Loading summary
Transcript83 lines
- [00:01]
Keith
If you are like me and you are sick of just prompting endless ChatGPT and Google Gemini to get an image or a video and you're just in that death spiral of making cool creative for your company, this is the show for you. I have the best expert in the world, Lore, who is the CEO and co founder of WeVee. He's joining us to show you some mind blowing workflows that are going to change how you build creative for your business. Let's get to today's show. Before we get into today's show, here's a quick word from HubSpot. Cutting your sales cycle in half sounds pretty incredible, but that's exactly what Sandler training did with HubSpot. They used Breeze HubSpot's AI tools to tailor every customer interaction without losing their personal touch. And the results were pretty incredible. Click through rates jumped 25%, qualified leads quadrupled and people spent three times longer on their landing pages. Go to HubSpot.com to see how Breez can help your business grow. I'm so excited to have you on the show today. We met a couple months ago and took a little while but we got you here to talk all about AI and creative today. Welcome to Marketing against the Grain.
- [01:30]
Lore
Thank you so much, Keith. I'm super excited to be here with you.
- [01:33]
Keith
You know, I'm excited to have you because you know, my meeting with you was one of my favorite meetings of 2025 because we just got to jam and debate and talk marketing and more importantly like how AI is changing, how marketers and founders create and tell stories and build creative assets. And I think part of the challenge historically has been it's like it's been very time consuming and very expensive to make really good videos, really good ads, really good website designs, product imagery, all of these things. Social carousel, social videos, all of these things. And you started this company, we AI. It got acquired by Figma and you were out on the journey to basically solve this problem and bring creative to the AI world and help make it easier and faster and better for everybody. Can you kind of just start off with like what have you learned in this process? What have you learned in building weavy in terms of what AI is really good at when it comes to creative work and where I really needs like a lot of human help and support.
- [02:39]
Lore
So I'll start saying that it's not easy. So one of the key things that we believe in and we started with it, is that it shouldn't be easy, but it should be much More robust. So kind of like there is a huge promise there of creativity, of speed, but I think that too much of the focus was at the ease of it. And it's not so easy. Like, I think it's. You still need the professional people involved, but their craft, the way they do things, have changed, so they can do much more. So the way we look at it is that the craft is shifting from kind of like making pixels into building processes or building systems that will guarantee much more speed and scale. And we've seen the ability to leverage all of these capabilities that this technology is able to offer, but it usually doesn't come without this massive trial and error and learning curve. And so I want to start by putting that on the table and then also saying that models themselves are kind of like, they're good at a lot of things, they're not so good in a lot of other things, but they are improving so fast. So whatever you can look at and say, oh, AI cannot do that. Okay, wait a couple of months, it's going to be able to do that. Right? So it's moving really fast. And you should be building for the future. So if you're shifting the way you build, the way you work, then you're always future ready. You're always going to be the first that is able to leverage these new capabilities that are always coming in.
- [03:55]
Keith
Well, I think that's going to be a big key theme of the show today, which is like, how do you evolve how you build and how you work in this AI world, especially when it comes to, like, creative work. Because I don't know, I've experienced, I've talked to tons of friends and customers who've also experienced. It's like, well, I'm trying to generate this image in chatgpt over here. Nana Banana over here. I. I did this VO3 clip, but, like, I couldn't get it to kind of sync up and work and it all lives different places, which is kind of hard. And, like, I have to retell it all the time what my brand colors are, what my brand style is. And so, like, I end up just iterating for an hour on something that isn't that good. And you become disenfranchised with the process. Right. And I assume that you're nodding because you talk to more people than I do on this topic. And so what we want to talk about is how you actually build processes to get past that and solve that problem today, right?
- [04:49]
Lore
Yes, definitely. It is a lot around, kind of like switching your mindset from trying to iterate on like an endless iteration to get to something that works, but it's never good enough. Because even if you reach the thing that you want, now you need to change it. Now you want to do it again, now you want someone else from your team to do it. Or you're a designer and you want the marketeer to be able to do it and you're stuck at the same point. So I think that the real mind shift is in switching from kind of like trying to do the same thing of like, I'm focused at creating this single asset into building a system that can then be reused that it's robust enough that you or your kind of consumers, like the marketeers or whoever it is that consumes your creative work, is able to then reuse your system.
