0:00:00 – (Matt Widdoes): Welcome to another episode of Growth@Scale. I’m your host, Matt Widdoes. And today we have a very special guest in Moshi Blum, who is the Chief Marketing Officer at the mobile gaming company Beach Bum. His past experience includes a stint as Head of User Acquisition at Viber and General Manager at Adjust.com. He also spent time as a mobile marketing analytics and ASO platforms leader and as a keynote speaker for global marketing and user acquisition teams. Moshi, welcome to the podcast. I’m so excited to have you today.

0:00:27 – (Matt Widdoes): Thank you for coming along.

0:00:29 – (Moshi Blum): Thank you, Matt, for inviting me and I’m also very excited to be here.

0:00:33 – (Matt Widdoes): Great. Well, I was able to give people a little bit of background on you, but I always love to hear it in your own words. So for people who don’t know you, tell us, who are you? What do you do? Where have you been?

0:00:42 – (Moshi Blum): Okay, so actually, because we’re doing an episode on AI, right? And because I use an interpreter with all of my work, I ask chatgpt, like, tell me, how should I describe myself? And the answer that he gave me was so accurate and so unbelievable. Yeah, unbelievably simple. I want to kind of read it.

0:01:08 – (Matt Widdoes): Yeah, let’s hear it.

0:01:10 – (Moshi Blum): Right? So I wrote, please describe me. And he wrote back to me back that you’re straightforward and prefer responses that are short, accurate and simple. Driven by curiosity and passion for learning, you challenge yourself to achieve excellence. You seem focused on specific goals such as performance, data analysis, competitor intelligence and content creation. You have strong management skills balancing efficiency with compassion.

0:01:38 – (Moshi Blum): Also, you’re not an Apple product fan, which is very funny. Right. Besides the Apple products fan, this is something that we will talk about later. But like when, when I heard a lot of podcasts and every time that I heard people present themselves, they present what they do.

0:01:56 – (Matt Widdoes): Yeah.

0:01:56 – (Moshi Blum): What they did and like what was there in the past, but nothing is today. They didn’t talk about themselves, what’s interesting to them and what, what they do, what represents them. And I think that from, you know, the numerous conversations I had with ChatGPT, including one that I asked him, which computer should I buy, to my house? And then he recommended Mac. And I said, I don’t like Mac, I don’t like Apple products, so bring me something else. I think this is where he came up with, you’re not an Apple fan.

0:02:28 – (Moshi Blum): I think that he kind of captured me very well in the professional aspects of things, in the personal aspect of things. I’m  also married. I’m 41 and I’m married. I have four children. I live in Israel, Herzliya. I always lived here. I love Israel and I enjoy learning new stuff every day. And this is, this is my passion to kind of discover new things. And I think this is what makes me good at what I do and also what makes the job that I do so remarkably fascinating for me.

0:03:06 – (Matt Widdoes): Yeah, you know, it’s funny you mentioned that I’d also, you know, it’s interesting to hear the history of ChatGPT and what I’m. Now I want to go to ChatGPT and ask a similar question. But you know, you mentioned, you like learning new things. It’s something we look for a lot. I mean it’s, it’s part and parcel to how we hire but it’s something I’ve always looked for in, in teams and, and when building teams and it was something that was very important at Red Bull.

0:03:33 – (Matt Widdoes): Of their kind of nine traits that Red Bull looks for is something that they call at least 20 years ago they called lifelong learners. So we’re looking for people who, you know, cruise back in. Back then it was like you’re, you’re spending hours on Wikipedia instead of like dome doom scrolling TikTok, you’re hitting random articles on Wikipedia and you’re just curious. And I think that growth generally is such an amazing opportunity to learn because things are always changing and what worked yesterday may not work today. And just because it didn’t work yesterday doesn’t mean we shouldn’t try it again.

0:04:07 – (Matt Widdoes): And you know, for me personally as many years ago when I got into gaming, I was playing some mobile games and I wasn’t really familiar with the category but I loved it. And I realized that the category is something called RTS or real time strategy. And I zoomed out and I was like, oh, so somebody was like oh, then you’d love Starcraft. You’d love all these other games that are like really more hardcore real time strategy games.

0:04:30 – (Matt Widdoes): And I realized kind of at that moment I’m like oh, what I do for work is real time strategy. And that’s kind of why I love it because you’re, you’re, it’s like, push one thing down, something else pops up, you solve something, it suddenly comes back up. You have to kind of plan for the future a lot and kind of get things in a row and yield those, you know, as they come to bear.

0:04:47 – (Moshi Blum): So oh, that’s so accurate. And actually I didn’t think about that right. But when I was young I used to play Red Alert 2. All the time I knew all the voices of all the different soldiers and tanks and planes, right? And yeah, you’re kind of building something, then you utilize it and then you see the results and you learn from it. For your next adventures, for your current adventure, right, how to fix it, how to maybe attack a little bit smarter, how to find those loopholes which you can kind of go into the enemy base and conquer everything. So that’s in the game context, but in the live context, it’s more, more or less that, like, that’s what you do. Like, you’re hiring the best person that you believe. You train them, you give them resources, you give them responsibility, you lead them into battle, right?

0:05:40 – (Moshi Blum): And you expect to get the best out of it. And if you’re succeeding with that, then your product is growing. If you’re not, then you need to try a different strategy.

0:05:49 – (Matt Widdoes): Well, and to make it even more interesting and complicated is that on all sides of the equation are humans. So we, whether we are marketing to somebody or presenting them some new experience or whatever that might be, not to mention, like the teams we have to manage internally to get those things over the line. But like, humans are insanely unpredictable in many ways and will. But. And, you know, can turn on a dime from their preferences or any number of other things. Or a new competitor could come into the market and suddenly everything has to change or some new piece of information comes where you’re like, okay, well, everything that we mapped was planning on this other thing being true. That is no longer true, therefore, we have to scrap 80% of that plan.

