YT YouTube2Text

YouTube2Text → Transcripts → AI in PPC: Where It Helps, Where It Breaks, and Where It’s Risky | PPC TH 122

AI in PPC: Where It Helps, Where It Breaks, and Where It’s Risky | PPC TH 122 — Transcript

by PPC Town Hall · 10,176 words · 1,418 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.
0:05I'm your host. I'm also CEO and
0:07co-founder at Optimizer, a PPC
0:10management software. Now, for today's
0:12episode, we're going to go in a slightly
0:14different direction. Usually, you get to
0:17hear from Fred, the AI fanboy, and all
0:20of the other AI fans. But today, we have
0:23Julie Bachini Freriedman. She is the
0:26host of PPC chat and that is a community
0:28where more often than not there are some
0:31people talking about fears around AI, AI
0:34anxiety, worried about using AI the
0:36wrong way and worrying about what it
0:38will do to their careers in PPC. So I
0:41think this is going to be an interesting
0:42one. We're going to get a different
0:44perspective on artificial intelligence
0:46in the world of digital marketing. And
0:48with that, let's get rolling with this
0:50episode of PPC Town Hall.
0:56Julie, great to have you on again.
0:58Thanks for joining us.
0:59>> Oh, thanks so much for having me. I I
1:01appreciate the opportunity to talk about
1:03PPC whenever I'm asked. Really?
1:06>> Yeah, you you do it all the time,
1:07several times a week. Um, so you've been
1:10on the show before, but for people who
1:12haven't had the chance to to meet you,
1:14tell them a bit about yourself, PPC
1:16chat, Neptune Moon, and what you've been
1:18up to lately.
1:19>> Sure. I'm Julie Freriedman Bachini. I am
1:21the president of Neptune Moon. So I do
1:24PPC consulting and I have been working
1:27in PPC since literally the beginning. Uh
1:30so I have seen such a huge just amount
1:33of change over all these years. So I've
1:35been in business for myself coming up on
1:3826 years. So I have really seen just
1:41about everything. And uh in addition to
1:44that, I have been running the PPC chat
1:47community for PPC professionals since
1:502017, which I can't believe it's coming.
1:53It's been that many years. Like it feels
1:54like it I took it over not that long
1:55ago, but yeah, it's been that long.
1:57>> And we're a free community where PBC
2:01professionals can talk with each other,
2:03ask questions, get moral support. We do
2:06a weekly live chat Tuesdays at 12
2:09o'clock Eastern where we talk about a
2:10timely PPC topic. And uh AI of course
2:14has been if not the topic, it floats in
2:17I think to a lot of topics that we that
2:19we talk about too because it's very much
2:20on everybody's mind.
2:22>> Yeah, that makes sense. AI is everywhere
2:24these days.
2:25>> Um and so yeah, thanks for hosting that
2:27community and taking it over back in
2:292017. And uh you're also known as PPC
2:33mom to some uh one of the nicest people
2:35in the PPC industry. Someone like you
2:37said, it's about uh mental wellness.
2:40It's about sort of like airing what's on
2:42our minds and not just about PPC
2:44strategies but also uh living good
2:46lives. And uh you're also an actual mom,
2:50right? And uh is that part of the reason
2:52why your glasses match your background
2:54and the shirts?
2:56>> Yes, I have to be very I mean not that I
2:58don't like to be coordinated, but I do
2:59have a teenage daughter who's very into
3:01fashion. So if I am like a skew or a
3:03miss uh in appearances and she catches
3:06wind of it, I will I will hear about it.
3:07So, yes, she keeps me on my toes.
3:11>> We're uh for everyone who's just
3:12listening to the audio portion of this,
3:14I highly recommend you take a look at
3:15the YouTube version of this or just go
3:18to the show page to see uh how great
3:20Julie looks with the glasses that match
3:21the galaxy in the background.
3:23>> Well, I am Neptune Moon, you know.
3:26>> Ah, yes, of course. Do you have one of
3:29the Neptune swatch? The moon swatch?
3:32>> I do not. I do not.
3:35>> Maybe a gift idea. Let's see who sends
3:37Judy one of the Neptune watches. Okay,
3:40but let's jump into the topic at hand
3:43here. So, artificial intelligence and um
3:45first of all, it's interesting because
3:46you said even if you don't have a PPC
3:48chat about that topic, it sort of floats
3:50in. Um so, maybe before we go too deep
3:53on AI, what are the topics on people's
3:56minds these days when it comes to Google
3:58Ads and PPC?
4:01>> Um, I mean, I'd say they fall into a
4:03couple different categories. AI is
4:04definitely a big one. Um, and when it
4:06comes to AI, it's sort of like how to
4:08utilize it. Um, where does it make
4:11sense, where doesn't it make sense? How
4:13do we make sure that we still have a
4:15job, you know, five years from now?
4:16There's a lot of of concern uh about
4:19that, what's everything going to look
4:21like? You know, the pace of change. This
4:23has always been an industry. It's tech,
4:25right? So, nothing stands still. It's
4:27always been an industry where there's
4:28been a ton of change happening. But that
4:30pace has really picked up, I think, in
4:31the last couple of years. Um, so we talk
4:35a lot about things changing. You know,
4:36you're have things working in your
4:38account and then
4:40one day you wake up two months later and
4:42it's not working anymore. So there are a
4:44lot of conversations around, hey, is
4:47anybody else seeing this? You know, we
4:48had X, Y, and Z and it was going along
4:50perfectly and now it's not. We have a
4:52lot of conversations about those types
4:53of things. How to set things up now, how
4:56to manage things. Um
5:00we talk a lot about what goes on in
5:01different platforms and how can you have
5:03efficiency across multiple platforms
5:06because many people are are managing
5:07accounts on just more than just one uh
5:09one platform as well. So where to spend
5:12your time, how to figure out how to
5:13prioritize, you know, where clients
5:15should be spending their efforts that
5:16that type of thing.
5:18>> And it's all getting more expensive. So
5:19how do we keep clients happy when the
5:23cost the click the cost of the clicks
5:25keep, you know, going up year after
5:26year? So they have to spend more money
5:29to really get the results that they're
5:30used to to getting like in the previous
5:32year. How do how do you manage that? How
5:34do you manage those expectations and how
5:35do you handle that and try to keep
5:36things efficient? I would say those are
5:38some topics we've talked about lately.
