Serge Kajirian - TNDE
===
[00:00:00]
mike-farrell_6_07-17-2026_123424: today's guest runs data governance and privacy operations at Universal Music Group, the world's largest music company, where the assets under his control care include the most valuable catalog on Earth and the fan relationships of hundreds of labels and artists and brands.
He spent more than 15 years leading data teams across media and entertainment, including Disney [00:01:00] Interactive and Fox Network Group, winning innovation awards at both, and he keynoted the International Copyright Technology Conference. He sits at the exact intersection the show is about, where rights, permissions, fan trust, and AI collide.
Please welcome Serge Khazarian
serge_1_07-17-2026_123424: Hey, thanks, Mike, for having me
mike-farrell_6_07-17-2026_123424: quick frame before we start. A lot of people treat AI like it's a new product category. I keep coming back to it thinking it's something closer to a medium shift. Uh, the way that Borders watched as Amazon rearranged what it meant for-- to be a bookstore in the online era back in 2000. The catalog, the fan graph, the permissions around both, that's the fight.
And Serge, you live in that fight every single day. so getting into the first, first topics here. The market used to price UMG per stream. Now every AI company that wants to build music features without getting sued needs your catalog. From your data governance seat, does that catalog feel more like content or like training data infrastructure?
serge_1_07-17-2026_123424: Yeah. [00:02:00] Um, you know, UMG is first and foremost a, a music company, right? Our, our-- historically we're valued as a content-driven company. largely now driven by streaming data and royalty economics. Uh, that whole aspect hasn't gon- hasn't, hasn't changed, right? Um, AI is definitely shifting the conversation. today, our catalog isn't just, you know, library and you go and you pick your track and you go and have a great You know, there's, there's consumption data to, to, to ingest. There's foundation of infrastructure around that so that things like AI can, can be leveraged. Um, from a governance perspective, and this applies to everyone in our industry, um, how do you value not just the recordings, but also the data, the metadata, the surrounded fandom information that you collect over decades now? Um, you know, as you, as you build a [00:03:00] rights-aware infrastructure for, for, for anyone in the music industry that'll understand that, you're, you're gonna be, bound by obviously the legal requirements that, you know, on paper what you're allowed to do for royalties. but then you also have to understand the access control, then you have to men- uh, build a semantic layer for, for something like an AI agent to come in and understand that.
And we'll be talking about that, I'm sure. The metadata layers have, have really increased with the music industry. Um, libraries now are filled with listening behavior. The advent of, of, of Spotify has changed our industry. Um, you know, platforms are exposing new metadata, enriched data, experiences of social media all around music. That's an, an, an awfully lot... It's an awful lot of, you know, involvement. Uh, you need permissioning around that, governance around that, infrastructure around that. Um, the asset around that becomes apparently [00:04:00] exponentially more valuable.
mike-farrell_6_07-17-2026_123424: Mm-hmm.
serge_1_07-17-2026_123424: And my role at UMG is really to ensure that things like the catalog, which, you know, I'm not, I'm not at that l- no, I'm not close to the music but I understand that, that, you know, there's catalogs that produce metadata and that data has to be safely, securely managed. it has to be leveraged in a way that, uh, not just for our own, you know, uh, um, audiences that need to, to expose that, but then new interesting AI audiences or, or the AI agents would now do- need to be exposed to that, right? How do you present respect those boundaries, right? Where they matter, right?
Like, you know, they n- we're gonna go into, um, areas where years ago distinctive separation was there, and now you have artist-specific requirements. You've got very tech-savvy artists and labels who understand the space. They have to be, um, curated with, with, you know, with all the different [00:05:00] distinctions like licensing and regional information and royalty programs, again, I mentioned. That is all... On, on top of that, you have the privacy consent world, which we now have to deal with. that is a changing, moving target. For, for someone in the music industry, it's a lot to handle,
mike-farrell_6_07-17-2026_123424: John?
serge_1_07-17-2026_123424: Do it at scale.
luke-arno_3_07-17-2026_123424: Serge, firstly, it's so nice to have you here on the pod, and thank you for taking us through your role and the scale of which UMG operates. When we talk about scale, when we're talking about thousands of websites and services constantly evolving, I would just love to understand more deeply how does this show up in your day-to-day?