- [05:31]
Keith
Part of the reason tools like we've and I've used we've and a bunch of other things are so important is because AI also lets you get a lot of different iterations and a lot of different versions of things. And sometimes we haven't. I think as marketers evolved our mindset past the like, well, I'm going to take a long time to make this one thing. No, make a lot of things to start. Pick the really good ones and then refine them. Hey, real quick, if you're getting value from this conversation, you're going to want to check this out. We're sharing Lore's exact YouTube thumbnail. A B testing workflow, the one his team spent a full week building plus seven more creative AI workflow blueprints and over 20 ready to use prompts. The goal is simple. Turn hours of asset creation into seconds. Grab it now. Hit the link in the description. Hey Lor, people are going to want to know can they go sign up for wevee? How much does it cost? Like give people the logistics before you show them cool stuff.
- [06:33]
Lore
So we is a self serve product anyone can go into Weave start with our free tier and then paid subscriptions are starting from $19 per month. And what it practically gives you is access to all different AI models for mostly around media generation and editing alongside classic editing tools which we'll see in a second. And then the ability to really build and scale creative systems which we'll walk through that as well. We were acquired pretty recently by Figma, so now we're part of Figma, which we're super excited about and we'll probably talk about bit as well.
- [07:05]
Keith
Sure.
- [07:05]
Lore
And yeah, I think that's as an intro.
- [07:07]
Keith
I'd love to walk okay, that's perfect. So anybody can go and use weavy. There's a free version if you want to go test it out. Then the paid version starts at 19. I think what's interesting is how you framed it is that there's a workflow component, there's an editor component, and then you also get access to a lot of the different AI models because certain models are better for certain tasks. Okay. With that all out of the way, now walk us through kind of the example that you have here.
- [07:32]
Lore
So let's start with a really simple and naive way of using weavy. So WEAVY is a node based editor and that means that each of the boxes that you see here is a node. It can get an input and it shoots out an output. Everything goes from left to right. So in this specific case we're starting with a prompt node so it can get a text as an input and we're putting in a text. And that text is kind of like a guidance for text to image model to generate an image. So you can see how this one node is connected to multiple different text to image models. I can choose and run all of these different models again if I want another version. But you can see how I have here Google's nanobanana and Reeve and Mystic and Minimax and GPT. So practically any major model that's available out there is going to be available in Weave. And one of the fun things you can do is to test them all together to do the same task. So the most basic one is generate an image from a text. So that's what we're playing around here. And you can see how each of them does something that's pretty different.
- [08:25]
Keith
Yeah, very different outputs from the same input. Exactly right. Which is also very emblematic of the good and the bad of AI. You know, generative AI. It's not going to give you the same repeatable output all the time. And this is like a perfect example of that.
- [08:39]
Lore
Exactly, exactly. And then what we're doing right here, again to showcase that a bit farther down. So taking one of these images and putting that as a first frame for multiple video models. And you can see how each video model added an octopus to the mix with a prompt. And you can see how each video model will do quite a different interpretation of that prompt. And then again, it's really like the most basic way of using weave. You just kind of comparing different models doing the same task. So that's fun. And kind of the reason people would be excited about that is that you have everything in one place because things keep on changing so fast and you don't want to change your subscription every couple of weeks. And if you're a big business, that's not. It doesn't take two weeks. Like you know that, right? It takes like months until you have things approved. So kind of like having everything in one place, it's really powerful, switching into something that's a bit more complex. So we're starting again with the same frame we had before. But the idea here is that you have all of the classic editing tools that professionals used to have. So our belief is that there is no reason to give it up. Like in so many cases you want to manually control things, you don't want to prompt your way through them. So for example, if I want to create the right depth of field, I can use a blur node and I can control it manually. And I'm not using any generative AI model, not spending any credits. But also I have full control over that. So kind of like isolating each of the different layers with masking and then editing each of those and then using a levels node to do color correction. So to get it exactly to the right coloring for myself. And then once I have the perfect image, then I can create the video. And so the idea is that there is no reason to give up control. And I think that's been one of the biggest challenges that creative have faced is kind of like you give up control. As you said, you do like endless amount of iterations and then if you need to go back and change something, it's like all over again.
- [10:19]
Keith
It's the false binary that has existed in the early days of AI where it's like, oh well, you can go and build this creative asset in these professional tools or you can go generate this image or video over here. And they're kind of very disconnected and it's like, well, they have to come together. And if you are a professional or even if you're a hobbyist and you know a little bit, because you could go and basically get feedback on a given image from an AI model, right? That would tell you, oh, we think you should do a color correction, we think you should make these changes. And what you're doing is trying to make it very accessible to make those changes.
