Veed's Sabba Keynejad: 99% of Companies Shouldn't Build Their Own AI Model

Speaker A: I think like 95%. Sorry, that's wrong. 99% of the time, it's definitely the wrong idea.

Speaker B: Hey, everybody. Welcome back to Agents of Scale. It's the show where we sit down with the operators and builders that are turning AI from experiment into infrastructure. I am your host, Wade Foster, the CEO of Zapier.

Speaker C: And I'm Dan Slagan, SVP of marketing at Zapier.

Speaker B: And today we are talking with Saba Kinjad, the co founder and CEO of VEED IO Video. Uh, VEED is an AI powered video editing platform used by over 10 million people a month, including 75% of the Fortune 500. It's a $40 million company that could have kept wrapping other people's AI models. Instead they built their own fabric 1.0 and then they've used it to completely reimagine what video editing looks like. So we're going to dig in today to why he made that bet, what it unlocked, what it tells us about whether you need to own your AI stack to win, and so much more. Saba, welcome to the show.

Speaker C: All right, Saba, so let's get into it. Would love to get your sense, Give us a state of video AI in terms of what's working, what isn't quite there yet, what's coming next?

Speaker A: Oh, God. That's a nice easy question to start us off with. Thanks for that. Um, the state of Voi is absolutely crazy, to be completely honest. So much is happening, so many new models are coming out, loads of open source models and yeah, I mean, everyone's just really trying to work it out, like how to fit it into their workflows, how businesses can use AI video. Individuals are kind of working out. So I think everyone is. I feel like everyone's kind of figured out that large language models are great at coding. I think that's like, everyone's kind of works that out. Now I think people are trying to work out where does video, like, how does it work with video?

Speaker B: You know, what, what is, what are the use cases that are working to your point? Like coding has taken off. Like the latest models from the Frontier Labs were so good at coding use cases and by extension anything that sort of is text based. Uh, but video, it's, it feels like it's on a different timeline right now.

Speaker A: Yeah, I think that like video is potentially like two to three years behind large, uh, language models and coding. And I think there's a few things that kind of keep it there. The first is how long it takes to generate a video. I Think the latency can just, I don't know, like, really get in the. The time to value is not 100% there. And I think the other thing is the cost, right? If it. If it costs 3, $4 every time you ask the query into ChatGPT, it wouldn't have the level of adoption that it potentially does now. So I think they're the two things holding it back. And when it comes to use cases, I mean, I think it's across the whole industry. Right. I think, like, people in Hollywood are definitely using it. Right. I think marketing teams are. Pretty much all our marketing content is done with AI now. So it's there. But I think the, you know, dissemination into loads of organizations and companies is. Is just still very much at the start.

Speaker C: Yeah, did I see. I was at Netflix just bought with it Ben Affleck's stealth AI company to help them do more of this kind of stuff or something like that.

Speaker A: Yeah, yeah, exactly. Um, it's funny, when a company buys a stealth company and you. It's quite hard to, like, evaluate. Do you know what I mean?

Speaker B: I'm curious, like, what are the smart people doing? There's a lot of, like, AI slop with video. It's kind of like the. The term of the, you know, the moment right now. Um, but you mentioned, like, all your marketing's being done this way so clearly, like, they're savvy people that have figured it out, like, what's different between them versus, like, you know, the average person.

Speaker A: Yeah, I think this. I think this, like, this term, AI slop is super interesting. And I think you can make slop with any single medium there is, whether it's your iPhone. You can just take loads of terrible photos. And I think a lot of people do it, but we don't decide to publish all of them to the world. It's that very small selection. And I think it's so easy just to churn out loads of very average or very below average bits of content. Uh, but I think, like, creation is becoming incredibly important. So I think that's like, a really big part of it. And I think the people that have figured it out are, I don't know, they've kind of, like, worked out these, like, crazy workflows of how to, like, pick the best model to do the best thing, right? And they have this workflow. So it's like, okay, midjourney will, like, capture the image. I'll then use, like, Nano Banana to, like, upscale it and kind of like, change the scene. I'll then Go image to video with a certain tool, and then the lip sync's not perfect, so I'm going to add a lip sync model on top to actually, like, get it up to the standard. So, like, very, very elaborate workflows.

Speaker C: How much does that matter?

