Old School PPC vs. PPC in the Era of Automation and Generative AI

Episode Description

In this episode, Sophie Fell, Director of Paid Media at Two Trees PPC, talks us through the big changes AI is making in paid search. As someone who’s been a bit skeptical about generative AI, Sophie gets real about her mixed feelings on tools like ChatGPT and the impact they’re having on the industry. She dives into how much control we’re losing with the rise of automation and what that means for PPC.

Sophie also shares some practical advice on how agencies can get ready for a future where automation is a bigger part of the game. She discusses how we can make the most of new AI tools without losing the personal touch that’s always been key to good PPC.

Episode Takeaways

1. Generative AI is changing PPC.

The introduction of AI tools like ChatGPT has created a divide in PPC. While younger PPC professionals are jumping in headfirst, seasoned experts are more skeptical.

“For my team members in their early twenties, ChatGPT is second nature. One of them even uses it for forecasting and audience targeting,” said Sophie.

But for those who’ve built their careers on manual analysis and control, it’s a harder shift.

“I still don’t trust it yet. I wish I did. I’m personally not ready for it yet.”

Fred pointed out that this skepticism is warranted. AI models are inherently predictive, which means they hallucinate answers.

“The whole large language model thing, everything is hallucination. It’s just that some hallucinations happen to be right.”

The challenge is knowing when to trust AI and when to double-check its output.

2. PPC automation is changing how agencies work.

PPC professionals have long relied on manual controls, but Google’s automation is making those controls less relevant.

“We need to get comfortable with losing manual control,” Sophie explained.

Performance Max, Smart Bidding, and Responsive Search Ads (RSAs) all rely on automation, and advertisers who resist the risk of getting left behind. But that doesn’t mean handing over the keys blindly.

“You don’t have fewer levers—you just need to know which ones to pull in automation.” — Fred

For agencies, this means shifting from execution to strategy—helping clients define the right goals, feed the right data, and let automation optimize within those boundaries.

3. Data quality is everything.

One of the biggest risks with automation is bad data. Your smart bidding strategy won’t work if your conversion tracking is off. If your lead quality is poor, your algorithm will optimize for the wrong things.

“If you’re seeing low-quality leads, assign values to each conversion stage—MQL, SQL—to guide the algorithm,” Sophie suggested.

Fred agreed, adding that even imperfect data is better than no data.

“Even if you don’t know the exact value, just indicating what’s better is helpful.”

For B2B advertisers, this means mapping the customer journey and assigning values to different actions so that automation has the right signals to work with.

4. RSAs and pinning: to control or not to control?

RSAs offer flexibility, but some advertisers resist because they don’t trust Google’s automation. Sophie prefers to let RSAs run freely:

“I try not to pin. If we’re using RSAs and smart bidding, I’d rather let Google do its thing.”

But many brands still struggle with the lack of control. Fred pointed to a recent Optmyzr study that found:

“Ads with lower ad strength labels actually tend to perform better.”

Google encourages advertisers to use as many headlines and descriptions as possible, but in some cases, tightly controlled ads perform better. The key is testing, not blindly following best practices.

5. SEO and PPC are converging

As AI-generated search results become more common, the traditional SEO vs. PPC debate is shifting. Instead of fighting over attribution, agencies need to think about how PPC can feed into AI-driven search experiences.

“SEO isn’t dying, but it’s changing. Brands need to decide now—do we optimize for AI-generated search results, or do we stick to the old way?” — Sophie

Fred pointed out that AI-driven search means fewer people will actually visit websites. Instead, they’ll get all their answers from Google’s AI summaries.

“The role of SEO is no longer just to be a blue link on the SERP. The future is about optimizing for AI chatbots.”

For PPC professionals, this means reconsidering how they measure success. If fewer people click on ads but more of them convert, is that a win?

The answer will depend on how well agencies adapt.

The bottom line: adapt or get left behind

PPC isn’t what it used to be. AI and automation are rewriting the rules, and the agencies that succeed will be the ones that know how to work with these changes instead of fighting against them.

But adapting doesn’t mean handing over complete control to Google. It means using the right tools to guide automation, validate AI-driven decisions, and ensure your data is working for you.

That’s where Optmyzr comes in. Whether it’s structuring data for smarter bidding, identifying optimization opportunities, or giving you the insights AI alone can’t provide, Optmyzr helps you stay ahead without losing control.