0:06:30 – (Matt Widdoes): And for me, yeah, that’s what’s really engaging, is that unlike work that is maybe more like accounting or something, where it’s pretty much on rails and you follow Gap and you do whatever and you put the paper here and you do this here, where it’s super predictable. Growth is filled with… It’s inherently unpredictable. Even if you just look at, like growing a plant, you know, for the most basic sense of growth, you’re constantly battling multiple things.

0:06:55 – (Matt Widdoes): So today we wanted to talk about AI and its kind of place in the world generally, but also as it relates to marketing, product, creative, you name it. And so there’s a lot to unpack from that kind of bucket. And it seems to be on the top of everybody’s minds right now. I’m curious, like, at a high level, where do you see AI in today’s state and how it can be leveraged today? I think we should also talk about how you think that may change in the future. But like where are you seeing opportunities right now where you’re, you’re getting great efficiencies and outputs and leverage with AI inside of growth generally.

0:07:32 – (Moshi Blum): So to answer the second question first, I can see opportunities for AI in everything that they do. Everything that we do in marketing can be done better with AI. And I’ll present some of the examples that we did and some of the examples we are trying to do and some of the examples that I kind of failed or still failing. And hopefully because I’m not the most advanced person in AI, I just like trying to do that.

0:07:59 – (Moshi Blum): So eventually I wish myself to kind of succeed with that. And to answer the first question right, where are we in the process of AI? So I think that we are just making our first step into the AI world and I think that today it looks amazing. It is absolutely fascinating, breathtaking. I don’t know, compare it to something that we used to do two years ago. This is a new world, but still this is just a beginning.

0:08:32 – (Moshi Blum): People compare that to the invention of the wheel or the electricity and I think that they are right. I think that within a decade of using AI, we will be transferring it to something completely different. But you have to kind of start using AI. I think this is maybe the reason I’m going. I agree to talk with you about AI. I think that everyone in marketing should jump on the bandwagon and start using AI. It doesn’t matter if it’s just the basic stuff or if you’re progressing over time to the more advanced stuff and it really doesn’t matter what skills you have today.

0:09:13 – (Moshi Blum): By utilizing AI, you can do everything better and to kick in, write some examples. So we used to have a copywriter, content creator in our team. She was freelancer and we were very happy with her and like she was responsible for basically everything that relies on molds and content. Right. So when we launched a new game, when we wanted to kind of write, copywriting copyright for our game for ads, when we want to kind of produce a script for one of our videos, we used to go to her and she kind of gave us head start or providing the task that she was assigned to.

0:10:00 – (Moshi Blum): And when time progresses, I think it was like two years ago, I think that our way of effort, like we, I think she decided to go somewhere else and we kind of decided to kind of go without a copywriter. And I remember that the first time that we saw ChatGPT3, it was I think one and a half years ago. I was so amazed by how well the model learned what she did, and then, like, how well the model provided us with some copywriting, which I think at that point I understood that, well, we don’t need any.

0:10:40 – (Moshi Blum): Like, we don’t need copywriting anymore. Yes, we do have some people in the team that are still trying to sharpen some stuff and make it more agile or compelling to the target audience that we have. But basically, from our point of view, I think that we kind of buried the copywriter position in art.

0:11:01 – (Matt Widdoes): Yeah.

0:11:01 – (Moshi Blum): Yeah.

0:11:02 – (Matt Widdoes): Well, and. And with any copywriter, I mean, the challenge is everybody thinks they’re an okay copywriter. From what I found, there’s. I was in. In college, I was. I studied advertising, and I remember a kind of capstone professor, he had some saying which was like, I don’t remember. I’m going to butcher it. But it was something along the lines of, like, there’s no. There’s no stronger desire in the world than for one man to edit another man’s copy, basically.

0:11:27 – (Matt Widdoes): And so even when you have a copywriter, you typically are making minor tweaks and you’re like, okay, that’s pretty good. Let me do a little like this. And like, let’s do that. Right? And you’re maybe debating the use of a particular word or. Or tightening it or whatever. And so, yeah, we’ve. We’ve seen that as well, where you were able to get feedback if you can talk to it like a normal person and say, hey, can you give me something that’s a little bit tighter? Or like, okay, I like that, but I want to change this word. And it’s like, okay, and it’ll remember like, all right, well, you like that instead. So I’ll remember that next time. Kind of like your discussion with. With your own agent about like, I don’t like Apple. And it’s like, okay, I remember you don’t like Apple.

0:12:03 – (Matt Widdoes): And. Yeah. And the speed and the iteration is just like, it’s kind of amazing. And so I. I do think there’s still a place for copywriters in an AI future in that they are probably training their own models. And. And. But now, all of a sudden, one copywriter could lead a hundred agents and could craft the kind of. Can pick. Like, you know, there’s like, no, there’s still. There probably will be, but there’s kind of still no substitution for the taste of a.

0:12:29 – (Matt Widdoes): Okay, this is a little too advertising. Like, let’s tone it down a little bit. And. But again, the model will remember that and be like, okay, I shouldn’t be so upbeat and like, this is the best thing ever. Click here to buy. Like, okay, we don’t like that, but sometimes you might.

0:12:44 – (Moshi Blum): So, so, so I agree, I agree with you with like, if you want to be very precise, let’s say branding or brand marketing, right. Then you have to be very picky and you have to get your human touch that will kind of maybe select the best one and maybe even edit it a little bit more. But if you’re performance marketing, you know the rule of like, let’s put as much as we can and the one that will survive will be the stronger one and then we will optimize according to that and grew another line of like even better copy for our text ads and even better store descriptions for our applications in that specific world, then the value of your opinion, I would say is weaker than the value of all the persons that encounter with your ads and choose whether to download your game or not download the game.