5:41>> Okay. So if that's on anyone's minds,
5:43know you're not alone in being
5:44frustrated with the ever rising CPCs. Uh
5:48but let's go back to the thing you said
5:49right before that maybe about the
5:51crossplatform management. And I think
5:54last year there was some serious concern
5:56over the future of Google in the AI
5:59world where Chad GPT was answering
6:01everything and Google's AI wasn't that
6:03good. But now Gemini is a leader. Uh
6:06Gemini is actually really useful because
6:08it ties directly into all of this
6:10amazing data that Google already has
6:11about us. So
6:13>> I don't have to do a Gmail connector. I
6:14don't have to connect it to calendar. I
6:16don't have to connect it to drive. It's
6:17already all in that Google universe. Um,
6:20so what what's your take on Google Ads
6:23as a platform and its viability? Should
6:25should people continue to feel safe in
6:28investing there or or is it really time
6:29to maybe look at some other places too?
6:32>> I mean, I think it's always wise to not
6:35have all of your eggs in one basket
6:37because as we've we know we are at the
6:40whim of any platform where we have where
6:42we have advertising going on. Um, but do
6:45I think that Google as a top destination
6:48for
6:50paid ads is gonna go away anytime soon?
6:52I I don't. Um, they have they've spent
6:5620 odd years teaching people
6:59how to find stuff, right? And a lot of
7:01that results in ad showing when people
7:04are are finding things. And of course
7:06they've expanded into and they've got
7:08the newer campaign types where you have
7:11um more im image based and video based
7:13assets. So you've got PMAX, you've got
7:16um demand genen, you know those types of
7:18things. So Google has been evolving
7:22and you know they just came out here uh
7:24very recently talking about how you can
7:27connect all your stuff you know into
7:29Google's AI now and you can have an even
7:31more personalized experience. So, I I
7:33don't think Google is um interested in
7:37falling behind really when it comes to
7:40the other, right? Like, they're going to
7:41pro I believe they're going to protect
7:43their turf and like why wouldn't they,
7:44right? They have a very expensive turf
7:46that they spent a lot of years building.
7:48So, I think it's going to take a lot for
7:51them to be pushed off the mountain. Now,
7:53that doesn't mean that there aren't
7:55other places. Open AI is is serious now
7:58and are announcing and talking about
8:02we're going to have ads. We are like we
8:03really truly are going to start having
8:05ads. So like things are moving forward
8:07in on other platforms as well, but
8:09Google has such an advantage at this
8:11juncture. It's going to take a lot to
8:13dethrone to dethrone them and make them
8:15a place where it's not going to be a
8:17priority for businesses and
8:19organizations who want to be, you know,
8:21seen.
8:22>> Yeah. And and I always find it quite
8:24interesting when none of these companies
8:27truly go into ads and I'm talking about
8:29the newer ones, right? Or or they make a
8:32a 360. Well, 360 is actually going back
8:34to where you were going before. So, they
8:36make a 180 on uh on ads like Perplexity.
8:39Uh but it's quite interesting because if
8:41they think about like what's the biggest
8:42monetization potential of this massive
8:45AI that they're building, sometimes they
8:47don't see ads as the thing. And that's
8:49perhaps a little bit surprising to us
8:51knowing how big and massive of an
8:53industry this is. Um but uh but yeah, at
8:56some point they certainly have to it
8:58just makes sense to support advertisers
9:00in all of this. So I can't wait to see
9:03what's to come in that. Now one uh type
9:07or campaign type thing that you didn't
9:09really mention was AI Max. Is that for a
9:12reason? Do you not see a lot of
9:13popularity on that one or what are your
9:15thoughts around AI Max? Um, I would say
9:18generally in the community people are
9:21experimenting with I mean it's still
9:23relatively new, right? It hasn't been
9:24around for that long. So people are
9:27experimenting with AIMAX. Um, and I
9:30would say it's like any other any other
9:32like Google product launch. Some people
9:36start with it and it works awesome and
9:38they love it and they're like, "Wow, I
9:40can't believe we didn't have this
9:41before." Right? And then you have other
9:42people who are like, "H, I tried it. It
9:44was very meh." And then you have other
9:45people who are like, "Oh my god, it was
9:46a disaster." So it's typical. I would
9:50say it feels a lot like when PMAX came
9:52out, right? Like same thing where some
9:54people are early adopters, some people
9:56and it like PMAX, I would say it
9:59probably performs better under some
10:01circumstances versus others, right? And
10:04so there's a pretty wide gamut, I think,
10:07in the experience that people have had
10:08when they've tested it and they've run
10:10it. So it's like everything else, your
10:13mileage may vary. So, if it feels like
10:16something that you want to try or you
10:17think could could potentially work in
10:19your account, like, you know, give it
10:21give it a whirl. See what happens with
10:22your particular um you know, your
10:25particular setup. And, you know, you'll
10:28know, I think, fairly quickly if it's
10:30going to be something that's going to
10:32work for you. And I'm sure it's going to
10:33evolve because like the Pmax that we
10:35have today is not the PMAX that we had
10:37when that baby rolled out, right? It's
10:39not.
10:40>> Not at all.
10:41>> Right. So I feel like a AMAX will
10:43probably be the same way. Like they
10:44rolled it out like here's what it is,
10:46right? And then it's going to morph
10:48over, you know, over time. I would I
10:50would expect. So we're still in that
10:52first that first release period really.
10:56>> And that makes sense. And so that's
10:57hopeful news then for the the people who
10:59are managing PPC accounts because
11:01despite all of the automation
11:03capabilities and the AI, it still sounds
11:04like it's not a guaranteed win. Um, now
11:08having been part of so many of the PPC
11:10chats and obviously you moderate the
11:12vast majority except the one or two
11:14weeks when you go on vacation and you
11:15have a guest host.
11:17>> Uh, but what what is sort of are there
11:19any patterns that you've seen around AI
11:22max and when it does work better than in
11:24other scenarios? Have you picked up on
11:26any of those?
11:27>> Um, I mean I would say generally it's
11:30more it things are harder for lead genen
11:32and I would say that's true for Google
11:34as a whole. So, if you're lead genen
11:37things like you're not the target market
11:39for Google advertising, like you're just
11:42not. And it's okay. It is what it is.
11:44>> Wait, did I just did Did I just hear you
11:47say if you're Legion, you're not the
11:48target audience for Google Ads.
11:50>> Yeah, I said it. I said what I said.
11:55>> Heard it here. Uh well, that's amazing.
11:58I mean, so uh go a little bit deeper on
12:00that.
