YT YouTube2Text

YouTube2Text → Transcripts → Mike Rhodes Shares Valuable Vibe Coding Lessons From Real Marketing Builds | PPC TH 121

Mike Rhodes Shares Valuable Vibe Coding Lessons From Real Marketing Builds | PPC TH 121 — Transcript

by PPC Town Hall · 11,030 words · 1,535 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Hello and welcome to another episode of
0:02PPC Town Hall. My name is Fred Vales.
0:04I'm your host for this show. I'm also
0:06the CEO and co-founder at Optimizer, a
0:10PPC management tool. For today's
0:12episode, we have a returning guest, but
0:14someone we haven't talked to in quite a
0:16while, Mike Rhodess. He's done many
0:19things in the PPC space, but currently
0:21he runs ads to AI and he has been very
0:24active building AI vibecoded tools. Um
0:28he's very well known for his Pmax script
0:31that many people use. So I'm very
0:34excited to bring Mike onto the show and
0:36just talk about how AI makes his life
0:39easier and what can we all learn from
0:42that. I know there's a lot of AI
0:43skeptics out there, but Mike definitely
0:45seems to be one of the people who's
0:46making the most of it. So uh let's see
0:48what we can learn from him and let's get
0:49rolling with this episode.
0:56Mike, thanks for coming on to the show.
0:58>> Thanks for inviting me back. It's
0:59wonderful to be here. Good to talk to
1:01you again.
1:02>> Yeah. Hey, since uh we last spoke, tell
1:05people a little bit about what you've
1:06been up to and also for those who
1:08haven't met you before, who is Mike
1:10Rhodess?
1:11>> Oh my goodness. Um so for 17 years of my
1:15life, I built and ran um an agency here
1:19in Australia. Um, loved that.
1:24And then
1:26a number of things came together. One
1:28was realizing that I had 134 Mondays
1:31left to walk my daughter to primary
1:33school. Um, we countdown every week.
1:37Some days I get two days, but at least
1:39every Monday we're down to 40 now. Uh
1:43but that realization and just seeing
1:46what was coming with AI. Um I got an
1:51offer for my agency. So I took that
1:53couple two and a half years ago and just
1:57got incredibly lucky once again. Right
1:59place, right time. And I have been
2:01building. I had just relearned to code
2:04before that. Not to your level, but went
2:08back and taught myself code probably
2:10four or five years ago now. I had a
2:12coding mentor that I would talk to on
2:14the weekends and all of a sudden GPT4
2:18arrives and that sort of level of
2:22curiosity was able to be almost all my
2:25questions got answered in instantly. How
2:28do we do this? Why are we doing it this
2:29way? Is there a different way we could
2:30do it? And that just sort of led me to
2:34play around. I I had this tool that I'd
2:38played with since I remember it's 2018
2:40because I remember Ralph interviewing me
2:42on his podcast at traffic and conversion
2:44in 2018. This thing called the profit
2:46curve and it was this tool I had inside
2:49of a Google sheet that I loved but many
2:53of my team didn't even know it existed.
2:56For others it was too hard to use or too
2:57hard to explain to a client. Um, it was
3:00designed to solve that problem of how
3:03much should we be spending this month?
3:05This is a long answer to a short
3:06question. Um, but I built this tool out
3:10and it was just this moment of like ah
3:13um and I was doing that pretty manually
3:15back then. This is kind of before fiber
3:17coding existed. But that led me to then
3:20go well okay then what can we build? So
3:23about a year and a half ago I built out
3:25this tool that's still live. It's still
3:27free. It's called 8020 agent. It does a
3:30fraction of what something like
3:31optimizer does. Um, but it just helps
3:34you see your data in different and
3:37interesting ways. It's it can be a
3:39useful tool. It's a bit clunky at times,
3:41hence the name 8020. It's not polished
3:43and finished. It does most of it. Um,
3:47and then that just kind of led me to to
3:50more and more things. I then people
3:51would start asking me, well, how how do
3:53I how do I build something like this?
3:54which then I ran some courses last year
3:58called Build the Agent. Um, and at the
4:00end of that first cohort, a lot of
4:02people were like, we what next? This was
4:04great, but this was only eight weeks.
4:06Like, we want someone to hold our hand
4:08and and and run. What I've always loved
4:11doing is running four or five hills into
4:13the future. And then coming back and not
4:16just reporting on what I've seen, but
4:18giving people useful, practical tools.
4:22And it's funny, I I see this arc now
4:24where I've always done this. I built the
4:26display grid way back in whatever that
4:28was, 2016, 2017, remarketing tools. Like
4:32I've always done this along the journey.
4:34I just didn't realize it at the time.
4:37And so I created ads to AI. So it's a
4:40community. I sort of curate and and
4:44provide content around AI and try and
4:48remove most of the noise and just give
4:49as much signal as possible.
4:52And just because I constantly spend six
4:55months in the future running over those
4:56hills, I keep looking around and finding
4:58these fun things like 8020 Brain, which
5:01I'm sure we'll talk about, and just um
5:04building it because I love to learn. But
5:07I've worked out that learning stuff
5:09without making it useful, without
5:12teaching it, without turning it into
5:14something, learning stuff just for
5:15learning it doesn't doesn't do much for
5:17me. So, I need to teach. I need to share
5:19it. This First thing I did when I found
5:22Google Ads in 2004 was like rush back
5:25and tell everybody I knew about this
5:27amazing thing called Google Ads. While
5:29you were toiling away in a cubicle in
5:30Mountain View somewhere, I was trying to
5:32like tell everybody that this is
5:33amazing. This is like stop doing yellow
5:35pages and radio ads. Like this is the
5:37thing. And it took me about a year and a
5:39half of people saying to me, mate, I
5:42don't care how it bloody works. Can you
5:44just do it for me? And I had to hear
5:46that a lot before I fell into starting
5:48an agency. Um, and it's kind of been the
5:51same again. I'm like, "This thing is
5:53amazing. This is going to change
5:55everything." And you're right, there's a
5:56lot of AI skeptics, and there's a lot of
5:58things AI shouldn't be used for, but
6:01there's so many amazing things that it
6:03can help us do, not replace humans,
6:06replace tasks, and help us with many
6:10other tasks. And I just I I'm just I've
6:14always been curious and this is just God
6:16what a time to be a curious learner
6:18because you can get answers to every
6:20question and build stuff which is the
6:22other thing I've always done. Lego
6:23blocks model airplanes. I've always
6:26loved building stuff and now you just
6:28get to sort of like go from idea to
6:30building to impact sometimes in like 10
6:34minutes. It's ridiculous.
6:36>> Having so much fun.
6:37>> It's pretty crazy. And now you seem as
6:39excited as I am about the ability to
6:41build stuff very quickly. But uh you
6:43mentioned an interesting point there
6:44where you're running one hill, but
6:46you're already looking four or five
6:47hills ahead.
6:48>> So let's maybe delve into that a little
6:50bit deeper.
6:52>> And what are those hills that we're
6:54going to be looking at in the next
6:55couple of months and years?
6:59>> I just think Andrew Ang said this and
7:01you know Andrew because you were at
7:03Google, but Andrew Ang is like the
7:04smartest person that most people have
7:06never heard of. founder of Corsera,
7:08basically the guy that introduced AI
7:10into Google back in the translate days
7:132010 by now deep learning. He said back
7:17in 2017, AI is going to be like
7:20electricity. It is just going to be
7:22ubiquitous. It's going to be in
7:23everything. It's going to be everywhere.
