“Leadership Is Leverage”: Motion’s CEO on the Future of Marketing

I think, back for the second time,

one of the friends of the pot and a reservoir motion has it going, man. Very good, very good. Good to be here, guys. Resort for motion.

At what point does it become like reservoir motion and run it or is run it the subset of motion? How are you, are you as a holding company now? Like, how are you thinking about that? You know, maybe this is breaking news here,

but I suspect by 12 months from now, it's only run this. Okay, there we go, see, we were looking for the hot takes. I'm glad we got one, 15 seconds into the episode. Yeah, you went, you went straight for it.

There is no question in my mind that, obviously, people love and use motion. We have no plans to like sunset that product any time soon, but I suspect within 12 months, none of our customers will use the core product.

And they will do everything they love to do inside of motion. We'll do it through run it and it'll be 100 times better. Is that like a terrifying feeling as a CEO to wake up and be like,

“hey, I think our business model or our main product”

is going to be obsolete and not a thing. Is that like an exciting thing? Like, because for me and Connor, and I'm not that I'm not just really like for a consumer brand. Like, I'd be terrified of like, you know,

no one needed to make up her while it's in a year for now. We were very early to invest very heavily into AI. We had just raised our series B in late 2024. And the first board meeting that we had, we were coming off the best revenue quarter of all time.

Like, the business was just like ripping and growing really fast. And I told our board, this is like a newly established board post series B, I told them that we have to reimagine the entire business and that for the for the time being, we're not going to have revenue targets at all. And the reason was that I felt that if we were going to have revenue targets for the existing business,

we were always going to just like hedge what we needed to do to build a real AI company.

And my, my reasoning was that like, if we don't do this,

“then the entire company is at risk because I think any SaaS company that doesn't reinvent themselves to be an AI native company isn't really big trouble.”

But on the other side, any SaaS company that quickly reinvents themselves to be an AI company, the market opportunity in AI is like a thousand times bigger than SaaS could ever be. And so you have two paths, one of like almost certain death if you don't act quickly. And the other path is like unlimited heights of growth. And so like to me, it was a no-brainer.

It was, it was terrifying, certainly. But, but I, I felt like we had no choice. And so from early 2025, the vast majority of our engineering resources gone towards like what has become run is today. And I'm very happy that we did it. And I think it's going to take a little bit of time for the market to realize what we've done here as like AI becomes more and more a part of the way knowledge work happens.

But I am so thrilled about what our team is built and we're, we're ready for the moment and it's going to be a lot of fun. So did you raise, because you raise a lot of money, did you raise specifically for this? Or were you like, hey, we have this money now, let's go and use it on this? Yeah, so, you know, part of the raise was obviously this idea that we have to build into AI.

“I think the thing that it wasn't, wasn't clear was like,”

do you just bolt AI on top of a core product? And I think many SaaS companies try to do that and like just didn't work quite well. And so we knew going into the raise that a big part of our roadmap was going to be AI. But I think very quickly, it was like, no, no, it's not like we need to add AI features. It's that we have to reimagine the entire company from first principles thinking about AI.

And one of the things that was concerning about it was like, it wasn't clear to me how much the market itself was going to change. How much like the core principles of the problems that we set out to solve with motion. How much of that was going to be in flux? But thankfully, like the core thesis of motion is like more and more true every day where, you know, there's a special type of talent, the person who is equally creative, equally data-driven,

and building, that's kind of who we built motion around the first time. It's like that individual is going to be a really important part of the marketing org.

And I think when AI happened, like that same person is still a really critical profile in the marketing org.

And so the core premise of the business never changed. The only question was like, our approach to technology, and I'm happier is the capital to be able to like have the time and resources to do what needed to be done. But like we basically, we looked at the entire product from scratch and basically rebuilt all of it as if we were competing with ourselves. The idea was I like if a newly minted seed company were to look at this problem with fresh eyes and a codex subscription.

And so that's what we did.

And the synergy though between like run-eth and motion is like, it's so one to one that I didn't feel like we've taken the business in a different direction. It's just everything that our customers would expect us to do from an AI standpoint is like baked into how we're thinking about run-eth. Maybe we could take like a very quick step back because everybody's been seeing the run-eth bought on X. Everybody's reading the X posts. But like fundamentally, you say you built run-eth in a way that competes with the core product motion or that you guys were competing against yourselves once you built it.

“Maybe you could just like briefly describe like, what is the work that's getting done within run-eth? How is it sort of an extrapolation of what was getting getting done in motion originally?”

And then where does it differ? Yeah, so if you think about, you know, motion was solving a problem within one part of the flywheel, which was like, you know, the creative workflow is you launch some stuff. You learn from it, you do research, you launch more stuff and like you do that in a flywheel. And motions core offering was around like understanding performance data in a way that creative people could like visualize it. So there wasn't just like pure data numbers, but fundamentally it was around like finding insights and signal among your data, right?

So like when you're doing competitor research when you're researching TikTok when you're like doom scrolling creators when you're looking at your own performance data. All of that are like inputs into context that you will process as a creative and marketing team.

And in some ways like the software industry is like always a couple steps ahead of like the rest of knowledge work.

But a lot of people in software are thinking about like the idea of a software factory where you have a bunch of like bugs and feature requests and ideas and like things like that that come in as inputs. And then you have the throughput of the factor which is actually like building the stuff and then quality control and then shipping those features. The way that we do knowledge work both in like software engineering and marketing kind of resembles like an assembly line of a factory, but obviously it takes exceptional judgment to make good decisions across that.

But the fundamental job is like how do we take inputs process them through our own judgment and then create outputs that we hope will do better than the last ones. And so like that that was the problem that motion solved. And with run it it's like it's it's the whole it's the whole thing it's not just like the performance data it's like running the entire system of the way we make decisions that we process context and.

“Most important like the way we compound learnings over time I think that's the missing piece that was very difficult for us to put into motion.”

I remember one of our earliest customer conversations like five or six years ago. We talked about this idea of like how can you memorialize your learnings how can you make it so that every turn of this fly wheel. We've processed inputs we've made decisions we've shipped stuff. How do we compound those learnings over time and I think you know people had like spreadsheets and whatnot whether you tried to like organize their learnings but it was. Fundamentally an unsolved problem and I think that's the difference with AI.

