Podcast: Building ad tech for the chatbot era (with Tal Shoham)

On this week’s episode of the podcast, I am joined by Tal Shoham, the CEO and co-founder of Velocity, which today came out of stealth and announced its $27MM seed financing round (note that I’m an advisor to the company).

We discuss the transition of AI-native applications toward ad-supported monetization models. Among other things, we cover:

• Whether AI-native apps will follow the mobile gaming trajectory by adopting hybrid monetization models to reach massive global audiences
• How the high operational costs of AI inference necessitate a move toward ad-supported tiers to maintain sustainable profit margins
• Why the granular intent signals captured in LLM conversations offer a superior alternative to traditional search-based advertising targeting mechanisms
• What new native ad formats will emerge to integrate seamlessly into AI workflows without disrupting the core user experience
• When the venture capital funding environment will demand more disciplined monetization strategies from AI startups over pure user growth
• How established ad-tech infrastructure from the mobile gaming era can be adapted to serve the unique needs of AI developers

Thanks to the sponsors of this week’s episode of the Mobile Dev Memo podcast:

• ⁠⁠INCRMNTAL⁠⁠⁠. True attribution measures incrementality, always on.
• Xsolla⁠. With the Xsolla Web Shop, you can create a direct storefront, cut fees down to as low as 5%, and keep players engaged with bundles, rewards, and analytics.
• ⁠Branch⁠. Branch is an AI-powered MMP, connecting every paid, owned, and organic touchpoint so growth teams can see exactly where to put their dollars to bring users in the door and keep them coming back

Interested in sponsoring the Mobile Dev Memo podcast? Contact Mobile Dev Memo advertising.

The Mobile Dev Memo podcast is available on:

• Spotify
• YouTube
• Apple Podcasts

Transcript

Eric Seufert: Welcome to the Mobile Dev Memo podcast. I am your host, Eric Seufert, and I am joined today by Tal Shoham. Tal, welcome to the podcast.

Tal Shoham: Good to be here.

ES: I was trying to think about how we first met. I actually do not remember. I believe we have known each other for quite some time. You were in leadership at IronSource and I will let you give the full intro in a second. Do you remember that differently?

TS: It was probably a casual connect in 2013 when I was at SuperSonic.

ES: That is right. They had that in Tel Aviv.

TS: That Tel Aviv in 2015 and I think we actually met even before that in San Francisco. Many years ago.

ES: Probably more than a decade here. You are doing something very exciting and new now, so I am going to let you talk about that. Please introduce yourself to the audience for those who do not know you.

TS: First, Eric, thank you so much for giving me the opportunity to be here. A little about myself, Tal, 42, live in Tel Aviv. I have been in the ad tech industry for quite a while. I started at a company called SuperSonic, which my brother started in 2009 and I joined him in 2012 on the business side. Then we merged with a company called IronSource and we were fortunate enough together with IronSource to build one of the largest platforms for ad monetization for gaming companies. We built an ad network with various ad units and we built a mediation platform. We were really one of the largest out there.

I spent quite many years there on the business side leading business development, relationship, and product on the mobile ecosystem. Then in 2020, I joined a company called Huge Games as CMO. I got to experience B2C hands-on. You think you know gaming or B2C when you are spending so much time at a company like IronSource and SuperSonic, but then when you actually do it, you find out that you did not know that much. That was awesome as a CMO. I ran around a 300 million dollar user acquisition budget over the course of two years. I also led M&A and publishing there and we took that company public at the Warsaw Stock Exchange for over a billion dollars.

After two years, I left Huge and for the past four years, I have been doing a lot of angel investing. I have invested in more than 20 different companies in ad tech, gaming, and cloud infrastructure. I do a lot of advisories and boards and I co-founded a few different companies without an operation role. I invested some of my money, brought a few angels and VCs, and am being very involved like a board on steroids with these companies. That is all gone because I started Velocity eight months ago and this is what I do now full-time.

ES: I actually had done a really big project with Huge right before you came on. I forgot you were the CMO, but I had visited three different offices over the course of two weeks. You came on right after that. We did not overlap at all with Huge.