7:26That's how I started my talk in traffic
7:28and conversion was with was his quote
7:29and talking about how electricity rolled
7:31out and how many decades it took for us
7:35to to use electricity and at the time
7:37lots of people went oh yeah this is
7:39going to be big but how we used the tool
7:42was not obvious and I think it's very
7:44very similar this time there are so many
7:46different ways this can be used it's
7:48going to take time to be adopted
7:51especially in enterprise but it is
7:55changing everything. It's just that most
7:58people don't have the time to put down
8:02what they're doing cuz they're running
8:03at 110% redlinining and I've got all
8:06this client stuff and I've got to
8:07deliver it and I've I've got to get it
8:09done. I don't have time to go and watch
8:11a two-hour YouTube video talking about
8:14the latest stuff. That's the stuff I
8:15love to I get to spend a ton of time out
8:18there and then just bring back the fun
8:20stuff. What specifically? Um, I think
8:24one of the biggest ones is AI as a
8:27thinking tool.
8:29Now, as soon as you say the word
8:30thinking, that implies work to a lot of
8:32people. I don't like thinking. I just
8:34want to get stuff done. I love thinking.
8:36I love going for a walk and
8:38brainstorming. And now I get to chat to
8:40an AI with that brainstorm to push me
8:44into
8:45areas of that conversation that I
8:47wouldn't have otherwise got to and to
8:50get there faster. and to help me find
8:53the signal in the noise. Um, a great
8:56example of that is is YouTube. I used to
8:59read a lot of books. I got about a
9:00thousand books over there. I've read
9:03most of them, not all. Um,
9:06and I realized at some point in the last
9:08two or three years that a lot of that
9:11was just in case learning. You know, you
9:13read a book about business.
9:16It's kind of like just in case I have a
9:18problem that this thing solves and maybe
9:21when I have this problem 18 months time,
9:23I'll be able to remember some of the
9:24stuff in this book about what Seth Goden
9:26told me and I'll be able to put it into
9:28action. And I've moved to more sort of
9:31like just in time learning where I'll go
9:35to YouTube or I'll read an email
9:37newsletter or Substack or something and
9:39I've got this specific problem. let me
9:41go search out an answer to that problem.
9:43I think that's probably what most of us
9:45do. Now, I could just do that at scale.
9:48I can grab transcripts from a bunch of
9:51different YouTube videos,
9:54stick them all into an AI, and say like,
9:56here's my specific problem. Help me
9:58think through how all of this
10:00information might help me. It's not
10:03perfect. It's not going to give me an
10:05answer and I'm going to go, yes, I'll
10:07just brush off blindly and do that. But
10:10it helps me think through and I think
10:11that's one of the biggest things that I
10:14see AI being used for. I'm not sure if
10:16I've answered your question there.
10:18[laughter]
10:18>> Yeah. No, I see it as a as a very smart
10:21partner in business who you can ask
10:24questions and brainstorm. I would
10:25certainly see that as well. Um you
10:28recently
10:29>> Go ahead. Sorry.
10:31>> No, go ahead.
10:32>> I was going to say because it has such
10:34incredible knowledge, right? it it has
10:36basically read everything that's ever
10:38been written. Whether that's a good or a
10:40bad thing will leave for another day,
10:42but it has all of this knowledge built
10:44in.
10:46Um, it's it's kind of how you extract
10:49the useful bits of that for you in your
10:52context. Um, and how you push the AI
10:56into those sort of unexpected areas.
10:59Garbage in, garbage out. If you if I I
11:02liken it to a solar system, right?
11:05There's this huge big planet in the
11:06middle called planet average because
11:08let's face it, the AI has been trained
11:10on Reddit and Twitter. That wasn't a
11:11very good idea, was it? But it's it's
11:13sort of like the average amalgamation of
11:15of everything. And if you ask a a quick
11:18little question, you get an answer from
11:20planet average. You have to really push
11:23it that there are millions of planets in
11:25this weird solar system that I've got a
11:28picture of in my head. And by asking a
11:31better prompt, by giving it better
11:33context, by setting up the AI, you're
11:35sort of able to push it to this far
11:38distant planet in this weird universe
11:40and get an answer that there's an answer
11:43on every single planet.
11:45And how you ask the question determines
11:47which planet the AI goes to to answer
11:50your question. And the the more you're
11:53able to push it and steer it and and say
11:56like that corner of the universe
11:58somewhere over there there's a there's a
12:00really good planet with a really good
12:01answer on nudge it over there rather
12:03than just getting an answer from planet
12:05average. That makes a huge difference in
12:08how much benefit you get from the AI.
12:12>> Yeah, that's an interesting way to uh
12:13visualize the vector space because
12:15that's basically the pictures that
12:16people see, right? It's all these dots
12:18in a three-dimensional space, but the
12:20vector spaces are many more dimensional
12:22than that. So, it's it's very limited
12:24representation. But the whole notion of
12:26what you're talking about is what we
12:28also see is the more specific you get
12:30into your request and prompt.
12:33uh you basically limit the box in which
12:35the AI has to respond
12:38>> and you point it in a direction of one
12:40of those boxes and so um you get much
12:43better answers and and usually we think
12:45about that the other way right we're
12:47like well we don't want to put too many
12:49constraints to a human because the human
12:51is going to struggle with answering
12:53things in the voice of say Albert
12:55Einstein but for the AI that's no
12:57problem I mean it just goes to that
12:59planet or that piece of the vector space
13:01and it h gives you the answer that
13:03actually makes a lot more sense based on
13:05what you were specifically looking for.
13:07>> I never thought of it in those terms,
13:08but I've always been a a fan of the
13:10power of positive constraints. Like
13:12artists will tell you this, um
13:14songwriters will tell you this, painters
13:16of like, paint me a picture. Uh okay,
13:20paint me a picture about this very
13:21specific thing and this specific style
13:23and make the picture this big. Like all
13:25of those constraints actually help guide
13:29the artist. So maybe it's a bit of that
13:31in my thinking.
13:32>> Yeah, makes total sense.
13:35>> Um, and so like a specific PPC example
13:37would be something where you say, "Hey,
13:40listen. I'm I'm writing some new ads or
13:42creative assets, but who are you writing
13:45them for?" And oftent times we think in
13:46terms of those averages. We start with
13:48the keyword and everything's about the
13:50keyword, right? Well, okay, if we're
13:51selling this, then let's mention this in
13:52the ad. But now with this creative
13:55partner that has all of this knowledge,
13:57why not say write the ad for someone who
14:00lives in Paris or someone who lives in
14:02New York because they will care about
14:04certain different things. They talk
14:05about different things and and this LLM
14:08will know that and so it pushes the
14:10creativity uh to a new level and and in
14:13doing so makes things more relevant to
14:15the person who's seeing the ad and that
14:18will boost your quality score which in
14:19turn makes your CPC cheaper. and we're
14:22back to the old fundamentals of how
14:24Google Ads works.
14:26>> Fundamentals aren't changing. But yeah,
14:27it's it's hard for us to get out of our
14:29own little bubbles sometime. And even if
14:31you don't use those ideas from the air
14:34because you look at them and go, "Ah,
14:35that's rubbish. That won't work." But
14:37every now and again, maybe you want to
14:38test one of those. But yeah, I it's very
14:40hard for me to get into the head of a
14:4222-year-old woman that lives in Paris.