You can really start to compound learnings in a way that was just not possible to do before. And so same problem but kind of just looking at it from an AI native standpoint and saying. Do we really need to load a bunch of dashboards all the time in order to get signal or could we articulate the signal that we're looking for and have AI like look out for those and then surface them or you know sometimes when people use emotion report. They're like creating a new report. Then they're applying a date range filter then they're filtering for some campaigns like really what they're doing is like they're asking a question.

And then trying to find the response to that question through like point and click the dashboard and like visually consume the information. But why not just ask the question you know. What's the difference between this campaign and that campaign or or this visual format versus the other one or like.

That's that's like probably the first fundamental differences instead of like point and click build a dashboard.

Ask questions in natural language in slack with your team you know.

“So I think it's like start one of the things I think about AI is that it just starts to become a lot more like the way that you collaborate with a human team member.”

And less so a software product that you have to like build workflows around is it AI should feel like a labor force at your company.

It watches every creative you've ever run knows what's inside of each one and connected to it actually drove revenue. You can plug it into your Clevio Shopify north beam your whole Google Drive it reads all of it. So instead of asking what work glass quarter anyone on your team can ask for 10 evergreen concepts worth testing before key for and get them back within minutes. We use it hex cloud use it Jones road uses it it's the best AI brain for e-com brands on the planet if you don't use it you're going to fall behind.

Go to run a dot com and book a strategy session mentioned marketing operator sent you and you'll get a thousand dollars in free credits. You told me over a year ago a long time ago you said.

Basically whatever process you can like lay out if you're onboarding into employee whatever process you can lay out you can basically replace with AI over time.

And I feel like you kind of just hit that point now reason a lot of what knowledge work is right now is asking questions via a point in click workflow. And it's like oh yeah all of a sudden we just take a step back and think about the intrinsic pieces of like what that work is. Then all of a sudden it becomes much easier to understand how an AI might replicate it it can select the dates and build the filters and do the comparison and then come back to you with an answer that's much clearer.

“And I think that's just such a big unlock and it's cool to see that sort of begin to get realized in products like money.”

And so like you know I remember when I was like when a lot of the AI changes happening like my first instinct was that like I need to be deep deep deep deep into the weeds myself which is kind of where the the chief marketer. Like do or and chief leader idea comes from is that I think it's very hard to do this stuff without being like deep in the weeds.

And one of the first things that I built with AI was this repo on my on my local machine called research.

And what I wanted to do with that was that like I felt like a lot of my job. Was just scouring for signal right I was like a context machine I'm just looking at signal I don't really know how I'm going to use it yet but I'm just collecting signal from slack from customer conversations from talking to our team from Twitter and I'm just like. Consuming a lot of context and then there's a synthesis synthesization that happens somewhere where I'm connecting the dots between all the context that I've accumulated and one of the first things I wanted to do with AI was that.

Ten X or a hundred X the amount of context that I consume right because I felt like when I consume high quality context I make higher quality decisions and I only have a certain you it's funny house similar we are to AI is like I only have a certain context window myself right like. If I consume too much I'm going to loosen it and so I need to be careful with like how many tabs I keep open but. My thought was that like if I can have AI scour all of this context for me and I can give it some instructions to say like here are the types of things I'm looking for here's like a question that I have.

Really good examples that like I wanted to know early on. Like what's coming up in our customer conversations right and. Very simple I like to talk all of our transcripts and I had like a fleet of a eyes just go and read through a thousand plus customer conversations looking for signal. In the past I would be like if I had like a couple hours free in the afternoon like let me watch a couple of gone calls just like.

“So keep up with the momentum of like what are customers saying like take that as like that's useful I think pre AI still useful for me to spend an hour watch a gone call but.”

If I can scale that. Research to like thousands of conversations.

Then basically it's like you can really scale good judgment and like I think a lot of what leaders do is is basically that like try to stay close to the trenches collect signal.

Have judgment about the future and make good decisions and I think what AI allows you to do is like really scale. That process of like collecting a large amount of context which is kind of again the difference between a product like motion and a part part like run is that you in motion you'd have to like go. One by one open a report look for signal open a report look for signal in run if you could be like I want to go and look through like every ad we've ever done from all time.

“Based on this the other thing and like let the agent run for two hours and you have three weeks worth of research done in like two hours right so I think it's like a different different order of magnitude of what we can do in terms of work.”

The problem becomes like. If you don't have good judgment and you point your AI at this like massive massive mountains of context. Then you're just going to confuse yourself and the AI right so it it doesn't mean that like it makes the work easier and necessarily on the leader of the person that's doing it's like it's so high leverage that.

What I learned how to use the tractor and like put it to good use but we got tractors now like what does that mean has that changed the way that we approach work. I want to like slightly switch gears and kind of explain some of this degree like I'm I'm very deep in like right now like context phase like I've been working on a new thing and just like building a lot of agents and so for me it's like. Everything and it's either going to be the best thing that I've ever done or it's like the models will get so good they maybe won't need but it's just like dumping as much context as I can you know like like I'm even in the place where all this stuff sounds so cringe kind of like what you're saying about like.

Treating it like it like an employee like that's how I'm thinking about agents and like you own my I cringe when I see people put like their agent or charts on like Twitter but like you kind of do have to think about it like that of like what are the roles and response about like how would you onboard a person and how do you think about onboarding an agent. So I'm like taking courses that I would use to watch and I'm like just giving the transcripts to agents and like having them you know ingest and analyze them and so a lot of people are obviously talking about that like.

Models may be not.

“Being the most important thing which I want to hear your take on and get into later.”

But talk about that like context layers and building a brain and how just like generally for people right we've probably listeners like very different levels of you know their their AI journey so like talk about that a little bit and then like. How you are doing that how you've done it at motion and how you think people should be doing it whether it's with the run and throw without. Yeah, so you know what's interesting about this this topic is that everybody intuitively understands the value. So like at the most basic level you could be having a conversation with chat GPT and you just upload a bunch of files and say hey.

Take these files into consideration and like help me answer this question. Okay, that's like one level of context feeding the next level up is. You create a cloud project and now you're like okay all the conversations that I want to have with you from here on out. Are going to be related to this context or here's a bunch of PDF files and whatnot and then you kind of store that in the cloud project. Then there's a leap that happens where you're like okay what the hell's a mark down file.

