Lessons From 20 Years in PPC: How Agencies Should Prepare For The Future

Episode Description

Frederick Vallaeys sits down with Aaron Levy, long-time PPC leader and now Optmyzr Evangelist, to talk about the future of paid search, AI, and agency life.

Aaron has spent nearly 20 years in paid search, from helping DuPont launch its early marketing programs to running a 115-person team at Tinuiti. He shares lessons from his career, including why “best practices” can kill originality, how to train new PPC talent in an AI-driven world, and how agencies can balance clients, vendors, and innovation. Key discussion points:

• Aaron’s journey from PPC intern to industry leader
• How to train junior PPCs to think beyond button-pushing
• AI Max, PMax, and how search behavior is evolving
• Turning client meetings into structured “prompts”
• Why agencies must invest in data infrastructure over flashy tools
• The risks of lazy “best practices” vs. the value of creativity
• Leadership lessons from managing 100+ PPC professionals

Episode Takeaways

Aaron’s journey from accidental intern at DuPont to industry leader at agencies like Seer and Tinuiti provides the backdrop for a wide-ranging discussion about how the fundamentals of training, managing, and executing PPC have evolved—and what hasn’t changed.

The conversation goes beyond tips and tactics. It dives into bigger questions about how to stay creative in a world full of copy-paste marketing, how to build real trust with clients, and what it takes to lead and grow strong teams at scale.

Aaron’s journey from PPC intern to industry leader

Aaron Levy’s journey into PPC began by chance when, as a student, he landed a full-time co-op at DuPont and helped launch their search program. The impact of that early work—and the attention it got from executives—hooked him on paid search. From there, he moved through several agencies, including early roles at Fanatics and Seer Interactive, before spending nearly a decade at Tinuiti, where he led a paid search team that grew to over 100 people.

In his early career, Aaron had to figure things out on his own, since there weren’t many resources available. He picked up skills across channels, from coding emails to managing affiliate programs. After years in the fast lane, he took a break to recharge and focus on hobbies like golf and gardening. When ready, he stepped back into the industry.

Aaron’s leadership style centers on autonomy. At Tinuiti, he encouraged teams to build their own approach within a shared strategic framework, balancing consistency with individual creativity. His approach helped shape a strong team culture and earned him recognition as one of the most influential names in PPC, with keynote appearances at major industry events. What began as a summer job turned into a career built on adaptability, curiosity, and a focus on empowering others.

How to train junior PPCs to think beyond button pushing

Aaron Levy points out a common problem in PPC training: too much focus on teaching people what to do, and not enough on why they’re doing it. He believes new professionals are often trained to follow steps and chase conversions without understanding the broader context—like what triggered a search or what happens after the click. To him, real learning comes from critical thinking, not checklists.

“Really bringing it all together is teaching people how to be more the ’take the TV apart to see how it works’ type of people. Not just looking at, oh, we have to try new broad match or in current state of affairs, we have to try Performance Max. Do you? What does it do? What are you hoping to accomplish? How is it different from what we’re already doing? Are you just trying a new toy because someone said you had to try it?" explains Aaron

Instead of relying on “best practices,” which he considers lazy and generic, Aaron encourages teaching through creative problem-solving and goal-setting. He pushes people to question tools, define clear objectives, and think strategically, especially when using platforms like Performance Max. For Aaron, the true value in PPC comes from understanding how tools fit into a bigger business strategy, not just knowing how to use them.

“Being the best at Performance Max in the world isn’t really a value proposition for a service vendor because everyone can kind of hit that button, but being able to explain what it will do to the broader ecosystem, explain what it’ll do to the client, and explain how it can help save the business money, I think is the most important part.” says Aaron.

AI Max, PMax, and how search behavior is evolving

Aaron and Fred also talk about how search behavior is changing, but not as fast as the tools. Platforms like Performance Max are pushing AI-driven automation, yet most users still search the old-fashioned way. That gap creates confusion, particularly for marketers trying to figure out what actually works.

“I would argue that we’re at a stage where everything is kind of new again. There’s sort of the old way of doing things or the way that we all know it around search terms and bids. But that AI shift is very real. People are searching in different ways. They are formulating what they look for in different ways. They have different touch points along the way to getting to that day of feeling googly and doing that final search.” says Fred

Aaron questions whether new tools are truly innovative or just rebranded versions of old ones.

“I think a lot of the industry is falling into that latter trap that you have to use AI, you have to try broad match, you have to try Performance Max, but without understanding exactly how the tool works, what it’s going to do to our accounts, and frankly what it’s going to do to our brand or our client as marketers." said Aaron

The biggest challenge now is training. With everything shifting, it’s unclear what new marketers should even learn. The core message: don’t follow trends blindly. Focus on understanding how people really use the internet and let that guide how you use the tools.

Turning client meetings into structured prompts

Aaron and Fred discussed how client meetings are often seen as a time sink. More talking, more follow-ups, and more work. But they suggest a smarter way to think about them: treat the meeting like an AI prompt.

“What if you could think about that 30 minute meeting as hey let’s really figure out what it is you want to do, how we’re going to go about it, and at the end we’re going to take that transcript, give it to the AI and the agent, and at least the simple stuff like adjust budgets for this thing that we just told you is going to be a promotional period like that should be doable by the AI.” says Fred

Aaron points out that the real value of client meetings isn’t in the slides, it’s in the business context you gather. The issue is poor preparation—too often, meetings lack focus, or clients fall back on “just show me what you did last time.” He believes success comes from entering the meeting with a clear objective and asking the right questions, just like you would when crafting a good prompt for AI.

