Snowflake Summit 2026 | Dennis Buchheim, Snowflake & Sherene Hilal, SiriusXM

Clips
More from Snowflake Summit 2026

Dennis Buchheim

Global Head of Marketing Technology, Media & Entertainment

Snowflake

Sherene Hilal

Chief Ad Product & Technology Officer

SiriusXM

play_circle_outline SiriusXM's Cortex Sense: Leveraging Audio Context and Snowflake for Personalized Listening and Precision Ad Targeting

play_circle_outline Exclusive YouTube/SiriusXM partnership for audio distribution and ad infrastructure

play_circle_outline Balancing Personalization and Discovery: Real-Time Audience Feedback to Surface New Content

play_circle_outline Safe Voice and Multimodal In-Car Interfaces: Integration and Standards with OEMs, Apple, and Google

play_circle_outline AI adoption: speed, standardization of content, and developer-driven innovation

Info
Transcript

Dennis Buchheim, Snowflake & Sherene Hilal, SiriusXM

Dennis Buchheim

Global Head of Marketing Technology, Media & Entertainment Snowflake

Sherene Hilal

Chief Ad Product & Technology Officer SiriusXM

In this interview from Snowflake Summit 2026, Sherene Hilal, chief ad product and technology officer at SiriusXM, joins Dennis Buchheim, global head of marketing technology, media and entertainment at Snowflake, to talk with theCUBE's Dave Vellante and Rebecca Knight about how data-driven personalization and open ecosystem partnerships are redefining audio media at scale. Hilal explains how audio — present in the foreground and background of roughly four hours of daily listening — has long lacked the structured data needed for real-time targeting and discover... Read more

explore Keep Exploring

How do you structure and leverage data (taxonomies, ontologies, semantic layers, and customer 360 approaches), and work with ecosystem partners, to enable media personalization, ad targeting, and measurement? add

How does the partnership between SiriusXM, YouTube/Google, and Snowflake support your audio-advertising strategy — including inventory, measurement, buying/selling, and the co‑development of technology? add

How can streaming services balance personalization with discovery so users are introduced to new hosts, songs, or shows without relying on manual search? add

How is SiriusXM leveraging "exhaust" data (including voice inputs) and collaborating with automakers and platform providers (e.g., Apple, Google) to create safer, more accessible, and standardized in‑car listening experiences? add

What is the primary application or benefit of AI for your organization, and is its adoption being driven top‑down, bottom‑up, or middle‑out? add

bolt Powered by CUBE AI

Dennis Buchheim, Snowflake & Sherene Hilal, SiriusXM

search

Dave Vellante

>> No.

Rebecca Knight

>> Hello everyone and welcome to theCUBE's Live coverage of the Snowflake Summit 2026 here at the Moscone Center in San Francisco. I'm your host, Rebecca Knight, alongside Dave Vellante. We are kicking off two days of wall-to-wall coverage. It's going to be terrific.

Dave Vellante

>> Yes. I'm really excited. This is one of our favorite shows, I think our seventh year doing Snowflake Summit.

Rebecca Knight

>> Indeed. Well, I would like to welcome our first guest of the day, first guest of the show, Sherene Halal, Chief Ad Product and Technology Officer at SiriusXM. Welcome.

Sherene Hilal

>> Thank you. Thank you.

Rebecca Knight

>> And Dennis Buchheim, Global Head of Marketing, Technology, Media and Entertainment at Snowflake. Thanks so much for coming on the show, Dennis.

Dennis Buchheim

>> My pleasure. Thank you.

Rebecca Knight

>> Sherene, I'm going to start with you because there is so much change taking place across media and entertainment and advertising. Why don't you start giving us a big picture in talking about how data and technology is evolving in terms of how you reach your listeners?

Sherene Hilal

>> Advertising for a while now has been a lot more focused on audience versus channel. And because of that, audio actually can take advantage of a lot of the data and technology trends that exist because audio has so much signal around it. When you think about it, it's in the foreground and background of four hours of people's time every day, but most of that time isn't really structured in a way within data systems that you can target and buy and make it easier to really personalize. And so as we've been starting to figure out how to make audio more modern, data and technology has been the recipe for that success, specifically how to get the context of audio, so the mood and moment of what you're listening to, the transcript from a podcast, and really making that data available so that it's a very tailored experience for the listener.

Dave Vellante

>> Dennis, amazing thing I find about Sirius is if you draw a power law, you think, okay, you got companies that are in the big part of the power law, the steep part of the power law, and there's a long tail. And Sirius is both, which is pretty amazing, the content. But do you see that in other media industries? Am I right that you typically see dominant media, mainstream media at the steep part of the power law, and then you got guys like us that are long tail? Is it unique to see companies playing in both? And how do you see that as a sort of a business model out there?

Dennis Buchheim

>> I don't think it's unique. I think it's actually pretty common across media, even on a global basis. I actually was just having a conversation yesterday about telecom, which is also in my broader patch. And I think it's even more pronounced in telecom where you have very dominant companies and then the long tail of different companies supporting the larger enterprises. And the biggest question then I think for us becomes, how does data collaboration work there? How does the interaction and integration work there? How do all of these companies or a lot of them work together effectively when it is that long a tail in some of the industries?

Dave Vellante

>> But your content touches both, and that's what I find unique.

Sherene Hilal

>> It does. I mean part of, and I love that you used power log, and that's probably going to be my favorite.

