Latest / Unharnessed / Will AI Agents Revolutionize DSPs or Just the Clicking? | Olyzon Founder Jules Minvielle

Transcript

Will AI Agents Revolutionize DSPs or Just the Clicking? | Olyzon Founder Jules Minvielle · Unharnessed

• Moffie: welcome to another episode of Unharnessed. We have Jules, who is the founder and CEO of All Eyes On. Jules, welcome. Jules: Thank you. Hello, Jo and Mo. Nice to be here with you. joseph hirsch: All right, let's just jump right into this dude. This is what I understand. I have been told that the name of your company, All Eyes On, is based off the Tupac song All Eyes on Me. Yes or no? Jules: Yes, true. you know, when you come to a name you need something memorable, nice, fun. And that's where it comes from. That's probably not how you transit it without knowing the story. You would probably say something like Olizon or like you would struggle now that you know the story, I hope you will always remember the name. All eyes on TV, like all eyes on joseph hirsch: Alright, are are you comfortable wrapping a small verse from the song? Jules: I am definitely not. I can't. I can't sing. I can't dance. It's just I just can't do it. I'm sorry. It's not a joseph hirsch: Right, but I didn't ask you to sing or dance. I asked you to rap. Jules: Still. I still I don't rap. I have many skills, but I can't rap. I can't sing. I'm sorry about that. joseph hirsch: Alright, tell us about the funding. Jules: Yes, we raised Series A a couple of months ago. done, it it's it's already it it looks like it's already a long time ago now. but yeah, we raised Series A with S4 Ventures, Sanya Partalo and Martin Sorel who are really experts in the industry. We just need more to expand in the US, spend more in in marketing and and be able to to keep building our products, the best products on the market. And we're happy to have them with us. joseph hirsch: Alright, here here's a burning question. Why is it that French ad tech is better than American ad tech? Jules: And that's a good question. First I think we have a few successes. I can name a few, like sticky addstive got acquired by Freewill, and then we had Creteo and we had Creative and Tids and few others, like Adulike and others and Vibe now. And I think it comes from the fact that we have great talents in France, engineers, data scientists. That are loyal that have a lot of experience in industry that come from all those successes, and that's a really cool pool of candidates. When you build a company in France, that you have really talented guys who know how to build strong ad tech. And then I'd say that we have a tiny market compared to the US, and we need to differentiate ourselves. It's a land of a lot of ad tech companies and we go into details. We have the same old codes as here in the US, but Publicis is French. And they are very they're paying attention to a lot of details. So if you wanna sell publicists in France, you need to make sure you have the best product before you go to market. And I think we are able to really do a good manufacturing like we do in luxury, but for our tech it's the same. And then we come here in the US to scale it. joseph hirsch: You don't Moffie: Yeah, but you guys are also you guys are also cocky, a little got swagger. Where's that coming from? Jules: I I I don't know if that's true. I I you you you tell me that. it's just that we come here, we believe we have strong tech. At least I believe I have strong strong tech here. joseph hirsch: had a different thought about the French. My thought was the French are revolutionaries. Jules: All right. joseph hirsch: And that's what Jules is. He's just like a renegade. He's got thousands and thousands of years of revolutionary blood coursing through his veins. And the moment that he sees the opportunity to disrupt, he takes the flag. He takes the French flag and he stabs it through the heart of his competitors. Jules: Love it. Yeah, maybe, maybe that's that's a good way to explain where we come from. isn't it all about the revolution in this industry anyway? What do you think? that's why we love it, no? It's like always changing. You need to understand the users' habits. now we talk about AI obviously, but then it was TV, it was digital, it was like everything is new every other year. You need to understand how people behave and then you need to adapt. And maybe yeah, create your own revolution. joseph hirsch: I couldn't agree more. the the the revolutionary force of ad tech is what creates the startup environment that provides, the fuel for ad tech to evolve. So I think I think we've really hit on something here. And I think I think the the obvious next question is, what what is Olazon revolutionizing in the ad tech space? Jules: Yeah, I'm happy to answer this one. But maybe one another food for thoughts regarding this. When you come from a small market and when you have less funding than in the US, you also need to have a lean team and to be agile and to do more with smaller funding. And I think that's also a way to be revolutionary is like you do more with less. And that's I think what I've been doing for many years now. When you come to the US you realize that you s you need you need more to scale, but the first Zero to one stage is is great