“So fundamentally in both cases it’s again it’s good prompt engineering or good goal setting and so I think as long as we are training people on—junior folks and senior folks alike—training people on how to go into these conversations with an extremely purposeful route, certainly in its current state then we can have something automate most of the presentation development, certainly all the charts and numbers and then we put the words on it, but as long as you go in there with something that you really want out of it then you’ll be very successful and then you’ll have your go forward instructions because you had your engineered a prompt or structured a meeting really well.” explains Aaron.

The takeaway is simple: if your meetings are thoughtful and goal-oriented, they naturally create the foundation for AI to take over the repetitive parts.

Why agencies must invest in data infrastructure over flashy tools

Aaron argues that building proprietary tools as an agency isn’t a great long-term strategy. Once an idea is out there, others will catch on, and it’s only a matter of time before competitors build similar—or better—versions.

Instead, he believes the real value lies in clean, usable data. Most clients don’t have their first-party data in good shape. It’s often scattered across spreadsheets, outdated systems, or stuck in silos. Agencies, too, tend to underinvest in data infrastructure, even though it’s the foundation for any effective automation or insight. Without clean data, even the best tools can’t perform well.

“I think if there was a place that I would invest as an agency owner, it would be in data infrastructure because building the tool, building the software, building the LLM, I think if you’re building it for yourself, things will catch up in a couple years.

Spending a ton of money in making sure that we can ingest and understand and clean up data. So then that could be fed in into something else. That to me sounds like an amazing investment. And funny enough, it’s something where agencies underinvest and where clients don’t want to pay because nobody has a data budget. And that to me seems like the most worthwhile investment for people.” Aaron said.

The bigger challenge isn’t technical—it’s human. Sales teams might input data incorrectly, or incentives may skew the numbers, like a call center inflating lead quality to earn bonuses. On top of that, agencies often get limited information, hearing only about the marketing side when they really need full business context. PR plans, product launches, even TV ads, to make smart decisions.

“Well, that’s back to the fundamental point of you know automation can only do as good of a job as the data that you put in. So fully agree you should get your data structure in place and make an investment in that.” Fred explains

The risks of lazy “best practices” vs. the value of creativity

Aaron strongly rejects the idea of “best practices,” calling them average and uninspired. At Tinuiti, he banned the phrase entirely, arguing that relying on them leads to bland, copycat marketing, especially with tools like Performance Max, where every ad ends up looking the same. He warns that automation and AI make it easy to lose originality, especially when people stop thinking critically and just follow what’s popular.

Instead, he pushes for creativity, experimentation, and strategic risk-taking. Great ideas, the ones that eventually become best practices, come from trying something new, not copying what worked before. He believes agencies should have honest conversations with clients about whether they want safe results or are willing to invest in testing new ideas.

Aaron encourages exercises that challenge people to think in unconventional ways, like giving students random campaign scenarios that AI couldn’t easily solve. His core message: in a world of automation, human creativity is the true competitive edge.

Leadership lessons from managing 100+ PPC professionals

Aaron Levy, who led a 115-person paid search team at Tinuiti, believed in giving people freedom within a clear framework. Instead of strict rules, his team got guidance, trust, and the space to work in their own style. That approach made it easy to spot who could thrive independently and who couldn’t keep up.

“Especially when I was managing a really large scale team. Yes, I had my super high performers, but I didn’t have to build stuff for them. They would go build their own stuff. I had to build stuff for everybody, and I had to build things with guard rails for—I had only A players, but some were A minus players.

Come on, everyone. Everyone has a bell curve. But so you have to make these things tangible for everybody and make them usable for everybody, including people that are not comfortable with technology.” explains Aaron.

He pushed for a culture where clients met the actual people working on their accounts, not just a polished pitch team. Internally, things like unlimited vacation and flexible hours were the norm—because if someone wasn’t pulling their weight, it showed quickly. For him, leadership wasn’t about controlling every step, but about setting clear goals and letting capable people figure out how to reach them.

Episode Transcript

Frederick Vallaeys: Hello and welcome to another episode of PPC Town Hall. My name is Frederick Vallaeys. I’m your host. I’m also CEO and co-founder at Optmyzr, a PPC management tool. For today’s episode, we have someone from Optmyzr, but he hasn’t been with Optmyzr for that long. He has been in the industry for a very long time though, and he’s a brilliant mind who’s been at a lot of agencies, managed a lot of PPC accounts, and really knows how to make things work in this constantly evolving landscape where the tools change, the team structures change, and recently, of course, artificial intelligence is also stepping in and trying to do a lot of the work that we’ve been used to doing. So Aaron Levy is my guest today. And with that, let’s get rolling with this episode of PPC Town Hall.

Aaron, welcome to the show. Good to see you. We see each other all the time and good to see you on the podcast.

Aaron Levy: I was going to say thank you for having me in apparently my own studio. It’s been exciting to get to know the team and get to know the product and get to work with you a lot more.

Frederick Vallaeys: Yeah, likewise. Very happy that you ended your little break from PPC and decided to come back. We spoke and here we are. One day, yes, you’ll have to be in studio in person. Right now you’re remote still on the east coast.

Aaron Levy: Yes.

Frederick Vallaeys: You’ve been in the industry quite a long time. You’ve been on the list of the most influential PPCers. You’ve given keynotes at all of the events that the audience has probably been to like SMX Hero. For those people who haven’t had the pleasure of meeting you, who may be newer in the industry, tell them a little bit about where you came from and how you got here.

Aaron Levy: Yeah, I think like most people, I fell into the industry kind of by accident. I’ve been in paid search for, oh, this hurts to say, 20 years now, 19 years, something like that. I fell into it. I was looking for—I was being an overachiever and looking for a summer job in November for some reason when I was in school and wound up landing a full-time co-op. So I got to help DuPont, the chemical company, get their search program off the ground when I was 19, 20, one of those. Really liked the power that it had and how effective it was, and frankly that it was new because me, a little intern, was getting attention from the C-suite, which was awesome to me.