You're you're in a new gear of basically AI adoption and one of our views internally we are lead creative strategist Alicia.

“Is not technical by any means but once you figured out what a mark down file is just like oh man like this stuff is cool and like and I think that's that's probably the first.”

Level up from you know building a context engine is this idea of. How to mark down files work what is a repository sounds very fancy and crazy but. The general idea is. If you ever look at like a software engineering project. It's just a bunch of files and folders even like.

Even the the the the most ambitious software products in the world like you take Google or you take meta and you're like what is this thing under the hood.

It's basically just a bunch of files and folders.

A massive amount of them they do a lot of interesting stuff and these agents have been trained on mountains and mountains and mountains of code.

“So one of the things about them is that they know their way around a file system very well like they can they can navigate.”

Files and folders as if it's like second nature to them. And so that is one of the things that. Makes the leap from like cramming all your stuff into a claw project to like basically having. Some kind of a file system it could be on your computer. This need to be in the cloud.

It'll talk in a second of like why it's valuable for it to be in the cloud.

And so imagine like if you guys seen this new format where people will post on Twitter and then it's the trying to explain a concept but it's basically a folder free structure. It's like a parent folder and like some folders like the lowest 10 meme ever. Yeah. But there's something interesting about organizing information that way where you're like okay here are the things that matter to me. These are my personas and each persona will have a mark down file.

These are like messaging angles that I'm testing like each messaging angle might have a mark down file. And so if you if you look at. Your system the way that you're doing work you can probably organize it in some kind of a file system. Even if it wasn't meant for agents like you might do that just like look at it and and be organized with you and your team. And so when when you do that and you create a folder and files structure like that.

With mark down files that are just giving context like rich detailed context about whatever that thing is. It turns out that agents are very good at navigating that. And this was one of the breakthroughs of last year during the holidays around like November December.

And it was the combination of Opus 4.5 and open claw basically emerging it around the same time and the pattern there I think there was like two big takeaways from that.

“Moment in time one was that. If you give AI access to a computer. They will do incredible things because they've been they've been trained on how to navigate files and folders in a computer.”

And the other is that coding agents were going to be the vehicle for all knowledge work so like you know if you took Opus 4.5 it was a really good. But it turned out that you can take a coding agent and apply it to any part of knowledge work and like fundamentally do it it will do a terrific job. And so anyways all of that to say from the the journey of like uploading files and folders into chat to BT to having some kind of repository which just means like a folder with a bunch of like sub folders and files of organized context.

That makes it easy for the AI to navigate because what happens when you just dump a bunch of files into cloud.

It's kind of probabilistic the way that the AI might decide like okay what information is relevant how do I know what to get but when you organize it a bit more of like a clean way. Then it knows how to navigate your folder structure and and find the right things when it needs to. And so that's really valuable and so I think a lot of people get to that point and they're like oh man I feel like I'm on top of the world I have. A repository on my local machine with all of this context and I'll use like my cloud or or GPT.

And they're just creating so much value but then it's like okay how do I. What happens when I turn my computer off and I'm walking away the now like everything's kind of shut down or how do I collaborate with my team and I think that becomes like the next level of challenge.

“People are the most important part of your business and finding real a players is hard and incredibly time-consuming that's why I want to tell you about propel propel is honestly the best recruitment agency that I have personally work with.”

They're marketing and operations specialist say source talent from US, Latin and beyond for top pcom brands like Jones Road like many others that you know and love. At Jones Road we've hired over 12 people through them we've hired video editors graphic designers we have multiple retention managers creative strategists from them. What I love is not only they easy to work with their quick but they work on performance so you only pay a one-offie only once you hire somebody you don't have to pay. Retainer no-on going fees plus every hire comes with a three month guarantee so I highly recommend them they work really great for us.

We've a really great team that they've helped us source had to link in the show notes and submit the form to get 10% off your first hire.

So before we get to the cloud portion because I do think like all of this becoming collaborative seems inevitable we're in like ining one of like sharing any of these files but I just want to touch back on this idea of like. Folders and files and that being the sort of the bedrock of any software we use today and as knowledge workers were like not really used to that. Most knowledge work right now is getting done within Google Slides Google Docs like spreadsheets things like that. And I remember having this conversation recently and I would I would love to get your perspective on it but I asked like oh yeah do we really think the future of knowledge work is like.

Mark down files like like like like like like files local on our I said at the time local on our computers and we could get to the cloud portion of it but like. But I guess would you agree with that do you think a year from now 18 months from now the law the way that we are documenting information as a knowledge organization is like. Is a different primitive than we're using today just because it's so much more conducive with the way that AI works.

“Yeah so I think there's two portions there's like there's raw unstructured data so for example you take like all your transcripts all your reviews like mountains of unstructured data that's like.”

That's like one piece of context I think with that it's probably less about like really rich organized Mark down files and it's more about. Basically being able to do semantic search or like vector search across a really large corpus of unstructured data so like I think that's one area even today that's the right way to approach so if in my like folder system. I'll have a folder called like raw corpus and in there I'm just dumping all of the all of the raw unstructured data in just one area of of the file system and for that you want to be able to ask like semantic search type questions that are like.

Of all the people that are frustrated what are they frustrated about right and and you're not going to get the answer to that through navigating a file system like you're doing to do like semantic search on that. But then there is like.

I do think those will happen in Mark down for some large period of time and the thing about Mark down is that it's just a text file right like the thing about it that is interesting there's a there's a tweet from.

Andre Carpothi in 2023 just like crazy way ahead of his time where he tweeted that English is the hottest programming language now.

And the idea is that like that what makes a mark down file special is that it is basically like it's a program for the AI to run it doesn't sound like that because it's so it's so natural language but.

“If you think about English is a programming language, I think that's correct and we are all going to be I think the future of knowledge work is.”

Building systems more so than like individual work a good example is like free industrial revolution. We were making like products my hand like artisans right and then factories come along we're still making products but now we're making products and like.

I think when people think about the scale of work from AI the first thing that goes into into their mind is like well it's obviously going to be sloppy it's obviously going to be terrible.

But if you take the physical product example just because we're making products in a factory doesn't mean they have to be low quality right like we might care about the assembly line the way it works the quality control and so on. And so I think the future of knowledge work becomes working in large scale systems so. You as like a leader of an org even today you're you're working in a system right you're like okay I have these teams and these channels and like this process and and in a lot of the way that we do work is.