Speaker A: Um, I mean, it's a good question. How much does it matter? And I think it's kind of back to Wade's question, right? Like, do you want to make great stuff or do you want to make stuff that's fine or below average? And I think we want to make great stuff. One thing that's, like, super interesting, I don't know if this is something that you guys see as well, but, like, people love elaborate workflows, right? Like, they feel like if there's an elaborate workflow that connects multiple tools together, that this is, like, a really secret source. But if there's, like, one button that does it all behind the scenes, it's not quite as satisfying, is it? Right. Like, it's nowhere near as satisfying.

Speaker B: It's definitely the IKEA effect. It's like, I built this with my hands, and therefore it makes it great. Set aside that this is a cheap piece of furniture.

Speaker A: Yeah, yeah, yeah, yeah, exactly.

Speaker B: So that. That sort of is definitely going on. Uh, you know, it does feel to me, and I would be curious if you agree with this is the. The folks that seem to be doing the best, they still have this, like, really sharp creative sense. And so they know the limits of what these video models can do, and they know how to fit that within their marketing stack or the story they're trying to sell or whatever, so that it. Even if it looks gimmicky like everything else, the gimmick actually works in that setting. It's sort of like it. Like, it plays effectively. Uh, and so they almost use the constraints and limits of the technology to their advantage versus just sort of passing it off as, like, you know, legitimately good art form.

Speaker A: Yeah. I think, like, you know, it's the idea of, like, hey, like, getting an elephant to walk across a beach is something you couldn't do right in real life, but with AI, all of a sudden, it's very, very easy to do. Right. Um, so there's interesting kind of like, using the technology in a way that is very novel and very new and can give you completely different outcomes. That's definitely something that's quite fascinating. I. Super interesting kind of point on this is right now, I kind of feel like the world buckets like video into two camps. One is AI content, and then non AI content, right. And I think that's just a very, very short period of time that we're living through. I'm pretty sure in one to two years from now. It's just video, right? Just like animation is video, right? Just like stop motion is video. They're just slightly different mediums of video. And I think the massive blue ocean opportunity right now is obviously in AI video because as I said, we're all trying to figure it out, right? And I think there's, it's, it's coming for everyone and it's uh, it's uh, it's a fun industry to work in.

Speaker C: And uh, to your point about sort of who, who's doing it best right now, do you find that it's the people that are willing to work within the current limitations of AI video or the ones that are trying to make it do things that maybe it can't? Because I go back, it's like last year, that big moment when, you know, Cal, she made that ad that they ran during the NBA finals. And I think they made it in a weekend or something like that. And you know, the way that they did it was, was like six minute clips and they went from, or six second clips. They went from one thing to the next and the next next thing. And they had done that within the current confines of where AI video was at the time, but they just did it in a way that no one had done it in the past. And it just really stood out from a creative standpoint.

Speaker A: Yeah, I, I, I think so. The I, I, I feel like there's like creatives are really a really strange breed. And I definitely fall into that category of strange breed. I studied art. Um, um, so that's, that's, that's my background. There is a lot of pushback from professional, traditional creatives about this technology. And it, within our circles within technology, it's everyone's very quite like liberal to like try new things. And they're very much early adopters. And so I think like the people that are really figuring it out are the people that are actually saying, hang on a minute, this is really exciting. It's early, but there's something there. And you know, like small communities that are like super hot, right? And like everyone's like super active, like really early bitcoin. Right. Or I'm sure like what large language model communities were like in 2020. Right. It feels like that's kind of happening within the video community right now.

Speaker C: Nice. Okay. So to that point, VEED launched its own video model, uh, fabric 1.0 last year. Walk us through. Decision to build your own model instead of wrapping it inside someone else's. What made you say we have to own this?

Speaker A: Yeah, very good question. And I think first I'll come out of a statement which is I didn't want to do it,

Speaker B: right.