From there I went to a couple of agencies. I worked at the agency that is now Fanatics. It was an ecommerce platform. So we got to manage both sides of the business. Then I worked at Seer for a long time. I think I was somewhere around employee number 12 or 13 at Seer. I’m probably best known for spending almost 10 years at Tinuiti, formerly known as Elite SEM when I joined it. While I was there, I ran the paid search team and my team was at its peak probably about 115 people. Then as you mentioned, I took a break. I took about a year and a half off to play a lot of golf, as you can tell from my hilarious tan lines. Do a little bit of learning, do a little bit of gardening, do a little bit of furniture rebuilding, and not enough home maintenance. But then, you know, you and I had talked almost as soon as I took a break. Graceful as you are, you gave me my space, but then when it was time for me to get back into the working world, I called and here we are.

Frederick Vallaeys: Yeah, that’s great to have you on the team. We’ll be traveling together to Hero Conference San Diego. Anyone who’s going to be at that show and wants to reconnect with Aaron or see myself will be there. I’ll also be in London in a few weeks for SMX London where I’ll also be doing a master class, and Aaron has graciously helped me prepare some of those slides as well.

But Aaron, you said something interesting and let’s jump into this. You were 19 years old, maybe 20, but young, green behind the ears, and you stepped into a big company, DuPont, and you helped them figure out PPC. That makes me think of Chuck Robbins who’s the CEO of Cisco. Just this last week he was asked whether he was going to reduce staffing due to AI because a lot of people like Dario Amodei from Anthropic and people like Sam Altman are basically saying listen, a lot of people are going to be put out of their jobs. Mark Zuckerberg says the same thing about marketers.

But Chuck takes a complete opposite point here and he says he actually loves having meetings with interns and younger people because they tend to be the ones learning these new technologies in school. They’re on the cutting edge of how things are actually done. They usually come in without bias and with a fresher perspective and they help the company move in the right direction.

I think that’s super relevant to the industry that we’re in, and you having been at so many agencies, talk to us about how you educate someone who comes in fresh and how you think that’s going to evolve in this new landscape.

Aaron Levy: Yeah, I think that’s a really good question. Certainly when I started there was not quite so much documentation. There was not quite such a robust blogosphere out there. I had to figure out a lot of it myself, which was of course fine at the time, and I think a lot of people have learned that way. But thinking back to a lot of the things that I was taught as I got more serious about my career. When I first started off, I did paid search, but I also did every other channel. I had to learn how to code emails and learn how to do affiliate and things like that.

What I found there, what I found in frankly my own training that I put together with my team at Tinuiti, a lot of our work with junior folks was teaching them how to do. We would teach them where buttons were. We would teach them what search query reports were and what we wanted them to look for. Like, oh, look for all the terms that spent $100 but didn’t convert. Let’s exclude them as negatives and all that jazz.

I think the challenge that I see in our industry and have seen for a long time—this is not a new challenge—but that very binary sort of transactional view of what Google Ads is like, oh, it’s easy, it’s just a math problem. Find out what converts, find out what words made it convert, let it go. But never taught us, never really taught us to look at outside influences.

I’ve made this joke quite a few times, but nobody wakes up and just decides to have a super googly day. Well, some people do and they work at Google, but something always happens before someone searches. Something always happens while they’re searching. Something always happens after. Google and many others of course like, oh, we can tell based on their browsing behavior or whatever. I think we all know that a lot of search engines can be a little obtuse on that front, but really bringing it all together is teaching people how to be more the “take the TV apart to see how it works” type of people.

Not just looking at, oh, we have to try new broad match or in current state of affairs, we have to try Performance Max. Do you? What does it do? What are you hoping to accomplish? How is it different from what we’re already doing? Are you just trying a new toy because someone said you had to try it?

I think a lot of the industry is falling into that latter trap that you have to use AI, you have to try broad match, you have to try Performance Max, but without understanding exactly how the tool works, what it’s going to do to our accounts, and frankly what it’s going to do to our brand or our client as marketers. Does it accomplish something that’s relevant? Do we have a goal in mind that we hope to accomplish? Frankly, most people don’t a lot of the times. And will this tool help us accomplish it? Being able to assess that sort of—I’ll call it a problem solving structure, but looking at the whole big picture instead of just how do I do my job as easily as possible so I can check it off a box.

Frederick Vallaeys: Right? And you made the point that when you did this, everything was new and so you had to be behind the scenes. You had to experiment a lot. I would argue that we’re at a stage where everything is kind of new again. There’s sort of the old way of doing things or the way that we all know it around search terms and bids. But that AI shift is very real. People are searching in different ways. They are formulating what they look for in different ways. They have different touch points along the way to getting to that day of feeling googly and doing that final search. So I think there is a need for experimentation.

But I like what you’re saying, which is don’t just use the tool because it’s shiny and it’s new, but try to understand what that tool can do. Then I would argue, and I’m curious if you’ve had this, but oftentimes these tools are designed by Google under the idea of, okay, it’s going to do this thing, but actually no code is perfect, and it always kind of does maybe something else or unexpected things. That could be an advantage, right?

So how do you strike that balance in an agency setting of maintaining your client’s happiness with doing the things that we all know we should be doing, but balancing that with enough of the new stuff where we could figure out, okay, here’s an opportunity that nobody else has figured out because we’re just better than everyone else at knowing exactly how Performance Max works and where to deploy it?

Aaron Levy: Well, and there’s a third, I guess, pillar in that little structure as well, which is your partnership with vendors, which put more simply is your relationship with Google. You don’t want to take a product like Performance Max and be like, “It’s stupid. We’re not going to test it for anyone.” But trying to figure out how to prudently test it and test it where things are more important is, I think, the tough part.