0:13:40 – (Moshi Blum): I think that’s. And this is something that we provided and yeah, you know, the copywriter example was 1 1/2 years ago, right when just started. And I do agree with you that today you can, you can do wonders and you can do whatever you want with that. I remember one time that we kind of took all the reviews that we had from a single application and we put that back into ChatGPT and ask, look, analyze it and let us know what should we focus, what should we do next, what are our main problems and what should we emphasize better in our store front? And the model did an amazing job, like highlighting for us some key things that we need to solve, highlighting what makes our application best by correlating styles with good reviews and gave us a very detailed picture that we couldn’t do by hand because we had thousands and thousands of different replies. Right. And different comments.

0:14:40 – (Moshi Blum): So different reviews. Right. So it did a fantastic job in understanding the narrative, understanding what’s behind the world and what’s important for our users and gave us an actionable plan of what we should do next in order to make things work well.

0:14:56 – (Matt Widdoes): And to that point, there’s nothing preventing anybody right now from essentially scraping reviews from the App Store live as they come in every second, having an agent that’s designed to review those, compartmentalize them, track them over time, report, you know, in the, in the past we might take something like that and like the best thing that we could do, you know, five years ago would be like drop that into a word cloud where it’s just doing counts and it’s making like fun really big and then it’s like exciting really big. And you’re like, all right, it’s fun and exciting, but it’s like, how helpful is that? And there’s no real analysis there. It’s really just like volume because somebody could be saying the reviews could be over and over and over again saying this is not fun or exciting.

0:15:37 – (Matt Widdoes): And then all of a sudden you’re like, they think it’s fun and exciting because the word not shows up only half as much.

0:15:44 – (Moshi Blum): I remember an example that like, you know, we analyze what’s the worst word that can hurt our reviews. I think it was back then when I was the head of User Collision Viber and you know, surprisingly enough, we found that ads are correlated with negative reviews. And yeah, thank you very much for your help.

0:16:06 – (Matt Widdoes): Yeah, they don’t like the ad.

0:16:10 – (Moshi Blum): Yeah, yeah. Now it’s something completely different, right? And you know, you can also mix and match between your competitor description, your competitor reviews and you can basically learn what to do better, how to differentiate yourself or even, you know, how to copy from the ones that are actually successful. If you go to application that your ASL tool is telling you that this is the best application visibility wise and stuff, you can analyze and see what’s make it so unique, what makes the description so unique and then like ask the same, the same tool to maybe write you your own description with your own for your own tool, for your own game. Now I’m not saying it’s working for everyone, but if you’re a small team.

0:16:57 – (Matt Widdoes): Right.

0:17:00 – (Moshi Blum): With a limited budget and you want to run fast, because every small team wants to run fast and you have to maybe get some shortcuts. And I think that using AI will be the perfect match in creating some valuable shortcuts and getting or extracting the value without wasting a lot of valuable time, I think that’s the establishment of AI. That’s what I-Scream is meant for, right? Helping small teams or helping individuals to become very powerful at what they do and how that, and a lot of.

0:17:36 – (Matt Widdoes): That work that we just talked about is like mind numbingly boring for a human too, right? So if you took a copywriter and said, okay, go read these 2,000 reviews and do a word cloud and come up with your own things and give me more, more copy, more copy, especially at scale and especially in situations where they’re oftentimes not really briefed on the performance of their last round of copy. And in some cases even if they were briefed, they’re like, okay, well I’m not like an analyst. Like, okay, this one did better. Like I can see that. But like the analysis of that may be like, okay, well I don’t really know where it is. And especially at scale, I mean we would have, even at Zynga and King, oftentimes the media buyer would be the one writing the copy. We didn’t have a copy team creating.

0:18:20 – (Matt Widdoes): And usually you’d have some bank of top performing copy in some notebook somewhere that you just copy paste in. You’re like, all right, run that again. Or you’re literally copying it, copying a previous ad, swapping out an image, changing some targeting or some bidding and like it just has the same line of copy over and over. And if you look at a spreadsheet for a game, spending a million a day and you just sort by copy, you’d have like 12 variants and those variants hadn’t changed for two years kind of thing. Right?

0:18:45 – (Matt Widdoes): And so being able to feedback that data. Same thing with things like Mid Journey or others where you’re like, or some, you couple that with something else where you’re saying, hey, here’s the data, here’s what we saw. Go create me more stuff that you think will be better. And like your mission is to give me better performance. And here’s what we mean by performance again, it’s just like it takes off some of the lowest hanging, least satisfying work.

0:19:09 – (Matt Widdoes): One quick example I’ll give on that, that was presented to me a long time ago is in the 1920’s, you know, bowling. Like you just go to a bowling alley in the past, whenever you knocked over all the pins, it was like an 8 year old that would just sit there and pick them all up and put them back like where they needed to be. There was no machine that was doing that. It was not great work.

0:19:31 – (Matt Widdoes): It was not something. So like when that, when, when AI or when technology came along to replace those kids, of course there were a lot of people who were like, well what am I going to do now? And it’s like, yeah, but that wasn’t rewarding work anyways and those kids found other jobs. But like, I think, you know, I think one big lesson generally in technology is like, and as you mentioned, like people should be jumping on the bandwagon. Is that like, just because it sucks today doesn’t mean it’ll always suck and if you ignore it, you’re going to wake up five years from now and still probably be ignoring it and being like, yeah, it’s not that great.

0:20:02 – (Matt Widdoes): And it’s gonna, you’re gonna find out one day that it does what you do. And if you had just been paying attention and you had been feeding it all the things that it does better than you, like copyright. When somebody comes and says, all right, well this thing can do your job, you’re in a position where you’re like no, no, no, it can’t do my job because I use like 15 AI things already to do my job. What I’m doing is the stuff that the computer can’t do because I haven’t, I haven’t found a computer that’ll do it yet. So if you know of some other thing that can do what I’m still doing, because I’ve already given 90% of what I do to something that can do it, you’re just ahead of that curve.