Speaker A: Like, the idea of building a model just seems incredibly daunting and very expensive. Um, but I think the current dynamics of the market back then were very much like, hey, we've got a video generation model and you can't have it, right? And when your back's up against the wall like that, you're kind of like, well, all right, fine, I'm going to have to play. And I think like, video generation is very different than, again, like, we will talk about large language models. They're very, very expensive, right? But I think within video it's a lot more accessible. So at, uh, Veeds, we kind of specialize in one type of content, and that is talking head content. The history of the business is people making videos, talking to cameras like we're doing right now, adding subtitles, adding music. And obviously the generative lens of that is to make a talking head video with a generative model. So that was very much our focus. So we assembled a really small team and were heads down for a year. And the model that came out, fabric, was incredible. The fact that it was actually public was actually a mistake. We ran a hackathon public, uh, one in London because we wanted to know what some of the use cases could be of using this model. And to make it accessible, we put it publicly on a platform called foul. And very quickly people started finding it without any marketing effort. And all of a sudden it became like one of the business levers for, ah, our, for our revenue growth this, this year. So it's been incredible.

Speaker B: How do you think about, um, who should build their own models versus not like, now that you've done it, um, do you have a sense of like, when it's a good idea versus when it isn't?

Speaker A: I think like 95%. Sorry, that's wrong. 99% of the time it's definitely the wrong idea. Um, and we tried before and failed. And I think the reason why it works this time is the team that we have are very entrepreneurial. They're like incredible research scientists, but the former entrepreneurs. So there's a certain mindset that you need and it's not an academic mindset. So I think it's like building that DNA in the company that kind of is a Little bit hacky and scrappy within a very much a, you know, very scientific space. Um, but no, look, to answer your point, I, I, I would very much better, yeah bet against it.

Speaker C: So the scary thing Sava, is I fully believe you when you say 99% of people should not do this. The flip side of it is where would VEED be right now or trajectory had you decided not to do this?

Speaker A: So this is and this and good um, question and I think this goes back to why we did it. Like when we were using third party models like other models that people created, they would often kind of use the model within their own platform to grow and then like three, four months later offer it uh, via public API. So it wasn't very developer first and they were then using that ability to have the model before everyone else as a way to like drive traffic to the self serve platforms and then so you're kind of like left to the side and the margins and the queue times and like, you know, not being able to inform the roadmap was something that was like really difficult for us and it just meant that we couldn't be at the cutting edge. And so that's really the reason why we did it. So if we didn't do it, I think we'd be in a really, really hard position right now. We really would. Um, and in a way I think it's actually made us more focused as a business because it's like hey, we just, we there's only one type of AI video content that we do and that is generative talking head. That is it.

Speaker B: What's the unit economics look like for training your own video model margins?

Speaker A: I wouldn't expect better than 30% to be completely honest. Uh, but that's kind of the world that we live in right, right now. Um, and we expect those margins to get better in terms of unique economics. Like training model is very expensive but it's not like crazy expensive. So like yeah, it's not horrendous data. That's another really big uh, constraint as well. But fortunately the company's been running for like six years and you know, we've been basically collecting data for a very long time from free users. So we had all the pieces of the puzzle were enough re scale that we had like enough free cash flow and that we could do it. So yeah, we took the leap for a second time and it worked out. And what's really interesting now is I think it's like we're now basically training the second version of Fabric Fabric 2 which will add prompt support to this talking model. And that's kind of building on the success of the last model. So I actually can see a world where like, there was a small window of time that we were able to like, build this, like, really rudimentary, quite basic thing, and then every six months be able to kind of go up the ladder and actually make it more, you know, incredible and also spend more money on doing it. So it's, uh, maybe that window's shut. I don't know.

Speaker C: I mean, how is it possible or how is it cheaper to build a video model versus an LLM?

Speaker A: Um, yeah, like completely significantly cheaper. Like not even, not even close. Like you were talking like a few million dollars to, to build like the basic video model. So one of our models is a lip syncing model, which is basically used mainly for dubbing use cases. So you just put in a version of Dan or Wade speaking a different language with the raw video and then the lips move basically against, uh, the new language. To give you an idea on, like that one, you know, the research on that has been going on for two years by one engineer. Uh, one year in, he shipped the first version of the model and we're about to ship the second version. And then there's consistent trading happening on a monthly basis anywhere around, yeah, like 50, 100K. So it is expensive, but it's not crazy. And the model more than pays for itself. Right.

Speaker C: So.

Speaker A: Yeah. Have I given too much away for a public podcast? I think I have. Guys.

Speaker C: No, because we were gonna, we were gonna, we were really gonna ask the why. Why? Why do you think that is? Like, what for those that maybe aren't as close video world, um, what's the main takeaway for them to understand the why here?