- [10:56]
Lore
Yes, exactly. So for example, like that layer based composition is like so basic. Right. But you just don't get it in generative media tools. And trying to prompt your way through kind of like move 2 pixels to the right, like that doesn't work.
- [11:07]
Keith
As somebody who's trying to do it, it's so hard.
- [11:10]
Lore
Exactly. Now switching into kind of like the system thinking. So I was talking before about creative systems. So here's an example of kind of like a systematic thinking. So that use case is like, I wanted to take an image from a scene and create new angles for this scene. So what I'm doing right here is starting with this image, using an LLM to analyze this image, and then to come up with 10 new ideas for new angles for that specific scene. So we have this person with a pink teddy bear in the desert, and not all of them are perfect, not all of them are consistent, but you get enough good outputs from that, and you can actually create all of these different images. And the thing is, once I've built this new angles machine, all I need to do is change that image and I can run this entire thing again and I'll get completely new results. And so the idea is that you're building your own toolbox. And once you've built your own toolbox, you can reuse it again and again. You don't need to start from scratch every time.
- [12:03]
Keith
Yeah, the big thing here is we're trying to move from the novelty of like, oh, you can enter some words and get an image or video, which is cool and awesome. But it turns out if you actually want to apply using those things, there's a lot of stuff that needs to happen. You know, you need a lot of derivations of them. You need to do some polishing, you need some editing. You need to make the. Make sure they're in the right format for the actual in place that they're going to live.
- [12:27]
Lore
And.
- [12:27]
Keith
And when you talk about building systems, that's really what you mean here. I think it's like, how do you actually do that in a way that from end to end, you can go from I've got this idea to a good final, polished output in the right format to go with you.
- [12:41]
Lore
Exactly. And so I think this one is something that a lot of people can relate to, and that's about kind of like starting with your own style. So that's about the illustration. But we're seeing a lot of brands building their own kind of style guides and guiding their generation based on that. So it can be imagery. In this case, it's illustration. So this workflow starts with these four illustrations that are in a very distinct style. It could be like a specific brand that does that and allows you to put in new guidance for what you want. And that's enough so let's see the output of that. So we wanted two characters shaking hands and you can see that the output is like the exact same style with new characters shaking hands. Now the cool thing about that is that once I've built a system I want to hand it over to other people. So I want marketeers to be able to now they have a new idea like there is an event happening and they want to create something, something immediately. They don't want to open a brief to the studio and wait until it arrives and everything. And the studio has actually already kind of outsourced that and allowed the machine to do it. So the studio has built this machine and they don't need to be involved in that anymore. So one of the things you can do is to turn this complex node based spaghetti into what we call a design. So that's a simplified interface and in this case the only input that we want is that what they want to see in the image. And you know that the output is going to be. So I create a few for the podcast and, and I created this Gummy bear watching Netflix.
- [13:57]
Keith
I like the gummy bear, he's good.
- [13:59]
Lore
And then that's the only thing you need to edit and you know that it always going to turn out consistent. So that's really the power of system building and the power of weavy, of expanding beyond that single person that's able to control AI is how do you turn that into a capability that the entire organization can then use and so you can think about your own kind of like brand style guide and what's important for you as a brand and turn that into your own brand machine with all the nuances.
- [14:24]
Keith
I think what's interesting about this, right is that you can have a specialist build a system. A system requires a lot of like basic understanding and know how some real experience and it takes some trial and error, right? You just don't build a system like you just showed overnight, take some iteration, but then you can expose it to people who don't have all that knowledge in a really simple way. And I think when you talk about building systems that's like what I think about because it's like right now people are in ChatGPT or Gemini or whatever and they're doing their own system and right. So if you've got 10 marketers, you've got 10 marketers doing their own system and it's gonna be really hard to be consistent, to be clear, to have like really compelling output. If everybody's doing their own thing, nobody's learning from anybody and everything's all siloed up. When you bring all that together and you're working, like off of one base system and then just like, cool, you can bring your individual prompt and instructions and vision here, but know that the output is going to be along the lines of what everybody else is doing.
- [15:21]
Lore
Exactly. And I think when we started, we're wondering if it's going to be kind of like, oh, we only need, like, whatever, 20 workflows and the entire world is going to use them, or are there actually going to be that many niche cases? Yeah, and I think it's the latter. Like, every big customer we're landing with, we are seeing them building dozens of different workflows, and they have like, their own kind of like, image expansion, their own image to video. And it's all built on what's important to them, the way that they work, the way that their brand is built. And that really makes the difference between, okay, I can get to a video that's like 80% and I'll iterate on that 100 times, or I'm going to work three days or a week to build this workflow. But once I've built it now, I can use it, like, endless number of times. And it's really helpful for like, 300 people that are now riding on that. And I think that that has been a huge unlock for marketing teams and kind of like changing the creative to be, instead of a bottleneck, to be an enabler. So they're building systems and the entire process of kind of like asset generation has been streamlined.