serge_1_07-17-2026_123424: Um, yeah, there's an operational, um, challenge there, if you can imagine. Um, I, I'm, I'm gonna, I'm gonna go on a tangent. W- you know, we're very similar to, um, the NFL in that way, right? We have, we have-- W- we're, we're a group of teams, right? I think the NFL's like what? [00:06:00] 32, 33 teams. the teams all play in the same level field. all respond to their requirements as a business, as a service, we have to then also referee and, uh, allow for a, a charter or a Magna Carta, if you will, to, to ensure that everybody has, uh, respect, not just amongst the teams, but the players. So if you equate that in the music industry, we have labels we have artists. imagine, if you will, we have to deal with, again, to your point, our scale is extraordinary. It's thousands of sites and responsible, uh, apps or, um, assets that have to be, you know, digitally, you know, updated. We have to control, um, obviously consent. Uh, it clearly is a huge part of that governing experience. Um, we have to understand also from an end-to-end perspective where the data flows, how does that data get handled, how is the transfer ship res-respected? then on, then on top, on top [00:07:00] of all that, we've gotta, we've gotta figure out the, the privacy components to ensure that the gatekeeping, the ring-fencing, and the modeling that we're doing to protect this data is secured. that, that requires quite a lot of moving pieces. So the day-to-day is, is really just juggling between those scenarios and at scale, you know, you, you, you have to be a universal problem solver sometimes. I like to tell people that there is some governance theater everywhere here. we've gotta be aware, we've gotta be pretty aware of what's going on, 'cause if you skip a beat, like, I tell people, like, "Hey, b- just be sure you've got everything covered. Ask the right questions before you start handing over," you know, "Oh, is this data AI prepared," right? Are we ready to go that step? Oh, uh, the federated music industry, "Hey, you know, you're supposed to play in, in the same team. You know, don't go across the other side." How do you handle the change management there? Uh, so, you know, artists come and go. How do you deal with some of those scenarios with the data space? [00:08:00] It's what we do every day. Yeah, very challenging.
luke-arno_3_07-17-2026_123424: I love that. I love that answer. It's theater management. You are problem-solving on a daily basis. The correlation to the NFL, it's, it's great. Thanks, Serge
mike-farrell_6_07-17-2026_123424: Yeah, that's great. Um, so I mean, we like to-- we talk about data as currency on this show, and at UMG, you have to segment and value those permissioning for the same person across multiple different labels. One person may be a fan of those labels, those artists. Um, what's your, what's your relationship with the data look like, um, across artist data and end user data, um, all within one single stack?
serge_1_07-17-2026_123424: yeah, you know what? We, we, we do look at data as a currency. I think the trust is the currency there for me, the data governance side and the angle of it. Um, you know, the data has value if everyone in our ecosystem trusts it well. And, and, and that to me is what's makes, uh, UMG and companies [00:09:00] like UMG very unique, right?
The, the- We're not just one monolithic sort of we're a business, this is our biz model, it's a one-stop shop. Um, we have a very clear roadmap for that direction. we collect labels' information, right? We, we collect artist information. Each of the artist information will have marketing strategies, royalty strategies, territory implications. the, the, the, the relationships that they will have specifically with the fans are completely, could be completely different, right? So that whole perspective, um, the way we look at that course, uh, or, or, or the course of that, that, that currency. So if you're a, you know, a single person, like if you're a Taylor Swift fan, you could also be a The Weeknd fan.
I'm a, happen to be a Rolling Stones fan. Uh, and by the way, I love my emerging in- indie, cool indie artists here in LA. So y- you wanna put that in perspective and try and understand how that combined meaning i- is, is important for a label, [00:10:00] right, and is important to the artist, should automatically be, um Uh, uh, ta-tangible.
It should, it should be, uh, you know, the, the fine, the fine balance here is our North Star for data governance is do you enable this, the, the data sharing to what degree, right? How do we, uh, enable, um, a label and how do we define their business requirements, uh, decide on their content strategy, allow for any activation rules? You know, do-- we, we play that middleware, right? I think back to the theater thing. I, I guess I'm directing here, right? And allowing the artist and the, you know, the, the, these, the, the moment to shine and let, allow them to go and enrich themselves with this data and then go build product on top of that, right? Um, now it doesn't mean, other side of this conversation, that, label automatically should have access to everything, right? Or, or, or an artist should have access to anything about their fans and all the fandom world. I think what we do a really good job on is we, we create a ring-fencing [00:11:00] model where we can maximize data sharing, if that's a thing, right? Maybe our North Star again is to say we, we allow that enabling, right? In a safe environment that requires, you know, the labels to, you know, build off of, again, a team comparison, build off of im-important level set data that can, you know, that can s- level up the playing field, right? So many labels, um, are very keen and very interested in, in, again, data sharing, uh, exposing the right type of information when it, it benefits the industry as a whole. Um, a lot of good, you know, data altruism here is, is a, is a, is a, is something that's real. and I think the artists now are known. Um, again, we didn't talk about it, but today's artists and labels are very tech-savvy, very data-savvy. Um, there's, there's no denying that their catalog is important to them, but so is the information surrounding it.