12:02I mean,
12:04okay, as a lead genen PPCer, I will say
12:08that um most of what Google comes out
12:10with is ecom first, right? It doesn't
12:13mean that you can't use it for lead
12:16genen. You you can we do, right? But the
12:20most of what they come out with is like
12:22it's led by lead it's led by ecom. It's
12:25designed for ecom. It's designed for the
12:27kind of volume that ecom brings. It's
12:29just like it just it just is. Um, and
12:32I've made peace with that because it's
12:34been this way forever. Um, and every
12:37year at GML, I get my hopes up. I have a
12:39square in my bingo card every year of
12:40like we're going to get serious about
12:42lead genen, right? And every year that
12:44does not get crossed off. And like it is
12:45what it is, right? Um, so I think that
12:49again when you look at like how I would
12:51say the same is still true for PMAX.
12:53Like the people who find the most
12:54success with PMAX are probably ecom
12:57versus I'm not saying that people aren't
12:59finding creative ways to to have it be
13:01successful in lead genen. There are
13:02there are plenty of people who are doing
13:04that, but it requires a higher level of
13:06creativity
13:08to make some of these things work.
13:10They're there just designed for the
13:13cadence and the rhythm and the way that
13:16ecom works, which is different from how
13:18lead genen works, right? It just it is
13:21different animals. Do you think it's
13:23about the richness of the signal where
13:25lead genen tends to just be lower volume
13:27in general or a longer sales cycles?
13:30So would you is that kind of a fair um
13:34point? Would it then be okay to say like
13:36maybe a lower volume e-commerce client
13:39is going to struggle in many of the same
13:41ways as Legion? I I would say probably
13:44because I do think the biggest factors
13:46that make it more challenging let's say
13:49to find the type of efficiency that high
13:52volume ecom accounts can find for for
13:54lead genen folks or for low low lower
13:56lower volume folks is that the internal
14:00machine learning AI whatever whatever
14:02you prefer to call it what that the
14:04mechanism that drives all of the
14:06automation inside of the platform it
14:09requires a certain baseline critical
14:11mass of of data Right? And Google's
14:13honest about it, right? Um, but if
14:16you're if you struggle to consistently
14:20like meet or surpass those thresholds,
14:23it is more difficult for the um the
14:27internal automation and AI to really
14:29find a groove and find a momentum in the
14:32same way it can when it has more to work
14:34with.
14:35>> And if you think about it, it makes
14:36sense, right? Like the way that the
14:37automation works is the more information
14:40it has,
14:42the better it is at being able to figure
14:45stuff out and figure it out faster and
14:47do it more efficiently. So, if it has,
14:50you know, 500 or more conversions that
14:55are happening in a 30-day period, right?
14:56If it has 5,000 that are happening,
14:59that's so many more data points for it
15:00to look at to find patterns, to find
15:03commonalities
15:04than if you're an account where you have
15:0730 conversions, you have 52 conversions,
15:10right? Like that's just math and it's
15:12data going into, you know, that's being
15:14put into the mechanism that's like doing
15:16all the things that it's doing behind
15:17the scenes. So, like it it kind of is
15:19what it is. You know, I'm not trying to
15:21throw shade in the way that maybe it
15:23seems like I am. I'm just being
15:24realistic about what it's like when you
15:26work on accounts that aren't the primary
15:29use case because we don't have massive
15:32amounts of data, you know, that that are
15:33flowing into the algorithm.
15:35>> Well, and so let's stop forcing the
15:37square peg in the round hole here is
15:39maybe what I'm hearing. So, uh where's
15:41the round hole or where's the round peg
15:43for legion advertisers? Where are you
15:45investing now to get those leads for
15:47Neptune Moon?
15:49>> I mean, you still have to be on you
15:50still have to be on Google. It's just
15:51it's harder. I mean, I think there are
15:53some options that can help you to
15:57I don't want to say like trick the back
15:59end, but it's they like portfolio
16:01bidding kind of like
16:05>> you kind of have to you have to be a
16:07little bit more creative. Like you have
16:08to you have to build things in a way
16:12where you're trying to lessen your dis
16:14your data disadvantage. I would say
16:16would be kind of the way I would if I
16:18had to put like a bigger topic over it.
16:21Knowing that the volume of conversions
16:23impacts everything that's able to happen
16:25inside of the of the platform itself,
16:29you have to be a bit more creative. If
16:30you're lower volume to try to get to
16:33that critical mass so that those
16:35efficiencies can start to happen so that
16:37that pattern recognition can happen
16:39inside of the automation for you in the
16:43same way that it would for an account
16:44that has like gobs of data coming in,
16:47right? You just have to you have to
16:49think about it a little bit a little bit
16:50more maybe in how you structure things
16:53you know and that that type of thing
16:55>> right and okay so that's the important
16:56nuance I think um when you made the
16:59initial point that I didn't pick up on
17:00is yes Google ads or Google search is a
17:04place where people are looking for lead
17:06genen sol or what lead genen companies
17:08sell
17:09>> however you're saying the Google ads
17:11platform isn't designed first to support
17:14those advertisers so you have to find
17:16these workarounds to make it work for
17:18you because the audience is there. It's
17:19just harder to connect with the right
17:21audience than it was if you were an ecom
17:23advertiser.
17:25>> And it's harder than it was five years
17:27ago. Um the automation, all of the
17:30automation, you know, which is sort of
17:32like under the AI umbrella again, it's
17:35it's designed for the high data volume
17:38accounts. Like that's how it's built.
17:40That's what it's looking for. So we had
17:43more options of things that we could do
17:46and we weren't our bidding and like all
17:50the things that happened in the
17:51background half of those things didn't
17:53happen in the background in the same way
17:54that they do now.
17:55>> Right. So, I think the challenge then is
17:57you're saying, okay, we could still kind
17:58of go back to the old ways of manually
18:00adding keywords, manually managing bids,
18:03but there's all this automation that's
18:04come in and and kind of derails. And
18:06that was also your earlier point, right?
18:08Things might be running fine now, but
18:10then all of a sudden, Google does an
18:11algorithm change
18:12>> or they launch a new feature in Google
18:14Ads and that messes with that perfect uh
18:16workaround setup that you had and now
18:18you have to figure out what's the new
18:20workaround to to continue making it
18:22work,
18:23>> right? And I think you also have to
18:24think about like how long can you do
18:26workarounds, right? Like I think that's
18:28the other piece where like the people
18:30who are in the lead genen space talk
18:32about and think about like all right you
18:35know it's not realistic to say like well
18:39you know I'm I'm manual bid forever
18:41right like in some ways you feel like
18:43well that would solve all the problems
18:45that you're facing with the automated
18:48bidding needing the the higher levels of
18:51data. You're like, "Well, just just do
18:53manual bidding then, right?" It's like,
18:54"That seems like that should be a simple
18:56solution."
18:58When you do the manual bidding, can you
19:00do that? Yes, you you can do that. And
19:02you may find that you have a perfectly
19:05accept acceptable level of success as
19:07far as like the advertiser is concerned.