when you come from a from a place where you you have to you have to deal with less funding. clearly we we started in CTV two years and a half ago. So to answer your question, Joe, it's it's about how can we help the brands to better spend on TV. CTV is bringing is changing everything. You can target differently, but in the meantime, it's very fragmented. You have hundreds of thousands of programs, it's a nightmare for the brands. and agencies to understand where they should run. So we just started by helping them to navigate across all the shows, all the programs available on TV, on CTV, by scoring them, ranking them different dimensions, and we had agents to do that at scale. And joseph hirsch: All right, well let me stop you there. How do you get access to show level data for C T V targeting? Jules: Well, it's getting more and more a thing now. So I guess it's not as opaque as it was two years ago when we started. And and first we started with with what was available. You you you go direct with the publishers, you ask what they can provide you with, what they are comfortable with in terms of show level data or channel level data, and then you enriched it, you gather data from third parties on what the show is about, what awards did they win. And it's not just about giving the name in full transparency, because if we give you the name like V Kings, what do you know with this? What do you do with this info? Is it good for your campaign? Is it a show? Is it a documentary? Is it the audience you are looking for? So it's not just about giving the name, it's about classifying it and making and and deciding whether or not it's relevant for your for your campaign. So that's why we are talking about decisioning. Because we just make sure that when you give me the name of your brand, I tell you if Vikings is relevant or not. That's what we bring to the market. joseph hirsch: And you do that in real time? Are you are you making decisions on the request level based off the show? Jules: Yes, but we classified the show before. So we have a a stock, we have a a database of all the shows that are classified. And when we see a request coming, we already know the data about the show and then within ten milliseconds we're able to tell a brand whether or not it's relevant for them. joseph hirsch: Alright, you can do it within ten milliseconds. So I gotta ask, are you using JE or a small language model to make the decision? Jules: No, it's very ba to be honest, it's it's very basic. As long as I have the show, I just need to I just need to to decide whether or not it's in the it's in the bucket that you need to target. So it's yes, we are partnering with the SSPs that allow real time data data enrichment, but the the piece of tech behind it is not that complex, to be honest. Well the most complex part is how can you classify, rank, score all the shows in all the dimensions. joseph hirsch: All right, so I'm Moffie: So you announ joseph hirsch: Go ahead, go ahead. Moffie: I was saying you announced your orchestration layer. one thing to clarify, I know you guys were focused on CTV and you're orchestrating buys across different DSPs, but are you now expanding? We have this notion of umps, these unified media platforms that was coined by Eperium. Are you looking to buy across more than CTV now? Jules: Yeah, that's a great question. Where if you look at what was the promise of the programmatic industry a couple of years ago now, more than 15 years ago actually, the promise was to buy everything from one platform and we realized that it's more and more fragmented. So when we wanted to work with our clients on a lot of managed campaigns, we work with more than hundred one hundred brands, dozens of agencies, all co-s independent agencies, we realized that the fact that it's fragmented is creating a lot of friction. And we do a lot manually. We onboard everything in Excel file ourselves. So when we buy from Netflix, from Disney, from Prime, on YouTube, we need to deal with hundreds of not hundreds maybe, but dozens of different platforms, logins, and measure is the same. You measure several KPIs and then you get data from everywhere. So for ourselves, first, a couple of months ago, actually nine months ago, we started to build agents that are plugged into the different DSPs. We started with the trade desk, Amazon DSP, D Versus 60, Vibe, and many others. And we realized that it were actually were saving us more than cutting the time that we were spending on the on the optimization by more than 80% on the setup, and then on the optimization, we're optimizing five times more than what we were doing. And then we went to market with our best clients on with a simple proposal which was do and test it for your own campaigns. So we went to agencies and they started to use it. And they needed more. They needed more than CTV. So we added Spotify. we plug direct inventory such as Universal Ads, which is still TV, but we also plug digital autofoam and I mentioned YouTube and now we are extending to other channels such as OpenAI ads. Because if you want to orchestrate all your media in one place, you cannot have one orchestration layer for TV and then another one for your digital audio and then so and so on. That's also