How AI Agents Are Rebuilding Marketing Agencies From the Inside Out with Pearmill's Nima Gardideh

Nima Gardideh: Yeah, yeah. So, so I'm Nima. I I run a company called Pearmill. We are a performance marketing agency.

Uh, my career started in tech technology, so I I'm a technologist at heart.

Um, and was an engineer for quite a few years and, you know, product manager of different sort of SaaS companies. And then somehow I ended up um, working on and marketing agency for really one reason, um, which is sort of my like mental model about like what my life should be about changed in my like late twenties and I wanted to be very good at something and started to think about like mastery as a sort of like a way to spend your time on and um, at the time my um, co-founder, we were working on a tech company, it wasn't working. He had done some marketing before and we're like, hey, let's let's focus on this because it was extremely new for me. It's generally exciting for me. I'm have a little bit of ADD and so I like newness a lot. And so it was like, oh, marketing, I've never done that before. And so, um, and now now it's like seven years later, we're super into it. We're forty something people and work with all sorts of brands, but um, yeah, my journey is sort of been like all over the place and uh, and I did not think of myself as a creative person. Now I do think of myself as a as a creative person, but you know, there there was definitely a a journey of becoming to be to get here.

Evan Lee: I love it, Nima. And and like I think creative person is the right definition. First of all, look at your space. Freaking beautiful. And then for everyone context, Nima plans like these dinner parties where he's just like curating every single element of it. Like the creativity is through the roof for sure.

Okay, Nima, the the this journey for the past seven years you had mentioned has been an interesting one because like started in technology, then you've jumped into the agency world and now AI is such a big thing in this in this process. So I'm wondering if you can shed some uh, shed some opinions and some light on like what is agency like before AI and what is it going to look like now that AI is here?

Nima Gardideh: Yeah, I mean this is a very hard question. I know you had put it up and I've been thinking about it for a while now. Uh, I think there are some obvious things that are happening, right? So like the obvious things that are happening is not and it just already happened to some the engineering field a little bit faster than everyone else and we're sort of uh, catching on in in in marketing where a lot of roles are sort of collapsing into each other somehow and no one knows like do you in the engineering sort of infrastructure, do you need an engineer or a product manager or a designer? Some of these are getting merged in our in our field in the in the creative realm and and media buying like, oh, do you really need a creative strategist and a designer and a motion person? You know, could that be really one person's job or like do they still need to differentiate and so on? It's it's sort of roughly unclear. At the moment, we still have all of those roles for instance, but um, we definitely don't have a copywriter role anymore. We used to have an explicit copywriter role. We used to have an explicit art director role. Those are sort of getting merged into the creative strategist and and and designer roles. And so, uh, my intuition is that we're going to continue seeing some like of this of happening inside of the organizations where we're going to see more and more being done by, you know, one person and lots of agents type of thing. Um,

Evan Lee: Yeah.

Nima Gardideh: and then the other part of this is like now this is happening now, so it's hard to like predict what's going to happen and you know, you guys launched Runeth and there's a lot to explore there. But like what how are agents going to enter uh, our ecosystem? So in in the engineering front that you know, agents are doing all the work, right? So humans are just steering where and what they do they're they're doing inside of their code bases. And so now the question is how is that going to happen in marketing? Marketing is not a fixed loop system in the way code is, so you can't really test it. So there's like some some things that are missing. It's not as cleanly obvious that um, how AI and agents are going to come into the into the system, but they're going to come and we we're using them a bunch and you know, we can talk about that a little bit. But I think organizations are changing. That's really what's happening. And organizations are going to be augmented with with intelligence in a way that hasn't been done before. And one of the things I like uh, thinking about a bunch and we talk about this a lot and I really like even like the way you guys talk about Runeth now where it's like your marketing team's brain. Like we we effectively are building a set of skills and a set of brain like a brain for our agency as well where we think like the way organizations are going to get built are roughly that they become these like sort of quasi-intelligent beings and AIs are doing some of the work, humans are doing some of the work and steering steering things. But that's sort of what's happening. And the outcome of that is that um, you will sell products still like sort of the same channels that exist, but there will be new channels and new channels are agents might be buying things for people. And so as marketers, like, uh, you know, it's very clear to us that at some point soon, we're going to have to market to agents and not just market to humans. There's going to be sort of like a new constituent in the way that Meta especially has had like our creative team when they're producing content, they're thinking like, how is a customer, potential customer going to see this, but also how is the algorithm going to sort of read this and and respond to it? And there's like this sort of second constituent of of the production. And so soon we'll have like almost like a third one, which is like, how are the agents going to read this and understand this and so on? And so, um, that's all very clear, but it's unclear how it's going to happen and what are the channels that are going to sort of spin up for us to work on and um, and how do how do agencies play within that? I think like, um, for us, like we think we're going to basically help brands do this transition. And so this is part of a new service model that we have where we'll help you become like AI native. And what that means is we'll help you build the agents, we'll help operate the agents that you don't have the time to operate, right? We'll do all of this work, um, in order to make sure that your marketing team is sort of catching up to to the current sort of like state of the art of things. Um, and like using things like Runeth, using Open Cloud or Hermes or whatever the different sort of agent harnesses that you have, uh, working in in your company, we can sort of be the operators of that or or like sort of shape them into solving the problems that you're trying to solve as an organization.