Because at the end of the day, it’s not about choosing between automation and strategy, it’s about making them work together.

Episode Transcript

Frederick Vallaeys: Hello and welcome to another episode of PPC Town Hall. My name is Fred Vallaeys. I’m your host. I’m also the CEO and co founder at Optmyzr, a PPC management software. For today’s episode, we’re going to talk to Sophie Fell from TwoTrees PPC. She’s a bit of a skeptic when it comes to generative AI, but she also sees that generative AI is fundamentally shifting how PPC and SEO are. So we’re going to hear from her how she thinks about the future.

What they are doing to prepare for changing to a more generative world and how to remain successful as an agency And as an agency employee, so with that let’s get rolling with this episode of ppctownhall.

Hello, Sophie Thank you so much for joining me today.

Sophie Fell: Hi Fred. Thank you for having me. I’m, very excited

Frederick Vallaeys: I am too. So tell people a little bit about who you are about Two Trees PPC, where you’re based, what you work on.

Sophie Fell: That’s a lot. That is a big introduction. So yeah, I’m Sophie. I’ve worked in paid media for around 10 years now. Fell into it by accident and then fell in love with it very quickly and have been here ever since. I’m based in the UK. Two Trees are based out in Sacramento in California. So you’re neck of the woods. And what started as a relatively small local agency has now turned into a bit of a bit of a global powerhouse really for paid media and PPC.

So I got the opportunity to work with them and haven’t looked back.

Frederick Vallaeys: Very cool. So how does one fall into PPC by accident?

Sophie Fell: When I was 21 or so I fell into sort of organic social media management. I was working for a startup and I was the youngest person there by about 15 to 20 years. And they were like, hey, you use Facebook.

Why don’t you run our social media? So started that then did some social media ads and they had an agency. And as soon as I saw Google Ads and Google Analytics, that was it for me never looked back. So fell into it due to my age. Yeah, and then that’s me ever since

Frederick Vallaeys: Well, I’m happy you ended up on the PPC and the analytical side because you know that’s the cool side of the defense.

Sophie Fell: always think so exactly

Frederick Vallaeys: but you know the fact that you were 15 years younger than the other people in the company and you were the one who oh my god she knows Facebook. That’s so cool. Do you think it’s like that nowadays or for Tiktok and because I don’t get ticked off

Sophie Fell: No, I don’t understand it either.

I mean the thing is at Two Trees, the team that understand TikTok are only five years younger than me. I’m not that old, right? But they’re the Gen Z people. They understand it and they know the trends. They know what to film, how to film it and stuff. It comes very naturally to them. So in the same way, I knew how to post on Facebook 10 years ago.

You know, there’s that evolution there now too, for sure.

Frederick Vallaeys: And so there’s that social evolution, but the bigger revolution that’s happening right now is obviously around generative AI, right? And I don’t know how you feel about it. Like, I’m so excited about everything in Gen AI, and I’m used to a fast moving world of digital marketing and Google always changing things.

But when it comes to open AI and cloud and Meta AI. It’s like it’s so much faster than even Google. Like there’s some weeks where my head just hurts like I’ve tried so many new things and I can still barely keep up. How do you see that for yourself and for the agency? Is the younger generation jumping on this more or who’s like taking the leadership when it comes to all this Gen AI stuff?

Sophie Fell: Yeah, that’s a really great question, because again, mentioning some of my team members that are early twenties, mid twenties. So for them, you know, their entry level and coming into this role and Chat GPT, for example, is just second nature to them. So I have a member of my team who actually uses Chat GPT for forecasting.

Like Facebook Meta audience targeting, he uses it for everything. And obviously with that human element too. But it comes so naturally to him to turn to ChatGPT to understand the right prompts to use and, and to get that information. And I feel like a dinosaur in comparison. I really do.

I still don’t trust it yet. I wish I did. I’m personally not ready for it yet.

Frederick Vallaeys: And that makes sense. And I think that skepticism is well placed, right? Because it hallucinates. And somebody actually explained to me that you shouldn’t be worried about hallucination because the whole large language model thing, everything is hallucination.

It’s just some of the hallucinations don’t make sense. And others are like, oh, my God, that’s spot on. And so it’s like that fact checking that you still have to do right. And then it’s also tricky, like your colleague who does forecasting with it. I’m betting he’s using more of the advanced analytics and the the code interpreter, sort of Python writing components, which work really well.

