrics into an AI 1:26:32 window. 1:26:34 I 1:26:35 don't quite understand that viewpoint, 1:26:37 but I understand that people have that 1:26:41 belief or that way of thinking. I'd say 1:26:42 more so in Europe than America and 1:26:45 Australia as a rough generalization. 1:26:48 I think you 1:26:50 first of all just have to be 1:26:52 thoughtful about it. You you have to 1:26:53 have some sort of data policy as a 1:26:56 business and have a way to make sure the 1:26:58 people are doing that. Now, everybody's 1:27:00 data policy is going to be different. 1:27:01 For me, I'm a bit more fast and loose. I 1:27:04 am very much of the beg for forgiveness, 1:27:06 don't ask for permission 1:27:08 ilk. Um 1:27:10 so I'm probably not the best person to 1:27:11 talk about data security, but 1:27:14 um 1:27:15 I think that 1:27:18 as yes, guardrails are the far more 1:27:21 important thing, but the 1:27:24 the huge benefit 1:27:26 of not blindly moving 1:27:29 forward, but moving fast and using this 1:27:31 stuff and learning it and seeing what it 1:27:33 can do and what it can't do. 1:27:34 And you have to implement it. You have 1:27:36 to experiment. You can't know That's why 1:27:39 it's called an experiment. You can't 1:27:40 know beforehand this is going to work or 1:27:43 this won't or this might be a little bit 1:27:44 out of the bounds or not. 1:27:47 I see far too many people moving 1:27:48 incredibly slowly, being super cautious 1:27:51 and super safe. 1:27:53 >> Yeah. 1:27:53 >> just I worry for them, for their 1:27:55 careers, for their businesses because 1:27:58 if you move that slowly and everybody 1:28:00 else Now, I'm not saying we should all 1:28:01 race ahead blindly and just you know, 1:28:03 YOLO everything. That's the other 1:28:05 extreme and extremes are never good. 1:28:08 But moving too slowly and too cautiously 1:28:10 means you just don't get the benefits. 1:28:12 You're not going to get that 30% 1:28:14 efficiency a year. You're certainly not 1:28:15 going to double your business moving at 1:28:17 a sales pace. 1:28:19 I think there has to be some sort of 1:28:21 middle ground. 1:28:22 And a lot of it comes back to trust and 1:28:24 guardrails and the appropriate backups. 1:28:27 And you can give the tools 1:28:29 read-only access and not write access. 1:28:32 Right? I 1:28:33 My AI can read everything in Gmail. It 1:28:36 can download the attachments. It can 1:28:38 create drafts in Gmail, but it's not 1:28:40 allowed to send any emails yet. David, 1:28:43 who I mentioned in my master mind 1:28:44 before, his is sending emails all the 1:28:47 time. He has a couple of members on his 1:28:49 team who think it's weird to get an 1:28:51 email from Claude. 1:28:52 And they are struggling with dealing 1:28:54 with that. Most of his team's totally 1:28:56 fine with it because it allows the whole 1:28:59 business to move faster and be that much 1:29:02 more efficient. But some people are 1:29:04 old-school and like "I don't want to be 1:29:06 told what to do by bot." I can 1:29:07 understand that, but 1:29:10 my example for that has always been 1:29:11 Uber. You got a quarter of a million 1:29:13 people and their boss is a computer in 1:29:15 San Francisco. Like, they're all being 1:29:17 told what to do and where to go and who 1:29:18 to pick up and where to drive and how to 1:29:19 get there. 1:29:20 >> Yeah. 1:29:21 >> You're told by bot. So, 1:29:23 you have to have a data policy. It has 1:29:24 to be intentional. 1:29:26 It has to match the organization's 1:29:28 appetite for risk, I suppose. 