14:44Like I just don't know what she's
14:46thinking about. But um yeah, so it's
14:49just it's just more grist for the bill,
14:51right? It's more ideas. You don't have
14:53to use it, but it's often that sort of
14:55thought starter that that prompt
14:57>> to push you to a different planet into
15:00different places and look at things from
15:02different angles and that change of
15:04perspective that makes you a better
15:07marketer or a better business person
15:08frankly.
15:09>> Yeah.
15:10So staying on planet earth about
15:12leveraging the many smart people that
15:14live on this planet. You recently built
15:16a tool uh that had quite a bit of
15:18traction on LinkedIn. Uh and I am
15:20fortunate to be one of the people in
15:22that tool. But you basically built a
15:24council of experts. Tell us a bit about
15:28what this tool is and and how that maybe
15:29connects to what we were talking about
15:31here.
15:32So, it it kind of goes back to the
15:34YouTube transcript thing I was talking
15:35about before where I found myself more
15:38and more or listening to podcasts,
15:40watching YouTubetubes and and they were
15:42just so dense, especially when I'm
15:43walking. There's a ton of stuff. I was
15:45listening to a wonderful podcast by Rory
15:47Sutherland. You've probably read his
15:48book, Alchemy. It's a wonderful book,
15:51but there was just so much in there. I
15:54sort of needed to pause for a moment.
15:56It's like almost God, I wish I still
15:58walked around holding my moleskid, which
15:59I hardly ever take anywhere now. and
16:02write all these ideas down. And I jotted
16:04down a couple of notes, voice notes.
16:06When I got back, I dug out the video,
16:09found the transcript, threw it into
16:11Claude, my preferred AI, and said, "This
16:14is the thing I was thinking about, and
16:15here's the transcript. What are all of
16:17the ways that these sort of ideas
16:19connect? How how would Rory solve this
16:22problem?" And it was really pretty good.
16:25So then I just sort of took it a little
16:27step further and I went, "Well, actually
16:28there's a whole bunch of really smart
16:30business people like Seth Goden that I
16:31would love to ask this same question
16:33to." And so I did that. I sort of
16:35collected a bunch of information about
16:36them and asked them, but they're all a
16:39bit different. I'm like, and then I
16:40found myself cutting and pasting
16:42conversations and like, "Hey, this is
16:44what Rory said and this is what Seth
16:45said. Could you like That was just a lot
16:47of cutting and pasting and a bit of a
16:49mess."
16:50Um, so then I just thought, well, okay,
16:54Claude, um, you know lots about all
16:56these people. You've read all of their
16:58stuff.
17:00Pretend there's like five of them in a
17:01room. Here's the question. What do you
17:04think they'd say? And it was, you know,
17:07it was okay, but it was a bit messy. And
17:09then I remembered this book that I read,
17:11God, 25 years ago, probably. this
17:14thinking tool from a guy called Edward
17:16Dabono, one of the best thinkers on the
17:18planet, I reckon, called six thinking
17:20hats. And I know that he's taught this
17:23everywhere from like eight-year-olds to
17:2488-y olds. He's taught this in primary
17:26schools. He's had company boards do it.
17:28It's a way of putting on your
17:29metaphorical hat, the six different
17:31colors. Just what we were talking about
17:33before that forces you into different
17:35perspectives. What are all the risks
17:37associated with this? What are all the
17:38good things that could happen? What's my
17:40gut say? These six different ways of
17:43doing it. and I just smooshed the two
17:45together.
17:47Six thinking hats is really good because
17:49it gives everybody around the table.
17:50I've used it in management meetings and
17:52stuff like that. It gives everybody
17:53around the table a voice so the quiet
17:55person in the corner actually gets heard
17:58and the guy, it's usually a guy who
18:00talks all the time doesn't take over the
18:03conversation and make everybody else
18:05just roll their eyes and want to leave
18:06the meeting. And so
18:10the facilitator says, "Right, everybody
18:12put on your black hat now. we're going
18:13to look at the risks and why this won't
18:15work. And you go around the table and
18:17everybody gets a voice.
18:19And so I just it was this iterative
18:21process of smooshing these things
18:22together. And I thought, "Oh, this would
18:24be fun to do with Google Ads experts."
18:26Um, so I said to Claude, "Right, there's
18:29a few people I've got in mind, but who
18:30would you recommend? If we were to pick
18:3230 people, who would you recommend?" Of
18:35course, number one, PPC Influencer two
18:37years in a row. Of course, it gave me
18:39your name and a whole bunch of other
18:40names and a few names that I hadn't
18:42heard of. And I said, I really want this
18:44person to be on. And and so we ended up
18:46building out a whole bunch of personas,
18:48just publicly available information. I'm
18:50not using anything private. I'm not
18:51using the contents of Kirk's wonderful
18:54books, for instance, because maybe not
18:55everybody has bought that book, and
18:57therefore they shouldn't benefit from
18:59that private information. You need to go
19:01buy the book if you want to benefit from
19:03that. but the freely available
19:04information that's out there. A couple
19:06of podcast transcripts and whatever
19:08else, blog posts and so on. Built these
19:10personas, threw it all together. I've
19:12got a Google Ads one, a marketing one,
19:14and a business one. And the first test,
19:17I remember the first test I did with
19:18this. Threw an idea in there. And it
19:21came back and went, "Mike, that's a
19:24really good idea." I went, "Yeah, okay.
19:25Here we go. But don't do that now. It's
19:28the wrong time to do this." I was like,
19:30"What? an AI that just said, "No, it's
19:34not just blowing smoke. This what?" But
19:37the weird thing was it sounded just like
19:39the two smartest women in my life, my
19:42wife and my old general manager. When I
19:45had the agency, I had this really,
19:47really wonderful general manager, who I
19:49occasionally referred to as my business
19:50wife, but that's just a bit weird. So,
19:52we'll call her general manager, Zoe, and
19:54my wife, Gabby. And it just sounded
19:56exactly like what these two really smart
19:59people would say. Mike, that's a really
20:00good idea, but no, not right now. And
20:03then it went on to explain why.
20:05And I was just like, this is amazing.
20:08And I just started throwing other
20:10questions at it. Built out the the
20:12counters. I think I've got about 60
20:14different experts on there now. Seth
20:16Goden, Rory Sutherland. Um, it's it's
20:19just it's fun. And then yes, so I built
20:21the little tool with the Google Ads one,
20:25created little avatars because I've I've
20:27learned and you know this, but show
20:29don't tell it. It's lovely running
20:31inside of an AI, but it's all textbased.
20:34So I just made a little sort of some
20:36pretty pictures and just a little
20:37demonstration of it and put it on that
20:4080/20 brain website just to to to
20:43visually show the idea. Um, and yeah, it
20:48kind of most people love it. couple of
20:50people hate it. Um, so it's probably a
20:53good thing. [laughter]
20:54>> Okay, good. Well, people are using it
20:55and giving feedback. That's amazing. Uh,
20:58now let's talk a little bit about how
20:59you created that, right? So, you said
21:01you're not a developer by educational
21:05background necessarily, but you you took
21:07this whole concept of something that was
21:09working really well on AI, but was very
21:12tedious, manual. You had to go back and
21:13forth, copy and paste a bunch of
21:14prompts. Now you've taken it to the
21:16level where anyone can go and sign up
21:18for 8020 Brain. They can play with this
21:21and it's basically a piece of software.