And I’ve used it for seasonality analysis, and it’s beautiful what it spits out. But where I’ve often found problems is if you sort of try to bridge the gap and you go from a Python statistical analysis to then interpreting it on a semantic language model there it falls apart because once it goes down to Python route, it stays in the Python run and Python is just not very good for semantics.

And semantics are not very good for data analysis. So like bridging that gap, that’s still where I find a lot of mistakes happen or where things kind of fall apart.

Sophie Fell: Yeah, I think there’s still, and this is why I’m not kind of full on into Chat GPT and stuff yet, because I think we still need human interaction. There’s the human QA-ing part of it as well. And obviously, you know, the language models and the learning, everything’s going to grow, everything’s going to get better.

I think some of it is I’m just rooted in sort of. The way I’ve always done things, which is not, not the best example to lead by.

So when my team member talks about doing forecasting with Chat GPT, the first time he told me, I was horrified. I was like, oh my God, what do you mean you’re doing it? But again, it’s just second nature to him and that generation and it, and it works well, so I’m aware that I need to put a bit more confidence into it.

Frederick Vallaeys: No, but it all makes sense. And what I find interesting too, is that you having done this for such a long time and you say you have your process, right? And you have a routine that you seems like you follow. The problem is if you went to generative AI, even two years ago, when Chat GPT was first released, you would give it a task and it would come back with something and it was like, well, how did it get to that point.

And you didn’t see the steps in between. You didn’t see the mistakes that it might have made in that, in that logic. Now with the, the strawberry model, model 1.0, the newest one that’s still kind of like being tested. It’s much more interesting because it actually breaks down the task into the steps and you can see it processing and saying, okay, now I’m like accessing websites to pull information.

Now I’m processing that. Now I’m running some mathematical calculations and it kind of lays it out so you can at least have an inkling that, okay, I get where it’s coming from. It’s also much better at breaking down these complex tasks into subtasks so you can actually go and ask it to achieve what I just asked you, which is forecasting.

Like, what are you going to do? And it explains it step by step. And then you can build confidence that you are not missing an important element. And then PPC, I would say something like, well, is it considering the fact that brand and non brand might perform differently, right? And if you see, oh, it’s, it’s just sticking everything together, then you might tell it, oh, please split it out based on brand versus non brand and then run the analysis.

So that’s one way that I see. Potentially us getting more confidence into what it’s doing.

Sophie Fell: Yeah, for sure. I think, if I could see those steps, I think I personally as well would have a lot more confidence in it because I don’t know if it’s me personally or a general thing, but I have a little bit of stigma if something comes from Chat GPT, I don’t quite trust it yet. Or like you said, I’m like, okay, but how did it reach that conclusion?

What’s it taking into account? And like you said, when you’ve been doing this for such a long time, you have your processes, right? You have your sources of data that you rely on to give you this information.

And yeah, I should be more open to it. But again, I think having those workings would build that confidence for sure.

Frederick Vallaeys: You just mentioned sources of data, right? And so that’s a big one. And you said you fell in love with the data and analytics. And how important is that data?

Do you think for the future of marketing, but I think there’s more automation, less manual control, like how does the data play into that?

Sophie Fell: I think it’s a critical function as PPC experts, people that run paid media, things like that, I think we need to get comfortable with losing those manual controls, right?

So, for example, things like PMAX, the response to PMAX and it being, you know, in the beginning, that black box of reporting and the lack of control, and that’s improved over time, but I think we’re so used to having all the levers, right. And, you know, going back 10 years, manual CPC, you track every bid, you know, there’s so much control we had.

And today the control is less, but again, there’s that comfortability with it. So with RSAs and things like you’re like, okay, I’ve got my inputs. I still have control. But I’m trusting the algorithm to kind of, you know, figure it out for me. And I think that’s a really big step and, and some people still aren’t there yet.

You know, some people still don’t trust RSAs, smart bidding strategies and things like that. So I think for anyone who wants that long term future in paid media PPC, I think it’s that comfortability piece that we’re going to have to get used to.

Frederick Vallaeys: Better get on board with the power pair

Sophie Fell: the power pair.

