He Built FaceTune's AI… Here's What He Says Marketers Get Wrong

20:44Nir Pochter:Like I don't care about that. And so I think just even that mindset shift is probably a good one to show teams. Like it's okay to get intense feedback from GPT or Cloud or whatever.

20:54Stephanie Postles:I really like the name Mean GPT. I'm gonna, I'm going to steal this one.

21:00Nir Pochter:Yeah, I almost named it Karen, but I was like, I'll keep it with me and GPT instead.

21:06Stephanie Postles:And we also have GPTs for products. So, you know, we just had a release recently. So the VP marketing of the model, he created the GPT with all the documents and the release notes and everything that we had about positioning, about branding, and he kept updating these documents. experience. So whenever it needs to write some marketing content or think about a brief or whatever, they can use this GPT and it's always updated.

21:36Nir Pochter:And it's shared across the whole org and everyone can.

21:39Stephanie Postles:Yeah, this one is shared across the whole org. And I think this kind of stuff is an example for AI usage that doesn't have to be complex. If I'm going back to the question about, you know, AI isn't helping me, I think people are always thinking about the most sexy and complex problems where they should start with the boring problems first.

22:01Nir Pochter:So when thinking about kind of going back to, you know, talking to the team, asking about the things that they hate, finding that, do you have someone who is managing that process and working with your team and then also helping implement, let's just say 80 % and be like the 20 we can't do right now? Like, is it who's working with these team members to really get these processes automated and tested and seeing if it's working and all that?

22:26Stephanie Postles:Yes, so we have quite a lot of resources on that. We were always a very technical marketing team. So my VP of creative marketing, he's actually the VP of creative marketing and innovation, and he has an innovation team that are really good with AI and are helping people with automating problems. So people, when they reach to a place when they can solve it by themselves, they go to this innovation team. The innovation team helps them solve the problem and then write a workflow that is wrapped with code and people can use. And in addition to that, we have a team of developers for the marketing department that are building internal tools.

23:21Stephanie Postles:So if the task is hard or has some security issues or data issues, something that is sensitive, we are using the developers to build it.

23:32Nir Pochter:Got it. Okay. I was going to ask you, earlier you mentioned the flop with trying to create a bunch of ads at once with AI. Were there any issues that popped up when you were rapidly trying to build up AI intuition and skill base on the team and giving them access to everything? Were there any hiccups that happened along the way?

23:53Stephanie Postles:I think that the biggest problem was with people not getting good enough results with AI, like you mentioned before, or they felt the need to show that they're using AI. And then this became their number one KPI. So they knew that I'm obsessed with AI. So, okay, now we need to show our goal is no longer the business. Our goal is to show Nir that we are using AI. And this is suboptimized. So this can cause problems. You can use AI for stuff that it's not doing well. And you start having like crappy copy for the ads and stuff like that. Yeah.

24:44Nir Pochter:At the bottom it says, do you want me to write this into five different ad versions? Hey, I know.

24:50Stephanie Postles:Exactly. So it takes time and iterations and obviously, you know, solving part of the issues when we try to integrate AI was people feel of AI. Right. It's, I think, a non-issue that we also had to address. Yeah.

25:11Nir Pochter:So, I mean, if we think about the stages we've gone through at a pretty quick pace, if I think about even like last year, I was talking to many companies who they were, you know, not allowing their team to have access, of course, to the LLMs, training, even consumers using ChatGPT. like there's so many new things happening that it felt like that was one big hump to get over of, okay, now most people have probably tried it out. Most companies are starting to allow access. Now people know how to use it. Where are we at? Where are we heading? Like what's next now? Let's just say for your team, like your team has been ahead of the curve.

25:47Nir Pochter:You've given them the tools, you've given them the training. And now it's also, you know, reminding people like, don't forget your human self. Don't just drop me a chat GPT response. That doesn't work. Like now what, What's next for a marketing team?

26:03Stephanie Postles:I think we're done with toys and are going to have tools, if I have to sum it. I think look at how people use diffusion models. They think of it as text to video or image to video, right? This is how most people call VRSR, for example. They say, hey, I'm using this text to video model, but this is not how those models are built. This is how they are exposed to us. I'll give an example. Those models, they take input and they create output. So one thing you can do and we're doing with our model is let's say that you have a problem. You shot an ad on portrait mode, and it was really successful. A great problem to have.