So how do you leverage that appropriately for cross-use [00:12:00] and determine with confidence that you're, you're giving them the right information, uh, but you're, you're giving them also the right access to the ecosystem so that there's no bleed, uh, or crossover, you know, risk. and, and these two worlds, you know, they shift, right?
Nothing is static. Then you introduce a new actor like AI, who is an actor, right? Coming in, stepping in, and wants to fulfill their role, And how do you treat the actor now, right? 'Cause we-- we're, we're, we're all-- I think everybody in every industry is trying to fathom that. And on your show, you talk about that as well.
luke-arno_3_07-17-2026_123424: Hey, Serge, I, I'd love actually to just go a little bit deeper if, if you're willing to take us there on the ring-fencing and more specifically, like, how do you keep labels feeling safe with their customer data or consumer data and metadata stay protected while still letting them play at the, at the level they want or at the right level with confidence?
serge_1_07-17-2026_123424: Oh, you think it's a free-for-all, huh? Um, the ring fencing, we use that word a lot. My team [00:13:00] uses it. We-- A lot of people in our industry use it a lot. It means something different for different organizations or industries, too. Well, ring fencing, uh, is a prerequisite for sharing, I think. Right? It, it is. Just know you inherently think about that. Labels believe that, you know, putting something in, in, in its fan data and, and sharing it on the enterprise means one label can see the other label's information and use it. and then, you know, if you do that and you, and you, you mistakenly build trust and then accidentally slip, you know, you'll, you'll-- everyone will stop participating, right? So there are some rules of engagement. So our governance models use ring fencing as a tool, as a mechanism so that we can just articulate the runbook better, right?
So we know these rules are, are, are, are, are baked in and they, they are, um, there are layers of ownership there. So we think about, first and foremost, who owns the data? Who owns the rights to this music? Who owns the rights to the marketing of this artist? [00:14:00] and if you find ownership, then you can start building your stewardship models, which everyone in our industry tends to do anyway by organically. So, and then the second part of that, maybe a separate layer of, of, of ring fencing is the identity model. And that's a more technical challenge. So we can solve for that technically, right? So we can do role-based permissioning, access control models, data lake sharing safely, things like, uh, uh, you know, you can create exchange hubs and all the modern universe of transfer of data. Um, and, you know, you have to understand when you bake those things, access isn't guaranteed, right? So we have to control how do you get access? Why do you get the access? Did somebody approve your access? Um, there's label influence there. Um, control and ownership of a label could be different than a territory that's doing maybe a one-off, you know, website that's figuring out they, they just need one piece of this puzzle but-- and, and don't need any, uh, [00:15:00] crossover with any other label.
So you have different dynamics. Um- The, the other layer of this is good, um, smart segmentation, the layer, the data itself. Uh, you, you hire really good architects, people who really know what they're doing. There's label-specific data that there's, there's control and ownership with. Um, again, I mentioned territories and regions.
You know, like a lot of other enterprise companies, not just a music industry one, but we have a pretty federated experience. Uh, maybe more than pretty... It's pretty severely federated. you have mechanisms there that have to be controlled. Then you have, um, oh, artist-specific maybe environments. People-- You know, we-- Artists come and go, and they move around or, uh, just like everyone else in the NFL. They got, you got players coming in one team. They're, they're next season, adios, you're on another team. So we have that whole dynamic to consider, around ring-fencing. So, you know, that's, I think, also part of the way we think about our package. Uh, when we talk about [00:16:00] ring-fencing, we have to do policies and, you know, do some po- great tagging. it helps, you know, to put, to put that in perspective. Um, we start doing that again in the consent layer. If the consent space and the fandom data we own isn't properly ring-fenced, then how can we e-expect anyone to really use the data upstream?
mike-farrell_6_07-17-2026_123424: feel like the more I learn about the UMG data model, it-- the more fascinated I get and immense amount of respect for what you all do. Um, so I'm, I'm curious, how have priorities around artist data and user data been shifting as AI has kind of rewritten the rules in, in this new economy? Once upon a time, I imagine artists didn't even really think of fan metadata as being an asset class at all that, that they had control over.