21:24So how did you go from not knowing how
21:25to code to producing something that is
21:29an amazing tool?
21:30>> We're going to need a bigger boat. Um
21:32the short answer is Claude code. The the
21:36longer answer is probably bit by bit of
21:40just oh like building this the profit
21:42curve that was one page and lots of
21:45cutting and pasting and I was the glue
21:47holding all of those bits together. I
21:49would get some HTML and some JavaScript
21:51from over there and some CSS and smoosh
21:53all that together and build a page. The
21:56AI coding tools are very intimidating if
22:00you've never learned to code. um Google
22:03Ads scripts. For the last two, three
22:06years, I've been teaching people about
22:07Google Ads scripts. I've got my scripts
22:09and sheets mastery course. We did build
22:11the agent. The the name from ads to AI.
22:15I kind of see as this bridge that I want
22:17as many freelancers and agencies as
22:19possible to like come on, let's get over
22:21the other side of this bridge because I
22:23don't think just offering Google Ads
22:24management services two or three years
22:26from now is going to be enough. For a
22:29few it will be, but for the person
22:32that's doing an average job of that,
22:34it's not going to be enough. So, we need
22:36some other skills, some other
22:38capabilities. Google ad scripts are the
22:40first step on that bridge. And then I
22:43had to think, well, okay, I've been
22:44saying that for a while, but what are
22:45all the other steps?
22:47What are all the other skills and
22:48capabilities and and
22:51mental models that we need to go from
22:53ads to AI? And yet, vibe coding became a
22:57thing. What was that like six 12 months
22:59ago? Um, apparently that's what I've
23:03been doing for the past two or three
23:04years is vibe coding, which is
23:06essentially
23:07talking to an AI quite literally. I talk
23:10to it all day and then it does stuff.
23:14I'm not writing any of the code. I know
23:17just enough to look at it and make sure
23:19that nothing's going horribly wrong. But
23:22you look at it as much as you want to. I
23:24wanted to learn to code. So in the early
23:26days I was looking at everything that it
23:29wrote and thinking why is it doing that
23:31and asking questions about it but you
23:33really at the point now if you were to
23:34start today you really don't need to do
23:36that at all. I know plenty of proper
23:39software devs that don't look at the
23:42code anymore. many will look at bits of
23:44it but they're trying to sort of the new
23:48way of developing and I don't know I
23:50don't know any like proper proper ones
23:52who are doing this in development right
23:53who work for Google and are building
23:55real stuff but a lot of them are sort of
23:58trying to code more and more hands-off
24:00and be the orchestra not the violin
24:01player you don't have to learn all of
24:03the different instruments and learn all
24:05these different coding languages and all
24:07of the minutiae syntax used to bug me I
24:10I could never remember the syntax, but
24:14now you've got to be the conductor of
24:15the orchestra, understand what the thing
24:17is that you want to achieve and
24:20describing that goal to the A, not how
24:23to do it. It knows how. You just have to
24:25explain the what and the why and let it
24:28rip. And the wonderful thing about code
24:30is it either runs or it doesn't. So, we
24:32started doing this with Google ad
24:33scripts. It either runs or it doesn't.
24:35And if it doesn't, you get the error
24:36message and you chuck it back into the
24:38AI and go, "This please fix." and it
24:41does and you go around that loop three
24:43or four times and now you've got a
24:44working script. But that automated
24:47reporting is just the first step on that
24:50bridge. So then we started thinking,
24:51well, okay, what next? How do I get AI
24:54to help think with that? Now, how do I
24:57automate all of that? So, I'm getting
24:58automated insights every Monday morning.
25:01Okay, what's the next bit? The next bit
25:03is really adding a bunch of business
25:05context. the more information you can
25:07add about your client, about their
25:10goals, about their strategy. Exactly
25:12what you said before, the constraints
25:14and the preferences, not just like, oh,
25:17the budget's 20 grand this month, but
25:20all of that information, the more you
25:21bring to the table, the more you push it
25:23away from planet average and get to
25:26something really interesting. And then
25:28you just keep iterating. The worst thing
25:31you can do with AI today, I think, is to
25:33sit down at the whiteboard and try and
25:35plan everything out. This is what
25:37exactly what I'm going to use it for
25:39because you don't know what you don't
25:40know yet. So, you just start build
25:42something really little, a budget pacing
25:45tool or something. And then you iterate
25:47and go, "Oh, wow. Well, that was easy.
25:49What's the next step on the bridge?
25:51What's the next step?" And we just keep
25:53walking across the bridge and keep
25:55playing. I'm really not sure if I've
25:57answered the question again. And I've
25:58wandered off on tangents here, but um I
26:00I am absolutely loving this ability of
26:03it to just and to help it. It also
26:05pushes back a bit and guides you along
26:07this path and says if you ask it what
26:08should be the next step here, it'll go
26:10well here are four directions we could
26:12go in. What do you think?
26:14>> Yeah, the question was how did you go
26:17from non-developer to producing
26:19software? So I think you did answer it.
26:21Um I'm curious about some of the the
26:23pitfalls though. So some of the pitfalls
26:25that we've seen is um if the code is not
26:28initially well written and it fails and
26:31you persist in telling the AI can you
26:34please fix it like this is the expected
26:35outcome eventually it'll just put an if
26:38then loop which says okay if Fred is
26:40asking for it then always give this
26:41answer because that's what's going to
26:43make him happy. Um so you do have to be
26:46a little careful there and then the
26:48other challenge that we see is sort of
26:49in the nuance of uh PPC. So, so when it
26:52comes to
26:55[snorts] I keep giving this example
26:57maybe too much, but like if you say uh
26:59suggest some keywords, okay, here's some
27:01keywords, but are these duplicate
27:03keywords for your account? And that is
27:05not just a simple is it the same text in
27:08one place as the other, but it's also
27:09are the campaigns geoargeting, day
27:12parting, same locations, same times. uh
27:15are there budget constraints on some
27:17campaigns that might make them actually
27:19not really competing? Um so there's all
27:22this nuance and so that's where the
27:24trick for us is how do you combine
27:26flexible code and AI and sort of all of
27:29this new stuff with the more
27:31deterministic which is like very hard
27:33and set rules that keep you within safe
27:36boundaries but then start doing more
27:38innovative things within these
27:39boundaries. Um, I'm kind of curious if
27:42you've run into any uh funny instances
27:45of where things didn't go exactly as
27:47planned.
27:48>> Oh, every day. Yeah. Like if if
27:50something isn't working and less so
27:52today than even four months ago, but if
27:56something isn't working, it'll just
27:57delete a chunk of code. Like you asked
28:00me to write a test to make sure that
28:01this was working. It didn't work. So,
28:03I've deleted the bit that wasn't working
28:04and now all the tests pass. It's It can
28:07be a really really dumb intern some
28:09days. Um, but on the nuance part, I
28:13totally agree and I think that's why
28:16there's a level of ambition, right?
28:18There's a scale between, I don't know,
28:20skeptic and complete like AI is going to
28:23do everything. I think you want to kind
28:24of aim for somewhere in the middle.