TS: You did a full analysis on the marketing side at Huge and helped us identify the strength and weaknesses and what we should improve. I have used a lot of the research you have done and insights you have given us exactly when I took office.

ES: That is cool to know. The occasion for having you on is Velocity. Maybe go into a bit more of a deep dive on what you are doing with Velocity.

TS: I will start with why I even thought that Velocity makes sense and the opportunity that I saw. Basically at IronSource and SuperSonic, we built a monetization layer that helps gaming companies monetize non-paying users. It is a big issue on the gaming side because you have 95 percent of users that are non-payers and you want to do something with that asset in terms of monetization and improving LTV, which obviously improves your CAC and growth capabilities.

We see something very similar in the world of AI-native apps. When looking at AI-native apps, whether it is a general AI search app like ChatGPT, a photo generator, video generator, or vertical AI for doctors or lawyers, everywhere that we see an interaction between a person and an LLM and that company is trying to monetize those users, we see that five or six percent of the users will turn into subscribers. The majority of monetization on AI-native apps is done via subscriptions, but still 95 percent of the users will never pay.

We see that pain and that gap is very similar to what we saw in 2013 on the gaming side and we thought this is a big problem for us to tackle. The crazy thing that we saw is that unlike in gaming where the non-payers are not that of a cost center on the company, on the AI world, everyone using AI for free is a massive cost center for the companies for inference and tokens. The problem is actually much worse than what we saw in gaming.

What do those AI developers do? They mainly put limitations on the users. You go to a general AI search app and you get two prompts a day or ten prompts a day. They are by design limiting engagement, limiting retention, and limiting the opportunities to convert those users into a paying user and to creating a habit for those users using those apps.

The problem is actually cascading into deeper down the funnel of the product and we thought this is something interesting for us to solve. That is how Velocity was born under the assumption that we can create another monetization layer for these AI-native apps to deliver ads within the AI experiences natively. Let us take an example of a chat app. We understand the intent, we take the prompt, the LLM response, and we understand what is interesting for that user and the context of the conversation. We collect all those signals in that conversation and we serve an ad natively within that environment and according to the user’s intent.

It is not intrusive, it does not harm the experience, and we are bringing value to the user because we might show him something that is actually interesting and according to what he is talking about. We are bringing a ton of advertiser value because the advertisers can now target users while these AI interactions. We are creating another monetization layer for the developers so they can generate money from these users and not only see them as a cost center on inference.

OpenAI is doing it amazingly. They introduced ads a few months ago for their free tier and they added a new tier called GPT-4o, which instead of charging 20 bucks, they charge eight bucks but with ads. They did not invent this, many companies like Netflix do it. That is the idea we want to introduce. We want to allow these AI apps to integrate ads into the experience, maybe introduce new tiers with a lower payment but with ads so the users are accepting those ads but they know they are paying less money, and to help them on their CAC to LTV journey that is always a challenge.

ES: I remember we spoke about this in January at PGC in London and I think there is a couple of really interesting threads I want to pull on here. The first is that a lot of people, and I wrote an article that was very declarative in May of 2025 where I said obviously OpenAI is going to monetize ChatGPT with ads, and I do not want to celebrate too much for that because everyone who worked in gaming knew this. It was not a question, it was an inevitability and it happened.

Everyone who worked in gaming knows that eventuality because you give away the stuff for free and free is ultimately the endpoint of the price point because that is how you get mass consumer adoption. You are never going to reach the potential scale by charging a subscription. So ultimately they have to go to free and that means ultimately they have to embrace ads and everyone in gaming knew that. But you called out a very specific difference and a fundamental difference between any company that is running inference and free-to-play mobile games, which is that the inference costs money. Usually this is just an operational cost for onboarding any additional user for a traditional consumer freemium app is essentially zero. But with any consumer-facing product that is running inference, there is a real cost. Talk to me about why advertising is probably even more necessary in this new product paradigm.