To me a big degree of it is—gross oversimplification here—but I think the way that most people use ChatGPT for search is essentially delegating tasks to it, and so you can look at it the same way that you might look at an outsourcing exercise or offshoring exercise, which I’ve been through a few attempts to do. Marginal success sometimes, not so good other times. If you can’t accurately describe what you’re trying to accomplish, how is a tool going to accomplish it for you?

You were kind enough to share a lot of the presentations you’ve done before, and a lot of it is about really robust prompt engineering. How to write a good prompt, how to write an effective prompt. I think prompt engineering is super important, but again, if you don’t really know what you want the tool to do, how are you going to engineer a prompt for it?

So yes, teaching it the semantics like that GPT is pretty bad at math sometimes and it forgets that CPA is a calculation and not a raw number. Okay, we can tell it that. That’s fine. How do we get it to look for things that are actually effective? How do we get it to look for things that are actually going to help our accounts? And as an agency, how are we going to get it to help power us strategically? You said being the best at Performance Max in the world isn’t really a value proposition for a service vendor because everyone can kind of hit that button, but being able to explain what it will do to the broader ecosystem, explain what it’ll do to the client, and explain how it can help save the business money, I think is the most important part.

So that goes again to unpacking exactly how these tools work relative to a foundation that we’ll have to establish in some way. To your point, I think that’s going to be the hardest part for agencies, and I’m frankly glad I don’t have to do it, is what do we train people on now? Because we know consumer behavior is changing.

We also know—and this is something that I realized during my little hiatus and coming back in—it’s also really important to take a look at the internet from an outside view and remember that frankly most people aren’t very good at it. A lot of people are going to get frustrated with AI overviews. A lot of people are going to keep on searching their same way. Others may evolve, but I took probably a little bit too much joy in watching my mom try to Google things and seeing how our parents use the internet and seeing how our older friends or less technologically adept—whatever the opposite of inept is—seeing folks who don’t use the internet for a living, how do they use it? Because that’s going to be most of the customers and how do they react to it? Behavior is changing, but it’s not all the way changed yet. So what do we teach people on?

Frederick Vallaeys: Exactly. And that’s maybe what makes this moment in time so difficult is that we have to be dealing with the past as well as get ready for the future. Because like having been at Google, people didn’t know how to search in the beginning and they would put in “the” and be very verbose about how they specify things, and then they started to realize, well, these words don’t actually help return different results. So let me use that limited space that I have for something that’s maybe a bit more meaningful that helps filter down the results to what I need. But people have gotten there over time, right?

What’s kind of funny now is that we’ve been conditioned to speak Google-ese in terms of how we search, but now we can actually start speaking human again and AI mode should presumably give us those answers. But it is a reconditioning and like I said, there’s thinking about prompting. Now, when I think about prompting, you can either be the expert and be very specific and explicit and basically the micromanager like here’s my task, here’s the five steps that I want you to follow and just go and do it exactly that way.

But what I find fascinating and maybe like with more junior PPC people is tell them what the goal is and ask them how might you approach this, and don’t say like here’s the goal, go and do it, right? But say here’s the goal, how might you do this? And then you see the thought process. Again, this is the benefit of that junior person. They come up with novel ideas and maybe approaches that you hadn’t thought about. If they come with an approach that you know is flat-out bad, well, they’ve told you, you can tell them that’s not how we’re going to do it and here’s why.

But I’ve certainly found myself being informed by these models and finding novel ways to approach problems that otherwise I would have followed the old methodology a bit more. Yeah, that’s how I like using it.

Aaron Levy: You know, it’s interesting. I might have seen my eyes raised and a little light bulb go off over my head. So I taught for a number of years. One as an adjunct and one just sort of like a survey level course. The survey level course at Drexel, I mean, we just had students come in one day a week. It was sort of an overview on all media types. I basically taught all media, which was tough to do in an hour. But the project that the students got was kind of a mad libs. They got a brand, a problem, and a delivery method. So you would wind up with the stupidest stuff in the world. Like we had one group of students once that had to do a contest to announce a recall of milk and they had to focus on social media. But I mean that one was hilarious.

But a lot of them, a lot of other ones were like, “All right, you had a PR snafu and you need to proactively get ahead of people, but also make sure that people are searching for it, that you’re in front of them. How would you tackle this?” And that sort of thought process—again, this little light bulb just clicked. I’m like, “Oh, there’s something there for present times, too. There’s something that we can put in there from a problem solving perspective. Just give students a really weird problem that ChatGPT certainly couldn’t figure out how to do a contest for milk recalls.”

But it makes it a little tougher. It’s almost like in academia instead of trying to fight people using GPT for essay writing, make them think more. Make the essays have to be actually valuable instead of just giving a book report. So getting people more creative ways to think that the answer can’t be accomplished just by putting a really simple idiot-proof prompt in.

Frederick Vallaeys: Yeah. And so that makes me think about education, right? And one thing that I’ve really enjoyed that so in the GPT-4o live stream, the announcement of GPT-4o coming out, one of the things that was demoed was explain visually the concept of how a wing on an airplane produces lift. And for folks who haven’t seen it, amazing—like it did take about five minutes give or take. GPT wrote a web page which had a visual diagram of a wing and you could change the angle and you could change the speed of the airflow and the little thing would show you is this airplane now taking off or is it going down?

And that’s so cool, right? Because if you think about the Bernoulli principle and all of these other things that make planes fly, like it’s very dry. It’s like formulas and it’s hard to visualize. You stick your arm out of the car window like ah, now I sort of get it. If you can visualize this on a web page, like how amazing is that?

And so I took that principle. I said, what if I could visualize how a Google Ads auction works? Because I mean, for 20 years, you’ve been explaining this too, but like, okay, well, it’s ad rank and it’s basically CTR and CPC, these two factors, right? But then how do you calculate the actual CPC that you pay? And how does the organic section come into play? And Google has these thresholds that they talk about for quality and minimum CPM. How does that fit in?