- [16:20]
Keith
Yeah, I think that's right. Making creative is hard. It's real work, and it takes real depth of thought. And businesses and brands you love don't get there by accident. They get there through a lot of hard work. It's just, how can you now scale that hard work? And the expectation is that that creative is going to only get better and more personalized and have more iterations in the future. And so you have to only do that by scaling through systems. Oh, I see a YouTube thumbnail flow. I love a good YouTube thumbnail.
- [16:47]
Lore
Yes, you're going to love this one. So this one starts with bringing in thumbnails. And the idea here is that once you've built that out, all you really need to do is kind of like throw in a thumbnail, decide what type of AB test you want to do. So we have, like, visual hook, we have headline, thumbnail pose, and gesture. So you decide what type of a B test you want to do for that thumbnail. And we both know that's kind of like YouTube. A B testing is available, but a lot of people are not leveraging it because it's like too much work.
- [17:14]
Keith
It's a lot of work.
- [17:15]
Lore
And so the idea here is that you throw in the image, like your baseline, so you are creating the first version of it, and then you decide, like, what type of a B test you want to run. And there's a really smart system. I'm going to make that available so anyone can kind of hop on over that and hopefully you're going to use it.
- [17:30]
Keith
Not pictured as producer Darren leaning deeply into the camera right now, being like, it's going to save me a ton of time. So it's awesome.
- [17:40]
Lore
So just a quick example, I just did it immediately before we started. So I took this image and chose pose and gesture as the AB test I want to make. And you can see that actually came out with pretty amazing results.
- [17:54]
Keith
It is actually pretty amazing. You know, people need to remember that these are like pretty complex generations that you're doing. So it's not like instant to do this. It's fast, but it takes a minute or two.
- [18:06]
Lore
Yeah. So to build that actually took like, it took us a week to build it. So it's not that easy to nail it down to something that actually works.
- [18:12]
Keith
Yeah. To build this workflow, took you about a week to build. Yeah, Yeah, I bet it did.
- [18:15]
Lore
Exactly. So it's like a lot of iteration until you get to something that's consistently that at least like a big portion of the result is useful.
- [18:22]
Keith
I mean, this workflow, though, is essentially an app that I'm sure there are probably a handful of people that are out there that are like building one to sell for like 10, 20, 30 bucks a month. And you're just built it in a week.
- [18:33]
Lore
Oh, we met so many kind of like one person agency that have built their own kind of like multiple systems like that, and they're just kind of like selling $10,000 projects. Most of them, they can run with a click of a button, but they did put a lot of like months.
- [18:46]
Keith
Of work into building and proprietary understanding and knowledge. This is what I really want to express upon people. Even if you look at this test like that, when you went and did the actual generation, you had the dropdown of all the different types of tests. Like, you have to know the types of tests. You need to know what changes visually for those different tests. You need to have best practices for that. Like, there's a lot of deep Understanding of the subject matter that comes through in this and like that's very hard.
- [19:13]
Lore
Exactly. So in this case we kind of like we built as you said, so we have like a description here of each of the types of A B testing. And maybe like you and the podcast, you do like a different type of testing. Yeah, exactly. You have different guidance and you do need to put in the time to build that in a way that works for you. So the outcome is actually useful for you because otherwise it's nice, but you cannot use it. You have here a marketing brief. So again we created something obviously using GPT for Marketing against the grain. Marketing against the Grain exists to challenge default thinking in marketing. Do you like it, Kit?
- [19:44]
Keith
I do, I do. Like this, this is good.
- [19:45]
Lore
All right, so here you put in your marketing brief, obviously you can edit it, you come up with something that you like that works well enough, and then you have AB testing guidelines. And then we've created this component that is just a list. And if you want to kind of like add new options, you would add the description here and you would add the options here to that list menu. So there is kind of like a plumbing work to get this thing built out.
- [20:07]
Keith
Sure.
- [20:07]
Lore
Now the second thing is that here we have a component, we call it Visual Design Extractor. It collects the inputs from these three. So one of the things you could think that would work is that you just use those as like a reference for nanobanana, but that turned out to not work well enough. So you use an LLM in the middle to provide a description of each of those. So you have a description of what worked for me. So there's a system prompt that asks the LLM to describe it in the way that I want. You are a senior visual design analyst and brand system specialist. Your role is to analyze a set of four existing YouTube thumbnails. Blah, blah.