0:20:35 – (Matt Widdoes): And those are the people that I think will have some much value. Not just now, but I think heading into the future where they’re like, oh yeah, yeah, I, I specialize in all of the automations and, and that stuff and I, I’ve built these things in the past and I. You don’t have that. I can do that. And yes, I, with my knowledge of all the AI and, and what we’ve done in the past, I can replace 20 people within your company because nobody in your company took it seriously.

0:20:57 – (Matt Widdoes): And that is like part of, I think that being that conductor and that scientist kind of behind the scenes is for sure where everybody, whether you’re in legal or marketing or product, like everybody should be leveraging AI in the area.

0:21:11 – (Moshi Blum): I totally agree. Like, we also have a good example when going to make creative with AI. Right. So you know, when Midjourney just launched, I guess like one year ago, we also were very busy in trying to kind of mimic our ads, the one that we created by ourselves into the Midjourney and how to create maybe a Beckham on board with Midjourney. And we’ve made countless attempts into prompting a Begum on board which is like one of our most successful games. Right into Midjourney and we failed. We felt miserable.

0:21:47 – (Moshi Blum): Too many hallucinations and too much creativity in trying to portray a simple bag on one board. Not every time you succeed but, but the fact that we were able to think or portraying what we need to do or what we need to kind of create in an AI gave us a lot of advantage because then when companies started to pop up, company that produced high value imagery and then like high value video creation. Right then we were in a position that we can choose the perfect partner to work with and to generate for us creating for our music acquisition. And by the end of the game, you know that in performance marketing, if you produce a lot of videos and a lot of maybe good videos or good enough videos, you can choose, pick your winners, learn from that, iterate on that and pick better winners and better winners until you will have yourself the perfect campaign until it fatigues, but the perfect campaign for a while. So by choosing the right partner for creating video, we will manage to cut costs dramatically while giving us more videos that we need.

0:23:09 – (Moshi Blum): And you know, our problem went from how do we produce more video to how do we test that enormous amount, enormous number of videos that we produced. And you know, just to kind of go back in time. I think it was also. Yeah, it was Viber. Back in 2016, 2017, we didn’t have any video ads, right? Because we didn’t have anyone in the team that creates video ads. So we paid a lot of money for an agency just to create something which I consider today a low quality video, right? And the amount that we pay was outrageous in today’s terms.

0:23:52 – (Moshi Blum): And we got ourselves videos and then we needed to edit them. So it was an additional charge. And I don’t know how we came up with like a couple of, I think tens of thousands of dollars for just creating a campaign and set of videos to that campaign. And today, you know, today from where we stand right now, looking to the past, that was embarrassing. Today we can produce videos with 10 times more quality or better fit to our needs at a fraction of that cost.

0:24:27 – (Moshi Blum): And I think that this part makes AI great for marketing. This is what makes AI great for a creative team, which is, you can create something, then you can iterate it with AI, you can maybe be concepting some production with AI and you know, it’s not perfect, it won’t be perfect in the near future. But if you combine, as you said, if you combine the expertise of a specific human being, a specific great designer, great After-Effect creator, with the abilities and the capability capabilities of AI today, you can create something really great at a fraction of the cost that it cost you.

0:25:07 – (Moshi Blum): And it can look almost cinematic in the quality of the thing you produce.

0:25:15 – (Matt Widdoes): Well, and the cost of that and the processing is going to come down and be faster. I mean, I actually think it will be very good very soon, like within a year maybe because there’s just so much incentive to do it. So like let’s say you and I had a hundred million dollars and we’re like, okay, well, then what might we do if we were trying to build a company like that? Okay, well, like, okay, let’s just start on Facebook and go to the biggest advertisers and scan. Go look. Because we have access. We can. You can access that to see what are they running.

0:25:42 – (Matt Widdoes): And. Okay, let’s scrape all that. Let’s do it for every vertical. So here’s a bunch of e-commerce stuff that’s low cost, low consideration. Here’s some stuff that’s high cost, high consideration, like a computer. Computer. All right, cool. Let’s do it in gaming, but let’s split that out in Match 3 and in RTS and casual. Whatever. Great. Got it. We just inherited all of supercell’s ads. Great, thanks. And we know that style.

0:26:03 – (Matt Widdoes): It’s all public. Okay. And you do that like a hundred times. And then you say, okay, now, computer. And we’ve tagged it appropriately. And you’re like, all right, give it to me in supercell style ads. Okay, now give me a blend for e-commerce. Give me a blend of all the stuff that you see on TikTok or all the stuff that you see on Facebook or Meta, and just do it in that style. Okay?

0:26:22 – (Moshi Blum): Ish.

0:26:22 – (Matt Widdoes): And by the way, like, here’s also my brand guidelines. So take these things. Marry that with what you know. And like, yeah, these look like ads because you can get creatives who’ve never done performance marketing, and they’re like you. They stand out. They look like a flyer that you would see for a karaoke night at a bar or something. You’re like, this is not a good ad versus kind of like rounded edges and bright colors and things that really draw the eye.

0:26:42 – (Matt Widdoes): And. And it’s not that far from being really good, as you mentioned, and cinematic, but certainly in GIFs or kind of slight movement and coins kind of bouncing around, it’s like, yeah, we see a lot of bouncing coins. Great run. Bouncing coins. And so like, that. If somebody had that today and they just said, hey, this gets you like, 98%. Like, these are pretty damn good. And it is a thousand dollars a month for an enterprise license. Like, okay, that literally puts every creative shop out of business.

0:27:09 – (Matt Widdoes): And you wouldn’t. You could charge 10 times that because the volume is unstoppable. Right. And so anyways, that.

0:27:15 – (Moshi Blum): By the way, I don’t know if you’re describing the future, right, but there are companies that do this today at the present. Like, we are working specifically with a company like this. We understood that, like, you know, a long time ago that we are too small to build ourselves very accurate models and then to try and to kind of integrate.