Nima Gardideh: And so this is sort of what's happening. And the other part of this is like, you know, here here is Runeth. Uh, we I just seated this brain. Um, so we we have a bunch of modules on Runeth right now, uh, when we work with brands. This is one of our clients that you all sent us actually, right? And so, um, as an example, there is a web of understanding about this brand, right? So they have a bunch of playbooks. Here are all the different playbooks. I'm not going to show them because this is all their learnings, right? Um, there is like reviews and comments coming from different sources. It understands itself, like there's different modules that there's a brain module, there's a concept briefs module and so on. It has a full understanding of what it can do inside of its brain. And we spend most of our time here where we first make sure the agent understands what is what it's role is, what are the learnings and institutional knowledge of the company thus far about how to uh, go to market for this product. And then we build tools to sort of self-manage this brain. So an an example tool I have is a competitive analysis tool where it can go and pull all the ads from the different uh, competitors or people that you're the companies that you're tracking in your space. And then distill the learnings that could be used for new briefs uh, through this this module. And so this thing is basically waking up every day, going and like pulling different competitor uh, ads, reading them, deciding, you know, is it steal worthy or not? Um, what what scores would I give it? It has like a a rubric that it uses. You know, Haley Rogers is one of our creative strategists. She has a pretty great rubric for figuring out if an ad is good or not. And so it's scoring all of these different ads and feeding it back into the brain. So there's this like sort of cyclical process to ensure that the brain is up to date. And then from there, you can ask it things like, hey, can you give me some brief ideas? Uh, and it's able to then distill all of this knowledge um, in order to be able to answer that in a in a way that quote Claude would not be able to just give you the answer or ChatGPT would not be able to give you the answer because the context is just not there. And so we spend a lot of time basically context building and and ensuring the agent can sort of help uh, understand in a way that you and I could understand after being onboarded to a team, how to answer some of these questions. And so now we're basically having these agents available in Slack with the clients and they're able to sort of use it to come up with ideas. Um, we have another team right now working on a version of this where they can go from brief end to end to a um, a version of the the storyboard at the very least that's generated using like, you know, other image or video models. There's a lot to do here and we're just starting on how to do this. And this was not hard to do, um, by the way, right? Like the the scaffolding took me a couple of weeks to do, but from here, the most important part is like, what have we learned? And can we really teach it that and then work with the agent to continue to do the work? And so this brand for instance is now starting to do this like weekly briefing that comes into their Slack with all the different things that they think they should launch given this new product launch that's about to happen because they have a content calendar. It knows about the content calendar. It's going to read that information. It knows what products are launching, right? So it's like, here are a bunch of briefs of ads to consider launching. And then if if, you know, the team likes it, they put it in their calendar and they start working on it, right? So a bunch of the parts of the work that you would have had to do manually, we used to have literally someone full-time doing that competitive analysis, right? Like they would go on on Motion and and Foreplay and all these different tools to to look at other companies' ads and understand what's going on and then distill that information such that our creative strategists could uh, come up with new ideas for the team. Now we don't have to do that. That like sort of super manual process, the agent can do it for you and can make it available for you to still see because you need to again understand the craft and what is going on. Um, but a lot of the nitty-gritty work can sort of be done for you in in whatever way you want to sort of shape your marketing organization towards. And um, you know, this is one of the things I I I like about Runeth because I know you guys had your own journey about how to do this and you had like an opinionated version at first, right? Well, the point of the opinion is what you're working on as a marketing org. Your whole job is to figure out how does your brand fit into the world, right? And that is unique to you. This is your job as a marketer and as a as a brand. And so that part no platform or no agent can figure out. It's your job, it's it's your collective knowledge and that is the IP you're going to build now. It's even more important than ever before. That has not changed and it's not not not going to change. The tactics may change, but the core thing of you're trying to do is like, how do I differentiate in this place and how do I tell the story there? And that learning can then be exemplified in your advertising efforts and so on. And the agents are just going to help you with that. Um, they're just they're not going to have the answers. Like we we do this a lot. We use a agents everywhere. It's simply that the thing that works is we have a point of view. We we there are ways to ensure that your agent generates ideas within your point of view and those ideas are helpful, but you have to apply the point of view. And we get, you know, this we we call it the rubric, but it's just an evaluator if you're technical. Basically, it's like a way for the agent to grade itself before it talks to you. And so we use these rubrics a lot to come up with um, content and uh, ideas that are within the framework that the brand has learned so far and wants to continue pushing forward with. Um, there is no sort of like silver bullet that the agents will just take over and do all the work. Uh, people have also learned that in engineering, it turns out that you still have to steer them and work on on how you're going to differentiate in the market as a company and as a as a product. So that's not going away. Uh, but this is kind of what it's it's starting to look like and and if I were to sort of speculate out, um, it'll be almost like end to end eventually where it'll even just produce the ad for you. But there is like, you know, multiple steps in that process where you as the brand would have created the guardrails and in that process, you're effectively showing the AI how to project your point of view. And so these rubrics that we apply for the different stages, like we have them for how to judge a hook or how to judge a script and so on. That is really our point of view. And that's how we're steering the agents towards what we want produced. And so this will become the craft in the future and so much of this is uh, going back to this understanding concept is not only understanding because historically a creative strategist, they have to understand to do the work. But now you have to understand to explain what the guardrails are. And that's like a new skill everyone has to build. Like you have to explain to an AI why you don't like something and why it's not the right thing. And so that's the new skill everyone has to build so that they can steer these agents. But otherwise, the core of the work will probably be very much the same, especially at this like more higher level thing of like, well, I'm producing creatives to get attention and so I how do I do that with agents now?