- [20:43]
Keith
So you have like a really detailed description.
- [20:46]
Lore
Exactly. And then it does that overall visual identity. The Visual identity projects atone authoritative expertise. All right, so you have this component here that analyzes the best performing images. And then here you have the engine. So it takes everything we've seen up until now. It combines it into a very detailed prompt. It's a concatenate. So it actually just connects text in the way that we built it. And then it comes up with five new outputs for the new tests and those are going into an array. So it breaks it down. And then the input eventually to nanobanana here, which is an image generation model, is the reference image for the original image we had here and then this very detailed prompt that relies on everything we've seen up until now. And this works pretty well. So we have like the pose changes. We kind of like, we have a wildcard that tries weird things. We have kind of like the visual hook. You have like the text change. And so eventually you have these image models generating things that hopefully a big portion of them are useful for this case. That's a workflow that I really like, and I think that it showcases kind of like a very niche and specific use case. And we're seeing teams build dozens of those internally and then use them. It's usually. It's not a one size fits all. Kind of like when you try to do a one size fits all, it will never get to the level that you need.
- [22:03]
Keith
Right. Well, look, marketing is about making the best guess possible, right? You're trying to make the best guess possible based on what you believe, the taste and style you have, and what your audience wants. And I think that in the post AI world, you have to have a lot of different guesses that you can test. Right? And that's where most of the upside is. And I'm sure we'll actually probably be using this to test the thumbnail for this episode. But what I always try to remind people is if you have something that you were doing on a very regular basis, you need to build a system for it. Like if you're running a YouTube channel like us, we know what you're always doing. You're always testing a thumbnail, you're always dealing with thumbnails. You're always testing a thumbnail. So you got to do that like that. That's something you have to make a system because you're just wasting so much time and effort if you're not.
- [22:48]
Lore
Exactly. And I think that one of the most challenging things to kind of change is the mindset that new technology means that it's going to be easier and faster. It is going to be faster, but not on day one.
- [23:02]
Keith
There's a learning curve.
- [23:03]
Lore
There's a learning curve and there is an investment. And so, yeah, I think that's what leaders need to understand, that they need to put in the investment to reap the rewards. Otherwise they're staying behind there. Like, they can tell the story of, like, I have everybody in the team using GPT, I have everybody in the team using this or that, and they're generating. But, like, the question is, are you able to get to production level? Are you able to shorten cycles? Are you able to really democratize creativity internally. And I think that most teams are still far behind on that because it's not that simple.
- [23:31]
Keith
I think I would go so far as to say every team is still far behind the opportunity of what could be done. Right. And the point I would just push on is that if you're leading a team, even if it's just like you, another person, the thing you have to focus on is, okay, do we have a system or not? Okay, if we need a system, because this is something we do regularly, do we understand what the next step is to make the system better? Right. You're like, oh, cool. Well, I actually need to actually extract deeper description of this image and not just let the LLM determine it. Right. In that last workflow we looked at just a couple of minutes ago, and it's like I have all the confidence, patience and time for any team where they're just like, they can clearly articulate, oh, this is the next step we're working on to make this thing better. And then you can debate if that's the right step or not or if something else is, but, like, that's the right area of focus. It's not just like, I'm doing some activity. It's like, I understand what to do, what the next opportunity is, and I'm trying to go take advantage of it.
- [24:33]
Lore
And I think it's so hard because when you're on social media, like, everybody's telling you how everything is agentic and they're doing everything in two seconds. And because this new AI model is just absolutely killing it. And I think that some of the challenges is also to be able to find, to identify the real stuff from the hype and be able to invest into them. It's not about an next.
- [24:56]
Keith
I think you're saying people lie on the Internet and that is, I mean, I will say it. I mean, people lie on the Internet and people are hyperbolic and they have incentives to do different things. Right. And it will feel impossible to keep up is I think part of what you're saying, unless you build systems and are clear on, like, this is what's important to me and what I'm going to focus on learning to make what's important to me better. And that's like, really what one of the core things we're talking about today is just evolving the mindset of how you use AI to actually build systems to do more durable work than just one shotting some stuff in ChatGPT or Google Gemini.