1:29:30 Um you could hash everything. You could 1:29:32 redact everything. You could anonymize 1:29:34 data. You could take that search term 1:29:36 report and map every search term to some 1:29:39 other term or some number or hash or 1:29:41 something. Do all the analysis, come 1:29:44 back through, retranslate it all back, 1:29:46 and then present the report. You could 1:29:48 do that, but personally, I don't see a 1:29:50 problem in putting search term data with 1:29:53 some metrics into a large language 1:29:56 model. One, Google's already removed 1:29:58 everything useful in there for privacy 1:30:00 reasons anyway, but also 1:30:03 these large language models are not 1:30:05 storing all of that stuff in a database. 1:30:07 Nobody's trawling through that and 1:30:08 going, "Oh, let's find some interesting 1:30:10 Google Ads data in one of these accounts 1:30:12 in Anthropic." 1:30:15 >> That we know of, allegedly, right? 1:30:16 >> know of, allegedly, exactly, right? 1:30:18 >> as if you stuck that data in Google 1:30:20 Drive, right? You know, it's you you 1:30:23 think, well, Google's not like 1:30:25 >> Google's got pretty good security. Like, 1:30:26 a lot of people are trying to hack it 1:30:27 every day, and no one's managed to do it 1:30:30 yet. 1:30:30 >> Yeah. 1:30:31 >> You put it all in a Google Sheet anyway, 1:30:33 right? It's 1:30:35 it's not that different. And yes, 1:30:37 Anthropic and OpenAI keep your data for 1:30:39 30 days, and but you can opt out of most 1:30:41 of that, at least once you start paying 1:30:42 them a little bit of money. So, yeah, 1:30:44 again, understand the rules. If you 1:30:46 really, really want, I wouldn't do this, 1:30:48 but if you really want, you go get a 1:30:49 lawyer, and they tell you what to write 1:30:51 in your policy, and they tell you what 1:30:52 you can and can't do. But, again, for 1:30:54 me, that is 1:30:56 going a bit overboard. But, I think the 1:30:58 main thing to answer the question 1:31:00 specifically is just you have to think 1:31:01 about it. You have to decide 1:31:03 what you as a business do, 1:31:06 um and what the rules are, and make that 1:31:08 really, really clear for everybody and 1:31:10 your agents, increasingly, to be able to 1:31:13 to follow. Had a really good thing from 1:31:15 Aaron Levie at Box the other day. 1:31:17 He was saying that, you know, 1:31:18 let's assume everybody has an agent at 1:31:21 some point in the future, right? Let's 1:31:22 assume just one for now, but each person 1:31:23 has their own agent. Just like we in the 1:31:25 agency used to give everybody a their 1:31:27 own VA. 1:31:28 It gives them leverage. 1:31:30 Your agent will never have more 1:31:32 permissions than you do. So, each person 1:31:35 at each level of the business, whatever 1:31:37 they that person has access to, their 1:31:39 agent can't have access to more than 1:31:42 that. If you're in a sub MCC, your agent 1:31:44 doesn't magically get access to 1:31:46 everything. If you're plugging into some 1:31:49 HR data, and you see your one-on-one 1:31:51 reports and your salary, your agent 1:31:53 doesn't magically get to see everybody's 1:31:55 salary. And that's a really good, simple 1:31:58 framing of let's just start there. 1:32:01 Build it so it doesn't get more than you 1:32:03 do. And then because we we've solved 1:32:05 that problem in organizations for 1:32:08 hundreds of years. If we understand that 1:32:10 there are different levels and different 1:32:11 people need to see different things. 1:32:12 Let's just start there and 1:32:15 iterate from there. 1:32:16 >> Yeah, I've yet I've yet to see anyone 1:32:18 who's like just throwing accounts to 1:32:21 Claude and saying go make changes and 1:32:24 and like it's I I so again, I think it's 1:32:26 a lot of just 1:32:27 fear posting for clicks and and debates 1:32:30 and and it's fine. It works for the 1:32:32 conversation, 1:32:33 >> But it's it's a t-shirt I want to get 1:32:36 built at or made which is common sense 1:32:38 isn't. 1:32:39 >> Right. 