TS: The issue that we see is that these limitations and this cost structure is causing companies to limit the product in capabilities, features, and how much they are democratizing their products. It is causing a lot of frustration for these companies because they are building amazing technology, but they are very limited in how much it can be adopted across the world, especially if you look at other geos where it is even harder to get users to pay.

Our thesis is that not only that with these types of tools you can monetize and create an additional revenue stream and maybe cover inference costs, you can actually expose your product to a much larger variety of audience. That alone for me makes a lot of sense and if you look at my gaming days, that was the business. We wanted to get as much adoption as possible and to get many users to talk about it, you increase your organic K-factor, and you have more opportunities to convert users into payers or you just have more users using your product, which is also amazing.

So that is one of the things that we really feel is really important in what we are building and solves a big issue. Not just the monetization side, but also the productization, the features, and the less limitations these developers can actually impose on their product. The monetization side is just as important. You get more money, your LTV is better, and you are going to be able to be much more aggressive and more efficient on acquisition and growth. If you look at the AI-native world, it is much behind of how sophisticated the gaming world when it comes to marketing or monetization. It is a little bit of an up and coming world and new layer of amazing technology and amazing products that solves amazing issues, but the monetization, ad monetization, marketing aspects are a little bit 2015-ish when you look at the mobile ecosystem.

It is amazing because we see that more and more of these companies are getting better at marketing, better at monetization, understanding the value of ad monetization, starting to do things that in gaming is like something that everybody does, like segmentation and treat the users a little bit differently in different funnels and predicting the LTVs of users and then deciding what price points to push down or should they show ads. That is what we are trying to bring to the table as a team that has done it and has many years of experience in it, but also we see the market is evolving there anyway without us or with us. We want to be there when that happens.

ES: And there is another piece here, which is there is an additional signal, which is the context that can be added to the bundle of signals that you can target against.

TS: Maybe I will just tell you my thought on the signal on intent. The way that I look at intent is that we have another piece of machine learning and algorithm targeting capabilities added to the pie. We always had behavioral targeting, contextual targeting, now we also have intent. When a user now speaks to a general chat about training to run a marathon, now all of a sudden, we can show that user specific ads that can bring him value like Strava, which is a running app, or specific supplements or running shoes or running sunglasses.

It can be super relevant to that user but not just like a regular search query which search is probably one of the most amazing things that Google has done with advertising in the world. The cool thing about intent and conversation is that we can extract deeper signals, more layers of understanding of the user, maybe budget, where he is from, what he wants to do exactly, how he wants to train, for how long he wants to train, and then we can target an ad that makes a little bit more sense than a more simple query or a simple search. The idea is to take all that in account with our machine learning algorithms and decide based on behavioral, contextual, but also intent. We see the results of intent already in CTRs and conversions even though we are just getting started. We already see those early signals of 5x or 10x better CTRs than what we thought we would see or what we are hearing from other formats.

ES: Essentially this natural language interface is going to be a consumer expectation for all the products that they engage with. It is just how they are going to expect to interact with stuff. It makes sense in a lot of use cases, it is more convenient, and they will just get acclimated to it. For that reason, there is going to be this whole ecosystem of new products but also retrofitted products that include that as an interface. You are going to have the ChatGPTs and Claudes of the world that build their own ads infrastructure for serving ads. Google has done that for Google AI mode, but you are going to have many more of them that just do not want to do that. Just as most gaming companies did not build their own ads infrastructure, they used IronSource.

So they will want to tap into some monetization infrastructure that is provided by somebody else. There is just two opportunities here. One is just this expansion of engagement surface area or there is going to be inventory that either exists now and transfers over to this conversational interface because all these apps are going to adopt that and probably new net-new inventory because there is a lot of new apps that get created to take advantage of this. The other piece here, which is the AI-native aspect to it, is that the placements are probably going to be different, they are going to look different. There is a new signal to capture here that means if you want to do this really efficiently, you sort of need to be native to this space and that is what you are doing. You are building this native technology, the native infrastructure to serve those new placements that are informed by this new signal in the most efficient possible way.