19:09That's absolutely a possibility. But
19:11what happens now is when you don't use
19:16an automated or you know smart bidding,
19:18automated bidding, it's called been
19:19called 19 different things. But like
19:21when you're not taking advantage of one
19:22of those technologies within the
19:24platform,
19:26then you also don't get access to a lot
19:28of the goodies that come with that,
19:30right, that Google's doing in the
19:31background for you. So I think, you
19:34know, you can try to hold on to like,
19:37oh, I'll just do it the old way.
19:40I feel like the days are numbered for
19:42that. I feel like you could make an
19:43argument that it's already over for
19:45that. Although there could be pockets of
19:47people who are like, I got tech I got
19:48extended text ads that are still running
19:50and doing great, right? Like that there
19:53are people out there who have ETAs who
19:55are still going like gang busters,
19:57right? Like they exist.
20:00So, but does that mean that like that's
20:02how you should think about your account
20:04now going forward? Like probably not,
20:06right? Like so I think that
20:09>> well you'd have to invent time travel
20:11for that, right? Right.
20:14But
20:14>> you pause them and you can never unpause
20:16them again. So
20:17>> it's all true.
20:18>> So that's not realistic, right? And that
20:20hasn't been realistic for a while. But
20:21like when we were in that transitionary
20:23period when you you know, so it you have
20:26to you have to be able to
20:29absorb what's happening and
20:33see what you're experiencing in the
20:35accounts that you're responsible for.
20:37And then you have to try to make
20:38responsible decisions as far as what
20:41things can we try. Um I think it's nice
20:44having you know like programs like this
20:46and you know the PBC chat community
20:48where people can kind of bounce ideas
20:49off of each other in like a a safe way
20:52too to say like hey has anybody had any
20:54success with X Y or Z. get a lot of
20:56questions posed in the community like
20:58not even necessarily during the chat
21:00time but like you know whenever people
21:02will pose questions of hey um you know
21:04has anyone had success with you know
21:06like this scenario running these types
21:08of ads or running this type of campaign
21:10or what's happen is has anyone like I
21:11haven't run you know I've never run the
21:14the demand genen um what do I need to
21:16know if I'm going to do that right so we
21:18have a lot of conversations that that go
21:20on like that where people are trying to
21:21adapt and they are trying to take
21:24advantage of the uh the technology that
21:27exists even if it can feel like you know
21:32maybe it's a little bit challenging it's
21:35a little bit scary it's a little bit
21:36like a I don't really know what I'm
21:37supposed to do with this you know in
21:39this account but I think people have a
21:40hunger to want to try nobody wants to
21:43get left behind nobody wants to feel
21:45like they wake up one day and suddenly
21:46they're you know a dinosaur who doesn't
21:48speak the language of anything that's
21:49going on like nobody wants that
21:52>> and what's that language of the what's
21:54the language language.
21:55>> Obviously, we're talking about AI here,
21:58>> but but what what are sort of the
22:00>> It almost leads into a question of what
22:02should people learn today to remain
22:04relevant in the future.
22:07>> Okay. So, I'm not the hugest fan of AI.
22:09I've been vocal about that. Like, that
22:10shouldn't be a surprise to anybody who's
22:12followed me anywhere. Um
22:15because I have questions about it and I
22:17think it has shortcomings. But I also
22:20realized that like that train has left
22:21the station.
22:24So
22:25my mindset and the mindset that I try to
22:27encourage people to have is that you
22:30know everything changes in in this
22:33industry and we are affected by changes
22:35that are outside of our industry. So
22:37like we have certain things that are
22:39happening in AI that are directly in the
22:42day-to-day stuff of what we do, right?
22:44Like Google has their assistant inside
22:48of Google Ads now, right? where like
22:50theoretically you can just have a little
22:52conversation with the assistant, you
22:53know, and kind of tell it what you're
22:55looking to do and it's going to like
22:57spoon feed you everything that you need
22:58to you need to do. Like that's new,
23:01right? Like that didn't exist before.
23:04Um, you have you can ask it questions
23:06like you would ask a person. You could
23:09we didn't have that before. Like, hey,
23:11why are my you know why are my CPC's up?
23:13Why are my conversions down? You can you
23:15can ask it now, right? You could do that
23:20two years ago. You had to figure it out.
23:21I mean, and I'm not suggesting that you
23:24only ask the Google Assistant why
23:27something's happening. Um, because
23:31you should probably verify what the
23:34Google Assistant is telling you. Um, but
23:37like that tech didn't exist before. So,
23:40how do we how do we manage all that? How
23:42do we
23:44how do we take the knowledge that we
23:46have for however long we've been working
23:48in PPC, whether you've been doing it for
23:50a year, whether you've been doing it for
23:515 years, 10 years, 20 plus years, how do
23:54you take the knowledge that you have of
23:55how all these things work and sort of
23:58what is needed to create successful
24:01campaigns
24:02and then how do you integrate that with
24:05the AI that seems to be everywhere? like
24:08figuring out which pieces are actually
24:10helpful. Like, oh, this is like a super
24:12timesaver. I could automate this, this,
24:14or this, right? I can have a script that
24:17does this. I can create a little um, you
24:20know, bot that does something for me
24:22that saves me time. I don't have to do
24:24it manually anymore. I can just skip to
24:26the part where it really requires my
24:28brain. Like, these are all the things
24:29that I think people are really chewing
24:31on right now,
24:32>> trying to figure out like
24:34>> where can they use AI in a reliable way.
24:37so that they're not potentially
24:40having AI generate something that turns
24:43out to be incorrect and then you're
24:45moving forward using incorrect data. Um,
24:47that's a real fear. But
24:51where does it fit in? Like where does it
24:52make sense? How can it make you be more
24:54efficient? Where can it fill in gaps
24:56that maybe you know you had previously
24:59that it it actually does like does a
25:01nice job with? And people are
25:03experimenting like crazy and they share,
25:06you know, they share their thoughts of
25:07like, oh, I created this or oh, these
25:08are the props I use for that or
25:10whatever. So, there's a lot of there's a
25:12lot of like testing and experimentation
25:14going on right now in general.
25:17>> Yeah. And I'd love to hear some of your
25:19favorite examples of what you've seen
25:20people do. But but before that, I think
25:22it also kind of breaks down into there's
25:24automation that comes out of AI where it
25:27builds scripts or it just does the thing
25:28you want it to do. And then there's the
25:30AI as the the thinking partner, the
25:33ideation platform, maybe pushing you to,
25:37you know, thoughts that you hadn't had
25:39in the past, right? If you're if you're
25:41going to be writing some ads, have you
25:44put yourselves in the shoes of every
25:46different persona that might be
25:47consuming those ads? Right? And and then
25:49that was really difficult because I
25:50can't put in the myself in the shoes of
25:54um an 18year-old black woman. um we
25:57don't have a lot of common experiences.