28:26Asking it to do everything inside of a
28:28Google Ads account just Yeah, that's a
28:31recipe for disaster. For close on 20
28:34years, I've been trying to build the
28:36ultimate flowchart of if this, then
28:38that, and if impression share is this,
28:40and this type of campaign, and I've
28:42never been able to do it. I've taught, I
28:45don't know, tens of thousands of people
28:46how to do Google ads at this point, but
28:49I've was never able to create like the
28:52ultimate ultimate guide. And then I
28:54found this thing. If you came across,
28:56god, what's it called? Um, the Drafus
28:59method. Nat Elias has a wonderful blog
29:02post about this. Just search Natalias
29:04Drafus method. There's a two images in
29:06there that changed everything for me
29:08about how I learn but also how I teach.
29:10And it's about going from novice to
29:12expert. And that last sort of 10% is
29:17really all about intuition. You cannot
29:19put it into a flowchart. It's the tacet
29:22knowledge. It's um the the learn by
29:24doing and the experience.
29:27There will come a point I suspect where
29:30AI might be able to do a lot more of
29:32that. But it's the problem with tacet
29:35knowledge is it's really really hard to
29:37get out of your head and into an SOP.
29:39And the AI needs input. If you can write
29:42it as an SOP,
29:44many agencies will have like a a VA
29:47maybe somewhere in the world helping
29:49them do stuff, following along a
29:51checklist, creating a search query
29:53report or something. download it from
29:56here and we put this heading and we
29:57change the column sizes that no problem
30:00for an AI but understanding the nuance
30:02between this keyword in this location
30:04and all the stuff that you just said
30:07don't use an AI for that stuff yet have
30:09it make suggestions sure but the human
30:12needs to stay in the loop the human
30:14needs to be that sort of doing that last
30:15bit and probably a few other bits along
30:17the way I'm not saying AI should be
30:19doing 98% of what all all ad
30:22professionals are doing but it should
30:24not be try. We should not be trying to
30:26ask it to do 100%. But we can ask it
30:30first. One great tip, there was a
30:32question I saw on the on the LinkedIn
30:34post about sort of future skills and so
30:37on. I think one of the best things you
30:38can do right now is before you do the
30:42task, have a chat to your friendly AI
30:45and say, "This is the thing I'm about to
30:47do. I'm going to do this and the
30:48client's asked me to do this and this is
30:50the situation." Maybe even show it a bit
30:52of data. You can take screenshots and
30:53throw those in and say, "How should I be
30:56thinking about this?" Not try and do it
30:58for me, but how should I be thinking
31:00about this? And it might just tell you
31:01what you already know. That's okay. But
31:04often it will say, "But have you thought
31:05about this?" Or, "Did you look for this
31:07reason for that drop in performance
31:09there?" Or, um, have you considered that
31:12this keyword might be in the account
31:14three times for a reason because there
31:16are different geographies or bid
31:17strategies or whatever it might be. um
31:20using it for that little extra bit of
31:22thinking which the first few times you
31:23do that it takes twice as long so no one
31:26does it because like I've just got to
31:28get this thing done but if you can get
31:30in the habit of doing that then you
31:32start seeing the shortcuts of oh
31:34actually when I do do this and I package
31:37it up in this way and I remember to do
31:38it this way actually that thing that
31:40used to take half an hour now now only
31:42takes me 10 minutes every time I'm going
31:44to save time but I'm going to have to
31:46invest a bunch of time up front which is
31:49why it makes it so hard.
31:52>> Yeah. And it's back to your example
31:53about using it as a thinking partner.
31:56>> Um, and one of my friends from Google,
31:59Clay Bour, he uh runs Sierra AI at this
32:03point.
32:04>> And he him and Brett Taylor were making
32:07the point that when AI fails, it is
32:11usually not due to a shortcoming of AI.
32:13is due to you not having been specific
32:15enough
32:16>> and having grounded it enough with
32:18context that it needs to do a good job.
32:20>> Yeah.
32:21>> Um and I think that's very similar to
32:22the point that you're making which is
32:24you know it's uh you need to think with
32:27it first to answer the questions that
32:29should be answered rather than just
32:31letting it go and do its thing because
32:33then it might make decisions that didn't
32:35align with how you thought about that
32:37problem and how it should be solved
32:39>> because it's a it's a Labrador puppy,
32:40right? it it is going to try really
32:42really hard to please you and it's not
32:44it's it's more interested in pleasing
32:46you and doing what you've asked for than
32:48it is about doing the right thing. So if
32:51you ask it to do the thing, it'll have a
32:54its best crack at actually doing that.
32:56But yeah, maybe it doesn't have you
32:59haven't given it enough context because
33:02that is all in your head. I was chatting
33:04to someone yesterday about how they
33:05choose negative keywords of all things.
33:07I'm like, you have an algorithm running
33:09in your head every time you look at that
33:11list of search terms. You you don't know
33:13it because it's just second nature to
33:15you because you've done this so often,
33:16but but there there is some sort of
33:18algorithm running in your head. You are
33:19thinking about all of these different
33:20things and comparing them. You're just
33:22doing that so fast and so intuitively
33:24that you don't know that you're doing
33:25it. But try taking that algorithm out of
33:29your head and writing an SOP. That is
33:31sometimes virtually impossible to do
33:34because there's so much nuance. But you
33:36can get the first 80% done. 8020. Quick
33:38plug. You get the first 80% done with an
33:42AI a lot of the time. But yeah, you I I
33:45really agree with that. I haven't heard
33:46it said that way before, but I like
33:47that. It's not usually a shortcoming of
33:49the model. It's how you're using the
33:51model. But usually context. Context, I
33:54think, is is basically everything.
33:56Garbage in, garbage out. Right. That's
33:58how we learned how to code a long time
33:59ago was that was the first thing you
34:01learned. Garbage in, garbage out.
34:04>> Exactly. Yeah, giving it the context
34:06which is basically grounding the model
34:08in the reality that is your truth,
34:12right? Which is often what are the
34:14actual numbers in a Google Ads campaign
34:16because we've all seen examples where
34:18you ask it for budget predictions and
34:20just make stuff up and it's like well
34:22how was it supposed to do this job if
34:24you didn't tell it what's actually going
34:25on? Even little things like telling it
34:27what today's date is and you're like hey
34:29we're currently seeing some like uh
34:32there's a lot of searches for roses.
34:33what's happening. It may not realize
34:35we're coming up on Valentine's Day. And
34:37so grounding it in the fact that hey,
34:38we're two weeks away or four weeks away
34:41from Valentine's Day. Uh that would be
34:43helpful. That is a piece of grounding
34:44that helps it better answer the question
34:46that you have.
34:47>> I I I go back to my universe picture in
34:49my head of like your first job is just
34:51to push it into the corner of the
34:53universe where a decent answer lives.
34:54Again, there is an answer on every
34:56planet. And that's how I think of
34:57grounding. You're just sort of giving it
34:58the facts, which in in six thinking hats
35:00is the white hat. You start by giving
35:03the council the facts. It's just that
35:06now with the brain and so on, it's it's
35:08able to go pull all of the data from the
35:11Google Ads account, look at the campaign
35:12structure, pull all of that to so you
35:15don't have to do a lot of that manually.
35:16You don't have to be the glue in between
35:18all of those steps necessarily anymore.
35:21Start there. Start with the facts and
35:22then say, "Right, this is what I've
35:24got." Yeah. Grounding is the is the
35:25proper way to call it certainly.
35:27>> Yeah. Anyway, just the technical terms.