TS: Yes, exactly like you said. What this new placements and creatives mean is first of all we are finding out as we go, which is very cool. It feels again like 2013 on the gaming side that rewarded video was not a thing and then we found out that is a massive powerful tool that every gaming company might want to use and today there isn’t a single company in the world that doesn’t use it on the gaming side.

That is exactly what we are doing. We are doing something that is a little bit different, it is in-chat or in-AI depending what it is. It could be an image generator, so for instance, while you are creating an image or creating a video, it takes maybe 20 seconds, 30 seconds. A very cool ad unit that we are showing now is that we show a large image or a large ad instead of showing just a timer or a spinner. After 20 seconds or 30 seconds when the image is ready, we shrink the ad that we just took the entire real estate and we show it as a banner below the image that was created. That is a loading phase ad that we’ve invented. We can show there an image, video, audio depending on whatever the AI does. That’s a cool example.

Other cool things that we show is we show mini GPTs or mini LLMs inside LLMs. We show video and carousel and images and animated GIFs and rich HTML. So everything around the creative side is very interesting and we try to do it in chat or in AI natively so it feels a part of the experience and not too intrusive and that it makes sense. Of course, the demand there needs to be according to the intent and context of the conversation or the session that the user does. Again, I’m saying a lot about conversation or image, but it can also be vertical AI in the world of doctors or vertical AI in the world of lawyers or whatever it is. But the idea is to be in the experience of the AI and native as possible.

ES: And then we talked about all this new stuff that presents an opportunity. Talk to me about how would you map what you are building to the existing ad tech workflows? How would you position what you are doing? Is it closer to an ad network, DSP, something totally new that has no prototype in the existing ad tech? Where would this fit on the Lumascape?

TS: It is definitely in the heart of it, it is an ad network that AI apps can integrate and monetize and advertisers can tap in and show their ads through our own algorithms and machine learning systems that decide what ad to show to what user. It’s definitely a monetization tool for AI-native apps and a distribution tool for advertisers in the world of apps, e-commerce, brands, and so on.

But it is of course much more than that because we come from an experience of also building platforms and mediations and so on. Our thesis is that this is going to become massive and a lot of different companies are starting to adopt it. Obviously, we’re also building a mediation platform which we come from that world so we’re building basically an in-AI native mediation for the players that are going to come into this world. There’s already players in this world, we already have competitors which by the way is great because if we didn’t have competitors, it means we’re doing something wrong and we’re in the wrong area of interest. I’m happy that we have competitors and competition was actually one of the things that I think made us succeed in IronSource, having AppLovin as a competitor is always good.

In this world of AI, we don’t have the AppLovins or Googles yet, we have other companies that we’re competing with. So we’re building an ad network, we’re building a mediation platform so any developer who wants to use multiple ad networks and manage them, control them, and optimize them all through a single technology can do it through our mediation. We’re also building a lot of other cool features like a conversation manager which is basically an abstraction layer that takes the prompt, strips it down of any PII or sensitive data, and basically passes only signals that are okay to pass to other bidders and other buyers and other networks so various companies in the world of AI don’t have to worry about their prompts or commands or requests from AI roaming around the world freely. We understand this is very important so privacy and integrity of data is something very important for us as well. It’s more of a platform that contains an ad network, mediation, AI specific tools like the abstraction layer and a lot of other cool things that we’re building as well.

ES: My operating theory for a lot of consumers is that gaming is at the bleeding edge and it creates standards that get adopted years later by other types of consumer products. I think that definitely was the case with mobile gaming. You saw monetization, UA, all of that got established by firms like IronSource who were really early in the space, or AppLovin. That infrastructure needed to exist before you could actually scale freemium games. That category didn’t really take off until you had AppLovin, IronSource, the company that Unity acquired that became Unity Ads. These were all founded 2011, 2012. I think IronSource was 2011. All that infrastructure needed to be in place and then the consumer category of free-to-play mobile games could take off. Then all of the monetization tactics, all of the user acquisition tactics that ultimately became the norm for the entire mobile ecosystem were established by mobile gaming. Talk to me about how your experience in mobile gaming is informing your approach here.