Um, are attitudes around data sharing or the concept of exchanging data in an artist network for the learnings across the industry emerging as new possibilities for how artists can connect to larger and more targeted or more, uh, hands-on experiences with their [00:17:00] fans?
serge_1_07-17-2026_123424: the music industry from, from an AI perspective is pr- very controversial, of course. We can pr- just touch on that briefly. like many companies who are doing creative content, the generative ca- AI concepts are, are all over our modern, um, discussions. I'm not particularly involved in that, right?
The, the, the world of AI is vastly different for me than it would be for the artist or the creator, um, and for the labels that are responsible for that. The, the mechanisms in which you have to protect yourself against generative AI are, are, are again being discussed all around the world, I think that's, that's moving in a direction, uh, it should organically.
Well, I mean, the, one of the biggest concepts for us to breach and to have agreement on is how do you attribute success and, and, um, offer, uh, credit where credit's due, [00:18:00] And I think once you start answering those questions to generative space will start evolving into something more grandiose than what it is today, which is a little bit of its-- still, I think, in its infancy, although we've been trying to fight this discussion since, since around COVID. So, um, the, the space in which I believe AI as an enterprise tool is coming to light is where, again, we're not alone in this, but the music industry, many other companies, is trying to use operational AI to present the challenges that we face real time and deliver, um Deliver, you know, answers faster than we could have ever done before.
mike-farrell_6_07-17-2026_123424: Yeah
serge_1_07-17-2026_123424: for, um, you know, to, to solve for that dramatic problem where it takes me six months to build a dashboard or a product, and now we can s- you know, scope that down to weeks, if not days. Yep. the, the s- the space that I [00:19:00] mechanically get involved with is very exciting. How do we leverage AI and navigate a world where DLP, enriched data sets are being dropped and built faster than they've ever been, ever been done before? Um, when I look at just even my own s- area, Mike, the, some of the DSAR that we have to be, um, responsible for, uh, on a timely manner, that presents a whole another level of challenges. How do we use AI models to help kind of effectively go find things faster and solve our problems? For me, it, it's just common sense operational 101.
We're gonna look towards using AI agents for that. Um, the North Star area for, for us is how do you build that semantic layer so that professionally I feel comfortable, delivering value so that AI rights, license rights, AI da- metadata, the golden records are all maintained and, AI agents can [00:20:00] understand that within reason. Um, you know, it's our role to present, I think, that strategy right now, and I think a lot of music companies are, are, are rushing into to, to discover what that means. I would love to-- My analogy is to people is like maybe 15, 20 years ago when, when there was a mad rush to build collective semantic layers for dashboards, it feels familiar.
This is the same kind of, you know, responsible building that we are-- we're kind of aware of. Now, again, back to the theater item, the actor is no longer you or me, it is an AI agent that can wear many hats. How do I present the answers in a way that I feel comfortable can be re-re, you know, producible, repeatable, um, that have no uncertainty to it or very little uncertainty to it? That evolves us to create a better data lake, um, and that is the challenges we're facing. I think we're not alone. I think it's quite a lot of, and I mean, not just for the [00:21:00] music industry, but a lot of other actors are trying to figure out how to play in that same, you know, same space.
mike-farrell_6_07-17-2026_123424: Yeah, totally. And I, I mean, I, I've always-- Like, when I talk to my friends about, in the music industry about all these things, we always talk about how the, the artist and the persona is what, you know, people really believe in, and that's, that's why they listen to someone's music, less so than, uh, maybe, you know, the, the riffs or whatever is going on in the music itself.
Um, and I think as I think about UMG and this, this new AI world, it's, it's really all of that metadata and, and kind of helping artists separate that out to understand, like, there's so much information, metadata that UMG has to learn and share across these different artists and labels, um, and helping them lean into that and learn more about their fans and have their fans learn more about them, um, and connect to just really unique pockets of the world.
Um, [00:22:00] like I-- there's always that Justin Bieber where he's like, he had that funny interview where someone in a small town in Sweden knew his name. Um, but UMG really is, with the metadata catalog you have, allows for making that a reality for smaller up-and-coming artists all, all the time.
serge_1_07-17-2026_123424: yes, the, the fact that, you know, again, UMG's catalog as vast as it is, breadth that we've had for, for, you know, years, decades, that's the, one of the founding principles of why we feel so, very comfortable in, in, in knowing that we've got data that no one else in the world has, right?
I, I mean, I could say that, uh, my personal opinion on that is, you know, it's, it doesn't-- it's irrelevant. It's there. The world knows UMG owns a vast majority of the catalog of the world. But we also know because of the past experience, that's the benefit and the promise we deliver to our new artists, right?