So I just made a tool. It’s visual. So I’ll put it in the show notes, but people can play with it. My idea is this a great way to explain it to your clients. They may not know the nuance of an ad auction, but if they were like, “Hey, Aaron, go and manage this campaign for me, but here’s my budget. It’s $10,000. You cannot go above it.” And you’re like, “Well, okay, let me show you the bids in this industry, and with that $10,000, I can only get you this many clicks.” And like look at this auction simulator like what do you expect? I can’t get you beyond position six. And you visualize it and you let them play with a slider. It’s like oh okay if I spend some more time improving my quality score ah my ad starts jumping and my cost doesn’t actually go up.

So that’s one really cool way that I’ve found that you can also use it for educational purposes.

Aaron Levy: I really like that. And going back to agency world, you just mentioned the thing that’s frankly the biggest time—I’ll call it a time suck, but that’s a little disrespectful to clients. The biggest time suck at agencies. The reason why we can’t just automate everything and go to AI is because of clients. The biggest use of time that we had was in client services, presentations, reporting, and those are areas where, believe me, I tried when I was at Tinuiti to try and figure out a way that we could have our QBRs automatically made. And certainly GPT and Copilot can make some pretty presentations. And there’s things like beautiful.ai, but they’re just spitting out book reports at this point.

And that’s the real hard part. And again, hearkening back to academia where I think yes, GPT and other sorts of generative AI can do a ton of work for us. It can also just as quickly call out the impostors. It’s pretty easy to tell someone who just copy and pasted a report or a bunch of insights from something they slapped into GPT because it will be—it’s almost as if you can watch it think. And so a lot of the reports, you know, “give me 10 insights about my campaign like, ‘Oh, my CPA went up. Why did that happen?’” Then a very rudimentary format would go like, “Oh, ‘cause CPCs went up and your CTR went down because you had less clicks and more impressions.” I’m like, “Okay, cool. That’s awesome. Great to know. Now what?” It’s like, “Oh, you should get your CTR higher.” Like, “Thank you. Thank you very much.”

It’s not good at providing insights. It’s not good at providing strategy. And that’s going to—that’s what’s going to be the biggest piece of work for real time savings.

Frederick Vallaeys: Well, I totally hear you on that, right? But one thing that’s—so I’ve been doing a little bit of vibe coding.

Aaron Levy: I was waiting for this. I’m so excited for you to counterpoint this.

Frederick Vallaeys: Yeah. Well, so in vibe coding, you basically just prompt the machine with the app that you want to build and then the app gets built and then you can say, okay, I don’t like this section or add more data like this here. But it’s all prompt driven. Now I’ve created something that’s somewhat complex. It has a lot of data structures. And what I find really fascinating is that in the past when I was developing and I got stuck on a bug, I would have to open a database, look at a table, maybe figure out how that table was related to another table, go look at that table, and then I had to open the file that was supposed to render that data, but that file would have dependencies, and I’d have to open those files.

And so just the literal process of looking at these 10 different pieces of information that I needed to understand what might potentially be going wrong, you can give it that to an agent, a large language model, and it very quickly opens these things in sequence and in a matter of seconds comes back and says I’ve read it all here’s what’s going on. Now it’s not always right, it doesn’t always look deeply enough, but just the fact that it—you know, say it takes five prompts so I spend three minutes doing this, that is still faster than me processing each of these files in turn and kind of losing my place and like did I really look at that, did I not.

So that’s where I think these LLMs maybe aren’t at the stage yet like you’re describing. So they get stuck in a very fundamental place, but if you feed them more information, that’s where they can go deeper. And that’s why I’m excited about Optmyzr, too, because Optmyzr has all of this deep insight and deep access into your reports. You can connect it with your first party data. And so we can start pulling these things together. We can even pull context across multiple accounts that you manage for the same portfolio, the same client. And so all of that research can now happen much more quickly.

So I think what you’re describing is a fundamental problem right now, but I see it as an opportunity to build something that’s better because the tech to do it is absolutely there.

Aaron Levy: I would agree with that. I mean, I do love toying around with a lot of stuff that you built and a lot of stuff you described. Not a pitch, I promise. But I mean I’m thinking even to some specific examples from my past like I worked on a large pest control company for a long time and they were having a slow start to the year and we spent weeks—well not literally weeks because we got the old phone call on a Friday “I need you in our office on Monday to tell me what’s going on” thing. So hours and hours and hours, 48 hour weekend to try and figure everything out. 10 people, so probably 300 person hours spent working to figure out that it was rainy in the spring, so there weren’t as many bugs. And so it’s literally—that’s literally all it was was that the swarms were slower.

Or we had won—this has probably happened to a lot of people—a fashion brand of some kind. We’re like, wait a minute, why did we just have our best day ever? Of course, we figured this out with Google alerts or whatever. Oh, Mila Kunis was wearing our bag. That’s why. Oh, because everyone wanted what she was wearing.

And those are the sorts of inferences that I know that generative AI can solve with a little bit of coaching. I don’t know that the everyday paid search person is going to get it there. So until the product is there, I don’t know that it’s going to be the instant time savings that folks are hoping for in agencies, right?

Frederick Vallaeys: Because it sounds like you still have to come in and ask the question. And so in the case of Mila Kunis wearing your bag, it’s like that is one image on social. Thank God for Google alerts where they identified the bag probably, but if they didn’t, how would you know, right? Until somebody told you.

The other example about the weather like that’s really interesting because that seems like could apply to a lot more scenarios and like listen ever since the first day that Google Ads scripts were announced it was all like hey bidding by—like okay there is a perfect example why that might actually matter and that’s why we support those use cases in Optmyzr. But that was interesting too is like we support you setting up weather-based conditions but we don’t necessarily tell you like what is the weather you should be looking for, what are the patterns that we’re seeing, and that’s where a machine learning model—it doesn’t even have to be generative AI but machine learning can sort of point you in the right direction.

