← DTC Podcast
DTC Podcast
Ep 589: 70% Different or Meta Buckets It: The Andromeda Creative Playbook (w/ Braydon Germain, Pilothouse)
00:25:54
Loading summary
Transcript42 lines
- [00:00]
A
One thing I learned during Black Friday is you don't want to put out something that's like, really obvious AI. I think that doesn't really help your brand, and I think a lot of brands are sensitive to that as well. And so I kind of had to find a strategy that a wasn't time consuming. Like, if I'm spending eight hours just prompting, trying to get the perfect video, it's a bit of a waste of time for me. So I really found ways to leverage these AI tools quickly and kind of let them flow under the radar of people being like, oh, is this real or is this fake? And so some of the stuff I
- [00:32]
B
was doing was chatgpt tips. So it isn't such a people pleaser how you, you know, Talk to your ChatGPT so it gaslights you less. This episode is brought to you by Contentful marketers. You know that feeling when your creative clicks, when that social post sends engagement through the roof, when your outside of the box campaign hits ROI positive, when a personalized homepage turns prosp specs into customers? It's utter marketing bliss. Contentful helps you create tailored omnichannel experiences without working overtime. No stress, no limits, only possibilities. Get the feels@contentful.com it's all killer, no filler. I'm Eric Dick on the DTC podcast and we have returning guest Braden Germain, who is sort of a resident expert with all things AI here on the Pilothouse team. I just feel like there's some absolute insane AI energy happening right now. D2C is, is. We're now SaaS developers. There's all sorts of things happening. What's. What's going on in. With AI in your world, Brayden?
- [01:45]
A
I mean, there's a lot, there's a lot of creative stuff that's happening, especially in the age of Andromeda. I think some of these static creation AI tools are getting really good. That's definitely going on. The video stuff is also super interesting and just getting better and better.
- [01:59]
B
Oh, my God.
- [02:00]
A
And, you know, amazing to see. I think a lot of these, like, After Effects nerds who just didn't have the raw materials to make cool videos, but now they can just, you know, prompt and get this, like, raw material and then mix it all together with, with After Effects. It's, it's pretty inspiring. Pretty cool.
- [02:15]
B
To kick things off, let's talk about the Clawbot situation.
- [02:19]
A
Okay. Yeah, for sure.
- [02:20]
B
Interesting. I read an interesting tweet and I want to parse some of the information I've heard. What's Your take on, on the implication of this for our industry, I guess. What is it?
- [02:30]
A
First of all, you can think of it as a autonomous AI that works continuously on a computer. So it's not so much that it's a chat interface or anything like that. You can talk to it through things like Type form or Slack, and that's how you communicate with it. But it will run on your machine. And so I don't know if a lot of people have heard of this, but there's like Mac Minis are sold out because people are buying these Mac Minis, setting up Multbot on their Mac Mini and running it on there. It's an open source platform and it's pretty crazy. Like effectively people are trying to, you know, make it into an employee. And I've seen some insane tutorials where, you know, answering emails, doing tasks on the computer, trading. Yeah. And you know, connecting to Zapier and people are trying to figure out how to make it safe. I think that's like the one big thing is if you just give it access to your computer, like it can open up Chrome, it can go into your Google, it can like do anything you want really. So there's a lot of safety concerns, I think, but it's effectively getting closer and closer to driving a computer. And if an AI can drive a computer, it's effectively what a lot of us, you know, as, as marketers do is, is we sit on our laptops and we drive our computers. So it's, it's getting close.
- [03:43]
B
Is it powered by Claude or is it, is it, is, is Claude underneath it or is that something else?
- [03:47]
A
Yeah, so it, it's, it's Clawbot, which is like, I think their logo is like a lobster. And then Claude got kind of, you know, miffed about that and so claudebot was changed to Maltbot.
- [03:58]
B
Okay.
- [03:59]
A
And so you can use Anthropic and you can use Claude as its large language model, but you can also connect ChatGPT, you can connect Gemini through APIs. And I, and I think really the ideal is that you would have a open source LLM that's not connected to any of these large ones. So you could effectively put like a, like an LLM on your Mac that just lives as is. It's like servers on the Mac. And so everything is kind of contained, self contained.