- [25:35]
Lore
Exactly. And ask yourself, like, what happens when, like, A better model comes out, like, are you going to go crazy and need to change everything you do, or are you in a position where like, okay, amazing. Now everything I'm doing, like, tomorrow is already going to be way better. Yeah. And I think that if it's the first, then you're doing it wrong. You're kind of like chasing the hype. And. And if it's letter, then you're doing the right thing. You're building an organization that's going to keep on improving and evolving with any marginal kind of tech change.
- [26:02]
Keith
You know, I completely agree with that. And what I guess I'd love to understand is, like, you guys are part of Figma now. What's the future of all of this? Like, you know, the models are moving fast. Yes. And they're going to keep moving and that's going to be good. But at the end of the day, I'd actually argue most of us shouldn't have to care about the models. We're trying to do some jobs and we just want, you know, the model to do the best job we're doing. So, like, get out the crystal ball and just kind of tell us, like, where is creative work going over the next one to two years with everything that's happening in AI?
- [26:36]
Lore
So I'm probably going to tell you a lot of what I said so far that I think that craft really matters, does, and I think that teams need to kind of change their mindset and craft is going to change. Figma was always around craft, and I think that's why that change has been so natural to us.
- [26:52]
Keith
Yeah.
- [26:53]
Lore
We've always looked at Figma as kind of like the. We had the same values, we consumed, we copied a lot.
- [26:59]
Keith
Yeah. Dylan team are the ultimate craftspeople. I think that they have done a.
- [27:03]
Lore
Great job and I think that in a way, kind of like Figma broke the silos in product making.
- [27:08]
Keith
Yes.
- [27:09]
Lore
And in the beginning that seemed crazy. Like, which designer is going to allow other people to see how they work real time? And I think that for us, one of the things that excites us the most is like creative marketing. Video editing work has still been siloed up until now. And while we've still early and kind of like the way you can collaborate on wevy, we see teams working together, being able to deliver from one person to another. So I think the future is a lot around changing the craft to system building, seeing something that's much more collaborative and seeing kind of like creatives evolve from kind of like pixel editors into system Builders and enablers, instead of just kind of like plant workers, they're going to be not just delivering their time, they're going to be delivering systems that are then able to scale up and deliver much more than that.
- [27:55]
Keith
Yeah. And you know what, what I think is interesting, Laura, is that when people hear the word systems, they're like, blah, blah, blah, blah. I don't know what, you know, what you really mean. Like, systems seem complex and process for the process sake and all of that. And I think where you and I are in like, deep, deep agreement is that the future of AI, for I think most marketing tasks and for most tasks in general at least, is way more collaborative than it is now. And you need humans working together right now. Those humans are very siloed apart on all of these tools. And because it's more collaborative, it's way more transparent, and it's not dependent on like one model or one technology or one workflow. Like you have people collaborating to build just better workflows. And those workflows eventually become systems and applications and all these complex things. If they're on the things, that's really, really important. Right. And I think the biggest thing for me on the future of AI, especially in creative work, is that I think it's going to move from an individual sport to a team sport. And it's not a team sport. Right now I'm out there, I'm talking to everybody. You are too. I mean, you can tell me if I'm wrong, but it does not seem like a team sport today.
- [29:03]
Lore
Now, you usually have like, few individuals that are doing brilliant job and people are struggling to follow and what they're building is like mind blowing, but the same time not really doing any business. Impact. Yeah. You want something that's going to impact the entire team. You want your team to progress. You don't want certain set of individuals to blow your mind and then be stuck at the same problem again.
- [29:25]
Keith
Yeah. And so I guess what's your advice for people who are trying to get over that hurdle? You know, and what I'm saying is because like, like you just said that thumbnail workflow took you all a week. And you guys are pros. So if you're not a pro, maybe it takes you four to six weeks, you know, and that's daunting because everybody I talked to is like, gosh, I'm on this hamster wheel. I got so much stuff to do. The thought of learning how to do things in a new way and actually getting less efficient to get more efficient it seems like a near impossibility. And so what do you tell people? I imagine there are a lot of people watching this that feel that way at this moment.
- [30:03]
Lore
I think it's one of two things. If you feel like you have too many things that you need to deliver tomorrow morning, you don't have the time to kind of invest in building stuff and learning new stuff, then go back to the old way. Don't try to hack it through the simple one size fits all tools, because they're just not going to work for you. I think that you should decide how much value you can gain out of it and you should spend time learning. The teams that are really successful are the ones that are willing to invest and they should look at it as kind of like R and D. Right. We're coming from tech companies. We know that not every effort you put into the R and D team will yield something substantial. In many cases, you'll get nothing out of it. I think that kind of mind shift of creative and marketing, they need to do R and D as well. And when they do R and D, you build your own proprietary capabilities that nobody else has and gives you like a real advantage over others. So I think if you're taking the simplistic approach of like, I'm going to hack it through gpd, I have this, like really long prompt is going to do something amazing that's not going to work. I think you need to be able to invest the time in the right set of tools and capabilities to keep on learning and to look for things that are deep that will give you that edge. And assume that some of your effort is not going to yield results, because that's how R and D works.