Yeah, exactly and I think PPC is moving so fast right now, evolving so quickly. If someone steps away from hands-on management for six months or a year, they can come back and completely not recognize the UI, Demand Gen, or any of the many moving parts at the moment

It’s sometimes a bit overwhelming how much there is.

Frederick Vallaeys: so let’s break that down. So one of the things you just mentioned was the levers. So You don’t have fewer levers necessarily, right? You can still do manual bidding and you have that lever of CPC but you probably don’t want to deploy that because it is going to be less precise than using some automation for it. So as you shift into automation and now you have the option of setting you know, you can maximize conversion value, but you can still set a T row out.

So that’s your lever. So, how do you think about these levers? When you want to deploy them and how do you figure out what they should be like? How does that client conversation go in this world of, hey, I just hired an agency and like, aren’t you guys just sticking everything in Google and Google does it all for us?

Like, what do you actually still do?

Sophie Fell: That’s a good question. And again, my somewhat old school approach, you know, that is apparently old school at this time. I think it’s about I still believe in the fundamentals of keyword research, solid campaign structures, all those things. And those are the things that we tend to kind of run through more specifically with our clients.

Similarly I think we get that response with RSAs a lot. People don’t trust. So not necessarily people within an agency, but maybe a client isn’t ready for RSAs, or again, losing the control of every single ad combination that may come up, and I think we have to address that with clients a lot of the time too.

Frederick Vallaeys: dive a little deeper on that, right? So I guess it varies, but you can pin certain components. Do you take more of an approach of let’s spin the whole ad so we know exactly what’s gonna come out? Or do you have a way of convincing the client that this does in fact lead to potentially better results and that it at the very least should be tested?

Sophie Fell: Yeah, I think so I’m still in the anti pinning sort of I really try if we’re gonna use smart bidding and RSAs I just I kind of want to let the algorithm do its thing which is not very control freak of me and I’m a control freak in any other aspect. But RSAs i’m like, okay, I trust you smart bidding.

I think You know a couple of years ago, I worked with a massive US SaaS brand who would not run RSA so they technically had RSAs but they only put three inputs for headlines, two for descriptions, and pinned all of them and that was their way of kind of having that control. I guess of the ads people were seeing and things like that.

We tried to convince them to use RSAs and they didn’t want to and that was fine. And I think there could have been a lot of value in doing that.

Frederick Vallaeys: well, yeah but you’re talking here about this was a major SaaS brand and my belief is that if you have a very strong and recognizable brand, you have a lot more leeway because so much of your CTR and the conversion rate is going to come from the fact that people know you, they trust you. So there’s less convincing, there is less experimentation that Google needs to do to get to that point of the click and the conversion.

But if you’re a lesser known brand and yeah, sometimes you don’t know what you don’t know, right. And that’s where it’s helpful to give Google a few options and let the RSA be an RSA. And in fact, there’s a new study.

So Optmyzr has been doing a number of studies. One of them has to do with ad strength. I’m betting your client who was spinning everything had pretty low ad strength labels, but they had good results. I bet as well. Right.

Sophie Fell: Yeah, I think, personally, I would have liked to see them do a little better. But mostly they were happy

Frederick Vallaeys: The study we just put out basically says that even so actually that the better performing ads on a CPA and robust basis tend to be the ones with the lower ad strength labels. If you have an excellent ad strength label, you tend to actually perform worse now.

The only point to make here is that don’t just trust everything Google says because Google is basically saying based on every advertiser who’s ever been out there. Like these are the things that tend to correlate to what we think is better performance, right?

So give us more variation, do less spinning, let the system figure it out. That’s what they’re pushing for but if you actually know what you’re doing or you work with a smart agency, then it’s fine, you know if Google tells you there’s not enough variations in the headlines, but you figured out the headline that’s like the killer headline that’s going to get that conversion and that click.

Then you don’t need to worry about the averages from all of humanity to get your results because you as an individual are different so that’s one study we put out. And then the other one, it was around pinning, right?

So basically saying the more that you can brain automation systems, the more that you have your personal bias and prejudice about what’s going to work. It doesn’t actually help the machine. It’s putting it in a box where it’s not able to do as good of a job as if you said, just go for it.

Sophie Fell: Yeah, and I can see that, which is why, personally, like, when it’s up to me, I try and stay away from, from pinning. I get it if you want to make sure your brand’s in a certain place, you know, I understand the kind of common reasons why you’d want to pin but I think that and the majority of cases again, if you’re using RSAs and smart bidding, I think let Google do its thing

Frederick Vallaeys: Yeah, exactly, now let’s talk a little bit more about the data, right?