- [04:29]
B
There I was following the social network, the Moltbot, Clawbot social network drama. And then I saw a really interesting tweet that followed that because to catch people up was essentially this idea that someone created a social network for these bots and they started talking and conspiring and they created a religion called Christafarianism or something like that. But then I saw a tweet just yesterday. I was trying to find it so I could reference it, but it was a guy who was taking credit for having written the Crestafarian tweet posing as a bot, being a human posing as a bot. And he said that almost all the posts that were referenced as, like, scary things that AIs were doing were actually friends of his that he knew were human. Which I thought it was really interesting because all these stories picked up and they told the story of these creepy bots doing these things. But all the examples they use were actually the human insights that were being generated. And so it sort of had me go full circle to be like, oh, good, there still is the power, the power of human creation. Creativity is still something that humans recognize and gravitate towards.
- [05:30]
A
Yeah, yeah, I think that was a scary moment. But if you think about, you know, we all work in DTC and we're all trying to get likes, impressions and clicks and stuff, and I think, like, you could go and instruct one of these clubbots to. To give the perception that it's coming up with its own language or, you know, that those. That sort of stuff, you know, coming up with religion. But I think most of the time it's probably human driven in order to, you know, get. Get lots of attention on. On Twitter and X and stuff like that.
- [05:58]
B
You have a note here to jump right into some of the Black Friday AI winners. I know it's a little post Black Friday, but I'm really interested to hear about the way AI was really successfully used. I assume that's for creatives during Black Friday.
- [06:11]
A
Yeah, I think, like when AI video and AI images came out, I think we had a lot of clients that were excited about it and just kind of thought, hey, now we can do insane volume, like just go for it. And one thing I learned during Black Friday is you don't want to put out something that's like, really obvious AI. I think that doesn't really help your brand, and I think a lot of brands are sensitive to that as well. And so I kind of had to find a strategy that a wasn't time consuming. Because if you're like spending with my job, like if I'm spending eight hours just prompting, trying to kind of get the perfect video, it's a. It's a bit of a waste of time for me. So I really Found ways to leverage these AI tools quickly and kind of let them flow under the radar of people being like, oh, is this real or is this fake? And so some of the stuff I was doing was, you know, really involved with Photoshop, using Nano Banano and taking like screenshots of influencers, taking existing content and then kind of repurposing it in like interesting ways. For example, we have a, a brand that has an eye mask. And one of them was like, let's put her on top of the moon. Let's put her in Everett. Like let's on top of Everest. Let's like have one of them of her standing with the sign that says Friday. And she's like outside of an airplane window. Just like, it doesn't look AI, but it definitely stops the scroll. And so that was a lot of like the processes that I did and they were like, quick. Was able to come up with lots of ideas, maybe create different age of people. So say we're going after like an older Persona. We could kind of get that like 50, 60 year old model that are kind of hard to source in the DTC world. And then the other thing was just trying to think outside of the box. And I think the cool thing with a lot of these video tools is that you can put the first frame and the last frame in. And so then I was able to kind of AI generate very organic looking content using existing content as kind of like the reference. So I'd get those statics and then I'd put them in. So one of them was the product was on like an airplane tray table and it started there and then that was the first shot and then the last shot was like the girl holding the sign up outside the airplane window. And so you kind of had this like POV shot of like the product and then like it looked out the window and you could see this like girl like holding a sign on the airplane wing. And that one was actually one of the better ones that we did. And it, it kind of flew under the radar probably. People were like, is this AI? Does this feel real? So it was kind of making it feel real, but also not wasting a ton of time trying to create that
- [08:34]
B
stuff because you can literally do anything. That's the challenge that you find yourself in. Whether it's creative or building, you know, new SaaS tools. It. Almost anything is possible. It's pretty wild. Talk to me about some of these static ad creation AIs that, that you used and to good effect. Right now, all of our ads for both the agency, the age, like our agency community, as well as D2C are all AI statics that we're hitting some really good numbers in terms of our CPAs at this point. What tools are you using in the static world?
- [09:06]
A
I've never really been too thrilled with, like, how AI tools have worked. Maybe like, six months ago, you know, I would have said, you know, the stuff isn't there yet. But I've been recently using a tool where you can plug in all of your brand's information on the back end, like their fonts, their logos, all the color palette. You know, you basically create this, like, brand guidelines for them, and then you chat with it, like ChatGPT, and you can access ad library and you can go and click on, you know, different ad inspirations and say, like, I want it to be like this brand's ad, but, you know, for this product, you can select different products on your website. And it does an amazing job. Like, you know, I was able to pump out, I time myself in two and a half hours. I did nine static ads, and no issues with the small fonts, you know, of the product. No issues with. With that sort of thing. It really nailed it. And, you know, to create nine static ads would have been a lot more time consuming to do the traditional way. So it's coming a long way. The tool is not perfect, but it's definitely. It's definitely on its way.