0:27:37 – (Matt Widdoes): It’s a totally different business. Yeah.

0:27:39 – (Moshi Blum): So we went to a third party, a third party partner that creates AI. Actually, we have two at the moment, and one of them actually provides the exact product that you’re describing.

0:27:52 – (Matt Widdoes): Yeah.

0:27:52 – (Moshi Blum): Look at the video that you like. This is our shop. Right. And they give you all the Facebook library and choose one of your competitors. Choose the video. Click on that. Within a couple of minutes, you will have Copycat, which is, you know, inspired by the original video, but not similar to the original video because you don’t want to violate any copyrights. Although ads don’t have any copyrights usually.

0:28:21 – (Moshi Blum): Right. And that’s it. And you have it and you can run it, and it’s that simple. Within a couple of minutes instead of within a couple of weeks, you can kind of do whatever you want, which.

0:28:32 – (Matt Widdoes): Is, you know, and this is still early. So. Yeah. So when I think about, like in a year, it’s like, okay, so now I have my own repository and they may be doing this as well, but I have my own repository. I’m at scale. It’s learning, it has my style. And meanwhile, I want it listening not to the market. I want it listening to four major competitors. And if they’re, if we’re seeing them massively change something, I want to be notified. I want to see what they seem to be running and what we think is working.

0:29:02 – (Matt Widdoes): I want to see proactively. Show me kind of what we would run if you had done that. But like, I wake up Monday and I have an alert that says, hey, we see some movement here. You should be aware of it. We. We recommend trying it. It’s like, okay, great. Whereas there’s no creative person on the planet that does that now as a human or can do that at that level. There’s no media buyers that are doing that because they’re so focused on the numbers and the spreadsheets. They’re not. Like, it’s very rare that they’re looking at massive creative output, like exports and. Or like, well, they might, but they’re not doing that super regularly. They’re certainly not doing it every hour, every minute of the day. And they’re not doing it.

0:29:36 – (Matt Widdoes): They just don’t have that much time. It’s a totally different job. Like, they need to be working on the media. Right. And they probably look at that stuff once every couple weeks. But the real deep dive competitive stuff in most companies is not being looked at in a serious way. And if they do, it’s a full time person that does that and that’s all they do. So I’m curious, like how do you think about just one last point on the creative side with the technology that we have today, how do you think about maintaining kind of authenticity and meeting brand guidelines and stuff like that? Is it just kind of sitting a human in between the bot? Is it training a bot? Is it a mix of the two?

0:30:09 – (Matt Widdoes): Where do you, where do you kind of approach, how do you approach that?

0:30:12 – (Moshi Blum): So first of all, we have humans, like humans. I have very, very professional employees which I adore.

0:30:20 – (Matt Widdoes): That runs the entire, just run-of-the-mill humans. Yeah.

0:30:28 – (Moshi Blum): You know, that runs the process and you know, that kind of examine every creatives and make sure that we are not doing any crucial mistakes and that we don’t have anything that is discriminating or offensive towards any type of population because we, you know, we cherish our users, we don’t want to offend them. But also like, you know, you asked about how we create our own authenticity and that’s something very interesting. Like I want to share with you a story about how we thought that our music on our ads is not that great.

0:31:08 – (Moshi Blum): Like probably as all other advertisers, we use the kind of open copyright music that we kind of find and items that we created with past content creators. And we kind of, we had a library of music and we also have like another music producer in our studio. So he kind of helped us with some music. But with time we thought, well, we need to maybe level up our game. So my creative team decided to kind of compose some music to our ads. Right.

0:31:41 – (Moshi Blum): As a differentiator. And of course, you know, thinking about that 10 years ago, if you wanted to compose the music for your ads, you need to kind of hire a studio, hire some, some band, write some songs. This process is very expensive and time consuming but for us we are just like hey ChatGPT, please write us a song about our games. And he wrote several songs and we asked it one time to be like rock and want them to be like pop and rap.

0:32:14 – (Moshi Blum): And he did all this mix and match of like okay, those are the songs. Then we kind of opened an account in Sono and paid for the account so we can kind of use it commercially. And we uploaded all the lyrics and here we go. You have some songs, right, that you can use in your ads? And the funny thing is that after running those ads with the song that we have, people start to kind of inquire who is the performer, who is the artist and where can we find a song outside of our ads?

0:32:50 – (Moshi Blum): Because I. They are really good, right? The songs are really good. So we open a Spotify channel and that’s like we have, we have one game which is called Spades Royale. And in Spades Royale you can play Blind Nil in which you don’t see your cards, but you gamble or you place a book that you will take, you will not take any cards. So we call the artist Blind Nil and we uploaded some, some songs from Sono and now we have, we are running an account for an artist on Spotify and that, you know, that we can hear or we can listen to our music, which is great. And every time that now a user is asking us or a follower is asking us on our Spotify.

0:33:41 – (Matt Widdoes): Spotify, yeah. Which is such a great, like, you know, you talk a lot about, in, in business generally, but certainly in gaming is like surprise and delight moments. And that’s like such a, it’s such a low lift for you guys, but it’s such a, like what you have a whole Spotify account and like that’s the kind of stuff that gets people. Your biggest fans get like because, you know, it’s your biggest fans that are looking for that type of stuff.

0:34:02 – (Matt Widdoes): Typically it’s not like I just saw you and I love this song. But maybe, maybe that does happen occasionally. But that’s, yeah, that’s such a fun thing. And like in the past that would have been impossible and you would have, I mean not impossible, but you would have had to have like it would have just been a hundred times the work.

0:34:17 – (Moshi Blum): And yeah, I remember one, like my friend back at high school wanted to kind of record an album and he said to me, well, I need like $20,000 to start recording an album. And it’s going to listen like it’s going to. The sound is not going to be very quality, like very quality sound. But for just like the initial steps of recording an album, I will need like $20,000.