- [31:19]
Keith
Assume a decent amount of your efforts. Not going to get results. The one thing I would maybe push and add a little bit is right now this learning and this investment in system building and understanding how I can build, in this case, creative. But there's lots of different things you could say here is like optional, and it's not really optional. Like, this is not a little bit of a transformation. This is a big transformation out in the world, which means you will hit a point in which you cannot compete unless you don't have the right AI first process and technology to win. And I think what you and I are here, whether it be CRM, whether it be design tools, marketing tools, what have you, like, you need to get the right infrastructure in place now so that you can go faster as the market is going to continue to speed up. Yes, right.
- [32:07]
Lore
Yes. I completely agree.
- [32:09]
Keith
And I'm seeing it every single day, every single week there. I know there are a lot of founders that watch the show, a lot of marketing leaders that watch the show. It's like, if you are not taking that R and D approach and that systematic approach to your marketing, especially in, like, how you're transforming to be kind of an AI first marketing organization, then it's going to be really hard to win a year from now, two years from now.
- [32:30]
Lore
Because the level keeps.
- [32:31]
Keith
Yeah. The expectations and level the game. Yeah.
- [32:34]
Lore
And it's like, as you said, like, if before that something was enough, like a year ago, that's not enough today, that's certainly not enough in a year from now. So, like, everybody's progressing, teams are becoming faster, teams are able to deliver things that are better, more creative, better tested using kind of like better data. Whatever it is, it keeps on progressing. And if you're not able to kind of keep on leveling up and pushing the boundary higher, if you find yourself using the same tools that your parents are using. So once my parents started using ChatGPT to create images, obviously a designer cannot use the same tools.
- [33:07]
Keith
Obviously, there has to be a better way.
- [33:09]
Lore
Yeah. So, I mean, you should not be using the same tools that everybody's using. You should not be be doing the same things that everybody's doing. You should make sure that you're always at the forefront of what's possible, and you should assume that it requires a lot of dedication to get something valuable out of it.
- [33:26]
Keith
I think what's really interesting about what you just said is that in the early days of AI, there was just these assumptions that the models are going to be so good that you just type anything you want in and get exactly what you want. And it would just commoditize any specialized skill or knowledge. And it's just not true. If it was true, we wouldn't be hiring software engineers at record pace. As AI has gotten really good at software engineering, it's like, no. AI is making the real specialists more. So I think one of the other things is if you're a marketing team and you don't really have a specialized creative, you still need that point of view, whether it be a freelancer, an agency hiring somebody, that you can't outsource that to AI yet what AI does makes your ability to scale that and the expense around that much better, but you still can't outsource that.
- [34:16]
Lore
And also, as you've seen before with the YouTube one. Right. It's like you needed an expert marketeer to say that's how we do AB testing. Like, you cannot give that step up and you need the expert creative to say, that's the brand guidelines and that's the constraints that I want to have here and that's how I built it and break it down. So the same skills you needed in order to do creatives or the same skills you needed to do marketing are still really important. So you have the right intuition, you have all the theory behind it, you have experienced it, you know what you're looking for and you know what's right and what's wrong. Because it's so easy to assume that GPT knows it, it doesn't know it. It will give you an answer that sounds good enough, but it's wrong. And so I think that having those experts is still critical. Like, you cannot do it without them. But those people need to understand that their job has changed. They're not doing the same job they've done a year ago.
- [35:01]
Keith
Yeah, I always tell people, like, the what you need to do hasn't changed nearly as much as the how you do it. You know, like a lot of what you need to do is very similar, but how you do it has changed massively. And the tools available to do it in better, smarter ways have changed dramatically. So Lor, one of the things that I think is very interesting that you also have done, in addition to like, deeply understanding how AI applies to creative, is you've run an AI native company, which is like this big buzzword out in the market now. But I do think there is something different about running an AI native company in this day and age. And, you know, tons of founders watch the show too. So I want to make sure that we're throwing a little bit of info to the founders out there. What do you learn about building an AI native company? How do you get customers in this world? Like, just how did it work?