In some kind of a system but the system is in our head and we are kind of following the steps of that system I think the future of knowledge work becomes translating that system into. Into it digital system through AI. And then you're doing quality control on that really large scale of of system and whatever the mechanism of that is going to be I think. It's going to be in the English language in plain text right and how we organize that I think can happen a lot of different ways but. My bet is that a big part of.

The context gets abstracted away until like the models get better and you're like okay you know what I'm just going to dump stuff and like it'll organize it will build these like graph databases and like. Join different concepts and I don't have to like. Organize context to the degree that I have to right now. But I view like there's a difference between there's some markdown files that are just really special like your system prompt inside of inside of your system.

“That describes your judgment that describes the way you make decisions those ones I think are going to be so important we're going to like.”

Go through them line my line as if they're like the constitution because everything downstream of that gets affected so. I think there's going to be.

Some text files that are really critical and we're constantly going to iterate on the way that we describe what we want in them.

The models are just going to get really good at. Organizing themselves and you know it's not just the model I think Cody brought this up but like the way that. Even even the model companies realize like the model itself is a is a commodity it's like the harness that you build around it that that creates all of value. And so you'll have a model harness combination that is tuned to toward a certain type of work like software engineering.

“It is really good at helping the knowledge worker of that domain fly right and I think that's what we're hoping to do for run it is that is that.”

As marketing leaders want to build AI systems for their companies. There's a lot of scaffolding that everyone's going to have to repeat and I think that becomes the value prop of vertical specific. That help everybody and at vertical just just fly in the same way that software engineering is is getting the benefit of that right now. Before Q4 hits the smartest e-commerce operators aren't just asking how do we spend more there asking how do we make every visitor more profitable that is where after self comes in.

After cell helps e-commerce brands increase revenue per visitor contribution margin and incremental profit by optimizing the highest intent moments. Car check out post purchase and even the thank you page brands like hexclad rich and pillar increasing their AOV across every cohort using these zero risk up cells.

It's just more free money in your pocket and with rocked thanks you can offset your after cell subscription.

No redesigns no risky funnels just better economics want to see how much money you can make with after cell go to after cell dot com slash rocket revenue forecast put in your numbers and see how much more profit your funnel will make before. BFCM starts. The models are so good there's and obviously there's all the open late ones so you know there's not necessarily that much. Leverage and being you know for mine six or five point out or something like that like they're all so good that they can do the job.

It's obviously like the the context it's the harness which is like the context and the skills and stuff like that.

Outside of the virtual machine like what makes run unique like how is it different than like an open claw or her knees or something like that.

Like are there vertical specific things that you've like does it have like the right connectors does it have like skills or parts of the harness that are like. Dedicated for like your ICP. Yeah, yeah totally the one of the ways I think about it is that. You know when when physical product companies decide to like vertically integrate their supply chain and it's like. One part of the supply chain isn't the thing that makes the whole difference it's the fact that we control.

So many parts of the supply chain so that at the end we can make a product that's exceptional so like. I don't separate the virtual machine from the harness from the data tools from the UI like to me.

“If you want to set up like a Hermes or an open claw one question is like okay.”

How do I tune the harness to be good at marketing okay that's one problem but then you do have to think about like the virtual machine or run it on a Mac mini and like how is it. How do you make it collaborative between different team members there's a lot of scaffolding that goes into that thing that's really funny to me is that. In some sense everyone is going to have to solve the virtual machine problem maybe by going and like grabbing a virtual machine subscription and like putting their open claw or the Hermes on it.

Because these things need to run on the cloud and even the Mac mini is is fun and exciting but like. The original agent was running on a mini the precursor to to run if was was a agent that I built called analyst it was running on my mini it was available to everybody in Slack it would get hammered by questions every day. It was a lot of funds like I built it and this again the idea of like what is leadership these days leadership is leverage and I was like if I can build this thing that gives leverage to my team.

It sounds like leadership to me right but it looks like I see work anyways we can we come back to that in a second.

And this thing is running on my Mac mini and one time I'm on I'm on vacation and the team's like hammering away asking analysts like all kinds of questions every day and the power goes out of my house. And then the Mac mini is off for like three days until I'm back home and boom like production grade agent just like stops and people can't use it. Obviously the Mac mini has problems from that standpoint so these agents have to run in the cloud and and so from that to. Things like the data tools that come with run if are very fascinating so we have a creator index of.

Four thousand creators that like run if just consumes like doom scrolls or content and saves it and so when you talk to run it about creators it just knows right like it just understands. Tens of thousands of creators and all their content is one example all the inspirational library of like what anybody else is doing.

“And like little things like if you want to get your Hermes agent to be able to watch videos.”

Yes, yes, I know the AI people can do it and it's like, well, I'm just going to grab a gem and I key and then I'm going to like make it so that when I upload a video it can work. Yes, that does work and then. But there's so much effort to do that and it's like did it watch the video correctly, did it look for the right things. Then you're like, okay, I also want to be able to upload a video and slack and I wanted to be able to take that video. And then process it and watch it and it's just like a bunch of unnecessary repeat scaffolding to do whereas for run if it just like it understands videos like built around video.

So you say, hey, watch this video and slack or anything like that and it just works. And one of the things that we are hoping to do. Is that.

“I get into a few of these fights on on on Twitter because people somehow think that what I'm saying is AI is the most important thing.”

And what I'm actually saying is that AI is the least important thing. The thing that matters is being a really good marketer and what my hope is that. I would love to level the playing field of the people on my team like Alicia, for example, who aren't necessarily the most AI-pilled.

Most of those people on marketing, right? And so my view is that like we need to abstract away all the scaffolding, all of the like technical engineering stuff.

And let marketers just be marketers, right? Like that's like that's the ultimate vision for run if this is to actually level the playing field to make AI not be an edge for anybody.

“Right now like the person who knows marked down files and knows how to set up a Mac mini, they have an edge against the marketer who doesn't, right?”

But that's not necessarily the person who's the best marketer. So to me it's like, how do you, how do you just make this stuff work out of the box so that even the people who are good at doing this? The real leverage is just being an exceptional marketer. Like that's what the job is, right? At some point you're like, wait a second, like I was trying to do marketing work. Like here I am spending three days like debugging my Hermes agent on like the infrastructure that's running on like why am I doing this again? Like why was this useful?