0:34:39 – (Matt Widdoes): $20,000.

0:34:40 – (Moshi Blum): And like now you can do it in like I don’t know, $20, $20 a month with all the subscriptions, all the different subscriptions. Right? Man, $30 a month. Amazing.

0:34:50 – (Matt Widdoes): Yeah. And I think. And it’s like a lot of that is in the ideation. Right. So even. Yeah. Using the music example and thinking of a band is like somebody has to in the past, you’d have to either share a bunch of songs that you like from other artists. So you say I like this song, I like this song. But you can’t just take those, they’re already copywritten. You’d have to write a song behind the scenes and then present it to a band and say, okay, what do you guys think about this? And someone’s like, I don’t really like this part. But I also don’t want to destroy your creative energy.

0:35:20 – (Matt Widdoes): Whereas now it’s like yeah, with something like Suno or whatever you could be like, okay, like this, don’t like that. Like this, don’t like this. And we own the rights to all of them. And we. It may be just faster instead of saying hey, change the baseline. Like no, okay, we, we’ve got the premise now we’ll just take that and we’ll do the minor changes in the room. But you could write at like a hundred, maybe a thousand times the speed and be before anybody put any real personal creative energy into something, you can already be aligned on what you want to create so you don’t end up with like a band that has, you know, a hundred songs and half the members hate half the songs and somebody really loves this one song. But because that would have just been killed in the ideation process when nobody has any energy put into it. Because it’s very easy to say I love this thing that Sono created.

0:36:07 – (Matt Widdoes): You hate it. Okay, whatever, move on. Like, you know, there’s some middle ground we’ll find where we both love it. And that’s just like not possible in the normal. I don’t know.

0:36:17 – (Moshi Blum): And by the way, like, you know, I’m not a creator. Creator can maybe visualize what they are going to create and by the end of the day they will have like this perfect picture before they’re recording.

0:36:30 – (Matt Widdoes): Yes.

0:36:30 – (Moshi Blum): But I’m just a regular person and for me to be able, you know, within five minutes to get three versions of my songs and then choose the end outcome of that song, that’s something that is remarkable.

0:36:44 – (Matt Widdoes): It’s the same on I think creative, you know, for, for marketing or whatever is that I don’t have the words beyond. I like that. I don’t like that more or less as a non creative, which is really frustrating to a creative because they’re like, okay, well what do you want? Like show me an example of something that you do like from somebody else and I’ll use that. Whereas you could very quickly get like a hundred concepts on a page that are wildly different.

0:37:04 – (Matt Widdoes): Circle the ones that you like. That creative person can then say, okay, here’s kind of what I think the consistencies are. Feed that into a model and you know, it’s just like it just say like so much of the effort in any sort of creative pursuit is in the ideation it’s in like maybe the strategy. If it’s, if it’s a, if it’s a business context, it’s. And it’s like now we can skip all of that and just get to final tweaks.

0:37:28 – (Matt Widdoes): And so, and that’s gotta be really empowering for a creative where they’re like sweet and yeah. To have that. If I was a creative, like a designer or something, I would just be using that all the time and be like, yeah, like, and here’s the ones I like and you’ve got some tweaks. Okay, great. I didn’t spend any energy on it. And I can get you another 100 in like five minutes. What do I care? It’s not like I feel like I wasted the last two weeks. It’s like that was just like that was what I did.

0:37:52 – (Moshi Blum): And by the way, like, we do need our creative, of course our artists, right. They are the one that kind of creates the seeds for then the AI to take that to the next level. So you actually need to have good seeds. Whether it’s like your cards, your gameplay. This is something that the AI still can’t understand fully. Like what’s logic, what’s not logic, what should I use, what’s the rules of the games and what you cannot do in a game. So you have to get this seed and from there, you know, changing backgrounds, adding some, some, some images, some, some, some figures or some characters, making it pop or making it stop or making it more musical or silent. That’s something that the eye can do better than a human being by suggesting what you should do and what you shouldn’t do. And then by the end of this process, you still have to have this like eye of a person that says, well, I like it, I don’t like it. I think that this is not up to its full potential and I think this can be a wonderful take and we should kind of try it on our, on our live campaigns.

0:39:09 – (Matt Widdoes): Well, and like not to scare the creatives in the world, but we’re not that far from that either. Where all it takes is enough smart creative people to share their thought process and say, hey, this is too, we don’t want stuff like this for this reason. And like, I’m making these choices because I think this is what you missed. And the AI can be like, okay, got it. And again, they can refine it over time, just like a human would, where they say, well, you told me last time you didn’t like big bubbly things.

0:39:35 – (Matt Widdoes): And you’re like, well, okay, sorry, sometimes I do, but not for that game or not for that channel. Or  I changed my mind, I like bubbly things now. Or. And it’s just like, okay. But like, over time, a human would also iterate and say like, okay, well, okay. So it went from you don’t like bubbly things to sometimes we do when these conditions are true. Yes, yes, great. Come again? These conditions are all true. I gave you the bubbly thing. Oh, I don’t like it because I forgot to leave out a condition. Oh, okay, well, great, I’m learning this. But that could be many, many years of feedback. And the human’s incapable of actually remembering all of history, which is actually 36 months ago, you told me not to do that. And it’s like, how did you remember that? It’s like, okay, yeah, I did say that. I forgot about that. But here’s some new rule that didn’t exist. And so you can build on recency of feedback, the full historical context of feedback with no loss of clarity or data.

0:40:26 – (Matt Widdoes): And so I think, yeah, I think that side is really interesting. And I know we spent a lot of time on creative, but like, that’s one of the most obvious areas, by.

0:40:33 – (Moshi Blum): The way, come to think about it, right? And you know, moving into user acquisition may be a part of it.