25:58So it's difficult for me to write ads
26:00towards that audience but AI has
26:04absorbed so much information that it is
26:06in some cases able to do that and so um
26:10how do you see that breaking down into
26:11in terms of the ways that people use it?
26:13Do they do they like it as a thinking
26:16partner? Do they like it as an
26:17automation partner? Or is it one or the
26:20other?
26:21>> It's kind of split, I think. Like some
26:24people love it and are using it for like
26:29everything, right? They're using it for
26:31strategy. They're using it for ad copy.
26:33They're using it to generate images.
26:35Like you have people who are in the like
26:38what else can I have it do camp, right?
26:41And then you have other people who are
26:43like, I don't want it to do anything
26:45because I don't trust it at all. Like
26:46those are the two like far ends of the
26:49spectrum. And I would say you have a lot
26:51of people who are sort of in in between
26:54who
26:55want to figure out ways that it can help
26:58but also not open themselves up to to
27:01risk if the AI is wrong. So I think um
27:07finding ways like one of Go ahead. Well,
27:09I was going to say talk a little bit
27:10about risk because I know you have a
27:12great example from your own business on
27:15ways in which risk plays out that I
27:17hadn't necessarily thought about, but
27:18where it goes pretty deep into uh
27:20talking to your insurance company,
27:23right?
27:24>> Yeah. So, I'm I mean I'm a big person
27:26who I I talk a lot about like what I
27:28call the business of PPC. So like
27:31obviously we have all the stuff we do
27:32where we take care of the the brands
27:34that we work on and in the accounts and
27:35all that but I also spend a lot of time
27:37thinking about like as if as you're a
27:40practitioner whether you're an agency
27:41whether you're a consultant whether
27:42you're a freelancer
27:45we're providing service to our our
27:48clients um and
27:52it is smart to carry various types of
27:54business insurance so if you're not I
27:57would think about doing that um but even
27:59if you are this technology moves faster.
28:02It moves faster than legislation. It
28:04moves faster than your insurance policy,
28:06right? It just it moves fast. And so
28:09I've been having some conversations with
28:11my insurance agent, which um they're
28:16investigating answers to these questions
28:19because they're like, "Well, that's a
28:20really good question." You know, so
28:22you're asking like, "Okay, for example,
28:23what if a client uses AI to dig into all
28:26of their internal data, right? and AI
28:29puts together,
28:31let's say, like a summary of this is
28:34everybody who bought what we're selling
28:36in the last year and these are their
28:37personas and like they they put this
28:39whole massive thing together, right? And
28:41they they give it to you as their PPC
28:43strategist and they're like, "Hey, this
28:46came from all of our data. We want you
28:47to use this." And you're like, "Okay,
28:49great."
28:50And then you have this data and you
28:52they're instructing you to utilize it
28:55when you are coming up with strategy
28:57when you're making decisions when you're
29:00prioritizing you know all the stuff that
29:01we do what happens if that data that got
29:05spit out of whatever AI that they use
29:06like is wrong
29:09right where does that liability lie is
29:12that solely with the client because they
29:14gave you bad information like I don't
29:16know these are qu these are you know
29:17these are sort of like existential
29:19questions questions that I'm pondering
29:21as
29:21>> well and that's an interesting
29:24>> I just signed you know our new business
29:26insurance documents and honestly for the
29:28first time in 10 years of running this
29:30business I think I understand what I
29:33signed because I took the PDF and I gave
29:36it to Chad GPT and I asked it does
29:38anything look weird in here like what
29:40should I be concerned about um in the
29:42past I would have just been like this
29:44150 pages of probably boilerplate stuff
29:47so here's my signature here's my payment
29:49let's move on to the next item of
29:51business, right? But uh but but I find
29:53it interesting where AI just opens up
29:55these opportunities and even like
29:56healthcare insurance um you know when
29:59when picking a health plan, put in what
30:01your plan has and like the medical
30:03billing codes and now you can ask your
30:06custom GPT a question. It's like hey
30:08I've got this this thing and like do you
30:10think that's going to be covered? And so
30:12it looks at my insurance it's like yeah
30:13that looks like a $60 co-pay but
30:15otherwise covered. Um, and again, these
30:17are things where in the past I would
30:18have just been like, "Yeah, I go to the
30:20doctor and uh whatever bill comes out
30:23like that it is what it is."
30:25>> Oh, see, I have a spreadsheet. I just I
30:27went through I did I I made a
30:28spreadsheet with all the options and all
30:30of the costs from the previous year and
30:32I blended them together to figure that
30:33out. But that's me.
30:35>> Did Did you make it or did you have an
30:37AI help?
30:37>> No, I made it.
30:38>> No, I didn't.
30:42>> So, is this a service you sell if anyone
30:44wants
30:44>> No. Oh my god, no. It was migraine
30:46inducing. No,
30:47>> that is not true.
30:49>> Um, okay. But that so that that's
30:51fascinating, right? How we have to think
30:53about the AI risk at these multiple
30:55levels and then how it comes down to
30:57insurance.
30:58>> Um, and it also then goes into one of
31:00the big roles that humans play is the
31:01the verification. Okay. So, we get this
31:03data that's generated by AI. Can we
31:05verify that it's grounded in the reality
31:07of the business? To what degree do you
31:10think agencies need to evolve into being
31:12more of a business partner rather than
31:13just being the ones that get that dock
31:15and then do something with it, but have
31:17no insight beyond, you know, where did
31:19that data come from? Was it correct? Um,
31:22was it the most exhaustive data? Should
31:24have have been more. Where do you fall
31:27on that? I think we're going to have to
31:30do more of that kind of work because
31:35um
31:37someone has to double check, right? We
31:40where we are in the evolution of of AI
31:42like somebody needs to double check. I I
31:44had a conversation with um another PPC
31:46pro and we were talking about the fact
31:48of like if you're working with a client
31:51who expresses an interest in having a
31:52lot of AI stuff like be part of the
31:55process, are you building into your
31:58pricing and building into your time the
31:59fact that like double-checking is going
32:01to have to happen, right? Like someone
32:04should responsibly double check stuff.
32:06Who is that? Is that somebody on the
32:08client's end? Is that like we get it?