And I think it's useful to do from like a learning standpoint. I think everyone should push themselves and learn and tinker and play with a bunch of stuff, but from like a productivity standpoint.

And surely we shouldn't all be like wrangling with the scaffolding and infrastructure. Like it doesn't, it's it's not a good use of time in my opinion. Especially because a lot of the people that we've spoken to that are really like knee deep in it, like setting up their own virtual machines and whatnot. In many, many cases like we've gone way further than them even than like oh wait like yes, it's going to take way too much time for me to build that level of infrastructure so that it works with my team and like permissions works and everyone has their own swim lane and like it just works out of the box.

And I think that's the difference is that we're building, we could when we built run if it was such like infrastructure bet that we could take run it to market is just like a cloud knowledge work agent for like general purpose stuff.

“That's how we use it internally for like all kinds of things like our sales team uses it or CSMs use it.”

We're like the biggest run if users in the world. We spent something like 70,000 dollars of AI cost through run if ourselves.

But my view has always been a lot of our team was like hey we should we should pivot this company to just be like general purpose why why limit it to marketing and I was like yes technically we could do that. We could build the best product in the world for marketers if we stay focused on that and there's going to be plenty of general purpose knowledge work AI that. That I'd I'd labor I'd rather stick to what we love and what we know and like build build the best product in the world for marketers.

So I just I think those two points. It's important to make that distinction a little bit because I totally agree about like the virtual machine component of it and like the the like raw general purpose infrastructure component of it. We explored this idea a lot and it's like yes, should we be working in get hubs should we be using Google drive like do it or like Google drive sink down to our local computers like we're trying to figure out this way to like how do we collaboratively work with the same files and compound learnings across team them or things like that run if out of the box fantastic.

“And that's like the general purpose component of it.”

But I also when we talk about it being a great vertically verticalized tool around marketing I feel like it comes back to that data component that like motion was founded on where I'm like. And I said this all the time actually our performance creative team is on the forefront of personal testing messaging testing content testing like we are serving billions of impressions and learning a lot at a really high rate. Nobody's better at like figuring out what those concepts are tagging that down and building the dimensions that we care about the motion.

Or I should say you guys generally as a company and that's really what sells me on the idea that. Not only is a great collaborative general purpose tool but also this like. Add creative beach head is the way that it becomes a great vertically vertically integrated you know marketing tool. Totally yeah and so like you know there's. There's all the data that we prepare that we do scroll that we watch that like is just available for people to consume but one of the things that's been really interesting for us is to like.

Help people see just how far you can push run it's like based on like all the stuff that it has and the fact that it could run 24/7. You can get run it you know this is one of the use cases that people do is like oh it'd be really great. If. I could build an agent that can go and watch all the latest like TikTok organic videos and then like summarize them and try to look for trends and like. The the scaffolding that it takes to go and build an agent like that is like it's not rocket science but it's not nothing either like build that and maintain it.

With run if you're just like hey every night I want you to go watch organic TikTok videos and do this that and like just explain what what you want.

Um us pre-processing the data and like making it available to you but then you being able to like.

Have an agent that lives in the cloud that has its own machine which gives it a lot of power and autonomy and you can give it instructions to say like from now on I want you to go and watch like. Every video from this creator and like just give it so much work that. It can just do out of the box so that basically what our hope is that if you have a task for run it that's like marketing specific. In the age of AI like okay I want this agent to do this task. And then you're like okay I'm going to do these like six or seven scaffolding pieces so that I can get it to do that task like watch organic TikTok videos.

Our hope is like no just tell it to watch organic it it understands what you mean it has done the prep work so that when you give it a marketing specific task it could just run with it because.

That's kind of the difference of making it focused on on on a specific vertical is that. You can give it a task that is related to that vertical and before you have asked that question we have prepared run if. To be able to tackle questions like that really well and obviously like there's a lot of these types of tasks are constantly like building more and making it. Making it better but just the current capability set of what run if can do is like we have like people can spend tens of thousands of dollars per month on.

“So it's just like putting run it to works it has the infrastructure to do that. Which is the other thing by the way I think that people are there there's like a mental model of.”

What is the correct level of AI consumption that we're all going to be doing as companies and I think people have not priced in this idea that.

We are going to be a thousand xing maybe 10,000 xing the amount of AI consumption that we all process and. And I think most people have not realized that yet and I think once you get to the point where you like. You know there's this funny thing when you buy a Mac mini and you step away you have this feeling of guilt where you like. I should come up with something for this agent to do while I'm away like it otherwise I've got this machine that's unproductive so like. Similar idea to running a factory imagine you have a factor that's not running half the time.

“You're like well this is not productive like we should find a way to like make this factory productive around the clock and I think that's the.”

Which that flips with AI where. It's really just your imagination in the way of saying like can we get AI to be productive 24 hours a day and if not why not like what what what what are we missing to not have this like super intelligence. Be productive for us around the clock why is it that it's just sitting there this like super intelligence just sitting there until I show up and ask it to do something no way that's like there's no way that's the right pattern. And I think once you get to that level of usage it becomes like oh okay yeah there's actually a lot of scaffolding that I need to do in order to have like AI be productive for me and my team.

24 hours a day and have that just like work smoothly. You know the difference between hitting your numbers and missing them clear signal on what's actually driving growth it can get really really noisy. So much noise you got platform data you got blended data MMM all the acronyms MTA experiments all of it all pointing in different directions.

“The more you're spending the faster you move the more the bad signal can cost you that's why we use house and we've been using it for years that's why the other marketing operators do as well.”

They're the best tool on planet earth for measuring what we call incrementality we talk about a lot on a podcast what is the true impact of your advertising dollars on your business. We have causal MMM for channel level budget calls causal attribution down to the add level and architect their AI agent tells you exactly where your next dollar should go. And the results speak for themselves dot kick saw 41% lift in I rose using house and you're not stuck with a help desk you get an embedded measurement strategies who actually helps your team make better decisions.

Their whole team is great. We've worked with a lot of them. They are world class there go to house dot IO slash operators HAUS dot IO slash operators and start backing your budget calls with real causal data. We've got a few of a con of so you're talking to two guys have minis and so know the pros and cons of them I it is very convoluted especially when you're like. When you're trying to set it up and you're not home if you have it at home you're like I have to log into this or whatever that's like the worst I have to like save things log in but Connor how's your money doing are you are you using it are you.