Aaron Levy: It certainly could, but then of course what data sources do you pull from and what level of do you have to give it? So in that particular example, like weather had to be unseasonably cold for an unseasonably long period of time, which would then delay the breeding period of termites and mosquitoes. So then our budget would spike back up in a month. We didn’t figure this out. Could machine learning, could generative AI probably, but who’s going to coach it and who’s going to build it?

Frederick Vallaeys: Right? And so one of the ideas behind large language models is that you make them as smart as they are by basically feeding it all the data and not getting in the way as the human. And this a little bit in Google Ads too, right? Like the more that we look at Performance Max campaigns, if you get in the way of it and say, “Hey, here’s keyword themes. Here’s negative keywords. Here’s audiences,” we don’t actually always see that correlate to better performance because you’ve made some human assumptions about what will and will not work. And you’re getting in the way of the machine doing what it does best, which is find patterns and set the correct bid for something.

And like I’ve always said, there’s no such thing as a bad keyword. There’s only such a thing as a bad bid. Right. Everything can be relevant at some level. Maybe you can’t bid below a penny. So in that case, maybe it’s a bad keyword, but it’s still a value question. It’s not necessarily a targeting question.

So, but you having been at so many agencies, like what if you were to start an agency today, thinking about this first party data, you have sort of the world’s more broad data set like Reddit, economic indicators, is that something you would build into an agency to be successful? Like, how would you think about that?

Aaron Levy: I don’t think that I would build it as a proprietary agency tool. And the reason being is that once it’s built out, once everyone will then have the idea. So they will see your pitch. They will see your web collateral. Yes. Could they—are they as smart as I am? Maybe, maybe not. Could they build it better than I could? Maybe, maybe not. But it feels repeatable to me. It feels much more like a software player or a vendor play or something that you would build into a tool.

You make an interesting point about first-party data as well. It would be amazing if every client had their first party data in order. They do not. The number of people that are still operating off of Post-it notes and spreadsheets and things like that, or their first party data hasn’t been cleaned up in 20 years, so nothing matches and it’s all backwards. And then some CMO tried to migrate to a different place and it got broken.

I think if there was a place that I would invest as an agency owner, it would be in data infrastructure because building the tool, building the software, building the LLM, I think if you’re building it for yourself, things will catch up in a couple years. Like we were just talking about at the beginning. We’re kind of at a crux or a nexus in terms of I don’t know the world if you will that these things are just getting started and they’re obviously moving very quickly. So investing all of that time and money into something that would be proprietary for me for a little while doesn’t seem like a good investment.

Spending a ton of money in making sure that we can ingest and understand and clean up data. So then that could be fed in into something else. That to me sounds like an amazing investment. And funny enough, it’s something where agencies underinvest and where clients don’t want to pay because nobody has a data budget. And that to me seems like the most worthwhile investment for people.

Frederick Vallaeys: Yeah. Well, that’s back to the fundamental point of you know automation can only do as good of a job as the data that you put in. So fully agree you should get your data structure in place and make an investment in that. I mean how did you see at the agency like people handling conversion data and offline conversion imports? Were they at least decent at that or is that a struggle and still an opportunity for agencies?

Aaron Levy: Most clients were competent at getting it done. They could handle it. Where we would run into brick walls is making sure that the sales people were doing it. We could connect the pipes, but we couldn’t make people put them in the right place. And so solving for the human problem was the biggest challenge. I mean, I went through this experience with a call center once where, oh, we’re doing amazing. Like, we have more MQLs than we’ve ever had this week. Our conversion rate is really high. What is happening? This is amazing, but is it?

Call center was doing bonuses for the most MQLs generated. So every SQL was an MQL that month because people wanted to get their bonus. And so it’s just like, okay, okay. That’s the biggest challenge that I think we faced as agencies is number one, people would only tell us marketing stuff and like, no, no, no, we need to know everything. We need to know everything that’s going on in the business. We need to know where all your data is. We need to know your PR schedule. We need to know your TV. So then we can use it. Then we can feed it into a proprietary tool or frankly anywhere and make better decisions as opposed to just here’s your marketing budget, go spend it.

Frederick Vallaeys: Yeah. All this stuff. And you know, you mentioned too that working with clients is one of the most time-consuming things. Doing the reporting for them. Sometimes the client gets in their own way of success. Here’s something I want to spitball. I’ve been doing this in other PPC town hall episodes, but one idea is why not see the client meeting as the master prompt for the work that needs to be done? Because most of the time, like agencies don’t like client meetings because it’s just time that you sit there and what happens at the end of that call? Well, you’ve got action items, which means more time. What if you could think about that 30 minute meeting as hey let’s really figure out what it is you want to do, how we’re going to go about it, and at the end we’re going to take that transcript, give it to the AI and the agent, and at least the simple stuff like adjust budgets for this thing that we just told you is going to be a promotional period like that should be doable by the AI. What do you think about that?

Aaron Levy: So first of all I don’t want it to come off like I hated client meetings because I love them. I frankly thought they were the best time ‘cause that was when you would actually get the business information. It was the preparation that was awful and very often the data format that people needed because they would need an analysis for their own budget. They would put their own budget tool, whatever.

It’s funny. I think you and I are speaking the same language but in a slightly different view. You—most people probably don’t know this. I’m a little bit of a Luddite personally. Like I tend to not lean into crazy new technology that often in my personal life, obviously in my work life, but so we’re saying the same thing in exactly different ways. What you just described is a good structure for a presentation. You just described having a presentation that’s purposeful, that you have things that you want to show them, that you have a goal, that you have something that you want a client to say yes to at the end.