26:58Stephanie Postles:And now you also want to advertise it on YouTube. It's portrait. You need it on landscape mode. So you need to reshoot it. This is what you used to do. The models, they don't just do text to image, text to video, image to video. They can actually do this. it's called outpainting and they just take the portrait mode and generate stuff around it that looks like it was shot on landscape and of course for every aspect ratio to every aspect ratio so that's a tool that is actually solving a business problem right and even when i look when we look on on ad productions um i saw recently uh a tick tock by someone i think his name was nate b jones he's doing lots of stuff on on ai and he was talking about the coca-cola ai ad he said they used 70 000 prompts um and he started doing the analysis he said okay if that's how many they used think of how many they tried and how much money they spent on this and the manpower it was probably cheaper to just shoot shoot the ad i i think we're heading to a place where we'll have tools that are trying for example to give more control because if the model take inputs part of those inputs can be controlled when you give them a start frame and then end frame and maybe a frame in the middle and maybe you put your brand guidelines and maybe you have a way that the character needs to move its hand, right?

28:45There's so many things that you can add.

28:49Stephanie Postles:It's not just, okay, I'm writing a text and I'm getting a five seconds or six seconds video. Usually that's it's helpful sometimes but it's not that helpful i like this okay so from toys to tools that's

29:04Nir Pochter:the viewpoint which i actually love how simple and like i understand exactly what you mean by that are there any other examples like that because i love the one about you know you used to have to reshoot but then also being careful about you know how many prompts are you using and manpower you putting in to get this video versus maybe you should just actually shoot this are there any other examples like this of where you see the future heading right now in this world of um video it

29:32Stephanie Postles:connects i think to the point of ai not replacing people but people who are using ai replacing the people who don't because when i look on flows and ai flows again let's take you shooting something you put constraints, I think you're going to find, you will start looking for the expensive shots and start by replacing them, right? And not everything because it's so hard to replacing all production. You will start looking for the, either take your production and decide which parts of the productions that you want to replace, or you're going to take a specific type of production that is easier to do with AI.

30:21Stephanie Postles:Maybe it's easier to, maybe animation is more forgiving. You start by doing animation production with AI, and some of it you still do without AI. I think it's augmented tool when you're using AI, but also doing some manual work. I think now many times people, when they use AI for the sake of using AI, they try to do everything with AI, and then Then you get those, okay, I'm using so many prompts instead of finding the hard tasks that are easy for AI.

30:52Nir Pochter:So I mean, I'm thinking about this future that you're talking about and people really leveraging AI for this creative process. And I do wonder, at what point does everything just become like same of same? Because people are using the same models, same like typing in prompts, kind of getting back similar responses. and instead of competing against human creativity and like basically what I'm putting out as my creative project comes from my lived experiences, my wisdom, all the things that have happened to me versus now we've got these models that people eventually will be relying on heavily to create output.

31:29Nir Pochter:Like how do you make sure it stays creative and doesn't just get poured into a soup of all the same type of like content or ideas or feedback or the prompts that are being used?

31:41Stephanie Postles:I don't see AI taking the human out of the loop. I see AI changing the way that people are working. Let's take developers now. So developers, they have now Cloud Code and they have Cursor. They started from copy-pasting their code into the LLM, but now they have these systems that save it for them. And now it doesn't mean that you don't need developers. You still need developers, but you need developers that know how to tell AI how to help them code. And you still can't have just one person come to work and say, okay, Claude, do this for me. Okay, now do this for me. and replace the rest of the developers.

32:35Stephanie Postles:And you're not going to get the same output from every developer. And I think it's going to be the same for creative. At the end of the day, we have many creative folks here. I know how to write prompts. I never get the amazing results that they produce because they have the creative mind. And lots of working with AI is about curation. It's about knowing how to select the best results and knowing how to explain why they are better. So in a way, AI is just helping people become better.

33:24Nir Pochter:Yeah, yeah. I love that last sentence. I haven't really thought of it that way of like, it's you knowing how to select the best results. I've seen that so many times where like put a prompt in there, same person puts the same prompt and, you know, I choose a different one, someone else chooses a different one. And knowing what's good, depending on the context, like it really depends on the person who's picking that and also on the input that you're putting in there as well. So I love that reframe of like, are you still a good picker? Like, do you know how to select good things? If not, and you never were, it still might not work out for you if you still pick the bad output every time.

34:02Stephanie Postles:Yeah, I think there's two skill sets here, right? The first is you mentioned you need to write the right prompt. That's an out, okay? Many times when people think that AI can't solve their problems, they didn't experiment They didn't learn how to write the correct prompt. It's something that you need to learn. That's one skill. And the other is the professional skill. It's unrelated to AI. If I'm a good designer, hopefully I can select the best assets and can iterate on them. And that's where the gap is going to grow. And that's why AI is actually increasing, I think, the gap between people. Yeah.

34:51Nir Pochter:If they're not using AI versus if they are, you mean? No.