Evan Lee: Even did those did those rubrics exist before the team started working with agents?

Nima Gardideh: No, it was all in people's minds. Yeah. So like we we had to basically get them out. And that was actually an interesting process because I would go on a call with our team and be like, hey, how how do you define a good creative? And they would like say things like, you know, the vibe has to be like this and you know, and then you realize like, you know, this is not a skill that anyone had before of like, you know, I used to I I call it like the thinking about thinking, right? It's like you have to sort of discuss what your internal process is and then be able to express that in a way that the agents respond to. And that feels like a new skill that a bunch of our creative strategists have to sort of like take on and build. Like and and it took a while for people to like get used to it even of like, oh, I'm trying to come up with a rubric like what do I look for? Because a lot of it is like outside of your um, language center. Like people, especially in creative fields, like I'm a musician, like I couldn't explain to you a lot with words what I'm trying to convey in the sound, but I I'm feeling it and I have a clear understanding of it, but those things are not always connected. You have to work on finding the words to describe what is otherwise an internal process to you. And a lot of creative folks have this. And we work with a lot of brands actually where the founders or the head of brand knows exactly what they want, but they cannot codify it in a document if if it was like the last thing they did, right? Because this is like a is an interesting problem, but now we have to like collectively learn to do that better.

Evan Lee: Yeah. Nima, this is so interesting. Honestly, because because there's even something to be said there of just like uh, if you can uh, organize it in a way that will be easier for the AI to understand, in a sense you can also troubleshoot if it's ever hallucinating. So it's like you know exactly where things are starting to go wrong. So those those play hand in hand there. So it's a super interesting perspective. I think one of the last things that's on my mind here is like we've talked a ton about the agency and like what you're building here. The other side of it are the brands that you serve ultimately. And we've uh, today we've we've had a couple different agency folks join us who are mainly working with like the larger D to C brands. And then there's always the question of like SMB and like a lot of people here are like founders of smaller businesses. And I'm curious how this starts to shake out in this relationship of agencies and brands. And uh, basically it's like based on the size of different brands, is this something that everyone needs to do and should they all partner with an agency or like where does your head go in this type of relationship at this very moment?