You get into it. Yeah, it's a good question I I've experienced all the you know trials and tribulation to get solved by running where I'm like I shouldn't be doing this I.

And it doesn't always work I had to go to Sonoma last week so I was like offline for a little bit so we're still doing a lot of our work where at the point where it's like collaborative in the sense that we're using.

“Like my performance creative team is using runneth right so we're like all working in the same workspace theoretically.”

I'm using Claude and then I'm using the runneth CLI so I'm picking runneth from there and then any sort of like yeah recurring tasks I just have scheduled and run through my computer so.

I'm excited about somebody else solving the virtual machine problem and not me trying to plug in Mac minis at home. Yeah, you know part of it is what when I when I built the analyst agent on the mini one of the issues I had was that some of our team would be like. Oh, could you get analysts to do this could I ask analysts the question it was wrong could you like can you can you can you ask it to do this. And all I would do is like I would take that screenshot and put it in my turn on we're like hey analysts like fix this like add this or change this and at some point I'm like why am I doing this like why am I like triaging all of this and when we switched to runneth inside of slack.

Because the runneth you speak to in slack is the same persistent agent. Like anyone can program it directly from slack to be like you know from now on change this or do that or like update your thinking and then we're like oh crap now anybody can program this thing.

From slack and get it to do anything and then we face people like bumping into each other and someone overriding somebody else's routine and so I'm like okay now.

We need runneth to understand when someone talks to it from slack it knows exactly who you are and it knows what you have permissions to do and what you don't have permissions to do. And so then we had it where someone was like. And then we had it where somebody else on my team built a routine and someone was like hey can you make this change and runneth we go back and forth with the person like what do you want it to do and so on. And then it'd be like okay this routine belongs to resa so like it would at mention me and be like resa give me the give me the go ahead and I'll make this change and I'm like yep go ahead do it.

Going from like me being the single point of failure for everyone's like future across to come me to then anybody can program this thing inside of slack to the chaos that comes from that to then basically like.

Collaborative programming of an agent all together in slack which is completely wild if people want to stress test this by the way we.

I started our slack community after the creative shudji boot camp and it now has like 10,000 members in it. I hope after this people don't blow it up. But we made runneth inside the slack channel. If you go to runneth in in the slack channel and you create your own channel with just you and runneth we've given runneth permission to like do crazy work for people like no token budget just go nuts. It'll have the memory for you and you can add stuff to it you can build routines you can like do whatever you want just from a slack channel inside our community and.

It just kind of shows the level of scale that runneth can go out like you have thousands of students that are like attacking it from different channels and it's able to like keep itself organized.

“And so yeah I recommend for people to try it inside the slack community if you want to get a taste of of what runneth can do made it available for people and and students are hammering it right now.”

Just be careful it's not it's not fully private obviously so like you know just do do work that is that is not sensitive but you can like do competitive research you can build like a dashboard of creators and like all of that stuff without. Without needing a motion subscription or or paying anything so feel free to try that out. So I've got a question for you I was having this conversation over dinner recently. And what this person was proposing was that there are only two companies now there AI native companies which he was proposing are basically exclusively new companies that you have to start from the ground floor to be AI native right now.

That's kind of what Cody's on right now building agents Cody whatever he's doing next is going to be like a single person nine figure business that that's where we're going there. I see a person I can guarantee that part. Yeah I can guarantee that. But but like there's that path it feels like to be truly AI native people who are at. In an advantageous position or the ones that are just starting and then everybody else is a legacy company and what they were also proposing is that.

Legacy companies are it's going to be extremely difficult to sort of shoehorn AI processes into your business and become AI native or to get the leverage that some of these more native earlier businesses have.

“I don't believe this I think there's totally like there are examples of.”

But I'm curious where you land on that scale reservoir for these businesses that are fifty sixty seventy people and are just trying to figure this out for the first time like what are some of the approaches.

Okay so this is this is where my view of the executive I see and we'll see if this works I'm trying to re-brand your job Connor from CMO to chief marketer is is is my aspiration.

“And I think it's related to this question I have thought about this a lot and I think.”

The leaders at these companies. They behave like ICs like these super high leverage ICs and then I think those companies can all transform the problem happens if if the leadership tier is like okay.

We're going to do like manager things and then we're going to tell our team hey team adopt AI is it is never going to work with no way.

I think if those companies no matter how big they are actually doesn't matter their size if the most senior leadership start to behave like super ICs where they are leading from the trenches they're like building systems with AI they're like. They are modeling the behavior that they want to see everybody in the company do. And then I think those companies can absolutely transform to be to be AI native but I think it takes a reimagination of what leadership means. I think the entire entire job of leadership is like fundamentally changed post AI because.

The job of leadership was to create leverage right that's the only reason we have leaders is that someone's really good at doing something let's say like. You know, Connor is the best marketer in the org and then people like hey this guys are really good marketer like we need to like clone him or we need to we need to if we could have like five more corners that would be terrific. Can't do that there's no clone machine so okay Connor you run.

This department to try our best to like clone you right ultimately some version of that.

“But what happens if Connor can create leverage in other ways what what matters is like can an individual who's really good.”

Can they create leverage and I think I think because you can do that with AI. The job of the leader is like the person who can create the most the most leverage in the org. But I think if if the leadership tier adopts that then I think they can totally adopt to be coming AI native and that's really all the difference that you'll see with the earlier stage companies that are becoming very AI native is that the people who are at the. positions of leadership even though it's a small team they are all behaving like ICs and then I have this argument a bunch of times with a few people who are like.

Okay and therefore we're only ever going to have five person teams. My counter argument to that is that. Suppose for a second somebody came to you and said. I have a staff of a hundred people who are all exceptional they're all like 15 out of 10 at what they do they're at your disposal. Make them productive.

Like a really talented leader will be like hell yeah I can do that like you're telling me you're going to give me a hundred people who are really good at marketing and engineering at like sales. Of course you can put them to good use and like create even more value than you could with five people so I actually don't necessarily buy the. Lean team argument as a like necessity because if you can create value with five really talented people. Then what exactly is the is the limiting factor that you can't create value with like 50 incredible people.

The problem is like when people grow their teams they're not always like hiring incredible people and I think that that's where the issue is I think there's there's like.