35:31Um, [laughter]
35:32but uh, yeah, let's just have it make
35:34sense for the the viewers. So, thanks
35:35for
35:35>> absolutely your planet universe.
35:38[snorts] Uh, so Mike, let's talk about
35:40another question. Um, so this was from
35:43Gabriella Benedeti who runs conferences
35:46in Italy, good friend of ours.
35:48>> He, uh, he's asking how to be good about
35:50reusing elements. So you developed a
35:53very popular PMAC script. uh you've now
35:56built this council of experts in the 80
35:58to20 brain. Um what Gabrielle is
36:02basically saying is like he also builds
36:04these solutions but often finds that
36:05they become standalone and not
36:07interconnected.
36:09>> Have you found good ways to turn this
36:10into a Lego block sort of system where
36:12you can leverage one thing into another
36:15tool that you're building? I remember
36:18saying gosh in a talk like seven, eight
36:21years ago like one day AI will be a
36:24whole bunch of Lego blocks and we'll
36:26just go to the shelf and pick off Lego
36:28blocks and put them together. We're not
36:29there yet. We are kind of there now
36:32eight years later. Um, I've done exactly
36:36the same thing that Gabriella did of
36:38build something and it's just sort of it
36:41works once for this specific thing, but
36:43there's really not much about it or you
36:44have to pull the whole thing apart to
36:46use it somewhere else.
36:50How would I solve that? I think thinking
36:52in terms of
36:54the smallest individual useful bits. So,
36:59uh, I don't know how to describe this
37:01without going down a huge rabbit hole
37:02and talking about skills, but
37:05a skill, all we need to know about for
37:07now is a skill in claude code terms,
37:09which is an AI. It's just a package of
37:12things. It's a a package of a a prompt
37:15and some examples, maybe even a little
37:17bit of code that explains to the AI,
37:20this is the way I want you to do it. And
37:22then once you have a few of those, you
37:24can start to chain those together. put
37:27the Lego blocks together.
37:29The good news is you can build anything
37:32you like at this point. There is no Lego
37:34brick that you can't think of that it
37:35can't build. The bad news is there is no
37:38Lego manual. There isn't that thing over
37:41there that you open up and it says,
37:43"Right, step seven, you put the blue
37:44brick on the red brick and then step
37:46eight, you put the window on the the
37:48manual doesn't exist. There is no
37:49manual. The only limitation is your
37:52creativity and your imagination. but it
37:55can build the Lego bricks for you and
37:57then you get to put the Lego bricks
37:58together in different ways. Um
38:02gosh, I've got so many competing
38:04thoughts in my head at the same time. Um
38:06on one hand, it doesn't matter that you
38:09built this thing and it only works one
38:11way because it is so fast now to say,
38:14"Hey, go look at this thing. See how I
38:16built it. Now build it for this thing
38:17over here. This is different. This is
38:19different." Again, all of the context,
38:21the constraints, the preferences, the
38:22guidelines, and so on. now go rebuild
38:24it. And at this point, um, it probably
38:28takes two minutes to build the new
38:30thing. So, you almost don't have to
38:32think, oh, how do I pull this apart into
38:34eight different Lego bricks so that I
38:35can use these Lego bricks just in case
38:38in the future? You just point it at the
38:40thing you built and go, I want one of
38:42those, but mixed with this and a bit
38:44like that, and it'll build it. On the
38:47other hand, having those individual
38:49pieces, thinking through what are the I
38:52guess in um James Clear terms, the
38:54atomic habits like the the little pieces
38:56that go into this
38:59systematic thinking like draw it out, I
39:02suppose, as a workflow like Niels has
39:04done for some of his scripts, like
39:05literally draw the boxes out on the
39:08whiteboard. First, we want to do this.
39:10We want to grab the data from over
39:11there. Uh, then we clean and sort the
39:14data because we don't need 40,000 search
39:16terms. We just need the ones that have
39:17spent more than $10. Great. Now, what do
39:20we do with it? What's the next step?
39:21What's the next step? Those are
39:23essentially the the Lego blocks. And
39:25then you have a chat with your friendly
39:28AI and say, "This is what I want to
39:30build." And just ask the AI, I want to
39:32build this in a way that makes it
39:34reusable in the future. I don't know how
39:36yet because I don't know what the next
39:37thing's going to be. But can you make
39:39sure when you build this that it's easy
39:41to branch off um and build something
39:45similar but different in the future. It
39:47it is so good now that we even the
39:50patterns from six months ago, never mind
39:52two years ago, kind of almost don't come
39:55into play now. It's I was building
39:57something yesterday and I just I usually
39:59have three or four AIs on the go because
40:01I got used to like okay you do this and
40:03you do I'm I'm the trying to be the
40:05conductor of the orchestra and I'd go
40:08back 12 months ago I watched a lot of
40:10YouTube because you'd get these things
40:12running and then I'll go watch some
40:13American football highlights or
40:15something else or something I've been
40:17meaning to watch on YouTube for a while
40:18some climbing video and it ding ding
40:21ding that all of them took less than 2
40:23minutes. Hang on. I haven't even got to
40:24the end of the first quarter yet and
40:26you've all finished. I need to come up
40:27with more work for you. It's it's
40:29incredibly fast and good now that I mean
40:34that's almost the hardest part is
40:36actually the hardest part is stopping.
40:38That's that's what I found because Go
40:41on. Sorry.
40:42>> Well, stopping is the hardest part and
40:44one thing that I've mentioned many times
40:46before on the podcast, but it's uh the
40:48most exciting time in my life, but also
40:49the most stressful time in my life
40:51because the opportunities have
40:54multiplied to an infinite degree. And
40:56it's like all these things where in the
40:57past I would have just said, well, it's
40:59kind of a cool thing, but I don't have
41:01time. Let me focus on the higher
41:02priority stuff. Now, like you said,
41:04because you have these four AIs running
41:05at the same time and they're so fast,
41:07it's like, well, should be doing this,
41:09should be doing that. Like, I've been
41:10thinking about that for 10 years. Why
41:11don't I do this now?
41:12>> Yeah.
41:13>> Um, and that's also a little bit of a a
41:15stress then, right? But, but I like your
41:17whole example of
41:19you don't think about
41:22a solution that scales to everything.
41:24So, historically, the based software has
41:27financially made sense was you build
41:29something that can be used by lots of
41:31people.
41:32>> Yeah. Yeah,
41:32>> because that's what pays the bill. So,
41:33but to make it work for lots of people,
41:35you have to make it at some level
41:37generic so that it fits many use cases.
41:40>> Now, in the age of ondemand software
41:43that's being built by Claude Code or
41:45Lovable or Replet or Base 44 or who have
41:48you,
41:49>> yes, take the thing that you built for
41:51one vertical and say, "Hey, here's the
41:53thing, but now we want to do it for a
41:55different vertical. What how would we
41:58change it for them to make it more
41:59relevant?" Right? And so that becomes so
42:01easy that at some level you don't have
42:04to think about building blocks anymore.