Nima Gardideh: Yeah, it's interesting because in my head, enterprise has also this like other problem where they cannot easily build this context graph that I just showed you. And some of it is security issues. It's like, can you like put this data in a system that's external? And some of it is like they need to re-architect their whole company to be agentic native. And so that that's part of the problem. And so I think there is like diffusion at the top layer of the market is just going to take the most time. Um, and it's unclear how it's going to get done and there are plenty of people focused on this part of the market. And I'm talking top like, you know, let's say Fortune 5,000, like the the folks that are spending millions and millions per month type of thing. Um, anything below that where, you know, maybe you're in the $50 to $100 million run rate and your team size of a marketing team is, you know, less than twenty people, it can be incredibly helpful to have these agents around. Um, because at the moment, as I mentioned, they're not inside generators. They are sort of like uh, lubricants to the work. And so the less the more resource constrained you feel, the more helpful the agents can be right now. There will be a point where the resource constraint is an irrelevant part. That's not that's not the the the bottleneck. And this is sort of happened in engineering. Like last year, two years, a year and a half ago in engineering in software, people were using agents and AI because it was a an accelerant to an engineer's work, right? But as of like December of last year, all of a sudden, uh, code has gone extremely cheap, right? Writing a line of code is basically almost zero dollars. And so the field has changed where all of a sudden, everyone should be writing software. So the volume of software building is just going to balloon and so everyone needs a software engineer, but the software engineers are using these agents. Um, and so in that realm, like everyone is going to use it. When it's not a tool to reduce the uh, the sort of complexity of things and be a lubricant, but it's a new sort of like shift in the the way the work is done. Until that happens to marketing, which could happen in the next year, by the way, right? In a year or two from now, and I was talking to our team about this, like, we might have a creative strategist that goes from idea all the way to execution, all the way to the brand the the thing launching on on on a whatever channels of the future, but Meta and and whatever, right? End to end without working with a copywriter, an art director, a motion designer and and or even actors potentially, right? So there there is like a world where this whole thing becomes end to end and and sort of content becomes very cheap. In that world, I think it's a little bit different where I think like, um, you want to use agents in a totally different for different reasons, which is kind of like how do you differentiate in in the space where content is basically like free. Um, but up until then, I think the mid-market and lower is best situated to use this to compete with the larger players. I think like a lot of brands we speak to are mostly uh, they're either bottlenecked on resources. So this is kind of why you come to an agency because you're bottlenecked on resources. You don't have enough people, like it doesn't make sense to hire two full-time designers to do the work. You have like a part-time person that you're working with historically and maybe you come to an agency and so on. Um, so that's like one reason like you're bottlenecked on resources. And then the other reason they come to us is because they're bottlenecked on ideas. Um, and so I think the bottleneck on resource people should be using agents as much as possible. Um, and we'll be happy to help them build those agents and so on. But you roughly want to come to agencies when you're bottlenecked on ideas because that's when we can be helpful because we see so much. Like the thing that is useful with agencies is like you work with one marketer, but that marketer is an ecosystem with 100 100 brands that have like done done work and there is like uh, knowledge that's institutionalized in our company that you just cannot get on your own in one one brand. And so that's why you would ultimately want to work with an agency because there's like expertise that's sort of horizontal in a way that is very useful. Um, so I think my my intuition is that if you're like resource constrained, agents is sort of like are like really, really helpful and you should use them, but they're not going to sort of necessarily break you out of ideas and how to how to sort of like change and shift the business. That's when you should like maybe work with agencies and so on. But almost everyone can basically at least use it in that in that way.

Nima Gardideh: Incredibly helpful, man. Incredibly helpful. I'm wondering if you have any final words to leave with our audience today.

Yeah, I mean, I think this is like can be somewhat shocking or like existential to people. So I I'll leave this with like to to everyone as well because this is like something our team struggles with a lot where um, the craft is changing. The craft is changing fundamentally. Like in a year from now, we might not be doing the work in the way we're doing it right now. Um, and so how do you think about that like existentially as as maybe a creative strategist or a marketer? And I'll tell you the following. So I I actually think of myself as an engineer first and foremost. And that happened to my field last year where like, you know, I I actually wrote this essay to our team where it's like, hey, the craft that I loved for years of my life just evaporated completely. Like the the thing I used to spend so much time on thinking about the beauty of the code and how it worked together and so on, it's completely gone. I don't do that. It's almost as if like, um, you know, you you wouldn't make a a um, maybe a knife without the tools anymore, right? You would like use all this like industrial things to make a knife. You could make it and some people do make the artisanal version of it, right? But you wouldn't do it anymore. Maybe a smaller percentage of the population will do it. And that's what happened to code. Like you could you could do it, like you can spin off your artisanal code where you write it by hand, but you wouldn't do it anymore. And so to some extent that's going to happen to creative production. That's going to happen to marketing. Um, and for me, like what happened was that like I felt extremely sad at first and somewhat depressed. And then I went through this other process where I was like, oh, but you know what? I have a lot of ideas it turns out. And now ideas are even more important than before. So I can just execute on ideas more than I was ever able to before. And so there is a light on the other side, especially if you're like the type of person who has a ton of ideas, which most creative people are, you know, you have an infinite set of ideas, just not enough time to execute on them. This gives you the path to execute on them. And so I would really think of AI as an extension of your mind's ability to sort of like come up with ideas and execute on them. And um, with that framing, it is the best thing that's happened to to my life and um, you know, my like capabilities have just gone through the roof because of that. Like I'm I'm building so much for our company, for our clients and for, you know, art reasons all the time and constantly hitting the limits on Claude's Claude's, you know, daily limits and so on. But it it is um, probably like the most interesting time to to be a creator in some ways, like as an artist, as a as a creative. So, um, I think like think of it that way and I think it'll it'll be um, a really like a beautiful change in your life.