34:55Stephanie Postles:I mean, first, this for sure. If you're not using AI, you have a problem. But let's say I'm a copywriter. Okay. And let's grade my skills. And my copywriting skill is seven. and you're a copywriter and your copywriting skills are 10. And then I think of AI as multiplying your skills. So let's say that it doubles for simplicity. So I was a seven, I'm 14 now, you're 20. I improved, but you improved more because you're iterating fast, right? You know how to write a prompt, whether you know how to select results, whether the gaps are going.

35:48Nir Pochter:Have you ever wished that your entire video production workflow lived in one place? Well, you're in luck because that is LTX, the creative suite that was built by Lightrix, all for AI video production. It takes you from idea to final 4K video, all in one workspace. And under the hood, you've got LTX2, the next generation creative engine powering native 4K, synchronized audio and cinematic quality fast. So if you make content professionally, this isn't just another tool. This is your new creative home. Go explore it at ltx.studio.

36:25Nir Pochter:Okay, so we've been diving a bit into the weeds around the video production process and all the efficiencies that AI is bringing to it and how you should and shouldn't leverage it. If I zoom out and think from the CMO perspective, are there new questions they should be asking when they're receiving creative, they're receiving new branding, there's a whole bunch of new processes happening behind the scenes. Let's just say with something like Lightrix that they might not even know how things are being built anymore. What questions, if at all, should they start asking or looking deeper into with all these evolutions in technology that are happening?

37:00Stephanie Postles:So every CMO will obviously handle it differently depending on their technical level and their priorities. But if I try to think of the framing, I think that, okay, there's a new technology that's here. And it's going to really change productivity, right? It already started and hopefully will change productivity even more. Let's think of when computers started entering the field. If a CMO would ignore them, it can be a disaster in the first few years. Because after five, ten years, everyone will adapt. But the difference between companies who will be very successful in the next few years and companies who won't is how quickly did they adopt AI?

37:51Stephanie Postles:So adopt AI. So I think they need, as a CMO, you need to make sure that your team adopts AI for the expensive stuff. That if your competitors will do them faster and cheaper, they will beat you. So it's either you're into AI and you ask people, are you using this? Are you using that? Etc. Or you have someone you trust. And that's their role, to make sure, not by the way, not that people are using AI, but people are solving business problems, that they are increasing efficiency and creating more value using AI. That's a huge difference. That's a great parody post by someone in X. I think it's Peter Greeners.

38:45Stephanie Postles:He had a great parody post on a CEO that's getting ordered by the board to implement AI and how they do the process of buying seats and showing adoption. What is adoption? People who we bought seats for. Great. We have a graph. It goes right into the top. We succeeded. No, you need to make sure that people are creating business value using AI. And I think that CMOs need to make sure that this is happening. Not just CMOs, of course, but everyone.

39:18Nir Pochter:Everyone. Yeah, I love that. Like being curious enough to be like, okay, awesome. You just put something in my inbox. Like tell me about how this came to be and like what tools are you using? If at all, like are you using them? And just like being curious enough to dig in to see is your team using AI? but that also takes you knowing about it to know the right questions to ask and to know the tools that exist right now of you know being mindful of that to even have the right questions to ask of

39:43Stephanie Postles:your team too it's not just about using ai but what are they solving what's the problem they're solving if they're using ai all you know is that they are helping the business of open ai or google or whoever you pay for your tools you want to make sure that they also helping your business Good.

40:02Nir Pochter:Okay. So one other question now that we're thinking about questions to ask. I'm hiring. I'm a CMO or I'm a marketing team. And I see this new title of AI marketer, AI marketeer, like people just adding AI in front of their titles. What are your thoughts on this?

40:18Stephanie Postles:I don't like it. I think it's LinkedIn fluff, to be honest. The title implies expertise instead of literacy. It's create this, oh, I have this AI expert, so I don't have to know AI. I'm going to go to this AI expert and this AI expert will solve all my problem. But if we use the framework that there's like two parts, right? There's the form part and there's the selection part. The AI expert who only knows AI, he won't know how to do the second part. He won't know if the AI actually creates something good. I think that everyone in the company should know to a certain level how to use AI tools. Because it's like, okay, let's say that we have a spreadsheet marketer and no one else have to use a spreadsheet anymore.

41:16Nir Pochter:Yep. Yep. I love it. All right, Nir, well, we are going to move on to the lightning round now. This is where I ask you a quick question and then you have under a minute to answer. Are you ready, Nir? And knowing that I might just ask random questions, by the way, so it might not be at all what's on here, which is probably better.

41:36Stephanie Postles:So I'll answer random questions.