32:10Like I feel like there's we're at a very
32:14interesting time where the way things
32:16have been done previously like they
32:18would send you a spreadsheet like oh we
32:20exported this from our CRM or oh you can
32:22we're going to give you a log into the
32:23CRM and you can pull whatever data you
32:25want to pull right like that's how
32:26things have been done in the past a lot
32:28or they just send you a spreadsheet an
32:30Excel sheet that has you know customer
32:34data or like summaries of X Y or Z like
32:36whatever you ask for. Um, now it's just
32:40as likely that what you would get from a
32:42client would be that information
32:45shoved into their AI of choice. Um, and
32:49then it would spit back
32:51>> either, you know, some type of
32:52spreadsheet or a narrative or a
32:55description or a summary of like, oh
32:57yes, our typical customer profile is,
32:59you know, x, y, and z.
33:01>> Yeah.
33:02>> How do you know that's how do you know
33:03that's right? I guess is my, you know,
33:05like I, and this is me, like I I look
33:07for places where there's going to be a
33:08problem because I want to make sure
33:10there's not a problem so that when we're
33:11doing the stuff that we're doing, we
33:12don't find out later it was based on
33:14faulty information. And that can come
33:15not from AI, too, to be clear. People
33:18can give you bad just as just as much as
33:21AI can.
33:23>> Well, and that's where your example is
33:25interesting because it's like this is
33:26our ideal customer profile. I I think
33:29some assumptions go into that and
33:31whether it's generated by a human or an
33:32AI. we don't really know every one of
33:35our customers, right? So, we're drawing
33:36some parallels. We're making some
33:39average decisions. And as we all know,
33:41averages suck because nobody is average.
33:44>> Um, right? But, but that's a very
33:46different sort of um misrepresentation
33:49in a way from well, I got the Google ads
33:53numbers and it had a thousand clicks and
33:55100 conversions. So the AI says, "Oh,
33:59that was a 5% conversion rate," which is
34:01incorrect because it did math
34:02incorrectly and it should have been a
34:0410% conversion rate. And so when I think
34:07about that, having written scripts for
34:08so many years, like one of the the
34:10things about a script is you can put in
34:11a script and you can say, "Okay, here's
34:13um here's a thing you need to do and
34:15then something comes out of it at the
34:18bottom and you can just take that
34:20spreadsheet with data and just trust
34:22that it's correct. But what if you
34:24pulled the incorrect field and maybe
34:26instead of pulling conversions, you
34:28pulled all conversions and so you're
34:29basing it on something that's not in
34:31line with what the business wanted. Um,
34:33and so the way that we I've dealt with
34:35that in scripts was as the tool goes
34:38through its logic, um, it spits out in
34:41the debug, okay, here's what I'm doing
34:43and here's a sample piece of data. And
34:45then I can go in and I can say, okay,
34:47let me verify not a thousand pieces of
34:49data, but let me like grab a few of
34:51them. And if they seem correct, then the
34:53assumption in deterministic code is that
34:55it's probably made some logically sound
34:59decisions that we can trust a little bit
35:02more. Um, right? And so, but but then
35:05the question becomes, how do you do
35:06these types of things when it comes to
35:09an AI? And for me, the way that I see
35:12that is
35:14ask the AI to build a piece of code for
35:16you. Ask it to build a piece of
35:18software. And so Gemini is doing more of
35:20this. Claude with their artifacts can do
35:22it. But basically, don't just answer it,
35:24but put it in a table. Maybe give me
35:26some filter so that I can say, okay,
35:28well, if you selected keywords based on
35:30minimum of five conversions, give me a
35:32filter in the UI where I can go and set
35:34that I want minimum of 10 conversions.
35:37Um and that builds the confidence
35:40through a UI that it is doing what I
35:42expected it to do. But but I do think
35:45there is a future where that uh the
35:48crutch of the UI goes away because we
35:51start seeing the the AI is consistently
35:54doing a good job. we we can start to
35:55trust it more and more and these
35:57scratches that are the UI go away and
35:59eventually we can just tell it, hey, can
36:01you find me high performing search terms
36:03that we should consider adding and it
36:05just knows what that means and it does
36:07it and then you can question it. You can
36:08say like how what was the math behind it
36:11and it'll explain it but you don't need
36:13to see everything anymore.
36:15>> I think that's true and I think one of
36:17the other things that you can do is you
36:20can ask it questions you already know
36:21the answer to.
36:23So in other words, what we're talking
36:25about here is like how are we using AI
36:26like going forward like something that a
36:28client wants from you or you know your
36:30stakeholders want from you a week from
36:32now, two weeks from now, a month from
36:33now, right? That's a different animal.
36:36If you want to get a feel for how
36:40accurate is a particular AI under
36:42certain conditions, you can ask it
36:45questions that you already know the
36:46answers to
36:48>> and see how correct it is, right? Like I
36:52feel like that's a decent way at where
36:54we are right now to be able to develop I
36:59guess I would call it like your
37:00confidence level in a tool that you
37:02might be that you might be wanting to
37:04wanting to use. So, in other words,
37:06instead of giving it something that
37:08you're still trying to figure out
37:09yourself and you you don't have all the
37:12information or you're you're working on
37:14figuring it out,
37:16why don't you ask it things you already
37:18know the answer to? And then you can
37:21see,
37:22you know, does it come up with the
37:24correct answer? Does it like that's a
37:27that's a better way in my mind to be
37:30putting yourself in a position to have a
37:32higher confidence level when you're
37:34asking it a question that you don't
37:36already know the answer to. And so that
37:38is one way that you might want to spend
37:40a little bit of time like pull some
37:42historical stuff, pull it from periods
37:44where like you know what the answer is
37:45on the performance. You know, like you
37:47already know that stuff. You've already
37:48done that work. Um, that's another way
37:50that you can work towards validating
37:53which technologies, what kind of
37:56prompts, like what gets you where you
37:58need to go where it's spitting out
37:59something that you you feel confident
38:01enough that like yes, I would use this
38:03to move forward on the things that I'm
38:04supposed to do, you know, going going
38:08forward from here. So that's another way
38:10that you can work with AI where it's
38:12like maybe it's not doing something for
38:14you in your immediate like deliverable
38:17workflow, but you're building you're
38:20building a way for you to feel confident
38:23about using it in in particular
38:25situations because you feel like you've
38:27you've vetted it almost, right? Like I
38:29think that's something that doesn't get
38:30talked about a lot, but it probably
38:32should. Yeah, that's a really good piece
38:34of advice in terms of how you should use
38:36AI to build confidence and make sure it
38:38works the right way.
38:40>> Um, and even though you're an AI skeptic
38:42in many ways, like what other tips do
38:44you have that people could put in their
38:47workflows to get the most out of AI when
38:49they sort of have no choice but to use
38:50it?
38:52>> Um, I mean, I think
38:54sometimes you don't have a choice.