So you made a point that if the data is good, then the power pair will function properly. But what does it mean for the data to be good? And I think we see this often with B2B lead gen advertisers. They struggle because a lead is just really someone filling out a form and the majority of that is not good.

You’re the system based on that. And like, I would say that’s not good data, right? So how do you as an agency think about good data and how do you bring it to Google?

Sophie Fell: So I think it’s again, sort of adhering to Google’s best practices. We don’t always adhere to Google’s best practices, but I think in this case, it makes sense again, particularly if you’re using smart bidding.

So again, it’s about obviously having conversion tracking in place, ideally enhanced conversion tracking in place. But it’s also about those sort of lighter conversion events too, right? So for example, like you mentioned for like a lead gen company, you know are there are newsletter subscriptions important to you are free trial demo requests, you know those sorts of things and I think in doing that you can help kind of again feed the algorithm with what it needs. And I think if you’re seeing like a very low quality off the back of your lead gen ads, I think it’s important to see where you can assign those values too.

So, you know, doing the math how many leads turn to MQL to SQL, you know, you end up doing and signing, assigning those values, I think can really help to in generating those leads.

Frederick Vallaeys: And so you’re talking about two things, right? On one side, you don’t have enough conversion.

So think about these newsletters signups and these micro conversions to give some signal to the system so that it can build up towards having more of the thing you actually want, which is not even the lead, but it’s the sale, right? But you have to feed the machine up until it gets to that point.

And then the second thing you’re talking about that I’m hearing is you have to value those conversions. So even if you’re now looking at actual sales, not every sale or not every lead has the same value. And I think the struggle has often been, how do you set that value? So what’s the process look like with your clients?

How do you get them to not be afraid? Perfection gets in the way of progress because a lot of advertisers, I see them saying, I’m not going to put in a value because I don’t know what the right value is. And so they don’t do anything. And the point that I usually want to make is, well, even if you can say that this is slightly better than that, even if you don’t know exactly what that number should be, that is helpful to the machine to indicate, okay, it’s 5 percent better.

That might not be the right number. But now it knows I prefer this over that. So how do you go about that with clients and with your data management?

Sophie Fell: Yeah. So generally we try and kind of reverse engineer from the end, right? So we’ve got a sale, we have a look at kind of what the average order value is from.

Leads generated by Google Ads, not necessarily from everything because it will differ, but generally speaking AOV from Google Ads, and then we kind of work backwards into so with an SQL and then MQL and again the percentages of people that make it to each of those things.

And if you keep going backwards eventually you figure out the value that you can assign to a lead based on kind of that math piece.

So we try as much as possible to do that. Some places will say that every lead is valuable or every newsletter sign up is valuable and, you know, but I think to give, again, if you’re using smart bidding, I think to give it the best possible results and to get the highest quality back from it.

Where you can use data and math to support that. I think that tends to help in my experience anyway.

Frederick Vallaeys: And I think this is one of the areas where generative AI can be quite helpful. So I think about generative and sort of like three levels of usage, right? You can use it to do things you already do faster.

So like we were talking about your process. If you can tell it to GPT, this is what I do. This is how I analyze keywords. These are the five steps. Just go and do it for me, make me faster. Or you can use it to learn something new and this is what we’re talking about, right? So, you know, you want to have a better modeling of what is the value of a lead, but you may not be statistically adept like myself.

And so I’m like, yeah, I can kind of put in values, but I probably could have done a better job if I understood statistics a little bit better. Generative can help me learn statistics. And if I’m willing to put in the time to actually see if it’s teaching me the right thing it can help me do that faster.

And then the third area where I think it’s highly risky is to say like, listen, I think I should have better values put in for these leads, but I have no interest in statistics. Like I’m just going to tell the system to spit something out and I’m going to go with it. That’s dangerous right because now you’re not validating the work.

You’re not actually learning and that’s when the hallucinations that are wrong could potentially like kill the campaign performance and that might be bad enough that your manager comes to you and is like, hey, sorry PPC management is not for you, go and find a different job

Sophie Fell: I hope not. I mean that would be hard to hear after 10 years.