- [10:10]
B
Now, you mentioned the age of Andromeda. So go one level deeper with those nine creatives, because in the past, you might have created two angles and three variations for each of those angles. But when you say nine static creatives, are you talking about, like, nine sort of like, flesh concepts? Like full concepts? Right. Talk a little bit about how you think about concepting.
- [10:29]
A
Yeah. So Andromeda is. Is different. You know, before we used to be able to do micro iterations where we would say, hey, we've got this winner. Let's change the color of the cta. Let's maybe change the product image, let's change the headline. And so we could effectively get more volume from our designers by saying, like, hey, I need five iterations of this. Subtle tweaks, and we could throw that in and maybe one of those would hit and we'd get a little bit better of a ctr or we would get, you know, a better roas. But now in the age of Andromeda, Andromeda and Meta's new AI is going to bucket those if they're too close. So I think, like, the general rule of Thumb is like, unless it's 70% different, it's going to bucket it into 1:1 creative. So you could put five iterations that look fairly similar and Meta will only see it as one. And so what you really need to be doing is just coming up with these new concepts. And that's time consuming from like a content briefing standpoint, but also from a designer standpoint and then also from a revision standpoint because now you've got five new concepts that like need the client to give revisions or give feedback on. So it's really forcing, I think, a change in the way that we think about how to deliver creative and tools like this I think are inevitable so that we can get the volume, we can get the. The contrasting concepts and then hopefully, like, because we're able to just kind of pump this stuff out, hopefully we can just speed up the process and then give Andromeda what it wants, which is just like a plethora of different ad creative ideas. So the way that I do it is, is we come up with a, like a strategic overview of what we're really trying to do. Like, like, what's the angle? And then from there I will kind of see how many different ways I can expand on that angle with different creative.
- [12:17]
B
I like your point. You have Next here about ChatGPT tips. So it isn't such a people pleaser. Like, I've never felt more gaslit than by my chat GPT. I just started using Claude and feel like maybe it's a little better. But I'm curious about how you, you know, Talk to your ChatGPT so it gaslights you less.
- [12:34]
A
You can go into the settings and you can give it custom instructions. And basically what I told it, I said, you are my mentor and I need you to pressure test all my ideas and I need you to call me out when they're bad ideas and tell me why they're bad ideas so that we can get everything solid and bulletproof. And just by doing that, it's takes a different tone. You know, it's not like, of course, Braden, or like, oh, I see. You know, it's just like hardcore mentor mode. Like, here's where you're going wrong. It's like the first thing it responds with and then it breaks it down. And I think that's just like a way that I found to improve its. Yeah, to improve its. Its outputs. Just really saying like, you're a mentor and like the. My. My success is based on what level of critical thinking you have. So that's one way that I've, I've done it and it's. And it's been really helpful.
- [13:24]
B
I love that you are Yoda. I am but a Padawan. The next point you have here is the future of AI and I'll just use this as opportunity to talk about our week here on the DDC side, which has just been absolutely wild. I just had a little memory come up of three years ago when I was using either ChatGPT or one of the other ones to create some images for this publication. We were creating and it was all creating these really great cinematic images and sometimes the fingers were messed up and, and I was comparing that to like what was happening this week, which was my co founder on DSC started looking at our onboarding flow which we're radically changing. Started looking at how content works in our newsletter and how we could make our newsletter, you know, more appealing to more people. So he ended up using Claude, creating a hall of 10 avatars who would rate every everything on our newsletter and give us all this feedback. Which then morphed into him to actually using Claude cowork and the Claude code to start building, building out HTML code for the newsletter which then led into him building a full content delivery system for the newsletter that allows us to both write real content, generate other content. This whole perfectly tuned SaaS product where he's built logins for us all and he's not a developer by any stretch and in two days he's built an entire sort of like newsletter system that could only be built for us with this. All these feature sets that he doesn't even know he needed. And it's just like, it's insane what is possible with these products.
- [14:55]
A
And yeah, it's you. Like you said, you just have to have the imagination and be willing to put in the time to kind of uncover what's possible. And these tools are, are, are here for it. They are down to help us do whatever we need to do. And that's pretty insane. So you're. So he just did that in two days?
- [15:15]
B
Yes, in two days.
- [15:15]
A
You feel like it's like functioning.