Speaker A: Great. So you decided to gang up before the episode and ask me the hard questions. Thanks, guys.

Speaker C: Don't worry, we'll end up with some softballs.

Speaker A: No, it's all good.

Speaker C: Um, interesting. Okay, so then back to VEED here. So with the models you guys have and the product you're putting out here, so 60x more affordable, 7x faster, like fabric clearly outperforming the competitive models. Um, when you see those types of results, I guess to go back to the model thing one more time, like, why isn't every company building its own model? I know you kind of said no, but if this doesn't become the new norm, where do we see the industry going? How does this evolve?

Speaker A: So why doesn't everyone do it? Like, the skill set is completely different than the skill set of building a software company. Um, software building software. It's like a very well defined process and easy ish to estimate. Right. How long things take and how and what the technologies are. And I don't know, everyone's kind of on the same page about it. Building a model is like putting something in the oven, spending $100,000 opening the oven and being like, oh God, is it good or is it bad? Right. And so like you have to be willing to kind of crack a few eggs to make the omelette. Um, so yeah, go for it. If you got the appetite to get it completely wrong for a couple of years and spend a lot of money in the process, a hundred percent. Right. And it could have gone very, very wrong. And I'm sure it has gone very, very wrong for a lot of people. So. And I think it comes back to like, who do you hire to do it? Because like the first time around that we tried to do it, we had like people with like really impressive resumes and really impressive CVs and backgrounds. But like, I think that entrepreneurial DNA is the same with actually building a software company in a way. Right. Like if you just get a bunch of people with like great resumes and great backgrounds, it doesn't mean they're going to make a successful startup, does it? Right. Because like, oh no, we're going to build from scale from the start and we're going to make all these assumptions like you need that hacky, scrappy energy that are just trying to work it out. And I think that's the, that's the entrepreneur hire, uh, that we got very, very lucky with our AI team that

Speaker B: does resonate like that's sort of this intersection of multiple unique, rare skills that kind of need to like collide inside of one person. And you know, as a result there's just not that many people that are, you know, particularly good at, you know, this hard, complex topic and this scrappy entrepreneurial way uh, of building and then sort of has the taste to build something that is like commercially like uh, exciting.

Speaker A: Yeah, I mean we pro, we probably forget as well that like the first video generation models basically took midjourney or an open source video, uh, image library and made two images and then animated between the two. That was the first video generation models. But that was enough to get people excited that a couple of companies actually built up enough revenue to make the next one and actually do that research. So it's that, it's that stair step approach

Speaker C: at the moment. It's still Pretty easy to spot an A.I. uh, generated video. Is that a feature? Is it a bug? And I guess from your standpoint, what are some of the signals that we're on a path where we've really solved video here.

Speaker A: So garbage in, garbage out. Uh, the biggest crimes are skin bad, like very glossy skin. Um, and I don't know if what creams you guys are using, but your skin's looking great, Mine is looking a bit more weathered.

Speaker B: Um, but we're actually A.I. that's the.

Speaker A: Yeah, right.

Speaker B: It's not real.

Speaker A: I think skin is like the most important one. Like people's skin looks very blotchy and there's marks and it reflects the light like we've lived.

Speaker C: Yeah.

Speaker A: The other one is audio. Audio I always got taught at art school when making films that uh, um, audio is 50% of the video and everyone forgets it's right. And so I think like great quality audio is the biggest giveaway. And what's really interesting about AI video versus normal video is with our customers that just want to take a pre recorded video and um, edit it. What they want to do is they want to remove all the blemishes, right? They want to remove all the ums, all the Rs, they want to remove the background noise, right? They want to do all of these things. And AI videos, people want to do the opposite. They want to add ums and ahs in, they want to add background noise in, they want to add static to make it feel more real. So there's this interesting convergence. Um, but to really answer your question, I think in, in 12 months you're gonna, you're gonna be very hard, very, very hard to tell m the difference.

Speaker C: And what's your, what's your recommendation to teams in terms of how j think about their video creation, in terms of what's good, what's great and what's absolutely best in class. Because sometimes, you know, maybe within product marketing you just videos out, you have a new product launch coming out. It serves a whole bunch of industries and if you can cut up 10 or 15 or 20 videos to address each industry, there's actually just more value in giving the specifics and perfecting the video. There's other times where you really, really need to make sure that every second of the video is perfect. But it can be hard to balance, uh, understanding how to approach the two. What do you see the best in class doing here from a frameworks perspective? What advice do you give?