So fundamentally in both cases it’s again it’s good prompt engineering or good goal setting and so I think as long as we are training people on—junior folks and senior folks alike—training people on how to go into these conversations with an extremely purposeful route, certainly in its current state then we can have something automate most of the presentation development, certainly all the charts and numbers and then we put the words on it, but as long as you go in there with something that you really want out of it then you’ll be very successful and then you’ll have your go forward instructions because you had your engineered a prompt or structured a meeting really well.

Frederick Vallaeys: Well now that you put it that way it’s making complete sense to me it’s like your perfect slide deck has an introduction about we are your expert PPC management team so you’re going to act in that role, here’s what we’ve tried in the past and here’s what we’ve seen, call to context, here’s what we’re trying to achieve, goal, here’s now let’s talk about that goal. So we put in place some limits and parameters. And then like I said at the end of the thing it’s like is that what you want us to work on? Yes. Okay. Go and do it. That is exactly how a prompt would be. So yeah I love that. I think it’s a great takeaway for anyone listening who maybe has more loosely structured meetings. Like think about it in terms of how you would prompt an AI. And if you get it to that point then eventually I think you’ll be at a level where you give it to the AI and it’ll do most of the work for you.

Aaron Levy: Well, and so many of these questions are—I mean, speaking to client meetings specifically, very often they would be like, “Oh, what do you want to see in this quarter’s presentation?” Just do the same stuff we did last time. Don’t let them do that. In the same way that you wouldn’t go to generative AI and like, “Just give me some ads that look fine. Here’s our website. Just go for it.” You just take them. You kind of have to press people. You have to press AI to get creative.

I was speaking with a former student and he graduated now who I mentor who’s from Villanova where I went to school and I unfortunately funny enough I do have Post-it notes in front of me and I did throw it away which I should have had AI remind me of what it was but I believe he called it the death of originality as the thing that we’re the most worried about which is very easy to fall into a trap with if you have someone who’s undereducated or doesn’t want to think very much and they’re working with a powerful tool like GPT or to a lesser extent, even Performance Max.

Every Performance Max ad looks the same. I know when I get targeted by a retargeting ad ‘cause it’s the same stupid thing with the same headlines that we all thought were best practices. And so I think if anything the most important thing that we can teach in current state of affairs for presentations for prompting for marketing in general is keep your creativity and remember it and don’t expect whatever tool to come up with that for you.

Frederick Vallaeys: Yeah. And remember too how the tool got to suggest those things. It’s because historically there’s been a lot of use of that and that may have correlated to good success. But for those things to become successful in the first place, at some point they had to be innovated. They had to be someone tried it and they saw more success than all of the other stuff and then everyone else followed suit. So if you want to think about leveling the playing field and having an edge, I completely agree. You need to be creative and be willing to experiment with something new. But I think that’s also a client conversation, right? Like listen, we can do the easy thing for you, not that hard. We’re going to get decent enough results, but if you really want to be best in class, here’s what we need to spend for a little bit of testing budget and here’s some of the things we need to try.

Aaron Levy: Right? And it might not work every single time, but we’re going to learn from it. Then we can learn, let that learn somewhere else, and we can let all of our idiot competition do the quote unquote best practices that they read on a blog somewhere, which by the way was a phrase that I banned from my team at Tinuiti. They were not allowed to say best practices. Best practices are lazy. They’re average. They’re the starting point. They’re what everyone else suggests. Like don’t say like it’s a best practice like no no no no let’s say what we’re trying to accomplish and why.

Frederick Vallaeys: Very interesting, hard to do. So Aaron, any AI tools that have some connection to marketing that you’ve been really engaged with that you found fun, some interesting stuff that you’ve seen?

Aaron Levy: Frankly I’m still relatively junior in my AI journey. Like I say I had a year plus of Luddite-ness which has been kind of fun. It’s been kind of fun to come back in as an infant. First of all, an obvious plug for Optmyzr Sidekick because love that little guy. For those who don’t know, it’s a natural language—what you call it—a plug-in tool toy that sits…

Frederick Vallaeys: A chat feature inside your Optmyzr account that basically uses LLMs and a deep integration into your ads data to help you with whatever you need.

Aaron Levy: Clippy for the 21st century. You know, I saw Clippy. I don’t know if people have seen the new Naked Gun movie.

Frederick Vallaeys: First of all, what? There’s a new Naked Gun movie?

Aaron Levy: Yeah, check it out. Good movie. And Clippy makes his return. And I’ll let you all find out if he’s super annoying or super helpful.

Frederick Vallaeys: Well, there’s my evening plans.

Aaron Levy: But no, beyond that, I’ve just been toying around with the kind of standard ones. Why did this just leave my brain? GPT, Claude, toying around with Gemini, toyed around with Copilot. We were toying around with some early integrations of whatever Gemini was called before it was Gemini and integrating that into some Google Sheets plugins. Doing a bunch of cool stuff that way. I did mess around with beautiful.ai a little bit, which is more or less presentation development, but I haven’t gotten as sophisticated as others, which I think if anything makes me a good candidate for a conversation like this because I think I represent most kind of mid-level people at agencies at this point that they’re going to learn enough to do their job. And so, how do we make them use it effectively?

Frederick Vallaeys: Right? Right. And it’s sort of you have these amazing tools, but not everyone’s gotten good at prompting or figuring out how to connect the pieces together. And that’s where you then see specialty tools coming out which very simply are built on top of GPT-4, but they figured out how to give better context, how to do better fine-tuning and really get it to achieve a certain task for you.

And yeah, I’m playing with a lot of it right now. Vibe coding these auction simulators. I’m coding a ghost blogger functionality so that I can write blog posts as if I’m talking to a reporter and it asks me about my thoughts on certain topics and then it constructs something that is much more relevant to who I am. And then of course all the things we’re doing in Optmyzr. So again not too much of a plug for Optmyzr but I’m super excited that Sidekick is actually starting to work across the whole account.