That's what-- I mean, I'm not pitching for UMG, I'm just saying it's natural
mike-farrell_6_07-17-2026_123424: Yeah
serge_1_07-17-2026_123424: learned so much, now we have the data to back it up. It, [00:23:00] it's, it's, it's, it's a huge library, right, of information that's, that's available, and we can grow from that and build new fandoms, right? That's the key for everybody is what is, what the listening behaviors and the whole, uh, mechanics behind it are very interesting. go ahead.
luke-arno_3_07-17-2026_123424: Well, I just, I find it so interesting. I, I mean, what we're talking about here is this incredibly rich intersection of data, which UMG has a, a catalog that's, uh, uncanny to anything else in the market. And speed used to be the advantage, and AI being the catalyst, speed is no longer the advantage. It is the baseline requirement to operate in this modern economy, and ubiquity and access of which is nested in permission and controls, and you're solving for this at UMG in incredible ways. I, I would love to understand, Serge, when you think about an environment where you're trying to make this AI-ready activation or semantic layer with AI [00:24:00] active in a federated music data model, like how do you do this? How do you go through? What does fact-checking look like in this environment? How do you operate with confidence and speed?
serge_1_07-17-2026_123424: Uh, yeah. We put together a framework first and foremost to, to familiarize ourselves with the, um, the needs of the labels and the artists there go, right? As they're representing them, the, the artist. And the fact-checking, the m- the model itself is just constantly being evolved. We don't have Uh, because we're a federated model, not one person's gonna answer your question, right?
It's gonna be a multitude of personas, of stewards, of owners who are constantly checking and rechecking. And the, um, point about kind of the prep- the prep work we're gonna have to have for our AI environments, common knowledge right now we're all attempting the same sort of signal checking models. We're, we're [00:25:00] looking at what works, what doesn't work. We're figuring out, um, even the models themselves are causing us havoc, right? And every time they do a version control change, we're gonna have to fact-check again. I mean, this is a modern problem to have. We, we didn't have this problem when we built a data lake.
We were rock solid with, you know, uh, exquisite control set, uh, in stone. None of that was a factor, and now we're having to recheck every model or we're supposed to, or we're told to, that's gonna have to happen. that's news to me, right? Like, oh, no. Um, the governance isn't just also about compliance. Okay, so with that part is another factor that it's baked in first and foremost.
So we are totally understand that, you know, the com- the, the compliance engine here is running and is never shut off. the, the enabler part of that is, you know, then you have to deal with the territories and what are their restrictions. What are my labels and the artists were, you know, res- responsible data sharing requirements need to be honored?
Again, back to that ring-fencing discussion we had earlier. the [00:26:00] framework the, that's the everlasting engine that goes... Oh, my, my team know me. I, I talk about runbooks a lot. I use that word like it's just everybody needs a runbook. You know, if you don't go to have-- Wake, wake up, have coffee, make a runbook, right, with breakfast. So you're, um, you know, the runbooks are gonna be trusted, and I think that's gonna also enable AI to understand their bou- of course, their boundaries, right? And so if you have really great runbooks as any company go-- Given the fact that you've sprinkled governance theater all over it, you know, the layers can be freely exposed, right?
Then you can train and expect and understand a different kind of AI language, and hopefully, that's what you need to build trust. Um, am I looking at every single data set and, um, you know, applying controls that, that, that fit a purpose? You know, no, it's, it's-- I'm doing it broad strokes, right? Again, at scale, this is very hard to do. If I had with single data [00:27:00] set, there'd be never, there would never be questioned. Uh, when you're dealing with thousands, tens of thousands or hundreds of thousands of data sets and models are all over the world, uh, you know, uh, structured or unstructured, you don't-- Where are you going to start, and how do you start peeling that onion?
The runbooks help, and that sets the, the, the precedence for us to avoid mistakes 'cause every time you fall, you-- every time you fail, if there's something, there's a, there's a risk, you need to be prepared to rewrite the runbook so that that doesn't happen again. And that, that's the constant challenge we're, we're doing and facing right now
mike-farrell_6_07-17-2026_123424: That's great. Appreciate the insight there. Um, something I think about in, in just the exchange between artists and labels and UMG, um, it feels like these artists, they get distribution, marketing, and compliance at scale, which we all know is, is quite difficult to do with all these regionalities. Um, but how do you see that relationship changing, um, in this new data economy?
serge_1_07-17-2026_123424: The exchange value proposition historically has been [00:28:00] really pretty straightforward. Uh, with, with, you know, you, you have artists, they make music, they brought the music to, to, to the label, the label then produces the distribution chains and, you know, there's a very clear sort of, you know, the marketing paths.