why our platform includes email management, because on TV and in some other channels you still have a lot of emails. Emails to deal with direct I/O management is taking a lot of time for the buyers, the traders. So we also look at the email data, the attached file. we can our agents can send email with all the tracking links, pixel, creative attached and everything to save hours to the traders and media buyers. joseph hirsch: That's fascinating. So if you if you're let's just say you're connected to 10 platforms, the the buyer, the agency, the advertiser, is there any awareness that the platforms even exist? Or is the net result just better performance and your your agents interacting with the platforms kind of on their behalf and and they don't care how you got the supply? Or are you hearing from your clients that are they dictating? you need to access the inventory through this platform. Jules: it depends who you are calling the clients here. the agencies clearly they are the first clients for this solution. And we are actually integrated with 35, more than 35 platforms now, and they clearly see the benefits and they come with their upfront, with their media strategy, and they want us to just replicate or to get just basically go as they already go. We are not joseph hirsch: Sure. Jules: here to challenge their setup. If you talk about the brand, the brand, sometimes some of them will be like, yeah, I need to understand what's going on with the agency KI. I want to run a POC with my agency. Others will be like, that's not my concern, that's the agency's concern, and everything that helps them to do a better job, I'm in. Or just sometimes they just don't even want to hear about it. But this is this is changing more and more. We see that brands now want to understand how the media is bought, if it's bought effectively. And what can be done better? Moffie: So we have agencies that are also building their own agents and orchestration. How are you handling that if they want to bring their own agents to the table? Jules: First there are different initiatives on different segments and blocks and then there is communication about those initiatives and when you get into more details with each agency you don't you understand that not everyone is as advanced sometimes as they as what they say. We plug into whatever brick they already built. So they already some of them have some agents around the data, others around the insights. Some may have already built some agents around the activation. We build a harness where everything can come together that works in their own terms and constraints. So there is a bit of onboarding, especially with the big whole codes, to understand what's their current setup. But we're not here to replace all agents. There are some great MCPs on the market and we cannot do everything in one place. So we just build the harness where all the agents that they are building or they already built all the partners they're already working with can work all together in one place and we build them the brain for their agency, the ontology for all their agents. Moffie: So you said our buzzword, we are unharnessed. You mentioned your harness. what differentiates the Elizon harness? Jules: Well it's purpose b first it's purpose it's it is sorry purpose built for media and we co-built it with our clients who are the biggest agencies and brands in the world based on their own needs. Then it comes with service. we help them for the onboarding, for the management of the first campaign. so yes, I will insist on the tech, but service I think is something that really matters in this industry. then we come with our own knowledge and two years of data on CTV buying on global campaigns. We run across more than 20 pay 20 countries. And and then it's the way it's the way it's built, it's modular. it's not just a meta DSP, it's not just agents plugged into the API of the DSPs. again, as I said, it works with email management as well, it works directly with some publishers such as Universal Ads. So it can really follow the the flow that the agency already has. Moffie: I wanna double click, you said email again. Is it just emails from the platforms or are you actually looking to get access from the emails of the traders using your agents? Jules: Yeah, you can create you can create an alias or mailbox that you CC for all your all your conversations with all the vendors and the agents can look at it and send emails on your behalf, look at the emails that you received and just do as your cloud would do in your own mail, you know, prepare the draft and do everything. It's just that you need everything to happen in one place. It's crazy the amount of time that is lost. It's more than ten hours a week for traders. to do dumb not dumb, I'm not saying it's dumb, but basic things that agents can do. So they could f focus more on optimization, on strategy, and bring more value. That's why we try to that's what we try to Moffie: Yeah, I'm interested in and how you're getting that access, especially with bigger agencies, because that is really getting email access for Muse, Instinct, and these really hot AI assistants is the unlock that gives them a ton more context on users to make it more proactive. So