38:56Sometimes you're it's dictated to you.
38:58This is what you're going to do. We've
39:00all been there. And I think you just
39:02want to document what you're what you're
39:05doing. Um especially if it's something
39:07that you haven't done before that that
39:09you know the expectation is that you're
39:10going to be integrating different types
39:12of AI at different points of what you're
39:14what you're doing. I think you want to
39:16be fairly meticulous in your
39:18documentation of like this is what I
39:20did, this is how I did it. Um because
39:25sometimes you can ask an AI the same
39:28question twice and it'll give you
39:30different answers. So I think you want
39:33to you want to be meticulous in
39:35documenting and it's a pain, right?
39:37Everybody who wants to have to document
39:39everything that you're doing. Like the
39:40whole point of of the benefit of using
39:43AI is to save you time to do things that
39:46like oh I don't have to do that anymore,
39:48right? I have this tool that reliably
39:51will spit out what what I need if I give
39:53it if I feed the right information into
39:55it. Um, but while you're developing
39:57that, while you're being asked to use it
39:59in ways that you haven't used it before,
40:01I me um I would document everything that
40:06I did. Like keep your like essentially
40:08like your own change log almost of like
40:11>> here's what I did. You know, this is the
40:14prompt that I put in. This is the data
40:16that it had access to. This is the
40:17instruction that I was given. This is
40:19what the results were. Like you just
40:21want to you want to track that because
40:25you want that historical information for
40:27yourself too so that again you can find
40:30your own pattern recognition, right?
40:32Like wait a minute, I asked it to do
40:34something very similar like three or
40:35four times and the results I got were
40:38very different. Like two of them were
40:40super similar and then the other two
40:42were like very very different but I
40:44followed the same methodology while I
40:45was doing it. Like that would make me be
40:47like hm what is going on there? Um, if
40:50you're not documenting it, you probably
40:52wouldn't necessarily know that because
40:54your your
40:55>> your focus is elsewhere. Your focus is
40:58on like, oh, I asked it to do X, Y, and
41:00Z, and it gave me this, and now I'm
41:01going to go on and do task A, B, and C.
41:04Right? For me,
41:07I like to document stuff. Just it's just
41:10a habit I have of just like covering
41:13yourself. if there are questions that
41:15come up as to well why did you do this
41:17or what happened here or what was the
41:19sequence of things that you did like I
41:21just I think that we're in a place where
41:23all of this is so new and it's changing
41:26so rapidly that that would be my biggest
41:28piece of advice is to just create a
41:29documentation system and it doesn't have
41:31to be anything crazy you can have a
41:33Google sheet where you like note you
41:35know this is what I was asked this is
41:37how I you know created the prompt this
41:39is the platform that I used this is what
41:41I fed it and this is what this is what
41:43it's spat out like you know just keep a
41:45record of that for for yourself.
41:48>> Yeah. It makes me think of the uh my
41:49analogies of like the the human role in
41:52the PPC automated world like the PPC
41:54doctor. I
41:55>> mean nowadays anytime you go to a doctor
41:56they're recording right? They got either
41:58a human assistant or they're recording
42:00device and everything's being
42:01transcribed.
42:03>> Why not have something like that running
42:04on your computer? And every time you you
42:06get something in or out you say, "Hey,
42:08grab a screenshot." Do you talk through
42:11here's what I think about this or here
42:12was my thought process going through it
42:15and then at the end you have this big
42:16transcript and and the AI actually does
42:19a good job of summarize that transcript
42:22in the cases where you give it very
42:24specific information it doesn't
42:25hallucinate quite as often
42:28>> plus you still have the recording right
42:29so if you go back to the summarization
42:31it's like I don't think that's what I
42:32said well go and listen to the recording
42:35and maybe you did say it like how often
42:37have I found myself mixing up words
42:39where I'm like, you know, that's not
42:40what I said. But yeah, actually that did
42:42come out of my mouth in that way because
42:44the human brain works in funny ways.
42:46>> It does. It does.
42:48>> So I I like that whole idea of um taking
42:51notes, copious notes, recording,
42:53documenting. Uh but it also seems like a
42:56great opportunity there to to to use AI
42:59to help you. Um and I'm a big VI coder.
43:03>> Then the AI's check then the AI's
43:04checking the AI. I don't know. Just you
43:06just make sure. Again, I'm a big like
43:10>> don't expose yourself to risk. You don't
43:12need to, right? I I feel like that
43:14that's that's a I could have that as a
43:15tattoo. I mean, it's a it's a mantra as
43:17far as like business things go. Like,
43:20just be sure that if push comes to shove
43:24and you have to you have to explain why
43:26you did something or where did this come
43:28from myself, I don't want to be like, I
43:31don't know, like, oh, I just I did the I
43:33did the AI like I that can't be the end
43:35of my answer for for me. So, you know,
43:38again, documenting stuff is not the
43:40sexiest work we do, but we do do it for
43:42a lot of things, right? We have like
43:44spreadsheets that have all the ad copy.
43:46We have all the images. Like, we do work
43:49like this on a fairly regular basis
43:51where we're trying to keep track of like
43:52what tests are we running this month,
43:54like what, you know, what's coming up
43:55next? Like if you think about it in that
43:58type of workflow, I don't think it's
43:59terribly different from a lot of stuff
44:01that we're we're sort of already doing
44:02when we're we're managing the flow and
44:04we're managing a strategy and we're
44:06managing like our testing, you know,
44:08what are what are we testing? Um what's
44:10our priority this month? You know, that
44:12that type of thing. Um it just it's a
44:16extension of that.
44:17>> Yeah.
44:18>> Um
44:19>> I'm curious, do you use a project
44:20management tool?
44:22>> Do I? No. No,
44:25>> I don't.
44:25>> Okay.
44:26>> But it's just but I but I work I'm a
44:28consultant. Like it's just it's just me.
44:29So like I just have to create stuff for
44:31me that makes sense for me. So that's a
44:34different situation than if you have,
44:35you know, multiple people who are
44:37working on things. I think if you have
44:38multiple people who are working on
44:39things, you probably need something
44:41that's a little more codified as far as
44:43like this is where we put it. This is
44:46the format we want it in, right? Like
44:47somebody should come up with a a
44:49template for this is this is kind of how
44:51we're going to do this. So, we're all
44:52going to document it in the same way.
44:54Like, for me, it's just me. It has to
44:55make sense to me.
44:57>> Um, so I'm a bit of a different use case
44:59than someone who,
45:01>> you know, is working with multiple
45:03people or if you're in-house, you have
45:04different documentation needs than if
45:07you're, you know, working as a
45:08consultant and and that type of thing.
45:10>> Yeah. I uh I built uh Founders Voice.