TS: It has a lot to do with my conviction on why we as a team even decided to do this because we feel that a lot of the things that we saw happening in gaming is starting to happen in this ecosystem as well and has to happen as well. The theory that the foundation models are only becoming better and better, building software is only becoming easier and even commoditized into the point that everybody can build a platform, everybody can build an app, everybody can build software, everybody can plug in whatever foundation model they want into their app or software, web, mobile and can have amazing capabilities. If that’s the thesis, then monetization and distribution will become the most crucial things for everyone.

Why? Because marketing, distribution, how do you rise above the noise when there’s a thousand that can do what you do? And monetization is how do you charge money when there’s a thousand like you that can charge a fraction of what you’re charging and maybe do the same thing. If those are the two challenges that we believe in, so we believe that what we’ve built at IronSource, my experience as CMO at Huge, our experience from coming from the gaming world is super applicable to what we’re building here now as well. That’s in the core of our thesis and the value that we feel that hopefully we can bring to this market in terms of giving them these monetization layers.

By the way, this monetization layer can be for apps, it can be for software, it can be for a lot of different things in the future that don’t even necessarily consider using this monetization layer on top of a SaaS model or subscription model or token usage model. Same goes for distribution. So that experience coming from building that hands-on, we’re very fortunate to become one of the largest ad networks and mediation platforms in the world. IronSource IPO’d for 10 billion dollars and was an amazing company. I as CMO managed hundreds of millions of dollars hands-on. Hopefully we can bring all that into this world and help this market as well and I feel that we are already doing it.

A lot of the times when we’re speaking to our design partners, we already have more than 12 different design partners which is amazing, when we’re speaking to them about monetization and bringing our tactics, our knowledge, our experience from the consumer gaming world to this world, they’re super happy about it. They want to learn, they want to implement, they’re super keen to experiment different things and try different things. Same goes for marketing. On the gaming side, creatives has become a religion. I have gaming companies that we generate 100 creatives a day. That’s the capacity. That’s not necessarily a known thing when you look at the AI-native world. They do much less and maybe they’re less aggressive in trying new creatives. On gaming, it’s gaming 101, everybody does it. If you don’t generate a ton of creatives and then test them out and have predictive models of what will work, what won’t, you have no chance. Just like a few examples that we do now on this segment that I think are super valuable.

The second thing that is very valuable is our experience when it comes to building SDKs and being in-app. Whether it’s a web SDK or mobile SDK, we believe in working direct. I love to work directly with the developers, with the companies, whether it’s an advertiser or a publisher, demand side or supply side. That’s something very big that I think because we’ve done it for so many years, I think we know how to speak the language of developers and the pains and the needs and hopefully we’ll be able to build the right technology for them to use as well in that sense. Just the fact that we’ve been in this massive shift called mobile, hopefully that would help us in this massive shift called AI. That’s how we see it.

ES: There are a couple of other relevant factors here particularly with the timing of this moment. One is Wall Street Journal just reported last week that OpenAI is considering very aggressive price cuts because they’re basically in a consumer attention war with Anthropic and they need to win it. It’s very important to them because they missed the enterprise strategic competition. I think you’re going to see that happen throughout the consumer space because I think it’s pretty obvious that for a lot of these verticals, it’s going to be winner-take-all and you’re going to need to have the largest consumer footprint and so you’re going to need to fight for that and the way that you fight for that is you drop prices and you’re probably going to have to take a loss on the inference. Well, you got to monetize somehow and ads is a really good way to do that. It’s probably the best way and they probably should have done that to begin with, but now is the second best time to start.

The other thing here is that the VC money is just going to run out. You saw that with gaming. People had to get a lot more disciplined about monetization when there was a free-to-play moment and when free-to-play wasn’t really seen as this exciting totally new category and it was just established and the VC money dries up when you stop seeing big content exits. Well, you got to get more disciplined about monetization and that’s going to happen with a lot of these AI apps. A lot of these companies are reporting basically fictitious ARR and they’re going to have to get disciplined about monetization because that funding is going to run out and ads is just going to be a really good way to do that. Are you seeing some of this happening already?