1:32:40 >> But it is just common sense. You 1:32:42 wouldn't just do that. You wouldn't hire 1:32:43 a new person and day three on the job, 1:32:46 you'd say just go do all of that stuff, 1:32:47 please with no training, with no 1:32:49 context, 1:32:51 with no systems. 1:32:52 It's it's just common sense. Just Yeah, 1:32:55 don't don't do that, please. 1:32:58 >> Yeah, I agree. 1:32:59 Uh Mike, this was uh incredible. It 1:33:03 uh you know, I can't I don't remember 1:33:05 the last time I actually have talked to 1:33:06 you, uh but it's hilarious cuz you know, 1:33:09 we just picked up right right on this uh 1:33:11 recording and uh 1:33:14 I I owe you to a lot of my success. I 1:33:17 would not have learned Google Ads at the 1:33:20 level I did without your group, without 1:33:22 your training in the early days and so 1:33:26 um you know, I think very highly of you 1:33:28 and uh I'm very grateful that I somehow 1:33:31 crossed your path in the weird world of 1:33:34 the internet uh 1:33:36 all the way you in Australia and me in 1:33:37 the you know, the small city of Fort 1:33:39 Wayne, Indiana. So uh it's incredible 1:33:42 what you've built and what you're 1:33:43 building with Ads to AI. 1:33:45 Uh I thought you might have been losing 1:33:47 your mind many years ago when you were 1:33:49 in the Facebook group always talking 1:33:51 about, you know, all of the uh this AI 1:33:54 stuff uh very early on and uh so I knew 1:33:58 when 1:33:59 um, you know, when it came to Google Ads 1:34:00 and and where AI's at today, I knew that 1:34:02 if there's anyone who's got it figured 1:34:04 out, it's you. And sure enough, you do. 1:34:06 So, anyone who's listening to this, 1:34:08 go do 80/20 skill. Uh, again, that the 1:34:12 the website to that will be uh, in the 1:34:14 description below. And just go buy his 1:34:16 course, spend the few hours it takes to 1:34:19 go step one, two, three, three, four. He 1:34:21 makes it very simple. A a guy like me 1:34:23 was able to go through it and you 1:34:25 absolutely will get hooked and you'll 1:34:28 realize using uh, these tools such as 1:34:31 Cloud Code is it's really not that hard 1:34:33 once you get into it. So, thank you, 1:34:35 Mike. 1:34:36 >> I appreciate saying all of that, but I 1:34:38 also appreciate you doing the work. You 1:34:40 took the training, yeah, but like anyone 1:34:43 could do that. Lots of people did. Not 1:34:45 everybody went and implemented it and 1:34:47 certainly not to the the level that you 1:34:49 have implemented. Now, you're doing it 1:34:50 again with all of the brain stuff. So, 1:34:52 thank you. That's what every teacher 1:34:56 always wants, right? Is for someone to 1:34:57 actually do something with it. So, 1:35:01 I'm just the the first step. You would 1:35:03 have absolutely got to where you've got 1:35:05 with or without me, but I appreciate 1:35:06 that you have um, 1:35:08 implemented everything so well and 1:35:10 continue to be so curious. I think that 1:35:12 is the most important thing right now is 1:35:14 just being curious and being willing to 1:35:16 play. It's you're doing it. Thank you. 1:35:18 >> Yeah, absolutely. All right, well, I'm 1:35:20 excited for how the future evolves with 1:35:22 this. So, I'll have to have you on 1:35:23 again, you know, uh, in the future cuz 1:35:25 I'm sure there's going to be a lot of 1:35:27 advancement and new things and exciting 1:35:30 >> fairly slowly, doesn't it? This AI 1:35:31 stuff. 1:35:32 >> Yeah. So, cool. All right, thanks, Mike. 1:35:35 >> Thanks, Bram. Good to see you.