Speaker A: Yeah, so I think the first thing is like your, every company is not publishing enough video. Um, to be honest, like, it's uh, you know, every day that you're not actually publishing stuff is like a, ah, missed opportunity to get more people seeing your content, your brand, your product, first of all. Second thing is like the first, there's, there's two checkpoints that I really look for when we launch videos. One is the first 0.5 seconds, right? The first half second of the video needs to have something. It's almost like the image, like the first frame, there needs to be some sort of hook or it's got to deliver something there. And the next thing is the first two seconds. If you can get someone watching for more than the first two seconds, there's a very high chance you're going to keep them. It's like retention of a product, right? The more people you get through onboarding, um, and successfully onboard to your product, the higher your retention is going to be over the long period of time. And it's the same with video. So they're the first things. And the second thing is like, this is just the muscle. If you do a video like periodically, once every like 612 months, you're just not going to be very good at it and you've just got to do it more. Right? It's as simple as that.

Speaker C: I don't know what you think, but to your, to those two data points, I would argue most companies are not publishing videos on a daily basis. And I would argue, and I would argue most companies are not thinking about the first half second to two seconds in the way that you just described. So given that, where are we in the evolution here? I mean, of course lots, uh, of video exists, but perhaps it hasn't been created or thought of in the way where the results really start to matter.

Speaker A: So this is a very interesting question, and this is, um, not a flex by any means, but we have like two YouTube channels. One of them has 150,000 subscribers and gets like 50 million views a year. The other one's got about 100,000. How many SaaS companies have YouTube channels that have play buttons right in their office? And I think the answer is very, very few. Um, and it's just such a missed opportunity, right? Like, YouTube is the second largest search engine in the world, yet everyone's got these massive SEO teams trying to optimize content, write content, get backlinks, and no one's doing anything on YouTube. It's crazy. It's absolutely crazy. And so I think, yeah, like, massive missed opportunities. And how's it gonna develop I think like right now video is like normally this like small agency within a company and it gets briefed now and then. It's the wrong way to think about it. This is a creation powerhouse, right? Like, you gotta let these people fly and make content. And then also, like, I'm probably boring you guys completely now, but like content per channel, like, you can't take an Instagram video and put it on TikTok or Twitter or whatever and expect it to do well. It's just not gonna work like the n the channel. They're very different. Like even just the small, small things. Like how does a square or 6x9 video show up in the organic X feed, right? Like, is it cropped? Is it not? Do you have to click it? Like, where, where are the safe zones? Like, there's so much nuance.

Speaker B: So all that stuff sounds pretty overwhelming, right? If you're a fixed. Well, I'm thinking if you're a small business and you know, you, you've got nine other things going on, or even if you're an enterprise and you're a CMO who's not done any, you know, video marketing before, like, what would you. What's your advice on getting started? Like, hey, just, just do these two things and you'll, you know. Yes, there's still like a masterclass that you can go learn later. But like start here and you're going to, you're going to get, you know, the 80 20, right?

Speaker A: So I think that, so this is why I think AI video is so important, right? Because to set up a camera, uh, hold a microphone, be happy how you look on camera, make sure the background, the lighting is correct, is a lot of work. Like it. I set up 40 minutes before this just to make sure it was all right. You know, check your Internet connection, all these good things. That's. It just takes time, right? So I think that's why AI video is so powerful. Because if you can just like have, you know, using a model, like fabric, have a version of yourself, be able to type out a script and get that video, I think that's really, really powerful. So I think that's the first thing that I'll say. And that's kind of where I think the puck is going right now. The second thing is when you are next doom scrolling Instagram, if you actually look at what people are doing, they're only doing two things. Everyone's adding captions to their videos and everyone's adding a bit of music and maybe some text. Like it is so light and I think everyone thinks about video editing as this like massive timeline, dark room, multiple monitors. Like that is not what editing and content is in 2026. It's just not right.