When people talk about the impact on like the job market and people losing jobs and whatnot I actually think there's going to be more opportunity than there ever has been with AI.

“But unfortunately you can't just like clock it in that's just not going to work like you you have to take your work seriously you have to be like good at what you do.”

And given that there's like unbelievable opportunity and I think that's that'll be the difference where yes you don't need to grow an org. Of people who are not that great but are just kind of pushing things forward slightly I think that that will not make a lot of sense but I think you absolutely have like large orgs filled with really talented people. So yeah I think it's possible to become AI native even if you're a larger company just the the leaders have to go first and get in the trenches. Quick gut check for the operators listening if you're spending on TV or CTV today can you actually say what it's driving incrementally.

This is exactly why we work and have worked with neon pixel at hex cloud for the last three years we've grown with them a lot over time in CTV has become one of our top growth channels they help us treat premium living room TV like a real performance channel.

Incrementality is the true north star on measuring channels and they're really measurement that agnostic they don't force us into their own black box dashboard they work inside the measurement systems we already trust and help us understand what TV is actually doing.

“If you want to check it out go to neonpixel.co and ask them to design a controlled CTV test on your numbers.”

One I love the discourse on X this week around your executive I see post it like ruffled more feathers than I was expecting because I read it.

You and Taylor Holliday are saying the same thing and I'm like I like both you guys I'm like dude we're all on the same page here I'm locked in on executive I see and I've just been saying. I feel like I have been so much closer to the work over the last couple months like even this past Monday just in like a few hours of work I did like. I just got so much deep insight across the org in moments when in historically I would have had to spend half my day plug in through intelligence trying to figure out what this AB test is exactly doing.

That's always my answer where this this person that I was speaking to is like legacy companies are cooked they're like it's over for these guys are not going to be able to adopt a I you're going to be they're going to have circles ran around them buy these more. Smaller lean team AI native companies I think it's an 80 20 thing we're like you identify the I don't even know if it's 20% of the workforce it could be 5% of the workforce and if you can get leverage around those individuals they'll help empower the rest of the team without everybody being as AI piled as as you know this core for you.

“And I think you end up getting just as much leverage as you need to be to continue to be competitive.”

I I totally agree and I also think that the the trenches work right now does involve a bit of AI piledness.

But I genuinely think that part just becomes solved so in the same way that in the early days of like the internet you'd be like hey you need like internet skills in order to do these like new types of jobs and but like now we don't. Think about it that way more it's like everyone knows how to use the internet everyone knows how to use like the basics of knowledge work and then the difference becomes. Are you good at whatever craft you're hired to do are you a great marketer are you a great engineer are you a great designer are you a great product person.

I think those are the skills that are actually going to matter and so if you have these legacy companies. Who are staffed with. People who are exceptional at the craft. I actually think in the long run they beat these like AI piled native companies who think they can like shortcut the craft by saying oh I don't need to be.

A good like product person I'm just going to tell a claw to make all my like product decisions but then they're going to have a software product with just like a million widgets on it.

And then what they realize like being really good at product means like deleting things how you're going to do that how how do you know. What to strip away and what to keep and that takes like really good judgment.

“And so I think the the best positioned people are the ones who were just really good at whatever craft they're doing because.”

The underlying AI piled layer is just going to get easy like we are not I remember when I set up open claw. It really felt like I was diffusing a bomb. I was like this is scary like this thing is I like I send it to our CTO and these like okay listen try it but. Be careful like this stuff is this stuff is dangerous it could really get hacked and whatever. But there's a basic level of it though like I don't think that anyone really needs to be doing that but like there are like I feel like sometimes we we forget like from.

Talking to other people doing this for John so like there are people that like even like a skill is a new thing like there I do think there's a basic level and and this is what I told my team and this is what kind of what you just said like. It's kind of like no one thinks about using email or internet or I like that's how we're going to think about AI in a few years maybe even now and it's just like a basic skill it's not less like advanced thing it's like this is just how work will get done. So I do think there's a basic level maybe not not you know 90% of the way they are cracked like you got at least like know the basics to be able to get the leverage.

I agree and I would I would actually say the skill that is important there is the skill of being like curious and high agency right people who are curious and high agency like they can't help but like. Go and poke around and be like what's going on here and like what like what what is happening and I think that if if they're not doing that there's something wrong at the level of like curiosity and high agency that is not you know pushing them towards this like obviously interesting thing.

Individual contributors the best leader is the best marketers like there was there was no there was no like it was a it was a crazy coincidence like there was a complete correlation just being able to see it like somebody would just like want to do things want to move faster want to like improve their department. And we'll just figure out how to use turn them on all how do you like they didn't know it at all but they're just like hey I want to go figure this out because I want to get this thing done and you know I'm a director but now I can go and do this test or whatever it is so I completely agree with that.

Most brands treat customer support as a cost but with AI every support conversation is now a chance to increase retention recover revenue and grow your LTV rich panel doesn't just use AI to answer tickets they allow operators to run revenue experiments inside your support you can test different windback offers on refund requests. Subscriptions save offers during cancellations and product recommendations that turn chats in details plus they guarantee you're going to save money 50% of your support volume automated in 30 days or your money back that's pretty good offer if you ask me.

So at a rich panel dot com slash demo they'll build an AI support team live for your business and show you how it can drive incremental revenue for you.

One of the things resume you say people that are good at their craft benefit from this.

“I think I'm going to try to segue your point into my point.”

People that are really good at their craft are going to benefit from it over the long term because they'll ultimately be empowered you brought up the example of the someone building a piece of software. So if it feels like what a lot of people are solving right now is just the ability to write code. But the people that are going to succeed long term are the ones that understand what problem are they solving with the software thinking about it from a product perspective and don't end up with a bunch of random widgets they end up with like a beautiful piece.

Beautiful product itself is a problem for people where I kind of feel like this relates to marketing right now is people are solving an opx problem when at the end of the day everybody has a cac problem. And many examples of building leverage so that you can free up more time make better decisions and solve a cac problem. But it feels like a little bit of what we might be doing is saying we're shifting the goal post and saying how lean can I just do all the motions of being a brand. When in reality it's like how are we finding fundamentally how to acquire customers at a higher rate and that we're going to maybe make that transition we have to make that transition again.