are you getting that level of access, like permissioning into an agency's Gmail? Jules: No no, we we that's not the purpose. The purpose is really they can create a dedicated email for the campaign. So we don't need to go into all history. You can only start when the campaign started, or then if you wanna bring more access you can. But the there is no access that is mandatory on all the all the email exchanges. joseph hirsch: Yeah, I think if I understand it right, you're gonna C C agent at Olazon or or client agent at Olazon, and it's gonna be a forwarded email with an attachment, and that attachment's gonna be the IO, it's gonna be the creative, it's gonna be the line line item details, it could be the tag sheet, et cetera. And then instead of the trader having to take that data and do the trafficking, it's gonna do that on the trader's behalf. I invented all of that, but I assume that's what's what's going on. Jules: Yes, and sometimes it's about setting up on behalf of the trader, but it's also about sending email on behalf of the trader. let's say you want to buy home screen from a partner that can't sell the home screen programmatically for some reasons. that the vendor is still using emails to receive all the information about the CPM and everything. Yes, agents to agents will happen and solve this, but we still have some vendors that that work on email. And then the trader will send an email with all the information, the CPM, the flight dates, the the the budget, the tracking links, the creative and everything. It's it just take like thirty minutes, twenty minutes to put all of that together in one email, nicely built if you are a human and it takes one minute if you're an agent, because you already have all the data about the company. But with that's not joseph hirsch: Well what Moffie: So Jules: all we do. I I don't Moffie: you've said Jules: wanna say that we are building email agents. That's one of the aspects of what we are building, but what some what we're building is much more complex. Moffie: You've talked about basically the ability to advertisers are paying a DSP and an SSP for transacting and now agents can handle that end to end. Do you see the kind of supply chain or the ad tech chain collapsing a bit? And where do you see that playing out? Jules: It's it's a risky question, right? Very pol ver very political question. many many legacy players will be challenged in the coming months and years, definitely. When you see platforms like Universal Odds and others will announce the same soon, allowing API access directly to their inventory. The question for the DSPs and the SSPs of this world will be: what can I bring here that that still brings value, carries value to my clients? Today, in terms of aggregation, in terms of bid optimization. Agents are not doing this. So I'm not I'm not trying to oppose and say programmatic and agentics. I think agents can help in programmatics. That's why the first platforms that we plugged in were the Tradesk, Amazon DSP, DV260, Vibe, DSPs and SSPs, such as Magnite and Pubmatic and others, because we believe there is still a lot of value in having beat shape clear, having just like RTB, RTB on a lot of inventory aggregated in one place. So for a long time I think there will still be value in SSPs and DSPs of this world. Moffie: So yeah, like you said, part of your agentic strategy is optimizing programmatic campaigns, just making it more efficient with agents. And then there's the more disruptive agent to agent where it's potentially bypassing programmatic and booking directly in an ad server. so far, agentically, what if you went from like percentage of spend, has it been more? The former or the latter, programmatic and just making it more efficient with agents, or the new, more disruptive and leveraging new protocols like ad C P or AMP from the IAB, et cetera. Jules: No, clearly for now, I think the best way to leverage agents is to go through the stack that is already there and where the volume is. The volume is still programmatic. And in TV it's still direct IO. Ad CP is an interesting project. We're happy to contribute. To be honest, it's not scaling yet, it's not scaled yet. And most of the inventory is bought programmatically or direct, not through Ad C P. And I don't think it will happen overnight. I think it will take time to go through agents to agents, because we still have a lot of challenges that we need to solve before we move away from programmatically from programmatic one hundred percent. Moffie: Well, that's my biggest question then. So let's look forward to twenty twenty seven. What has to happen to really scale meaningfully? We can just focus on CTV, the agent to agent buying. Jules: Well, first, to be honest, I'm agnostic here. I'm not that's not my objective. I'm not relying on it. I grow my business, whether or not programmatic stays or or dies, that's not my that's not my battle. I'm not as revolutionary as as as Joe said. I'm still happy to work with programmatic players. it's all about the efficiency. if you can buy direct inventory from a publisher. And it makes you save 20, 30, 30%, then maybe the the frictions that you have by moving away from programmatic to full agentique, you can spend a lot