Speaker B: So, so maybe, maybe like if you were a small business owner, like a way to get started would be to, uh, go to, I don't know, ChatGPT or Claude or whatever and say, hey, I want you to interview me about my business. Pull your iPhone up and just press record and just start talking. And then once you're done, do the thing you said. Add little music, add some captions and now you got some clips.

Speaker A: Yeah, bang on, correct. Simple as that. Um, it's just, yeah, more people need to do it. I think, I think we're getting there. I think we're getting there and I think there's a generation of people that are growing up doing this natively and they're going to bring it into the workforce.

Speaker C: Where do you think the enterprise is right now?

Speaker A: Oh, I mean this is a, this is a, this is the bane of my life. Um, in, in a couple of places, right? And it's really interesting. Um, so like traditional news media, right, like the BBC or Fox or whatever that used to have these kind of like live channels have all moved to social and these guys are adopting it really well, to be completely honest. Like every single news company has basically got a social first strategy and that's it. And that's where it should be. The enterprise, like traditional software enterprise is. I think it's so, so, so behind. It's so, so, so behind. And I don't know, I don't like why. I mean, where do, where do you think you guys are with video right now?

Speaker C: Well, it's interesting when you ask the question, because is it specific to human created video or AI generated video? And so the company has a good history of sort of human created video, but we're on a similar spike right now of trying to think about what the AI strategy might, might be. And so when you're seeing enterprises that maybe historically have done a lot of human LED video, is AI an accelerant or a blocker? Like, what is it doing to teams at the enterprise level? Because it comes with both so much opportunity and then also a lot of risk.

Speaker A: Yeah, I mean, I think traditionally we're not going to see the enterprise be at the forefront of this market. Right. I think something that is very well accepted across the enterprise right now is what we're doing right now. Right. Can we get a couple of people onto a call, record it, have A really great conversation. That's a really nice long form piece of content. It benefits from the fact that I'm going to share it with my community. You guys share of your community. So we get dual reach and then also on the back of that we get a bunch of clips as well. So this is a very well run playbook right now and I think it's a great start to get leaders speaking publicly.

Speaker B: Yeah, yeah. You know what I think is interesting, like you asked, uh, you know, how, how are we on our journey? You know, uh, for us, like we came up and we founded the company in 2011 and so that was the era of blogs and content marketing and the written word and getting started there was. It seemed a lot easier, right? You literally just like bang out words on a keyboard and you publish, you put it on a site somewhere and that's what was rewarded. Like that's what drove distribution, traffic and you know, ultimately customers. You know, it feels like what we've seen, you know, in the last five years is uh, a much stronger shift to video on these various social platforms. Whether it's YouTube, TikTok, uh, Instagram, uh, you know, you name it, pick your social platform of choice, like the, the video has become a higher percentage of that content and it seems like the engagement rates are much higher as well too. And I think for companies like us, you know, we built up our skill set in one era and now it's trying to figure out, okay, how do we add this new dimension, uh, to the fold. And yeah, I think partially, you know, we'll also see like a new era of folks enter the workforce that will help with that. Like, you know, uh, I didn't grow up with like an iPhone and video is like the default thing. And it's like the idea of just like making funny videos for my friends was just not. That just wasn't available to me. Whereas like, you know, the kids that are like 22 years old now, they've been making funny videos like for themselves for, for ages. And so they have like a, they just have more practice at it. Um, you know, which comes back to your prior point where it's like, this is a skill like anything else. You just get good by putting in the reps. Uh, and that just wasn't a thing we had access to, or at least I had access to growing up. And so, you know, it does feel like as that generation ages, as the social platforms have started to create video as a more dominant platform, it's ripe for just like being much more pervasive across, across all videos or all businesses,

Speaker A: I should say there's a few interesting things. If you go onto your uh, screen analytics on your phone or the average person's phone, I'm pretty sure Instagram and all the video platforms will be up there. Like if you see someone in public looking at their phone, nine times out of 10 they're looking at a video. They're not reading a blog, writing an email, they're watching a short form video, right? So it's just where people's attention is. I think that's the first thing. Second thing is attribution. And I think like blogs. Oh, uh, attribution heaven, right? Optimization heaven. Amazing backlinking. Oh, such a great playbook. And like you can, everything's so easy to attribute but with video it's much harder. And I think there's two things that you can track. One is just like views, right? Like how many views did that video get? How relevant is it? Uh, about the really potent things in our products that we really care about. And what's super interesting is it shows up in brand traffic, right? That's where you see the lifts, right? So like if you're doing really, really well, you'll see your brand traffic lift. And then I think the final consideration, which is something that's like also super interesting. And I don't know if you guys have seen this as well, but because most of these videos are consumed on mobile, you need to make sure your mobile experience is excellent, right? And when I say mobile experience, whether that's an app or the web, like it's all got to connect. It's all got to connect. So yeah, that's the game.