45:14It's a content management system for
45:17myself to to figure out what to blog
45:19about or what to put on social media.
45:21And um even though I'm really the only
45:24user of it within my company, it it was
45:26very important because I struggle with
45:28the same thing that you're saying, which
45:29is okay sometimes uh we do a data study
45:31and the analysts send me the the CSV
45:34file and then I process the CSV file and
45:37it becomes different iterations of it
45:39and then I go to Claude and then I go to
45:41Gemini and then I go to GPT and I ask
45:43him different questions about it and
45:45eventually I'm just like wait I said
45:47this was the conclusion but like where
45:49did that conclusion come from again?
45:51and like good luck finding back that one
45:54chat because I might be like I remember
45:56it was Claude who said it but Claude at
45:58some point ran out of memory. So then I
46:00started a new conversation and now I
46:02spend an hour digging through old
46:04conversations. Um and then and so so
46:07then I was like okay within this tool
46:08that manages my projects in my brain I
46:11do need a place to put in what was the
46:14the AI chat that I had right so it keeps
46:16it all in one place but it makes it much
46:18more findable and that's kind of why I'm
46:20asking like how you go about it um I
46:23know everybody's brain works different
46:24ways but for me it's helpful to have
46:27these things connected in some sort of a
46:29a structured manner even though I don't
46:31have other people necessarily looking at
46:33it.
46:34>> Yeah. Yeah, and I think that's a great
46:35point. Everybody sort of thinks about
46:37things differently. So, you know, I I've
46:40say been saying forever there's hardly
46:42ever one right way to do anything in
46:45PPC. And I would say all of this AI
46:47stuff definitely falls into that
46:49category. Like, is there a right answer?
46:51And no, there's not a right answer.
46:52There's not a wrong answer. There there
46:54were a myriad answers for how you might
46:57do things, how you might utilize a tool,
47:00how you might apply it, how you want to
47:02manage your backend, how much of it you
47:04want to retain. Like there's no right or
47:07wrong way to do it. I just I always like
47:09to bring up the sort of the business
47:11side and like the risk management piece
47:13because there's nothing it it pains me
47:17to think that people in the community
47:18could not think about that stuff because
47:21it's not necessarily top of mind,
47:24>> but have them end up in a space where
47:25they're they're in hot water um because
47:29it just didn't occur to them that like
47:31you have to document this stuff or this
47:32might be important or somebody might ask
47:34for it or god forbid you get sued,
47:37>> you because of something that happened
47:39or you get pulled into a suit where even
47:41if you're not the per the direct person
47:43like you touch the account like you're
47:45going to get pulled in like there's a
47:46lot of things that potentially can
47:48happen and I hope they never happen to
47:50everybody who's listening here today but
47:53um I have had an experience like that
47:55and that um very early on in my uh
47:58career as an independent business person
48:00and that has molded me into somebody who
48:04thinks about all this stuff like very
48:05forward
48:07>> in everything that I do. Um, so I do try
48:09to share that with others because I'm
48:11like, well, my pain should be able to
48:13benefit other, you know, other people to
48:15not have to go through some of the stuff
48:16that I had to go through at one point.
48:18>> Yeah. Thanks for doing that. And so, um,
48:20we're coming to the end of the episode
48:21here, but, uh, Julie, is there like one
48:23story, and you can take this either into
48:24the like the happy place or the not so
48:26happy place, but like, do you have one
48:28specific AI story where it either did
48:30something amazing for you or it was like
48:33a classic AI fail that illustrates why
48:36people should be cautious?
48:38>> Um, again, I don't use AI a ton. Um, but
48:43I did have a client provide like a
48:47massive I mean massive document um
48:53with
48:54we'll call it like guidance based on
48:57sort of like this is like here's our our
49:00sales data, here's who bought stuff, you
49:02know. Um,
49:05and there wasn't really any way to ver
49:06verify it. So, like you ask, you know,
49:08questions of like, okay, well, can I can
49:11I just see the raw data for, you know,
49:15the actual things that were purchased?
49:18Oh, you have demographic data. Can I
49:19just see the raw demographic data,
49:21right? Um, and that wasn't possible. I
49:25only the only results I got was that was
49:27the AI. um that made me uncomfortable
49:31because
49:32I you know it's it's putting a lot of
49:35faith in something that
49:38I didn't have any way of verifying.
49:41So you know that that piece for me
49:45I would say in the last year that's
49:47probably and that's happened more than
49:49once. Um, that's the space where I I get
49:53really like, ooh, um, I'm going to need
49:55I'm going to need more than that, you
49:56know, and and some some clients are
49:58happy to provide more and others are
50:00like, but I gave it to you in the AI.
50:02And then you're like, okay,
50:04I guess we're going with what the AI
50:06said.
50:07>> Yeah.
50:09Uh, no, that absolutely. I think the AI
50:12handoff problem uh whether it's AI
50:15handing off from one AI to another AI to
50:17continue work on this previously
50:19developed piece
50:20>> is a big one uh but also to humans then
50:23right so you should be able to say okay
50:25what was the underlying data set and
50:26even going back all the way into the AI
50:29chat to the beginning like what was the
50:30PDF what was the CSV that was attached
50:32let me look at that let me
50:34>> verify um certainly very needed
50:38>> okay well Julie uh this has been Great.
50:40So, uh, if anyone's interested in, uh,
50:42working with you or learning more, where
50:46should they go?
50:47>> Uh, you can probably the easiest way to
50:50connect with me is you can find me on
50:52LinkedIn. So, Julie Freriedman, Bachini,
50:54Neptune Moon. Um, obviously you can find
50:57me through PPC chat as well. So, we are
51:00everywhere. Um, we have Slack, Twitter,
51:04Discord, we have a group on LinkedIn.
51:06Like, we're we're everywhere. Um, you
51:08know, as far as PPC chat goes. So, I'm
51:10I'm highly findable. If you Google me,
51:12there's a bazillion results. So, you you
51:14can definitely find me if you're
51:16interested in chatting about anything.
51:18>> Great. Hey, well, thanks everyone for
51:20watching. Thanks Julie for being on as a
51:22great guest today. Please subscribe if
51:24you want to know when the next episode
51:26comes out and also engage with us in the
51:28comments and the questions and we'll try
51:30our best to answer those. And with that,
51:33thank you Julie. Thank you everyone and
51:34we'll see you for the next one.

About this transcript

This page contains the full transcript of AI in PPC: Where It Helps, Where It Breaks, and Where It’s Risky | PPC TH 122 by PPC Town Hall, generated from the public captions YouTube serves with the video. The transcript has 10,176 words across 1,418 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.