This all the engine there was pretty well-defined, and it is a very matured industry, right? So it has understandings of the scale and the infrastructure, and so one day it's CDs, the next day it's, you know, digital streaming. So that part of the music industry is the same. It, it, you know, uh, you, you, you got infrastructure to steer, uh, artist's career and you, you package it and it's, it's, it's, it's ready to go. in today's data economy, right? Um- Artists are increasingly aware, again, mentioned earlier, the value proposition of the fan data. And, and, and again, how that, that plays a role in their own catalog is a, is a factor out there, right?
mike-farrell_6_07-17-2026_123424: Mm-hmm.
serge_1_07-17-2026_123424: [00:29:00] And AI experiences is, is now part of that new discussion. So, um, how do we train the AI, uh, and power new experiences, right?
It's, it's a common problem for everybody in the industry, as we talked about. How do you see the partnership there evolving, right? So we have to get the artist involved, right? This is the creator. Um, you know, does data become a new bargaining chip? Yeah, of course it is. It's, it's like your, your metadata and the AI agent is gonna understand the success of your track or not, or understand how to build a brand new s- uh, know, what, what's the best, uh, um, you know, list at the moment, right?
How do I con- dynamically create that? Um, I, I think artists acknowledge that that's the world we're going into. And it's not just generative AI I'm talking about. I'm talking about understanding and letting AI really make decisions about your, maybe a marketing campaign or your music's values or your taste distribution, right? Um, and I think like there is, there is [00:30:00] a key part of that, like, you know, where, who do you then give credit, where credit's due? We talked about that. I'm sure, sure we're gonna solve for this, right? We, we, we have to because at the end of the game, there is a monetization discovery here. Um, opportunities around AI is n- it-- again, AI is a consumer of music. That's what it is. It's another actor listening. It's not just people anymore. So
mike-farrell_6_07-17-2026_123424: Yep.
serge_1_07-17-2026_123424: Feel like now you have to include the AI discussion because music itself is still the foundation of everything. So if it's now in the metadata, right? And, and there's more value add proposition because the metadata's never been able to get consumed that way before. And how do you-- you know, those new threads of information like the fan relationships between different musics, different genres, different decades of consumption that we already have, that opens up some pretty cool opportunity. Um, and, and I think that becomes an asset, right? So do the labels help artists capture that information and then provide the package where it's valuable and while making sure the creators feel [00:31:00] trust in that and they get the credit they do, the control. Like if some may not want it, so how do you start navigating control? And, and ultimately, you know, that comes all the way down to the, the compensation model. What do you need to do to change that? are the things we naturally have to bridge together, um- And, and again, I, I, I'm not in the attribution space.
I'm not an AI specialist at all. I, I don't do licensing. I'm a privacy and governance guy. I'm happy to stay here, and that, that impacts me tremendously because I get nervous when I hear, you know, well, how are we gonna shift? How are we gonna trust? How are we gonna now change stewards? Um, you know, it's not just simply a distribution deal. It's what do you want me to tell you? If the AI engine's gonna come back with something we've never experienced before, how do we know and how do we prepare for that?
mike-farrell_6_07-17-2026_123424: I'm sure we'll figure it out, but it, it is a very exciting idea to, to think about just, you know, new ways of folks, um, just new, new mediums of connecting with your, your favorite artists and how that world's gonna look in five, 10 years from now
serge_1_07-17-2026_123424: Yeah. [00:32:00] Yeah, right.
luke-arno_3_07-17-2026_123424: Well,
serge_1_07-17-2026_123424: You'll figure it out for me
luke-arno_3_07-17-2026_123424: UMG's artist playlist is one of the best you can put on, whether it's Apple Essentials or Spotify. It, it runs through our house constantly. Um, when you think about the fan experience and the fan side, are, are you at the point where you're exploring letting someone understand other artists while still maintaining and keeping consent and fairness across your different labels?
Are you guys there yet?
serge_1_07-17-2026_123424: Well, I'll say what I can. Uh, listen, Luke, they-- you're, you're absolutely right. I'm a big fan of my playlist as well. So the-- You know what, you know what a playlist is, right? It's an opportunity for you to listen to more music. While now that's not just, you know, we've been doing that for what now? Decades, right?