Speaker B: How do you think about? I, yeah, I'm curious. Kind of shifting the topic less from like what business issues do so to maybe what society should think about. You know, one of the challenges you mentioned is, you know, you see someone out in public looking at their phone, they're probably watching these short form videos. There's this whole big debate. I'm like, hey, is that good for you or not? I would say it's not very good for me. I don't like it when I get sort of sucked down the like YouTube short rabbit hole. I'm like, ah, this is, I need to like, I need to clear my brain, uh, for a minute and try and refresh it. Like how do you think of how to use that, you know, addictive mechanism for good versus bad?

Speaker A: Look, it's a really good question and I completely agree with you. I feel a Bit icky after getting stuck in a hole, um, on Instagram for a while. Um, I do, I do. Like, how do I feel as a tool builder? We just gotta make the best tools possible so people can actually make great stuff. But like, any tool can be abused, right? Like if you want, you can sign up to any male client and spam people until you get blocked. Right? Like, it's as simple as that. I mean, I do think one thing that I do quite like about social media is that, uh, it doesn't actually reward bad content. Right. Like, I mean, I get a lot of, a lot of, um, peace by the fact that if I do a really bad tweet, the algorithm is just not going to send it to me.

Speaker B: Use 37. Like, well, that didn't work.

Speaker A: Yeah, it's the same with video. It's just like, cool. Yeah, that one didn't go do too well. I'll delete that tonight.

Speaker B: You know, I mean it is, that is like a. You know, I do think there's a lot of folks worried about sort of looking dumb on the Internet. And I think this is actually a good counterpoint to that where it's like, hey, if you do something bad, like, you won't look dumb on the Internet because no one's gonna see it.

Speaker A: Right?

Speaker B: Like, so you might as well try.

Speaker A: Let me. Actually, this is a, this is a really good, good story actually. So my, um, I hired uh, our former chief product officer, Sam, and when I hired him into VEED, it was like a terrible platform. Not terrible. Early. It was early, right? You couldn't edit multiple videos together. It was just like one video. Trim, crop, add some text. He left because he actually raised funding to start his own startup. And the way that he did it was by making a video of VEED about a side project that he worked on, put it on Twitter and it went super viral. The next day he had a term sheet to start his startup. And that just shows the power of video, right? It's just like it's communication mechanism. So I think that's when it's, it's done really, really well, right? Publish more video, good things come to you. You're just planting seeds, you know, I

Speaker B: mean it certainly is the trend these days. Like every product launch video, like launch is now a video. And that didn't used to be the thing. Like now it's, you know, founder walks into frame, sits down on couch, you know, nice background behind them and says, today we are changing the world because of da da da da da.

Speaker C: Ah, so I'm curious, you, you mentioned sort of video being, uh, a little bit, you know, behind, maybe a step or two behind in the AI adoption curve. What are things that sort of worked with written content as we saw AI really become more popular, that maybe won't work as well as we say AI play into video as much.

Speaker A: Yeah, I mean, I think the first thing is just like your script is your message and it's always going to be the most important thing, right? So I think like, just get a great script together. Um, and the other best practices is just like don't overcook it, right? Like just, just do that light edit. Keep it really, really short. Keep it really, really simple. It's, um, it's. It's. That's, that's it. It's not too hard. I think the biggest blocker is just recording yourself. Like I really do. I think people just don't like looking at themselves and they feel weird talking to camera. And if you do it in public, it's like five times weirder. Right. Like it is.

Speaker C: That's great. So then, okay, so then to go back to the human element of it. If VEED succeeds completely the way that you want it to, what human skills around video becomes more rare and what becomes more important?