TS: I think it’s a great point. Yes, we do see a lot of it happening already. Especially on the apps that we work with and the companies that we work with are much more focused on monetization, being profitable, managing a healthy business than probably the rest of the foundation models which first of all have endless amount of money and they’re fighting for attention and adoption. I think that’s part of the interest that we see from the market in adding ads. The fact that more and more companies are focused on that versus market share or market attention I think is massive.

And I think VCs’ money and investors are looking at that as well. On the gaming side, today people look at your game economy and monetization before anything. You could be a low DAU application but with amazing KPIs and you’ll get funded. That wasn’t the case back then. I think that happens on the AI world as well. A lot of the companies we speak with are very, very intrigued on what will be the impact on KPIs, what will be the impact on retention, engagement, conversion to payers, how much money I can make out of this, can this cover inference cost. We have a lot of answers to already because of the design partners, but we are seeing that more and more companies are intrigued and open to try, want to do experiments, and they understand.

Also funny, when you look at the gaming world in 2012, 13, 14, people were very skeptical about ads. They weren’t as keen to integrate ads. No, it wasn’t the case like today. That’s the notion that I’m getting from this world as well. With the fact that you have other comparison industries that you can compare to, so people see the value and they’re intrigued by the value so that’s how I experience it.

ES: Well, the value on the publisher side but also the value on the demand side. If OpenAI needs to race to hoover up all the attention, what’s a good way to do that? Buy ads. Buy ads to promote ChatGPT, get installs. I imagine that these large companies that want to be the absolute category leaders are probably very intrigued by placing ads in other apps that do feature this natural language engagement model.

TS: Definitely and the cool thing that we see is in gaming we saw that gaming advertiser or gaming demand works really well on gaming supply. We see something that is very similar to that on the AI world. We see that AI demand, so AI advertisers, different tools of AI, work really, really well on AI supply in terms of conversion, in terms of click-through rates and so on. That’s something already that we see in the data and I would say doubling down in terms of getting it more and more of these type of advertisers and definitely works really well. So we’re bringing a lot of value to these advertisers and the fact that AI on AI works really well is amazing with intent or without intent, but of course with intent we can take it even farther. If you’re doing a task related into image generation and I can show you something in the world of image generation as a tool, then why not? Of course we’re not only showing AI tools, we’re showing other campaigns as well like e-commerce and fintech and gaming. I thought that’s a very interesting observation that we saw lately that reminded me again of the gaming world of gaming demand on gaming supply and how amazing it works. We see a similar trend.

ES: How are you thinking about measurement? How are you approaching measurement?

TS: For us measurement is as of now it’s exactly the same as it is in in-app gaming or in-app other experiences and so on. We’re performance marketers by nature so we look at performance, we work with all the regular attribution methods and companies and we report everything on the event side to the advertisers. For us it works the same. The interesting thing about measurement in my mind will be what will happen in the future when we’ll be able to complete the full funnel of the transaction within the AI experience. I think this will be the most interesting part and also OpenAI still don’t do that on ChatGPT but I think we’ll have a more of a what we call agent-to-agent capabilities where you’ll use Agent A and Agent B will serve an ad maybe to that agent without you even being part of that experience and will complete that transaction, or with you being part of that experience, clicking on that ad that the agent showed you and completing the full transaction through whatever AI application you’re using. I think this will be the future of what we’re building and what we’re striving for, this agent-to-agent capabilities that we’re already experimenting quite a lot with and I think there measurement, attribution, the full funnel and events will be, I don’t want to say more challenging, but different than it is today.

ES: Talk to me about the size of this space. How large is this space? Are there some scaled AI-supported apps that people might not be aware of or that they’d be surprised by in terms of their scale?