So we're getting close. Um, the fan side, yeah, um, it's a huge win, right? Uh, you're gonna help... I can help people discover the music they've never h-had before, right? I can train [00:33:00] models whether they love them or not. It's an exposure experience. Um, you have to kind of understand the, so the metadata behind that, I will say is, is pretty rich. Um, so like, like everything that we can help you with, we're managing all this information and throughput finding out what segments work, what don't. Um, and whether the relationships me-matter, right? That's, that's the key, I think, signal for, for not just our industry, but for everyone who's dealing with lists and recommendations.
Um Yeah. Um, look, each data set is different. Yeah. Um, are some competing pr- business priorities, of course. We've gotta throw in some business direction here. Um, but at the end of the day, if we're, if we're really being true to ourselves, we're enriching the fan experience. So if my playlist is cr- is, works, is, is, is cr- it crosses boundaries and lets you explore and... Then I think we've done a good job of it here. And, and, and really, I [00:34:00] mean, it does come down to letting the, the fan have some control over that as well. So we, the, we're talking about if we step back a few, few moments of that, that great experience you just, just listening to, you know how it took a long time to present that list, and if you go really early on in its, uh, in the discovery phase, we allow the fan to let us do that for them, right?
You have to give personalization updates, privacy controls. You have to, to balance the tension between that a little bit, all right? Sure. Um, it's important to allow the fan that responsibility. You know what? That's, that's, shout out to Mike. We need to do that, right? We need to allow the fan to say, "Do you allow us to, to explore this with your consent further? In order to get you this really cool fan experience, I need these things. oh, by the way, I'm gonna ring-fence the heck out of it, so you-- I, I, I wanna get your trust first. Know that I'm not only protecting you, the fan, but I've promised some, you know, some really important responsibilities with [00:35:00] my labels and my artists," right?
And your labels and your artists, right? So, um, I mean, that, that, that, that's the, that's the cool side of the exploration engine. Um, I don't think it, it, it doesn't end with just, hey, a business, you know, model. I need to just go and, you know, find some guy and make a click engine run. It's beyond that. I, I think I, I like to think about it like music science is involved here, although it's not my team.
Uh, I love the fact that there's, a religion to understanding music, right? And
mike-farrell_6_07-17-2026_123424: Sure
serge_1_07-17-2026_123424: respect to people who really take on that challenge. It's, it must be fa- it's fascinating work. Um, so yeah. It's hard to understand your taste, right? That's not just my industry, but for everyone who, who's in the business of, of understanding taste. That is key. And then from there you can learn how to govern it safely monetize off it
luke-arno_3_07-17-2026_123424: What I love most about that answer, Serge, is you're 100% right. It's, it's tough to understand your own personal taste, let alone someone else's, and [00:36:00] discovery is so valuable. But at the center of everything you just mentioned, you've used the word trust several times in ensuring we establish and maintain our customer's trust before we do anything else. And I think that's an, an understated requirement that has to be amplified with how you talk about what you're doing, and it's great
mike-farrell_6_07-17-2026_123424: We love it. And, um, I think there's a, a really unique experience that, that UMG can, um, build at the center of this in, in a highly ethical way. Um, but yeah, Serge, the first half of the show we learned about-- well, we learned from you. Um, the second half we would like to learn a little bit more about you.
Um, could you take me back to the little days, little Serge? Um, did you know you'd end up running data governance for the biggest music company on earth?
serge_1_07-17-2026_123424: no. No, not at all. I, I, I was always interested in the entertainment business in LA. you know, there, [00:37:00] there's, there was a few stepping stones before I ended up in the music business. Uh, so I, I will tell you that I, I, like many other people who were before me, I, I joined the music company thinking, "Oh, it's exactly the same as the entertainment company."
It is not. It is an industry, separate category on the list, uh, distinctive, unique, and, and there are a lot of overlaps, right, like a Venn diagram. But it is deservingly so a unique industry, and it's a lot of fun. Um, I do tell people working in a music company, it, it's like you get... You are introduced to music in a way that, uh, I took for granted, right? I'm just very excited to be working with artists I admired for the rest of my... You know, for, for a long time. And you, you just get to see, you know, transformative information come through, and g- it's been great
mike-farrell_6_07-17-2026_123424: It must feel really rewarding working, um, with, you know, the, the music you listen to every day. [00:38:00] Um, I think that's something we appreciate here at Transcend as well, where we serve companies that we use every day. And when we get to work with your, your amazing artists, it's, it's such a, such a great experience.