As marketers we talk about this all the time, but you know, that data-driven decision making, coming back to data, all these things, I think, as a whole industry, we’re still not there. I would love to say that we are, and we all think statistically, and we’re just not, and not always all the time, it’s not always possible, right?

So that’s what makes it so difficult, I think. But in an ideal world, we would, we would have accurate lead volumes and values and things like that to really help feed the algorithms.

Frederick Vallaeys: Makes sense. So let’s shift gears here a little bit. So you’ve worked on SEO and you and you have SEO teams obviously at Two Trees.

What do you see the interplay being between SEO and PPC going forward as automation takes hold more and more?

Sophie Fell: Yes. I think this is a really interesting one because I still see PPC and SEO teams are siloed and almost butting heads or trying to take the credit or, and I personally, I don’t understand it, but, so I think the place we need to get to is right now we’ve got generative AI for creating ad copy.

It’s in its very early stages, right? There’s a lot of human feedback. There’s a lot of manual input still, you know, to some extent to kind of correct what the generative AI has come up with. That’s not going to be the case forever, right? All the feedback that we’re giving. And again, as the language models kind of evolve and learn, we’re not going to always be in a place where it needs manual input.

I can kind of see, this is my, like, dystopian future, I can kind of see a position where the generative AI kind of, again, uses maybe that single source of data, so whether that’s web copy or, again, a huge manual input of copy, but I can imagine they won’t want that kind of human interaction at that point, right?

So, I’m thinking there’s that single source of information, maybe a website that PPC and the generative AI used within PPC can kind of pull that as it needs to for kind of each individual person at auction time, that’s what I’m picturing eventually. So I think that again, for PPC success, there needs to be that input into the content that’s on the website or into the content that’s being fed to the AI to make ads that really work and say the right thing.

Frederick Vallaeys: Yeah, that makes sense. And so is this the death of SEO? And I think it’s the death of SEO as we know it. The purpose of SEO is no longer to be one of the blue links on the search results page and regurgitating known information is no longer good enough.

You want to bring something unique because that unique thing is what the generative system is going to pick up on and is going to make part of its summarization and its answer. But we have to, and I think this is both, true on PPC and SEO, but we have to come to terms with the fact that the way the consumer comes to buying something from the companies that we support is going to be fundamentally different.

But that generative system needs a ton of information. In fact, there’s some, some people are saying it’s already ingested the whole internet. So how does it get better? Well, one answer to that is we actually need to make better content, more content. So there’s more work for SEO to help the LLMs become even more sophisticated, but we have to be okay with the fact that that is not necessarily going to lead to people visiting our website.

It’s actually going to decrease time on website. I think the importance of a landing page, I think is just not. that huge anymore. It’s the landing pages for the transaction, because if someone has a conversation with generative, they’re going to stay there and they’re going to ask all the questions.

They’re going to get all the information. They’re going to know which company they want to work with because you fed it that information. By the time they come to you, you don’t have to convince them anymore. You don’t have to have a pretty landing page. You just have to have an easy checkout process. I think that’s the gift.

Sophie Fell: Definitely. Yeah. And I think again, it doesn’t mean like you said, SEO is dead, but it’s new. Right. And again, there’s this, this whole new thing happening. And again, for advertisers and brands are like, you’ve kind of got to make that decision now. Very, very soon, like, are we going to start optimizing for AI chatbots, you know, generative AI, LLMs, is that what we’re doing?

Or are we going to continue doing things the way we’ve done? And you know, when we all get sort of forced down the road of adoption or die, you know, it’s, you’re not going to have that wealth of data that other brands have. So I think we’re like at a very interesting point right now because the generative AI isn’t perfect.

And the chatbots aren’t perfect and the new search experience isn’t perfect, but it will be very quickly.

Frederick Vallaeys: I mean, think about Google 20 years ago, right? Like they would re index the web every couple of weeks. So all the data was always outdated, it was always stale. But now look at Google, like you post a change to your webpage and within a second, they probably know and picked it up.

It’s crazy. And we’re going to see that same evolution in the generative systems. And so, and then there’s also Amara’s line. It says that we overestimate the short term impact and underestimate the long term impact of a technology. So we’re going to have like all of these thoughts about, oh my god, like our industry is changing, like in the very near term we’re all going to lose our jobs, we’re going to do SEO, PPC differently.