0:40:41 – (Matt Widdoes): Optimization, part of it, 100%.

0:40:43 – (Moshi Blum): You’re creating a video, but then you feed it into the network Facebook, just to be. Just for an example. And then Facebook gives it to their algorithm. And what you’re basically trying to do is trying to crack the algorithm. You’re not trying to do the best creative for the best audience. You’re trying the algorithm. So the algorithm will say, well, that’s the best creation for this specific audience.

0:41:11 – (Moshi Blum): So what you’re doing is like, and this is maybe funny, you’re doing ping pong between humans and machines, right? You have your human artist that creates something with AI, which is the first ping pong, right? Then you’re going back to the creative art director, which will say, well, let’s go in. That’s not going in. Then you upload it into the machine again. And the algorithm on the other side examines, well, whether that should be good or not for the target audience.

0:41:46 – (Moshi Blum): And then coming back to the user position manager, which kind of decides whether, should I upload it into the main campaigns and whether that would be my next best creatives or not.

0:41:58 – (Matt Widdoes): Well, and it’s not far where that’s a closed loop where it’s just like it’s all just taking place on a platform. Yes.

0:42:03 – (Moshi Blum): So like you have humans and by the way by the end of the day you have the user which is also human or at least we hope it’s a human that is downloading the app and then like start playing the game. So it’s a ping pong between humans and machines. I guess like in the future we will cut down the interaction that we have between humans and machine and like, you know when we will know that we kind of made the entire AI transition, when the AI artist will talk to the AI algorithm and they together they are going to decide whether this is going to be a best creative or not without running it on our target audience and they will choose just the best creative to just upload it to the campaigns.

0:42:52 – (Matt Widdoes): Yeah, it’s exciting and yeah, and there’s, I would guess that like maybe from an adoption standpoint where people are really taking a serious crack at this versus kind of dabbling in it and being like, oh yeah, we put that we do some copy and chat GPT versus like no, we’re, we’re running full production, we’re, we’re feeding back the data, we’re letting, we have multiple things that are working. We have like three agents working as a team where one is our analytics bot, one is our creative bot, one is our copy bot, one is our strategy bot, whatever.

0:43:19 – (Matt Widdoes): And those are all talking together and like they’re still a little wonky but like we’re figuring that out because it’s worth figuring out. I think if you take that type of approach, that’s maybe still sub 5% of the market. If I had to guess, like there’s not really many people super taking it seriously. If you look at the people who were, you know, they’re like, yeah, we’re creating seven mid-journeys and we have a human doing it kind of lighter past that maybe 15, 20%.

0:43:42 – (Matt Widdoes): But like I think no matter how you slice it, there’s like easily 50% of people who were just like, yeah, we’re not using it. It’s not very good. They may be at, I don’t want to throw any companies under the bus, but it may be somebody like Cisco who’s like, yeah, yeah, not important. Or somebody who’s really removed from technology.

0:43:57 – (Moshi Blum): Maybe, you know, you know that AI adoption is something that is interesting. So I’m asking all of my friends. Ask. Yeah, ask all of my friends, right? How are you using it? How much do you use on that day basis? And it feels like not a lot of people are using it or less than you would have expected them to use. Maybe it’s like Bitcoin, right? Everyone is talking about that, right? Not a lot of people actually have Bitcoins to hold.

0:44:29 – (Moshi Blum): And you know, everyone says to themselves, oh man, I should have jumped on this bandwagon when it’s costing us like hundreds of dollars, right? Not now. It’s nearly $100,000. It will pass $100,000 within like a week or two. And I remember very clearly, I think one night 10 years ago when one of my friends during a dog walk tour, he told me, well, there is something new which is called Bitcoin, I think.

0:44:58 – (Moshi Blum): And he said, well, buy some. It’s just like $300 per Bitcoin. I said, well, I don’t believe in fake money.

0:45:08 – (Matt Widdoes): But you do. We’re all fake money.

0:45:10 – (Moshi Blum): Yeah, it was back then, right? I end up buying a Bitcoin, right. But it was like five years later and I think 40 times higher than what he kind of advised me to buy. So, you know, it’s the. Is the thing that you’re missing in the present and then like your. When time progresses, you kind of feel sorry you didn’t adopt them earlier.

0:45:38 – (Matt Widdoes): Well, and I think, I think with Bitcoin, like, I can very easily understand and having been in similar spot where there’s a lot of questions on the feasibility, the longevity, the durability of it and like, okay, what is it? And it’s like every time somebody explains it doesn’t really make much sense. AI for me is like a totally different camp for me, AI is much more like somebody in 1997 being like, nobody’s ever going to use their credit card on the Internet.

0:46:03 – (Matt Widdoes): It’s actually past that. I understand in like ‘97 you might believe nobody’s going to use their credit card on the Internet. Okay, but nobody’s going to have a. Not every company is going to have a website that’s probably a little bit less believable. But like, the cat’s out of the bag on this. And like, I think anybody who’s like AI and I, granted, I think we could very well be in a bubble with AI. And whatever. It’s like we’re too excited too soon.

0:46:24 – (Matt Widdoes): But there’s no denying it. Like, unless we delete all the computers in the world, like, it’s already out. And so the idea that in like a hundred years, I actually have even been questioning, like, do you even really need a high school education anymore? I’m not sure. Like, you need, you should know about, like, you should have some program. Like, you should know about various civilizations that have come and gone and wars, and you should know geography and you should know politics and economy and all these other things. But do I really need my kid sitting in a classroom eight hours a day learning at this, like, snail’s pace where you could just say, hey, why don’t you just keep exploring what you’re interested in on ChatGPT? And we put some filters.