TS: I was actually surprised when we did our market research before going to do our fundraising on the size of the market. It’s actually quite massive. You have the foundation models, the foundation models are probably around 50, 60 percent of the AI usage today and on top of that you have almost double of what they’re doing in terms of daily usage, impressions and applications. You have a ton of different AI-native apps, software, you have vertical AI which is exploding in their world. You have Open Evidence as a good example, it’s a vertical AI for physicians and doctors in the US making around 200 million dollars a year just from ads. You have vertical AI for lawyers, you have vertical AI for all kind of different things and of course you have everything around the agents, embedded AI, docked agents in websites and so on. That’s massive. And it’s only growing as time goes by. So the opportunities and market as we map it is really, really big in terms of the size of supply and the advertisers are super keen to get into that opportunities. We have a good match of a lot of supply and a lot of interest from the demand side to be there as well.

ES: Do you think the reality is that the TAM is everything because basically every consumer-facing app will adopt some kind of natural language interface?

TS: I think definitely yes. Every website, every app, even gaming apps, even a lot of consumer apps have added some conversational layer to their platform or some AI capabilities. Doesn’t always have to be a conversation. We do see a lot of even traditional platforms, websites, applications adding those capabilities into them and then we can help them monetize those opportunities as well.

ES: What was it like going to market to raise money with this? Advertising historically has been less exciting from a fundraising standpoint. You had a fantastically successful fundraise. What was the experience like? Did the AI updraft collide with the historical investor skepticism of ad tech?

TS: For us I think it was a combination of our passion and experience in ad tech. We’ve done it, we’ve built massive platforms, we’ve built great success at SuperSonic, IronSource later, my partners at Unity. I think that experience and our understanding of the ad tech market was key. The second thing is that we look at what we’re doing as almost like an index bond for AI. This is how I saw SuperSonic when we did SuperSonic. Basically we said okay if gaming is going to become large and become massive, SuperSonic, IronSource we’re going to grow together with it as an ad network, as a mediation platform. We’re going to have a lot of clients using us, a lot of clients spending money with us, monetizing with us, marketing with us and so on and we’ll grow with the market at least and maybe we’ll be better, we’ll grow faster.

How I see what we’re doing now is the exact same thing only for the AI world. If indeed AI will become as big as everybody thinks it will and take over every part of our lives in software and applications and be embedded in more and more interfaces, then Velocity will grow with it because hopefully we’ll be able to monetize with a lot of these opportunities and to market a lot of these opportunities. I think the combination of our experience, the index on AI growing with AI and I believe in the thesis that building software is going to become a commodity and our investors really feel the same way. They all feel that if that’s the truth, everybody’s going to need another monetization layer at some point and maybe even more than just one. So we’re going to offer one and hopefully going to be adopted by millions and millions of users, developers and so on. That was the core success reasons for the raise: the team, AI, and this thesis that monetization is going to be a need and distribution.

ES: And the distribution point is pretty key. I published this long-form podcast series called The Prosperous Society and that was the whole basis of the series. It’s like all the binding constraint moves to distribution because there’s going to be this just massive baseline increase in the amount of content available. Well, the good news is advertising systems are really, really efficient at matching interests with the most relevant place to express them. That’s probably the most efficient way to discover things is advertising because there’s the commercial component of the bid, but there’s also just the ability to measure that intent. I’ve seen a lot of startups that are just saying well we’re going to do ads for chatbots. That’s kind of like a bad model. I don’t think that’s that interesting. Talk to me about the difference between ads in AI output and truly AI-enriched advertising because that’s what Velocity is doing. It’s actually capturing the value of the signal to do routing the most efficient way.

TS: I think AI in the ad tech world has basically three layers that it’s impacting and helping. One is on creative and messaging. You can use AI today to create a creative on the fly for Eric based on your location, your language, based on your interest with a message that says even your name or your hobby and you’ll get something custom made for you as a creative and Tal will get something completely different. This is something everybody can use, not just AI-native like what we’re building. The second thing on AI is just better optimization. When you’re looking at machine learning models and matching and bidding and overall performance, AI is dramatically improving that. The last layer which is the most relevant for us is everything around intent understanding and having the ability to see the intent, understand the signal, have a long-form conversation from a user and extract whatever is interesting there in terms of really understanding what can be incremental, what can be relevant for that specific user. You can understand goals, context, preference, budget, a lot of different things that you are unable to get without this AI interface. That’s the relevant layer for us and that’s where we’re focusing as well. We’re going to do creatives in real time as well, we have better optimization in machine learning algorithms because of AI, but that’s something that I think is becoming a commodity for every ad tech company in the world. The third layer hopefully will be something a little bit more unique that we can have as an advantage or as another layer of targeting and understanding intent better.