Speaker A: Look, I think we're in this. I think we're in a. I think we're in a really interesting time and I wish I had all the answers for you, Dan. I really do. Like, you know how I think it's. We're asking ourselves a parallel conversation which is like, how good can an AI be at writing a poem, a song? Right? And I think the people that are going to be really successful are the ones that like jam with it. Right. If what I found super interesting, you know, the company Suno, who does the music generation, it turns out, and like their revenue numbers are crazy and their growth is absolutely crazy. And I was like, what's going on here? And I dug in a little bit and it turns out that actually musicians are using it to make music and like prototype ideas. I just thought that was really interesting. So like, this isn't replacing the artists at all. This is giving the artists really incredible tools to be able to do stuff that might have taken them like four or five hours before. So it's just making them way more productive. So I think that's really what it's going to do.

Speaker B: Uh, that to me is what's most excited. I think what most people miss. Like, if you think back to music or art in the Renaissance period, Or things like that. Those were high status activities. You had to be wealthy, you had to have access, you had to have instruments, you had to have paint, you had to have access to all those things. And some of our best creators, things like Mozart Box, you know, uh, all these folks still uh, managed to come out of that era, but how many of the population could have been just as good if they had access to those tools? How many of those folks had the ideas, had the capability to do that? But they did, they just didn't have more accessible tools. And that's where I think, you know, what you do, what Suno does, what, what any of these like creator tools do is it brings it to the level where the, the getting started is so much simpler. Um, and then it can help you, you know, just get those ideas out of your head. And what I hope for is that we'll be in a world where more people create, you know, world class art, you know, whether it looks like old art or not is kind of, you know, besides the point. It still is great by whatever modern definition that is.

Speaker A: Hundred, um, percent. It's like the, hey, to build a startup 20 years ago involved like server racks in your office.

Speaker B: Oh God.

Speaker A: Right, right. And now.

Speaker B: And you got no, the customer got no value out of that. Like there was no, no value to the customer for the founder having to go rack servers.

Speaker A: No. And now look at like vibe coding. I mean like the, the, the how, how far we came in 20 years is like crazy. And just on like building a website. And so I think that yeah, you can use a vibe coding platform and make something that's absolutely garbage that no one's going to use. Right. Or you could make, you know, the next zapier. Right. Like you just don't know.

Speaker C: Awesome. All right, to bring it full circle, we got one last question for you. In a world where dropping out of college sometimes is popular in the tech world, we see people starting their own companies. You are going the opposite direction. You are headed back to college. Uh, we saw, ah, you just enrolled in Stanford's Graduate School of Business as a CEO of an eight figure AI company. What are you hoping to get out of this formal education right now? Now, uh, what, what should people know?

Speaker A: So I actually did this, I uh, think about a year ago and I had this opportunity from the team at Sequoia to go on this like three month mini mba. And I've always loved education. I think it's some of the happiest times because it's such an indulgent Opportunity to, like, get lost in your own ideas and explore and things don't matter, like board meetings and quarterly numbers and KPIs and OKRs, and I think that's so fantastic. So I had to do it, and I absolutely loved it, and it was an incredible experience. Very thankful.

Speaker B: What. What. How did it. I'm curious how it helped you. Like, what. What did you learn from it? What, uh, what'd you get out of it? Because it's such a rare choice to have done.

Speaker A: Yeah, what do I get out of it? I mean, I think, um, I would. I would put it out there. One of the, like, the best. Like, you know how sometimes, like, the days and the months go by, you're like, oh, my God, how did it get to October? You know? And I think, like, when I look back on a year and think about, like, what did I achieve and what do I do and what am I really proud of and what. Like, that was what. Definitely one of the highlights of the year. And I think it's just important to look back or plan every single year that you have those kind of moments. And we've gone massively off topic, guys, but, like. Yeah,

Speaker B: well, it's. I mean, it kind of makes sense. Like, if this is something that you find rewarding, like, you know, uh, everybody's got a hobby, everybody's got interests. Why not that, Right?

Speaker A: Cool.

Speaker C: Awesome. Saba, anything else you'd like to share with our listeners to close us out?

Speaker A: I mean, loads. It's been a great conversation. We could talk for hours, but the camera, uh, is not going to last that long. Uh, so, you know, I appreciate you guys and thanks for having me.

Speaker B: Thanks for coming, Saba. It was great having you. Thank you so much for joining Saba, Dan and myself on this week's episode of Agents of Scale. If you like the episode, subscribe, leave a review, tell a friend.