- [15:17]
B
Yeah, we're gonna send out Friday's newsletter. If you're not subscribed to the newsletter, maybe subscribe on Friday because you might first ever AI HTML newsletter from us. But we're. Yeah, we just were on a one and a half hour long meeting this morning where we were like fine tuning and tweaking it and we're going to send the first one generated this way. And again all the content is still written by human for now, but it. But we have it, like, helping us frame it in a way that's better based on these 10 avatars that we created. We're also going to then create. We have already created on ChatGPT last week, like a DTC podcast brain. So it's got all of DTC podcasts in it. So we write a newsletter and we're like, hey, come up with examples from the podcast for each of the things that we talk about. And then it pulls from all of the different podcasts we have. So what we got to port that over to Claude. We just had a tech meeting where we're going to port these giant brains over to Claude. And then it's like this, this comprehensive system of building our newsletter, referencing past newsletters, referencing podcasts. We're going to eventually put in performance data from each of the newsletters or each of the podcasts. So Claude now knows where. Which newsletters are performing better, which content performs better. All. The only thing that I'm thinking during all this is it's like, you can literally do anything now. I'm like, should I be like, the things that people will be using AI to build are going to be absolutely insane. Like, we're building a newsletter for D2C founders. Why don't I ask it to figure out a newsletter that would pay me 10 times as much, you know, and figure out how to execute that? Because I probably could at this point.
- [16:44]
A
It is wild. And it just makes me think that, like, you. You would never hire an employee to do all that work of indexing the transcripts and reading through the transcripts and searching through that. You couldn't hire an employee to do that. And now I think that's sort of the tasks that are like, it's really excelling at, at this point is just these. You could never hire enough people to do that. It's just not cost effective. And now you have this person, this, like, brain that effectively can, you know, really steer an organization or steer at least your life. So, yeah, it's awesome.
- [17:14]
B
What a time to be alive. What. What else were you going to offer about your thoughts on the future of AI?
- [17:20]
A
I think there's a lot of doom and gloom out there. And I think, you know, part of me really thinks we should be optimistic about it. And. And so I just. I just think I would encourage everybody, you know, as the future comes quicker and quicker, just get your hands on some tools. I've. I've spent some time, you know, on some AI boot camps and Just learning the stuff myself just to get, you know, better at the tools. And then if a piece of software like comes out with an AI integration, like just making sure that you're using that stuff and, and, and seeing ways to, to leverage it, I think is just important in the, in, in the future. And I wouldn't be, I wouldn't be afraid of it. I think I would just embrace it. And I think the future is going to be pretty cool. And obviously if, if they create an AI that can drive a computer better than I can, obviously that's going to be a little scary. But, but I think if you just stay on top of it, you'll just be able to be a controller of AI and just create more value.
- [18:19]
B
I am a techno optimist generally as well, and I think you're absolutely right. And yeah, seeing what is possible is just so eye opening. I'm trying to figure out when this podcast will come out, but next week on Monday we have a podcast from New York and co the CMO there, Laura Cantor. And a lot of the podcast is actually about a post she made on LinkedIn that went crazy viral, getting millions and millions, millions of views. And literally it was, the post was about how she switched her site from hydrogen to liquid. They went off headless onto like a pretty standard Shopify build for this huge e commerce site a week before Black Friday. And how she ended up using this tool that's called Zenit. Z E N Y T A I kind of does what Jeff kind of built for our newsletter where it scans your website as a customer or as different kinds of customers would view your website and gives you like 5, 500 suggestions about whether it's like you have a lot of a big product catalog and your images don't match up. Maybe you used a lot of AI images and there's some problems there or whatever. But the point of her post was in this age of AI, it's super important that we also keep, keep connected to our community because people are constantly solving these problems in different ways that you didn't even know your website has. Maybe, but this post just resonated and got like 2 million views or something like that. Sold dozens or hundreds of these, these Zenit subscriptions. But again, every person that hears that could go and fix their website, which again is a job that you couldn't hire a Mechanical Turk or 100 Mechanical Turks to do, but now it's possible.
- [19:52]
A
Yeah. And so, so I guess what you're saying is staying connected to a community and being exposed to all these ideas that are happening around us. Because if you kind of shelter yourself out from being on accident and, or being a part of these like, AI communities, then you really don't know what's around the corner or you don't know if there's like a solution that actually could help your business. That's just like a few clicks, like you're saying two days away to transforming your business.
- [20:16]
B
Embrace the fomo. There was one other thing you had on here and we do have a little time, so let's shout it out. What are some. I know we as a shop use motion app or a lot of the team members do. What are some of the new AI features that are moving the needle for that platform?