ES: What’s the sophistication bar for starting an ad tech company now versus like 2013 era? Because it feels like ad tech has always used machine learning for optimization. That’s not new. But it seems like looking at the kinds of things that Facebook is doing, you need much more capable technical talent to really launch something tractable now than you probably did in 2013. Is that the case?

TS: 100 percent. It’s actually funny because we spoke about how much AI is actually building software and technology easier. I think when it comes to ad tech, it might even be the opposite because you need to have the right type of people in terms of the tech capabilities, experience, understanding of machine learning algorithm and so on. But also if you look at what we’re building, we’re going direct. How do you build the SDK? How do you actually build technology that someone will be willing to integrate your SDK? You know how hard it is. Nobody wants to use a third-party SDK because of what it can potentially do to your application. Whether it’s web or mobile, I think you need to raise even more money than people think today because you need to be able to have that capability of learning with real traffic and maybe even bleeding some money on performance understanding and model optimization and so on. The experience of actually doing that in the past, understanding the machine that you need to bring and using the resources to bring the right talent and to actually go into market aggressively and build and bring supply and bring demand, that’s much, much, much harder building this two-way marketplace as well. You can’t commoditize that. You need to just go and do the work and it takes time and it takes a lot of heavy lifting and the right team. But that’s what we love to do and what we’re passionate about. You have to fight the algorithms of Google, of Facebook, of AppLovin, of Moloco, of Liftoff. These are insanely amazing companies with super smart people with all the money in the world and all the data in the world and that’s what you need to go up against. Again, it’s a different world we’re going into in-chat and native AI, but you want to have the technological capabilities in the end of what these companies are putting the bar at. That’s what you need to build.

ES: Finally, how do you see the space in 12 months?

TS: What I see is there’s a gazillion more AI products being built every day. Every day that we do our market research and we’re on top of all the different developers and all the different apps and we speak to them, we’re very much in the market doing hundreds of conversations every week. We see an explosion of these AI apps in every category. In terms of adoption of advertising, I definitely see more and more AI companies adopting advertising. We see it ourselves with our own technology design partners. Developers are telling us I would never even consider doing ads a year ago, now let’s test it, let’s see how it works, let’s see the impact on KPIs and so on. I feel that advertising is going to become a more and more meaningful part for the demand side. So more and more advertisers are going to want to target and find users within these AI interactions. OpenAI is doing us a lot of work on it and helping us and a lot of people are already buying on ChatGPT and then when we approach them to buy on us, they already know what we’re doing and how it looks like and what’s the value that we can bring and what’s intent and so on. So I think that is going to continue. Maybe another point is that AI-native ad formats is going to continue to emerge and going to continue to develop and we’re going to see a lot more different ad units within those AI experiences that we haven’t seen before like mini ChatGPTs or agent-to-agent and more experiences. I do think that the performance of the monetization with the right segmentation on monetization is going to help cover inference cost for a lot of these developers and then they will adopt more and more. That’s how I see it. Again, I can only say from my experience what I see now and let’s not forget that I’m super biased of course. But that’s how I see the market. I see it all like it’s just the beginning. It’s gaming 2012, man. That’s how I feel.

ES: Tal, this was fantastic. Thank you so much for sharing your insights today. How can people learn more about Velocity?

TS: They can go to our website, velocity.io. They can reach out to us on LinkedIn, they can reach out to me. We’re always happy to speak to anyone that is interested in learning about what we’re doing, about the technology that we’re building, the problems that we’re trying to solve. Reach out, we’ll be happy to speak.

ES: Cheers. Thank you so much for your time.

TS: Thanks so much, Eric.