42:06>> I um
42:07>> def I want to add something to that
42:08because it's so so important. Um, a lot
42:12of members of my community have found
42:15this quite difficult and and I was
42:18surprised that they found it difficult,
42:19but I mean how you would have, I
42:22imagine, coded with your team back in
42:25Google back in the day is you'll stand
42:27around a whiteboard for three months
42:30sometimes and build this huge waterfall
42:31plan and it's all very and you know
42:33before agile came along. You plan out
42:36like certainly how Microsoft would have
42:38worked 30 years ago. You plan everything
42:41out for a year or two when we're going
42:42to do Windows 95 and this is what it
42:44looks like and you and then you throw
42:46that plan over the wall to the geeks who
42:50unsurprisingly
42:52come up against new questions and new
42:54blocks and and things that nobody's
42:56thought about because you can't think
42:57about everything and stuff runs overtime
43:00and over budget and that's how a lot of
43:03project stuff used to work and then
43:05somebody invented agile and ah actually
43:08maybe instead of planning everything
43:09out, we'll plan out the next little bit
43:11and we'll do that and we'll learn a
43:12bunch along the way. And once we've
43:14learned that stuff, then we'll plan the
43:15next bit. And that is by far with the
43:20tools we have now a better way to go of
43:22just iterating. I'm pleading with people
43:24all the time. I know you don't know
43:26exactly what you want to build. Have a
43:28rough idea of the direction you're going
43:30in. Just build version one. Just get
43:32some sort of proof of concept, minimum
43:34viable. just get it to grab some data
43:37from there and put it in there. And then
43:39when you see it do that, you go, "Oh,
43:41can we have a little chart of that
43:42data?" Yes, you can build the chart. Oh,
43:45well, if I've got a little now, can I
43:46have a little thing that changes the
43:47date range? Let's do that. And it's just
43:49you just iterate and iterate and iterate
43:52and then it becomes hard to know when to
43:54stop because you have another idea. And
43:56it is literally a hold down a key on the
43:58keyboard, talk to your friendly AI, can
44:01you build this with a bit of that and
44:02some of this and oh, can you make it
44:04pink? Great. And by the time you've
44:06thought of the next thing, it's it's
44:08it's done the thing. And it's just so
44:11many times Gab will be like, "It dinner
44:12time. Dinner's ready." I'm like, "Yep,
44:14hang on." Right. You do that. You do
44:15that. You do that. You do that. Um Oh.
44:18Oh. Oh, hang on. It's finished
44:20already. Um, okay. Now do this.
44:22[laughter] It's just
44:24giving it more of stuff to do. I had a
44:26Trello board with some of those ideas
44:29that you're talking about, the someday
44:30maybe and and things I wanted to add to
44:32the PMAX script and things I wanted to
44:34build and a website that I'd never got
44:36round to over Christmas. I just got
44:39enough's enough. Took a screenshot of
44:40the Trello board. I didn't even try and
44:42figure out the API or the MCP or
44:44anything. I just took a screenshot of
44:45the Trello board, dropped it into Claude
44:47and went, which of these can you build
44:49first? I went, right, I think we should
44:51do this one, this one, this one, this
44:52one, and here's why. Um, these things I
44:54don't think you need anymore. this one
44:55you've actually done. Would you like me
44:57to get started? Off you go then. And it
45:01just it it I've had 80 90% of my Trello
45:04board done. One of my members I've got
45:07to tell this story. This is so good. One
45:08of my members, David, who didn't know
45:10what the word GitHub was
45:14eight, nine, 10 months ago, and he won't
45:15mind you saying he's in his 70s.
45:17Wonderful guy. Was a client of Web Zavi
45:19back in the day. One of the first
45:21members of my agency Zavi group way back
45:23when. He's now in my mastermind
45:26and he demonstrated what he does for his
45:29business. So he started the business 40
45:31years ago. Now his daughter and
45:32son-in-law run it. He's sort of just
45:34like been sitting back building all of
45:36this stuff for the last six, eight
45:38months, vibe coding, I suppose. And so
45:41they had this offsite the other day.
45:43He's in Scotland and he demonstrated to
45:45the team some of the stuff that he's
45:47been doing. And they all just like
45:48looked at each other, right, you're not
45:50allowed to do anything else except this
45:52now. like, "No more fiddling with the
45:54Google Ads account. No more coming out
45:55on on on site with us and doing
45:58installs. You just do this." And then
46:01the next day, his head of installs
46:03called him and said, "I've been thinking
46:05about what you I won't try and do the
46:06Scottish accent. I've been thinking
46:07about what you showed us yesterday. Do
46:11you think you'd be able to build an app
46:14for my guys on site?" And David was just
46:18about to say like, "It's not as easy as
46:19that, mate. Like, there's a lot that
46:21goes into this." and we have to think
46:22about things. Yeah, I'll see what I can
46:25do. And he sent me the conversation.
46:28He said to Claude, "I'm just about to go
46:30to bed. Here's what I want you to do. I
46:33want to wake up to a fully functioning
46:35app for the installers." Because he had
46:37the brain, so it's got all of the
46:39context in there. There were some PDFs
46:40lying around about installing a bunch of
46:42stuff that was in there. He just said,
46:43"I want to wake up to a finished app.
46:46Go."
46:48He said he sat down the next morning at
46:507:00 a.m. He'd forgotten he'd done this.
46:52He opened up his computer and went, "Oh,
46:54what's this?" And a fully functioning
46:57app with links and PDFs and all he had
47:01to do was drop it into Versel and deploy
47:03it, send it to his install guy and went
47:06like this, you mean? [laughter]
47:09And it all worked first time on his
47:12guy's phone. Now the whole team are able
47:14to use it on site. And it was literally
47:16a twominut can you build this because he
47:21knew how to ask and he knew all of the
47:23files were there. All of that grounding
47:24that you've talked about all of that
47:26context. I I picture it as like the AI
47:28model in the middle and all of this
47:29context is the sort of the next ring
47:30around that that is so so important. It
47:33built the whole thing for him overnight.
47:36It just
47:39>> crazy.
47:40>> Yeah. It's so bloody useful. It's it's
47:43if you embrace it and don't
47:47expect it to do everything, it's not
47:50going to do that nuance. But it can help
47:52with so many other things. And stuff
47:55that that used to take us well stuff
48:00like you were saying, stuff that used to
48:01take so much time that we probably
48:02wouldn't do it. We wouldn't have time to
48:05do it, that stuff now becomes
48:07accessible. And some of that stuff is
48:08now accessible at scale. There's
48:10something that maybe you would do for
48:12your top 10% of clients once or twice a
48:15year because it was worth it for them,
48:17but you never had the time to do it for
48:19the bottom 80%. Now you can do that for
48:22every client every month. There's it
48:25just removes so many of the the limits
48:27and the and the barriers and the
48:28constraints.
48:30>> So [snorts] let's bring it back to uh
48:32one of the questions we had in digital
48:34marketing. I mean obviously there's a
48:36ton of excitement that you and I share
48:38around this but if you could say what is
48:41one skill and this is Sharon's question
48:43but what is one skill that you think is
48:45currently underrated in paid ads
48:48>> and that's that would make the biggest
48:50difference in the next few years.
48:54>> My flippant answer was going to be
48:56thinking but I won't go there. Um I
48:59think business skills business skills
49:02business context. I think what I always
49:03sort of had in the back of my mind for
49:05web savvy was to create
49:09create McKenzie for the small businesses
49:12that couldn't afford Mckenzie. But
49:14really quickly I worked out that all the
49:16really really smart people got hired by
49:19McKenzie for an amount of money that I
49:21would never be able to afford to pay
49:23them. And so it never really happened.