41:40Nir Pochter:Yep. Answer random questions. It's easy. It's like everyday life. All right. First up, what is a book that you've read or listened to in the past year that deeply impacted you?

41:51Stephanie Postles:It can be personal, business, marketing, AI, whatever you want.

41:57Stephanie Postles:I listen to books, so I just need to open.

42:01Nir Pochter:Yeah. Open up your Audible or whatever you listen to them on.

42:04Stephanie Postles:Yeah. Let me see. Which audible book did I like? Oh, I read Amp It Up.

42:11Nir Pochter:Who's it by?

42:14Stephanie Postles:Frank Slutman. Great book on execution and how to win. Stuff we have.

42:22Nir Pochter:I will check it out. This does feel like a book right up your alley. I like it. Okay. What was a line, a piece of advice, something someone said to you that really stuck with you and influenced your career?

42:35Stephanie Postles:So when we launched Facetune for iOS, it was a great success. And I think that we did really good marketing and we got good reputation in marketing. And then we launched Facetune for Android. And I was certain that it's going to be easy. And we launched and it didn't work. And I started telling many excuses to myself. and then at some point I remember I was in the kitchen and our CEO Zev who's my friend for 25, 26 years now, long before Light Weeks, he came to talk to me and I started giving an excuse and he's like, listen, these are our excuses and you need to understand when you are wrong and making mistakes so you can fix them.

43:31Stephanie Postles:And I thought about it and I was like, fuck, he's right. Then whenever I started, I started feeling that, you know, I have this feeling of trying to tell excuses. I remember that and it connects to something that one of the VPs here who's working here for 12 or 13 years, told me once, he told me that being a marketer is about thinking of a great idea, executing it, seeing it failed, going to sleep, waking up, having another great idea and trying it again with enthusiasm. So it's like...

44:13Nir Pochter:I love that. Yeah, I really like that. And I love the directness. Like you've had your own mean GPT over here just giving you good feedback like this for 25 years. This is what a blessing.

44:27Stephanie Postles:It is.

44:28Nir Pochter:Oh, that's so good. Okay. Next one. What is a marketing hill you're willing to die on? You get in your team meetings, you're like, I will not back down from this thing.

44:41Stephanie Postles:I'm not willing to get any document that didn't pass GPT or Claude or whatever LLM and got at

44:49Nir Pochter:least 85 so you have a rating system do you give it to your team for this where it's like it'll

44:54Stephanie Postles:give the rating on it no i don't send them the rating but they know that they have to uh when when we started people didn't always do it so they sent me something i was like did this pass gpt no okay so i'm not reading this you know you have to you have to do it first um it really helps but did you give them your prompt of like near's idea like did you have your own custom gpt that you built up with your brain inside of it that then they can run it by i i didn't send when when we started i sent people my prompt but now i think people know how to write their own uh but i think people i think people who work here know me by now every single has the things that they care about so they know what i care about yeah okay amazing because if not i would probably just

45:40Nir Pochter:take any recording you've ever done and just take the transcript and upload it to understand what you care about. Amazing. Okay. Last one. If every marketer listening could just master one AI skill this year, 2026, what should that skill be and why?

45:57Stephanie Postles:I have, the thing is I have two, but - Fine. Give two.

46:01Nir Pochter:Give two.

46:01Stephanie Postles:All right. The first is the out of prompting. It's so important. When you come to, So you can just get totally different results, and it's really, really easy to get the right prompt. First, you can search. People have done tons of guides on the right prompts. You can search on X, Google it. Or the best thing you can do is go to your LLM and tell it, hey, I want to solve this problem. What is the best prompt I should use? Usually, you're going to get a great prompt. Then you put a new chat, put it there. You're going to get results. The second one is if you come in and try to solve a very large problem, it's going to be very often hard for the AI.

46:51Stephanie Postles:So, okay, as we said before, I want to create this ad, one minute full ad, create it for me. It's hard. So there's an art in, you want to bring it to pieces, break a problem to pieces, that each piece is complex enough, but also solvable. And then the AI can solve those for you.

47:13Nir Pochter:Yep. I love it. Okay. Well, Nir, thank you so much for joining Marketing Trends. It was amazing hearing about the space that you're in and more about LightTricks and where you all are headed. so until next time can you tell me more about where our listeners our viewers can find you and Light Tricks

47:30Stephanie Postles:yeah so you can go to ltx.io where you will find links to our model and our products and if you're a creative marketing person I encourage you to try it and let me know if it doesn't work well

47:47Nir Pochter:there you go I love it I love it alright well thanks so much Nir appreciate you joining we'll see you next time Thank you.