- [20:30]
A
I just think, like, creative strategy is so important in the age of Andromeda. And some of these new AI features that motion has been rolling out are so impressive. You know, they've got this AI tagging feature that will, you know, it'll actually look at your ads. It'll. It'll read the transcripts. I think it even looks at the video. Like, it'll snapshot the videos and so it gets context of what your ads are doing and it will tag your ads and then you can go into like, different patterns and you can see like, okay, this is like our intended audiences. This is who we're going after. These are the formats that have been working, you know, and then they have this like chat interface. So then you can go create a motion report that shows you, say, your target audience that you've been working on. Maybe it shows you other audience that are. Have potential that. But, but maybe you're not really servicing with ads and you can click on that, create a report. Then you can get the AI to analyze it. Then you can get it to say, like, how are we not speaking to these people? What are other ways that we could create ads? It's just phenomenal. It's like it really has come a long way in six months. And I think with Andromeda, with like the speed at which we, you know, are to be creating new concepts and coming up with new angles, I think, I think it's a really cool tool. So, yeah, just the, the fact that you can have a conversation, a dialogue back and forth and it's integrated into your ad account is like, so cool.
- [21:47]
B
And I think it's a huge area. I actually just saw a tweet from someone that was heard through the grapevine that Meta is going to be putting a bunch of investment into creative analysis for their platform and the post was lamenting the fact that creatives are what drive the biggest revenue generator of the biggest ad network in history. And yet there's almost no you know, native intelligent ad intel creative intelligence built into the metal platform.
- [22:14]
A
Yeah, there's always like if you went into business ads manager you could find like creative insights and you click on it and it's like wasn't it didn't help very much at all, you know and it, it is a huge, it is a huge hole that I'm surprised like, like you're saying I'm surprised they haven't filled. I think it'll be cool like if they can, if they can beat motion. Like you know those are big shoes to fill. Motion is really on top of their game. And then the, you know I think too in the future I would say that Meta is probably going to come out with a tool that creates ads as well. Like I mentioned earlier, the static ad creator tool that I've been using. I think, I think Meta will just be able to do that as well on platform in, in a few years.
- [22:56]
B
I'm. Have you done anything agentic? Have you done anything where you're turning something loose to act on your behalf? Because that's, that's everything that I've done is sort of hands on keys, you know, cutting pasting into different platforms. For me a big change will be when I, when I bridge over to you know, giving something permission to act on my behalf.
- [23:14]
A
Yeah, I, I've just used the agent mode in chat GPT and it's just slow and, and clunky. You know it has to explain every move that it does and it takes a long time. So I've got it to do things like you know, hey, like you know, while I'm working on something I'll send the agent to be like go through and find me like five really good testimonials off this website and it'll go through and it'll click through the reviews. I've done stuff like that where I can kind of multitask in that way or, or CRO stuff like can you just like look through this page, analyze it, give me some feedback. So that's the sort of way I've used it. But watching Claude Bot, watching some people use Multbod or Cloud bot, it's fast. It like opens up things and things are happening on your screen. It's crazy. It's like it's really. So I think I would be interested to play Some people are using utm. It's like, it's like an open source piece of software to create like a, an isolated Mac within your Mac so you can have your, your computer and then you can have like a window that's like its own Mac within your computer. And then, then they'll give Maltbot access to that Mac. So it's not actually getting access to all your stuff. It's just this kind of like empty shell of a Mac os. And that could be interesting, but if I were to do it, I'd be really cautious with security and what I give access to. I think that's. It has infinite memory. It can just like it just remembers everything. And then it also just gets better and better and better. And I even saw one where the guy gave it access to his, his computer and he woke up and it had done a bunch of work on his, on his piece of software that he was building that. It was like, hey, like, I noticed, I noticed that you were having issues here. Like, I just went ahead and made some changes for you. Like if it works, great. If not. So it's like it's doing things behind and then people are saying they're instructing it to add in my calendar all the time that you're doing stuff. And so they'll see at like 2am it's like it did like an hour of work and like, you know, while you're sleeping, it's just like doing stuff in the background. It's crazy.
- [25:15]
B
Brave new world techno optimists. Thank you for coming on the podcast again, Braden. We'll be in touch again soon.
- [25:21]
A
Cool. Always a pleasure.
- [25:29]
B
Thanks for listening to today's episode. If you're not getting the DTC newsletter, you can subscribe for free at directtoconsumer. Co. And if you want to learn more about Pilothouse's all killer no filler services, take off to Pilothouse Co. I'm Eric Dick and this has been the D2C podcast. We'll see you next time.