0:47:06 – (Matt Widdoes): When you run out of stuff to look at and you’re finding yourself not wanting, you’re like, I don’t know what else I want to learn. Okay, did you look at the Roman Empire? Did you look at, you know, whatever? Did you look at any of these traditions? Have you learned algebra? Like, you should learn algebra. That’s going to be a pain. But like, it’s helpful generally, but you don’t technically need it in application. It’s just maybe from like a system standpoint there are certain things, but like, I’m, I’m like, I’ve, I’ve had thoughts, I’m not there, but I’m like, should I just like teach my kids everything there is to know about AI and like, just train them to be, you know, developers now? Because I feel like the amount of time they have to wait to get out of high school and go to college and do all these other things, I’m like, this is a total waste of money and time. So not all the way there, but I, I am like of the camp that there’s zero chances of going away.

0:47:53 – (Moshi Blum): I agree. Like I, I told you before, right? I have four sons. The oldest one is 11. And yeah, I, I, I, I sit with him, you know, every, let’s say once, once a week or once a month maybe, to be honest, because we do other stuff like solving math problems and learning. But rather than that, like, I, I’m sitting once every month and I, let’s do some magic and we designed some cards or alternative cards for Pokemon on Dall-E. Right? Or I tell him, look, I know that you write some presentations decks to your school, so let’s try to kind of create a deck in less than 10 minutes and we do that right on ChatGPT and some, some other AIs blocks AI which is like creating everything that we need and copy paste it into a Google Slide and boom, voila, you have what you need.

0:48:57 – (Moshi Blum): So I totally agree with you. You have to invest now in order to be professional later. And we’re at the forefront.

0:49:07 – (Matt Widdoes): I mean this is like the beginning. In five years, if you spend the next five years, literally all you do is you just focus on the latest technology and AI and learn how to harness it and how they interplay together and what does and doesn’t work. In five years you will be one of the leading experts in AI because nobody, it’s like we’re all at the start line right now so deeply, deeply believe in going in, going heavy there.

0:49:29 – (Moshi Blum): Like one thing that you know is still missing in this ecosystem of marketing, something that I haven’t done or that I’m not feeling comfortable yet is data analysis with AI. Right. I know you can upload some, some stuff to your AI and you have some models and you can train some agent and I try to do that. Like very basic stuff but I try to do that. So I don’t know why. When it comes to creative it does an amazing job. When it comes to content, it does an amazing job. What. But when it comes to this kind of data analysis, I’m still stuck.

0:50:07 – (Moshi Blum): So maybe, maybe I’m just experienced. I just need to wait and be more patient and within one year that will also be amazing and maybe I’m doing something wrong. But I think that once we will also check on the data analysis part that will signal the beginning of the end for human user acquisition managers. Like or maybe maybe not the beginning of the end but also like the new job or new definition of what is a UA manager.

0:50:47 – (Matt Widdoes): Yes, I agree and I think I am both right.

0:50:49 – (Moshi Blum): We both grew up as UA managers, right. We started as a media bias and we progressed with the job and it kind of changes throughout the years from let’s think about what’s the interest of my audience in order to target them. Now it’s more like okay, let’s be very efficient on what’s the goal, what’s the goal of my campaign and how well it serves my maybe further progression of the user.

0:51:25 – (Moshi Blum): But I think that we have to say, unfortunately I think that we were the first and the last human user acquisition managers in our company.

0:51:40 – (Matt Widdoes): Yeah, it’s a different paradigm and yeah yeah we’re in all we’re a dying breed for sure. Well, and I think that even all of the data analytics stuff, that will all come, that can’t be that far because if you take 10 of the smartest data analysts and you marry them with 10 of the smartest media buyers from a context in different forms, right? People who were here 15 years ago, people who are here today, and you were just in there with a thing that’s listening and learning.

0:52:07 – (Matt Widdoes): And again, you can create multiple agents that work together. That’s not. And, and you’re like, okay, here’s what we want for statistical significance. Here are different Bayesian models or whatever that we might use for xyz. And we’re going to give you room to operate and learn. And like, we’re going to correct you just like we would correct anything that we’re teaching. It’s like, okay, like, but like, I can data, I can transfer all of that knowledge to you and I can realize through your errors what I forgot to transfer because I’m like, oh, yeah, I didn’t mention that. But like, this is some edge case thing.

0:52:37 – (Matt Widdoes): And here’s what I would view. This is what I read from this data. And what, by, by the way, what they see might actually be unique. Where I’m like, oh, I didn’t see that. Like, that’s interesting. Like, yeah, let me go do something else behind the scenes to kind of pursue that theory of some pattern or something that maybe didn’t jump off the page to me or where we’re looking at weekly reports and. But you didn’t forget like the last 15 years. And you’re like, wait, but here’s this pattern that comes every October.

0:53:02 – (Matt Widdoes): And I’m like, oh, that is a cyclical thing. Like, yeah, I agree with that analysis. Right. But it’s not that far to have that. And then you have. We haven’t even touched on things like in product and how we can have things that are learning as you go through onboarding to kind of streamline that at an individual personalization level based on billions of records of other people just like you who clicked just as fast or didn’t or whatever.

0:53:25 – (Matt Widdoes): So I, I know today we spent a lot of time talking about creative. I’d love to do another one where we go into the other areas where AI can be applied. But just being respectful of time. Really appreciate the time we had today and look forward to doing it again in the future.

0:53:39 – (Moshi Blum): Yeah, I actually, it was really nice. I enjoy a conversation. I definitely would like to kind of go back or return to the podcast within, like, one year and maybe update on different stuff that we’ve managed to do and succeed to do and, like, what has changed.

0:53:58 – (Matt Widdoes): I love that.

0:53:59 – (Moshi Blum): It was a pleasure, Matt. Thank you very much for hosting me.

0:54:02 – (Matt Widdoes): Yeah, same here.

0:54:03 – (Moshi Blum): And lovely to kind of meet you again.

0:54:06 – (Matt Widdoes): Yeah, for sure. Thank you. Thank you for the time as well.