49:24Our best clients were those clients who
49:27we were the strategic adviser for that
49:30called us for to ask about other stuff,
49:33not just other marketing stuff, but
49:34other business stuff. But that didn't
49:37scale because there were me and a few
49:40other people that could answer those
49:41business questions. But a lot of people
49:43were like, "Well, I was hard to do
49:45Google Ads. I know Google Ads. I don't
49:46know anything about business. This is my
49:48first job." And they're asking me about
49:50this help? Can can can someone that
49:53knows this help me? just it didn't scale
49:54very well. But I think that
49:58that business context and understanding
50:00more about business, it's never been
50:03easier to learn, but that is going to be
50:05the thing that yes, the fundamentals are
50:08still very important and that's
50:10obviously like table stakes. You have to
50:12know the fundamentals of of running
50:14Google ads. But knowing more about
50:16marketing, knowing more about
50:18advertising, understanding different
50:19channels, but then the level above that,
50:21understanding
50:23the client better and why are they
50:26asking for this and what do they really
50:28need and what is going on and how else
50:30could we help them with these other
50:32areas of their business. Now, that won't
50:34work for all agencies and it won't work
50:35for all clients because some clients
50:37will be, "Hang on, I I hired you for my
50:39digital advertising. I don't want you
50:41telling me about my strategy. Like,
50:43we've got that under control. Thanks
50:45very much. Off you go. But others will
50:47be delighted that you're showing an
50:49interest in that, that you're thinking
50:51about that, that you're wanting to help
50:52with that. And I think more and more
50:55that will make the difference because,
50:59and you've written about this very
51:00eloquently for a very long time, right?
51:02That there's more and more automation
51:03inside Google Ads accounts. Those that
51:05automation is available to everybody.
51:07That's table stakes.
51:09Everybody's using the same tools. It all
51:11kind of converges to the same point.
51:13Most of the wins happen outside of the
51:16Google Ads account. Whether that be the
51:18landing page, the offer, how the person
51:20answers the phone in the legal office or
51:23whatever it is, it it it's outside of
51:25Google Ads. That's where most of the
51:27wins can be had. So, understanding more
51:29of that part, I think, will be more of a
51:32differentiating factor in the future.
51:35>> Yeah, that makes sense. Um, and he kind
51:38of said like doing the thing that's
51:41going to become automated,
51:43you know, looking for your search terms
51:45with $100 in spend, no conversions,
51:49right? Why do we need to pay someone to
51:51do that? Optimizer automates that,
51:53scripts automate that. Like that's the
51:55easy stuff, right? But understanding
51:58what do we do next? Why did these things
52:00not convert? Was it the landing page?
52:02Was it the offer?
52:03>> Yeah.
52:03>> Was it just bad keyword selection? maybe
52:05this doesn't work anymore in the age of
52:07prompting instead of doing keyword
52:09searches, right? So those questions
52:10become much more important. And like you
52:13said, it's about thinking and the
52:15strategic partnership that you bring,
52:17not the uh the final tactical execution
52:19because that is the stuff that
52:22automation will do. So I think to sum my
52:26part of that up more succinctly and not
52:28wander off on 15 tangents again. I think
52:31the skill is asking better questions
52:33because as AIs get smarter and smarter
52:35and smarter, the answers get better and
52:38faster and cheaper. We need to ask
52:40better questions.
52:42>> Great.
52:44Well, um, fantastic stuff, Mike. This
52:47has been very helpful. Um, a lot of
52:49people will probably be running to their
52:51computers and starting to build some
52:53cool stuff. So, if you do, please put it
52:55in the comments. We'd love to see and
52:57comment on it.
52:58>> Uh, Mike, uh, aides from ads to AI and
53:018020 Brain, uh, I'm sure people can also
53:05connect with you on LinkedIn. What else
53:07would you like people to do?
53:09>> Let's just Yeah, go build. That's what I
53:11would love people to do to not be scared
53:14of it. In fact, I was talking to a lady
53:15yesterday. Uh, I shall leave names out
53:18of this, but we were talking about all
53:20of this stuff and she said, "Oh, but
53:22Google Ads scripts still scare me." I've
53:25got the solution for you. I actually
53:27called it my scary challenge. It's been
53:28up on the This is another website,
53:30another URL. Sorry, I'll have to link to
53:32things better, but 8020agent.com/scary
53:36and it's like a little 14minute
53:38challenge that you go through to make
53:40you realize that Google Adcripts aren't
53:42scary. It just starts to show you what
53:44is possible. And so it's the same thing
53:46that all this stuff that we're talking
53:48about. And yeah, Claude Code looks
53:50scary. And if you happen to be on a Mac
53:52and you happen to be already playing
53:53Claude 100 bucks a month, you now have
53:55access to Claude Co-work that came out a
53:56couple of days ago, which is giving you
53:58the power of all of the stuff that us
54:00geeks are playing with, but making it
54:02less intimidating. That I think will
54:04definitely trickle down. That will
54:06become more widely available. The key is
54:09just get over being scared of it.
54:11That'll only last for the first five or
54:1310 minutes. It looks intimidating. It
54:15was built by geeks who don't have the
54:17best UI sensibilities in the world. But
54:19don't worry about it. It's basically
54:21just a chatbot. And that won't be the
54:23final form factor of all this stuff. But
54:24start there and ask better questions and
54:27just try it. I think far too many people
54:30try it and go, "Yep, told you it was
54:33crap. This thing will never be as good
54:34as me at doing blah blah blah." And
54:36therefore, I will forget all of AI and
54:38put it away for another 12 months and I
54:41was right and I don't have to look at it
54:42again. It's like, but it's so much more
54:45than that. It can do so many different
54:47things. I don't want to start a whole
54:50new conversation about jagged
54:51intelligence and but it's just try it.
54:54Play with it. That would be my my plea.
54:56Don't worry about my stuff. Sure, come
54:57adbrain.com. That's the demo that shows
55:00you some of the stuff that that's
55:01possible and as to AI is the community.
55:05Um, but just play just just start. Open
55:09up the open up Claude and say, "Build me
55:12a working snake game." And watch it
55:15build on the right hand side. It'll pop
55:17open a little window on the right hand
55:18side. You may have to change a setting
55:20to do that. Just ask Claude, "How do I
55:22make sure this thing pops up on the
55:23right hand side and something called an
55:24artifact?" Like, "What's an artifact?
55:26How do I get that to show?" And just say
55:27like, "Show me a game." Ask your kids,
55:29"What's your favorite game?" Right?
55:30Let's build that. We built Asteroids
55:34two Christmases ago in half an hour. a
55:37fully working game of asteroids that my
55:39daughter absolutely loves. Can we play
55:41that game again, Dad? Like, don't forget
55:43the Google Ads bit. Build something fun,
55:46but just build. Just see what it can do.
55:48That would be my my play. Keep on
55:50building. Great advice. Keep on
55:52building. Okay, keep on building. Keep
55:54on watching. Subscribe to our podcast if
55:56you want to know about future episodes
55:58coming up. Make sure you connect with
56:00Mike Rhodess, our guest from today. And
56:03with that, thank you everyone for
56:05watching and we'll see you for the next
56:06one.

About this transcript

This page contains the full transcript of Mike Rhodes Shares Valuable Vibe Coding Lessons From Real Marketing Builds | PPC TH 121 by PPC Town Hall, generated from the public captions YouTube serves with the video. The transcript has 11,030 words across 1,535 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.