That’s probably overstating it, but what we’re not looking at is like how fundamentally Commerce and doing business will be changed 10 years down the road in the longer term. Because our minds can’t even grasp what that’s going to look like.

Sophie Fell: No, and I think that’s like, it’s almost scary, isn’t it?

It’s exciting, but I think it’s a bit scary too. So I think about, so I have a 10 year old daughter and I think about, I have no idea what job she’ll have. I have no idea what phone she’ll have. You know, by the time I’m old, I can’t even imagine what her world will look like. And I think, we’re in a very interesting time right now.

Both PPC, SEO, marketing in general, I think we’re almost on the cusp of, of something really big. And again, it’s almost like Y2K, like, do you believe that this is going to change? And are you on board with it? Or are you going to say, no, I’m not adopting this. This is never going to happen and get left behind.

Frederick Vallaeys: Exactly. And I agree with you there, right? So if you don’t believe in it, or you’re not going to jump on board, you’re going to be that 95 percent of people who lose jobs. I was listening to Bill Gates in an interview and, and he made the point that when the phone came out, we could sort of like, you can envision, okay, that device is capable of doing these things, but it’s always going to be a phone.

And so there’s less fear because it’s, it’s utility, it’s useful and you can see where it’s headed, but with something like generative AI and artificial intelligence and potentially general artificial intelligence, we just can’t envision what that means. And so that’s why. It’s exciting and scary at the same time.

We don’t know what that future holds, but I think right now the message would be get on board like your colleague who’s doing all of these amazing things with GPT to to be more effective, to do new things they couldn’t do before. Like that’s the game plan, right? That is the type of person I think who’s going to be the most valued at the company, and they’re going to be in the best position to survive whatever this this big transition is going to be.

Sophie Fell: Yeah, exactly. And I do think some of it is slightly age related too, who knows? I’m only 30 and I’m not ready to pick all this stuff up. Yeah, and I think, again, what’s that going to look like? What’s the adoption of it going to look like? I mean, are we always going to have PPC traditionalists, you know, that will always do it in a manual way?

I can’t imagine that Google will let that happen forever. You know, I think at one point there will be that cusp of, okay, no more manual input, but your input instead is this, and it’s, you know, something completely different. So it’s it is really an interesting one.

Frederick Vallaeys: So when you talked about the fundamentals and a PPC traditionalist, will that still be useful? And so I’ll argue that no, you don’t want to be like the way that you don’t want to be the person who insists on doing it the old school way.

However, my thought is that you still want to know the fundamentals because at some level that tells you what the levers are that you can pull to get the generative AI.

To move in a different direction. And so I’m curious at an agency, when you get this younger generation joining the company and they’ve never worked with a manual CPC and they don’t know how the conversion rate might be part of the cost per acquisition formula, and they just come in and they’re like, turn on smart bidding, turn on RSAs, all automation.

Like how important, to what degree do you want to instill fundamentals in them? You teach them fundamentals or are you teaching completely new skills within the agency because there’s a new skillset needed in this new world.

Sophie Fell: Yeah. And I think, again, it comes back to those two different approaches, right?

So for me, I would say right now, I guess I am a PPC traditionalist. I mean, I’ll do smart bidding and RSAs and performance max and things like that, but I know it’s not always going to be this way, but I’m also not ready for that next step yet in terms of always using or relying on generative AI or GPT and things like that to kind of help you with my role.

So for me with the 20 year olds and the company I still want like you mentioned those fundamentals that understanding of why so when we talk about smart bidding again there’s that understanding of okay there’s the auction process and there’s the data and there’s the historical data and I think yeah, having that fundamental knowledge is important, however, I’m also learning from them, right?

When they’re saying, hey, you can actually do forecasts the manual way and it takes five minutes, not five hours. You know, there’s that piece of it too, so I think You know, if we can forecast seasonality without again, manually fishing all this data, putting it together, putting it into a chart, and we can do that in 10 minutes, then it’s really going to change a lot of things.

So I think it goes both ways. I’d like to instill the almost traditional values of PPC, but have them brave enough to Tell me what else there is, you know.

Frederick Vallaeys: Yeah, I mean, you’re making me think this is a little bit like the generations and families, right? Like, there’s not a clear cut, okay, there’s the people from before and there’s the people from after.