- [35:50]
Lore
So the interesting thing is there is no playbook. And I think it was true before AI and I think it's even true now. The market is like, much harder, I think, than it used to be in a way, because everything is so noisy. But then at the same time, it's kind of like the best time to build a startup because you're building the fastest and the wallets are open. Like, companies are looking to adopt new tools. And I think there was no time in history when that was so true. What worked really well for us in our case is that we've built something that's really appealing for the most advanced users in the market. So we actually haven't tried to capture the entire market. We just tried to do something that the top 1% of creatives are going to get really excited about. And the cool thing is that these people are happy to also talk about how they work. And because we've so much about how you build things and not just the final output, that makes it something that's really fun and easy to share. So that created that really cool flywheel for us and I'm sure it's not applicable to any startup, so it's like very specific to our way. But for us that thing worked really well. I think that what hasn't changed is that as a founder, you need to find the thing that works for you. You need to understand your market, you need to understand where your customers are, you need to understand your beach head and like, who is that audience that you're reaching out to first and what's the best way to reach out for them. So for us it was a lot around kind of community building. So identifying that audience and putting all of our efforts to bring them in as a community into wevey, we call them the Artist Collective. We gave them a lot of support, they gave us so much into like what we should actually be focusing on. So like most of the best workflows that were built over the past year were not built by us, they were built by the community. And you can see them on social media. And so that's the thing that works for us. But I mean community is probably not for everyone and that's a play that you should lean into if that works for your audience and your product.
- [37:35]
Keith
Well, community really works these days because you need to learn quickly and you need that close knit group of people that's sharing all the good and the bad with you. And you were very focused on who you were building for. I think the other thing that's interesting is even though you were solving for like maybe a more advanced creative, you had a free version of the product. You know, I do think there's a motion for every AI native company out there which is like your community makes cool videos of stuff they've built and somebody else can go build something like that in a low friction way because they can sign up for free, they don't need to talk to a salesperson, it's not super expensive. And like my thesis all along has been the best companies will be those who spend a higher percentage of their cost of customer acquisition to like API calls to the LLMs than ads to Google and Meta. Right. And I think you all are really good example of that's exactly the path you took.
- [38:26]
Lore
Yeah, I think that makes a lot of sense. And you see much more than me on kind of like how the market behaves and how it works. And yeah, I agree that people are looking to learn and there is an opportunity to build something that scales up pretty fast through content and how you can achieve things. And then I think it's also about the audience. Right. Because not every audience is kind of like distributed through enthusiasts and champions, I think. Kind of like developers are there, entrepreneurs are there, creatives are there. It's probably not true for, I don't know, like HR people, I would guess. Kind of like if you build an HR software, not sure it has like a self distribution on, you'd be surprised.
- [39:00]
Keith
Gusto. And a lot of those folks have really high NPS products and word of mouth is a big part of it, you know, so it's like building a good product that is valuable to a specific set of people is timeless. Right. And you've figured out how to do that in kind of the AI native way. But like there are some things about building really great companies that just persist and exist the whole span of time. Laura, this has been awesome. We're running out of time, so I want to say thank you on behalf of all of our viewers here at Mark against the Crane for coming on, showing us some amazing workflow. I think we're going to drop a link or two to some of the weavy workflow flows below so that if you want to go and customize those or use those yourself, you can go and do that and we'll see everybody really soon on Mark against the Grain. Lord, thanks so much for being here.
- [39:42]
Lore
Thank you so much. I really enjoyed it.
- [39:58]
Keith
Foreign. You know, Kieran and I have been doing the podcast for a while now. We've been at this for a couple years. We love it. We could not be happier to be doing this, but we wanted to take things to the next level. We want to level up the impact we're having with Marking against the Grain. So the next step of our journey is something we're really, really excited about. We're going to launch the Marketing against the Grain newsletter. And Marketing against the grand newsletter is going to be amazing. If you are a marketing leader, practitioner, you're in the trenches doing marketing every day. This is for you. We're going to deliver right to your email inbox and you're going to get all the behind the scenes frameworks practices, tutorials from us, from the guests we have on the show, and from people even beyond the podcast that we think are going to be helpful and really have an impact on your day to day, week to week doing marketing. You're going to love it. It is something we've been talking about for a while. We're really excited to have it out in the world. We've already got a hundred thousand marketers who are on this newsletter. Please join. It's completely free. We'd love to have you as part of the Marketing against the Grain community. And it's easy. You can click the link in the description below, or you can head to marketing against the grain.com subscribe.