So, if you as an account manager need to go in and say, “Listen, I just need to make a report for my client, and I don’t want to use the custom reports that are in Optmyzr, but I’m just going to click through the interface, and as I look at stuff, I’m just going to talk out loud about what I see and what I think that means.” And so, at the end of the session, the Optmyzr Sidekick has heard you talk about like a variety of tools, and then it compiles that and says, “Okay, well, let me group it. Here’s some keyword stuff. Here’s some budget stuff. Here’s some strategic long-term vision and it actually spits out a narrative that’s not just generically generated by the LLM, but it is actually you as the account manager and your expert opinion combined with what the data is showing. So this is sort of that future vision that I’m very excited about and I think is going to be hopefully very helpful to people.

Aaron Levy: Well, and it’s tangible. It’s easy to work with and frankly that’s a lot of the way that I’ve been—my own again I’ll call it AI journey has been all right I’m going to try something a little bit more complicated this time I’m going to—I have had an idea in the back of my head for an app forever that I’m going to vibe code now because now I’m getting more comfortable with it. But just toying around with simple things and figuring out what it gets wrong.

In my keynote for SMX which was gosh probably over a year ago at this point. It was around this time last year. I sort of peppered in bad AI throughout which was just single prompted which was at the time my perception of AI and I think a lot of people’s. But now as we’re getting much better and much stronger at it and the models are getting better and learning faster. All right. Now this is a toy I can play with. Now I can build an app for people to find other people to go play golf with. Don’t steal it. It’s mine.

Frederick Vallaeys: It’s funny too because you did that keynote dissing AI and I probably did a keynote the one before at the previous SMX like all hyping AI. So we’re the yin and the yang here. But I think that’s good, right? And that’s why I’m happy having you in the company to sort of bring in that agency practitioner perspective and maybe boots on the ground and like how does all of this actually play out because because I living in Silicon Valley I mean I’m so surrounded by it and that vision where we think we could go but like what’s happening in most of the world. Let’s bring it back to reality at some level.

Aaron Levy: Yeah, I’m in South Philadelphia where nothing has changed for 150 years except my corner store is a house now, but besides that. No, it is I mean, I’ve always tried to take that approach. Especially when I was managing a really large scale team. Yes, I had my super high performers, but I didn’t have to build stuff for them. They would go build their own stuff. I had to build stuff for everybody, and I had to build things with guard rails for—I had only A players, but some were A minus players. Come on, everyone. Everyone has a bell curve. But so you have to make these things tangible for everybody and make them usable for everybody, including people that are not comfortable with technology, which I think makes us kind of good foils here.

Frederick Vallaeys: So, and that’s kind of an enviable position, right? Like having A players and maybe a few A minus players. How do you do that as an agency? How do you get those players?

Aaron Levy: Well, number one, we gave our teams a ton of freedom. And so in turn, if there were people who required straight up direction, direction, direction, direction, they would get found out quick. It would be people asking for step-by-step directions for doing everything. Instead, we would sort of, you know, we would have a framework and here’s a rough thought process, but you take it, you build it in your style, you build it in your own way.

So, I always like to say that at Tinuiti, you could look at an account and tell it was a Tinuiti account, but you could also see that it was like Alana or Jeff’s account because they would have their own style that would take off our own start point. And again, speaking to large scale agencies, almost always the person clients almost always want the person who’s going to be working on their business to pitch the business. So in turn, the person would get to pitch themselves just as much as we would get to pitch the company. And so that would make clients feel a lot more comfortable because then a person wasn’t kind of forcing or maybe neutering their own style because it had to fit into a company’s best practices.

Frederick Vallaeys: Right? So give them autonomy. Tell them what you’re trying to achieve and not how to go about it.

Aaron Levy: Autonomy, a lot of trust. And then you’ll find that the lower level players will shake out quickly. You know, we had at Tinuiti, we had unlimited vacation. I don’t think I’ve ever had a company where I had to count my hours. I mean, even you and I working together when we started, I was like, “So, what hours should I work?” And you said, “I don’t know. Whenever you need to do your work.” That sort of an attitude has always been viewed as sort of less affair, like, “Oh, how do you make sure that they’re not slacking off?” If anything, I think you find that out faster that way with freedom and trust.

Frederick Vallaeys: Yeah, good advice. Okay. Well, Aaron, you did a keynote at SMX like a year ago. People can go and find that, but you’re also going to be speaking at the next SMX, I believe. What can people look forward to?

Aaron Levy: More me, more contrarian stuff. I mean, a lot of what I’m excited about in working with Optmyzr is use a super strong team of analysts and obviously a ton of data to play with. And you know, in the past, I certainly had that available to me, but not to this level and not to this volume. So, what I’m most excited about is to start dissecting some of these things that are coming up. I’m really excited to poke into Performance Max and find out if it really is something special or if it’s just DSA with different buttons.

And trying to figure out how many of these things are incremental, how many of these things are driving value. That’s really what I’m looking for is not new tool. People tried tool good, but try to figure out what did the tool do? Is it doing something new or is it just shoving food around your plate? That’s the sort of stuff that I’m going to be diving into the most because those are the questions that we get the most.

Frederick Vallaeys: Great. Okay. Well, everyone sign up for that. And then if you’ve enjoyed PPC Town Hall, do subscribe, give it a like so that other people find it. And Aaron. With that, people can find you on LinkedIn. You’re part of Optmyzr, so they can reach you there as well. Go and see Aaron and myself at the upcoming conferences. But with that, we’ll wrap it up here. Thanks for watching, and we’ll see you for the next one.

Aaron Levy: Thanks, boss.