These people, they live together, they learn from each other, and the kids usually are a bit more resistant to listening to the advice from the older ones. But still, I mean, the parents repeat it again and again. And so eventually some of that gets absorbed by the kids. And I wonder if that’s the same at an agency, right?

So you get that younger generation coming in and it’s not about like us versus them. It’s like, Hey, you did interesting stuff. Can you teach me how that worked? Oh, and then you can be like, well, based on my fundamental understanding, like that might be that that was a great way, but how about thinking about this other thing that you could have asked generative to do.

And so as a team, everybody transitions to be hyper effective in this new world

Sophie Fell: Yeah, definitely and again I think you know you mentioned earlier with with gen ai how you can get the the process and the steps and I think again that pleases the traditionalist in me the person that wants the the workings, right?

I want to understand the workings and I want to teach the workings so that you know my team understand how it works or why their input brings this an output. So again, when he said, he did a forecast in Chat GPT, I was horrified, but understanding the steps involved and the data that’s involved again, we can be much more efficient if I just start adopting those things.

Frederick Vallaeys: Yeah.

Sophie Fell: Yeah.

Frederick Vallaeys: And then that’s interesting. So as a SaaS company founder, I think about this too, right? GPT is obviously useful, but the problem is you see when you go to GPT and you want to do an analysis on ads data, you have to go and get the ads data. You have to give it to GPT, then you do the modeling, then you get the Python out of it.

Now there’s this concept of drift. So. If you ask GPT the same question a couple of times in a row, it’s not going to come back with exactly the same response. It may lead to the same result that the Python code would be written differently, but it gives you the same output, but there’s not even a guarantee of that.

Sometimes the Python code may work. Sometimes there may be an error in the Python code. And so. That’s where I think about like, how do we take these methodologies that people are discovering? Like, Oh, it’s really easy to do a seasonality analysis in GPT. Yes. But what if you could combine that into a tool like Optmyzr?

And now you can say, okay, pick my brand campaigns. And we already know which are the brand campaigns because you’ve labeled that, or we’ve identified that through keywords and you do the analysis. And now you’re like, Oh, that’s really interesting. Now, let me do it. With two clicks of a button for a specific set of products or a specific geography.

And that’s the thing where, you know, GPT is such a great tool for ideation, but I think at the end of the day, people do still want to work in the tools, the email software, they have the task list management system. They have, they want that to have generative built in and use all of these cool new technologies, because at the end of the day.

Using generative by itself can actually be slower. And I experienced that very much when I started writing scripts. What was it like 15 years ago? These scripts were huge time savers. Individually, but then eventually I had so many scripts to maintain and I had to install the trip in each account.

And then I had to that Google would change something. So now I had 50 accounts where I had to go and like individually change that script to make sure it kept working. And that’s the thing that I think people will soon realize is, yeah, GPT is great, but like you have to deploy it into your corporate software, your enterprise software that you already have.

Sophie Fell: Yeah, I agree. And I think, you know, I’d like to see Google ads do more of that too, right? I want to see, you know, if my CPCs are down, I want you to consider every possible impact, external and internal, and have it in one place. Whereas, you know, now, for example, you know, we’ll get a report from Google and it might be, you know, you did the, you did this, you edited this one thing, and now your CPCs are high, or it was a new competitor, and it kind of stops there.

But I would love to see us get to a point where all those external factors are covered, like, oh, it was raining, so. Everyone was in a bad mood. This happened and, and having that, and all like the impact of like things like consumer confidence or the news. And I would love to get to a point where Google ads could tell us that, but again, like you said, in the platform, not after the fact

Frederick Vallaeys: Google is a bit slow, so I’ll see if I can build that functionality into Optmyzr.

You will Google, either one. Well good Sophie, this has been very exciting. Thanks for bringing so much energy and sharing all your thoughts on generative PPC. Anything else you want people to know or how can they get in touch with you?

Sophie Fell: So you can find well, you can find us on twotreesppc.com. You can find me on LinkedIn and X @sophiefellppc.

I’m always talking about PPC and whinging about PPC sometimes too. So you can find me in those places. And yeah, twotreesppc.com

Frederick Vallaeys: Very good. Well, thank you, Sophie. And thanks everyone for watching. If you’ve enjoyed this episode and want to see future ones and get notifications, please hit the subscribe button and we’ll let you know when new episodes are coming up with that.

Thank you for watching and we’ll see you for the next one.