of time trying to solve them because you see that there is clear value. If the cost of programmatic is a couple of percent and the cost of changing your setup is high, maybe you will you will stay with programmatic. So I guess it's all about finding the right balance, the right in cost the right cost. Programmatic D SPs, SSPs, they bring a lot of value. They just need to find the right pricing for the value that they bring. joseph hirsch: Yeah, I think that's a fascinating take. I think the two things that I think are critical there is one, does agentic trading push down take rate and programmatic as a mechanism to reduce atrophy, or not even atrophy, but like to reduce, you know, I guess just a transition. And then the other thing is like if transitioning, as you said, is more work than it's worth, then you know, why do it? So I think I think those two things are critical. One, is it easier, is it less friction? Is it more efficient? And then two, is it is it that much cheaper? and I think I think in twenty twenty seven we'll see we'll see both of those things play out at scale. Jules: Is it worth it? At the end of the day, it's like, yeah, it's nice, but I'm not gonna change everything just because it's new. It's just it's not because it's new that it's better. Like, what does it bring me? it's clear to me that a harness of agents dealing with everything that you do, orchestrating your media across all the channels, that brings value. Replacing the DSP or the SSP. Is still something that everyone is looking at, interested in, but the conclusion of whether it's the future or not, I mean we don't have the answer yet. joseph hirsch: Yeah, just don't tell Brian O'Kelly. Jules: Maybe, maybe No, I won't understand. Moffie: So you mentioned orchestrating across multiple platforms. And Jules: Ha ha. Moffie: basically they all have like their own agents, like Amazon has now for the Amazon DSP, has its own agent. Jules: Yes. Moffie: is there a world? Well, let's just stick with Amazon, where they've blocked Meta's Muse agent because they need to get better terms if if any, rather than let Muse just go wild shopping. On behalf of Amazon customers on Amazon. are you concerned at all that the walled gardens that are building their own agents may block you and orchestrating into their platform to benefit their own kind of agent? Jules: What's your what's your take on this one? joseph hirsch: Well, my thought is is that if you look at Google and you think that there is some kind of risk, and the risk is anti competitive, the hope is that other ad tech companies look at that and they're like, I'm not going to make the same mistake and prevent other companies from competing with me by doing something that's anti competitive. That's like That's like the answer that I wanna believe. But, if you're a big tech executive and you wake up one morning and you want to play God, you get to make that decision now and face the repercussions maybe ten years later. So, you know, it's it's really like, well, you know, have I have I spurned the ad tech gods? Did I pray to them loud enough? Was my prayer heard? at least for me, when I'm praying I'm always trying to get the volume right. Jules: Okay, I let you pray and then I will analyze the last couple of months and what happened. I think I don't have the answer for the future, but clearly the last six to eight months showed a very interesting trend, which is more opening. Like the Tradesk opened their API more than they ever did in the past. it's easier to get access to universal ads than it was a year ago in direct. And you see these trends over and over, and I'm sure that Here, no one will be surprised if Netflix announces an MCP soon to allow buyers to plug directly into their inventory. So I would say that the trend in the past six to nine months, and not only from the publishers, because we can understand why Netflix or Universal Ads are doing it, but for the trade desk it's a different question or DV360 to come back to your point, Joe. But I th I see clearly a move towards more headless DSPs. Yahoo DSP were the first mover or the first one to really Go public about it. And I think that's a trend. They realize that if they wanna we are talking about the value that they bring to the market a bit earlier, if they wanna keep bringing value, they cannot force people to go the old fashioned way and click on buttons when everyone is just using agents for everything. You use Instinct to book your next tickets, you use Cloud to prepare your presentation, and then you need to go and click in DB360 to do your campaign. It doesn't make any sense. And they know it. Whether or not they will make the change now or later, we see that they already started to do it actually. Moffie: I think that optimistic note is a great place to end things. It was great having you on, chatting all things agentic. Jules, anywhere folks should try to find you. Jules: Of course, yeah. everywhere on LinkedIn. we are in New York, in Fi Dai. Everyone can come and grab a coffee. And yeah, we'll be on every single industry event as well, of course. Moffie: Well, great to have you and this has been another episode Jules: Thank you. Moffie: of Unharnessed. Jules: Thank you, Mo. Thank you, Joe.