Um, what is your, your first memory of the power of first-party data? Um, and what did it enable, um, when you, when you experienced that?
serge_1_07-17-2026_123424: well, Disney Online was-- I got it, I got into Disney y- years ago, and, um, I was involved in the search engine technical space for, for disney.com. And I think that I, I came from a small... You know, we built little SEO company, web company, building little sites for medium to, you know, enterprises. When you're, when you're experiencing the volume that explosive, that things like trend analysis and search results and, you know, um, usability testing, all of that at scale, just, just, just it's impossible to, to pick up for the first time. [00:39:00] that... Now, now, now, you know, I know I, I could, I could build a website, right? I got that part of it in my career. But when I went to Disney Online, I couldn't believe how one small change, minute little thing that you would have imagined is gonna be, you know, needs to go through user testing, approval processes, and all the different regression and, uh, U-UAT and, uh, oh, heat map testing, uh, all of that usability scale, uh, usability factors at scale come into play. So one small change can make a dramatic difference for better or for worse. Uh, and at that time, you know, there was little analysis that I've ever been exposed to. I, I didn't know. You know, I went, "Oh, but thousand hits. Oh, that's a huge number." When you go to Disney at scale, millions is a joke
mike-farrell_6_07-17-2026_123424: No, that's, that's great. I, I can only imagine what kind of, uh, insights and trends you would get from that. Um, you've done some data at Disney, Fox, a Korean startup, a tool retailer, and now music. What does music teach you about data that [00:40:00] nowhere else did?
serge_1_07-17-2026_123424: the consumption of music, uh, we, we touched a little bit about on this earlier. The, the music consumption business and listening habits, as data is unique. I, I really thought I got a handle on that with video consumption, but I think listening behavior is its own religion, right? I mentioned earlier. Uh, the m- the, the day-to-day microcosm mood and, and, and, uh, you know, the trends, the genre, and evolving, you know, music habits around the world i- is insane. Uh, so I got-- Uh, first eye-opening experience for me was like, you know, uh, how do you-- Why, why do you even listen to a track? How do you listen to the track?
How did you even get to the track? All those are the questions that I never-- I took for granted, right? Now I lo- I open up my Spotify account and I'm like, "Wait, wait, should I be calculating what's going on?" Um, a lot of that is back to taste, right, Luke? So, you know, understanding that taste is a, is a very important, um, uh, and very [00:41:00] difficult, uh, quantifiable product, if that's the word I'll use. Uh, and it can't easily be put into a box, right? Like maybe a purchase habit can be or a purchase trend could be. So we... Yeah, that's pretty, that was pretty cool stuff and, and again, very hard to, to fathom
mike-farrell_6_07-17-2026_123424: Yeah. It feels like it's its own, its own thing that just takes over and, um, you know, it can have an effect on your mood. It can change, change your mind. Your-- It, it's very, it's a very, um, interesting to think, to think philosophically about. Um, yeah. So introspection question here. Um, what advice would you give yourse- your younger self a few days after graduation, um, with everything you've learned, um, through now?
serge_1_07-17-2026_123424: would embrace the, the cliché is coming. Here we go. I would embrace the risk and fail, uh, mantra. Um, but fail early, right? Okay. So, um, I think the goal, the goal for me would've been, hey, if [00:42:00] I, if I walk back in your shoes, uh, you know, don't, don't, um... It's, it's the, you know, the goal isn't to avoid failure, uh, but it's to learn quicker, right?
And then pick yourself up and just don't do it in 30 years from now, Serge. Just do it early and, and, you know, don't, don't wait till last minute. and, you know, sometimes it'll work out. I'd also tell... I'd take some bets maybe. I'd also tell myself, "You know, it's okay. Go mine some Bitcoin while you're at it."
mike-farrell_6_07-17-2026_123424: Oh yeah, if we all could do that. Um, well, sweet. Well, Serge, we, we really, really appreciate you taking the time today. This is incredibly insightful. Um, where, where can people find and connect with you?
serge_1_07-17-2026_123424: Um, Mike, I'm always available for a chat. Uh, you could go to my LinkedIn profile if you really have questions. Um, and, and I, I'm responsive on LinkedIn, so I think that's a really good way to get ahold of me. Um, and you know, it's a small world here in LA, so, I bump into people all the time. So, you know, shout out if you have [00:43:00] any questions
mike-farrell_6_07-17-2026_123424: Amazing. Well, thank you, and, um, thank you to our audience as well. If you learned something today or laughed, um, tell someone about this podcast. Uh, this has been another exciting episode of the "New Data Economy." We'll see you all next time