“Post is like AI adoption wave totally you know one of the things this like I think to be a marketer is.”

Genuinely one of the one of the rarest and most difficult skills in the world and it is the skill that will survive post AI if everything gets automated. The two things that remain are like starting a company and marketing the company. I don't know if maybe maybe maybe AI is going to like consume all the information and run it but I think people are going to want to hear from.

The exceptional ones are like doing the work so yes I think I think the talented people always have that avenue as well.

But like marketing is. Marketing is a very special skill set because the job itself is high leverage like if someone is good at marketing. They create leverage on the company basically means like that their work can have such massive impact in terms of value creation. And it's similar to software in that sense like one software engineer could write software that impacts a lot of users. One marketer could do work that like creates an incredible amount of value for the company.

But I think the job of the marketer is. A value creation job it is not like a private equity let's look at the process is here and see what we can slash and whatnot like it is. Figuring out what can you do that is going to resonate with the world and. And so it's a very very rare skill set in my mind and so like yeah the people who are not great at the craft. In the same way for like software people are not great at the craft like oh great we have like a free code machine let's like point this to do.

Create like tons of code same way like yeah you can get. You can get AI to like build you. Hundred landing pages and a thousand ads and like great wonderful like now what what is that. Is that going to be helpful is it not so the bottleneck really does come back to the people who have really great.

“I think there's a backlash on the word taste so I won't use it anymore.”

The word judgment seems to be landing a little bit better and I think.

I was going to say you're like going for like the tech bro bingo card you got first principles in here.

We talked about context layers we talked about taste like we I think we're getting like three for five probably three. Yeah well you know it's like. It I think memes are mean for a reason you know like there's there's there's a reason behind it and so.

Where's the bottleneck you know it's like it's the judgment to decide like what's good what's bad what to do what not to do.

Yes people can go and ask AI hey what do you think right and like outsource their judgment to AI but there's no edge there because everyone can do that. And the whole job of marketing is to it's a competitive sport and the job is to find an edge against somebody else and so if everybody has the same like intelligence machine then. You need the people who can maintain that layer of judgment on top of it and like man that's a rare skill which is why these individuals who are good at that. In the years ahead I suspect.

The world has not priced in this idea that. What does it mean for a single individual to be as potentially create as much value as like. Twenty five people in the past. It must mean.

“The that eventually gets reflected into the comp of that individual and I think the comps get ridiculous over time not just from the sense of like hey I can automate the work of 25 people it's like I am.”

Creating as much value for a company.

As potentially 25 people. What's the salary of this individual how do you how do you compensate that individual it it's. It's not clear to me that the salaries will stay the same at all like I think I think this is why I say that. AI is going to have unbelievable amounts of opportunity for a single individual who's good at stuff. Can probably earn the salary of like five six people.

Alex Lieberman from like morning brew he's got like 10 exes his company have you heard it. Yeah. I don't know. I've seen a comp structure I've seen I've seen his business but what the crazy I don't I'm probably speaking out of us but it's like they have like a bounty system almost so like an engineer can just like pick what they work on. And just they like you can make up to million millions there or you can make you know normal engineer salary by just like seeing a task and like yeah I'll go and do that and so you can kind of like choose your own value that you're creating for it and they have like a completely like no cap upside.

It's pretty unique and I wonder if something like that in marketing would ever be you know helpful.

“Yeah, I think marketers are going to be among the highest paid individuals in the world very very soon especially as like.”

Software engineering becomes like cheaper and cheaper to like. Produce stuff and like just in general execution becomes easier. So like to me that the the the the chief marketer idea is not a step down for the executive like all I have to go do IC work again I think. The amount of value that talented ICs can create. They will be compensated the same way as like teams used to be so like we're I think we're entering a very interesting era and it's going to be a lot of fun anybody who says like doom and gloom with AI is like.

It's just going to be a complete blast and a lot of fun. This is great news for all marketing operators listeners you heard it from Resa where I'll get in raises. You are in the right industry because think about it if if if you're in something that's like automatable right that has like a a closed system. Then like AI is really good at that so like the mathematicians might be in trouble because you know there's like it's it's a very verifiable discipline.

But marketing is competitive it's dynamic the world changes culture changes and like there's always going to be a bleeding edge.

“And that's why I think it's like it is it is it is genuinely one of the best places to build a career right now and.”

But it's because of the value that you can create on top of AI so don't get pulled into the to the AI shortcuts like that's not that's not that's not where the value is from marketing it is like. Read the old school books like read David August be try to become an exceptional marketer without AI and. And the leverage is just going to start like showing up in your life and it'll all just work when it comes to AI adoption there's a lot of like. Rules of thumb of like what's what's a good way to look at it.

One of the ways that I think makes a ton of sense is thinking about AI is a percentage of payroll and. I actually made this comment to somebody was like I think people should have a hiring freeze until they are spending at least 10% of payroll on AI. It's like. The mental model I think is like that starts to be the right way to think about things is that. AI is a productivity on your on your labor force and so like.

I think we're all understanding by a massive degree on AI and the cost of AI are going to fall off a cliff. But I think cost of AI are going to fall off a cliff and we're all going to spend like 10, 20, 30% of AI on payroll and so there's like an order of magnitude thing of.

That's a lot of money and you're like wait, how am I even going to like spend that and I think that's the right kind of questions to be like.

“Yes, that is the order of magnitude change of the way that we're going to be deploying AI.”

I'm saying 10% Jensen from Nvidia said that it should be around like 50% and so like somewhere in there is your somewhere in there is your range.

Got it, so marketers making up you know 60, 70, 80% of payroll AI doing another 10% like that.

Yeah, that's the way that's the way. Awesome, this was good one.

“Reset, thank you so much for coming on where can people follow you.”

You got a lot of hot text recently, so tell people where they can follow you.

Where can people, you know, try try motion, where can people try running, what's going on there.

“Yeah, so I'm I'm trying to get into more fights on Twitter so you can find me there.”

We are also creating a lot of content, either motion app.com or runeth.com, a lot of our content these days are going to be around like AI and marketing so you can follow us there. And if you, if you go to runeth.com, you can find our Slack community and join it and just start talking to runeth and like put it to work. We have given runeth no token limit inside of that Slack community for our students. So go and give it a try and put runeth to work. Awesome, and this is great.

Thank you so much for coming on. I think people are going to love this one.