Dennis Buchheim

>> Sure. I'm off to a good start.

Sherene Hilal

>> We're off to a good start.

Dave Vellante

>> You have stern listeners and then you got deep tracks.

Sherene Hilal

>> But this actually ties to the data and tech piece. That power log is about reach. And so how do you make sure that that long tail of reach allows you to create niche content for audiences that might not be mainstream? A lot of what we talk about is taste maturity and taste making. How do you take the preferences that someone has and translate it into media in the exact environment that they care about? And no one is one dimensional, and so part of that power log is understanding that everyone has attributes of listening that overlap with other people. And I think the big heavy hitters in the mainstream media piece can often underestimate the value of that long tail data. And so part of what we partner with together is how to make all signal available because you don't want to eliminate even one person from how your media is able to reach them.

Rebecca Knight

>> So how do you figure out this personalization and real time audience intelligence? Because as you said, we contain multitudes. We like our comedy greats. I like talk radio. I like the advice columnists. So how do you slice and dice and figure out exactly how to get the kind of content that that listener wants?

Sherene Hilal

>> I think it really starts out with a data taxonomy. I know that doesn't sound like maybe the coolest, sexiest thing, but it is structuring your data in a way that doesn't eliminate key signal. So for instance, we don't just look at the genre of a song. We're looking at the lyrics of that song. We're looking at what proceeds and follows that song. We're looking at what people name things. We have the transcripts of what you listen to. And all of that is contextual signal that allows us to say, "Hey, in the morning what someone might prefer is this." Maybe it's news talk and sports. But then as they transition to the car, maybe it's like a power ballad or something to get you ready for your workday. And then over time, maybe you have more people in your car later in the afternoon and you can't have explicit lyrics or you need something a little bit safer. And so it's a combination of adding structure to media and having deep expertise in media. It's finding partners with similarly deep expertise. We try to look for ecosystem partners like a Snowflake or a YouTube that are committed to the role of making media more personalized, and then working together to co-develop a solution.

Dennis Buchheim

>> Let me add to that. I mean, yes, I think taxonomy is sexy again, right? If it ever was, right?

Sherene Hilal

>> Power log and taxonomy.

Dennis Buchheim

>> Again, off to a wonderful start. Actually, I just came back from a trip to Korea and Japan for work and had a bunch of meetings there where the biggest topic that came up time and time again was ontology, so getting even more specific. And we talk about the semantic layer. We actually just announced Cortex Sense. So there's an intelligence in just defining the semantics of what do all of these words ultimately mean for your business, and then building on top of that. So first it's understanding the whole kind of corpus of everything that you do with your customers, everything you do across your business. And then it's really understanding your customers, I think, on top of that. So we talk a lot about our Customer 360 and really fully understanding everything you can about your customers from the data you have, and from third party data that can be connected to the data you have, and then building on top of that to really essentially fully get to personalization, to get to ad targeting and personalization of content to get to measurement. But it's all of these. You keep stacking these blocks basically to build that out.

Dave Vellante

>> Well, it's interesting. We throw around these terms like context and taxonomy and ontology. Sanjeev Mohan just wrote a piece, he's on LinkedIn, Sanjeev Mohan, and he explained all this stuff and how it relates. And he used a laptop as an example, Mac, the model number, et cetera, et cetera, but then got into deeper concepts like ontology and how you can then use that to personalize. So maybe three or four years ago we would think of SiriusXM, Snowflake's vision was the app store for data, for data apps. It's dramatically evolved now into sort of AI and trust and governance. All these words were thrown around like ontology and being able to personalize. So over the last couple of years, how has that affected your ability to meet your mission?

Sherene Hilal

>> I think the first one is being more open about what other data inputs into our own. So in a world where we just operate in a closed environment and we only observe the media and the listening behavior that happens directly in our cars, directly in our app, you are eliminating a huge percentage of time. And it's not just time spent listening with audio; it's just time. Time on devices, time in the world. And so our belief is that audio is ubiquitous. Because it's foreground and background, you need to have the ability to partner with the rest of the ecosystem that grabs that attention. And so probably the number one thing you've seen from us over the past 12 to 18 months is that commitment. So named marquee partnerships, first of their kind relationships with partners like a YouTube, like a Snowflake to say, "We are committed to making the audio data accessible. We're committed to working with partners like the Interactive Advertising Bureau, so the IAB, to build the standards together and let others be the judge of the quality of our decision." So I think that's the big one.

Dave Vellante

>> Okay. So I get the Snowflake connection; that helps you make better decisions and better targeting. The YouTube piece is a distribution channel for your...

Sherene Hilal

>> It's a combination. So when you think about YouTube, oftentimes people don't think about audio inventory. They think about watching a video. But there is a huge surge of audio listening happening on every channel where you have what we call multimodal media, so the ability to lean into content or lean back. Maybe you want to see what someone is wearing and so you lean in for a second, but then you're back into just listening in your headphones. And so YouTube and Sirius XM, we came together and said, "How do we make audio easier to measure, easier to buy, and easier to sell?" And we said, "Well, who has the largest supply outside of Sirius XM?" And it was Google. And so we decided to partner with them for an exclusive, what we call an ad rep deal, so the ability to not just sell the inventory, but co-develop the infrastructure. So this is really us going far beyond owning the audio media into building the technology with big partners to say, "This is really going to be an open ecosystem."

Rebecca Knight

>> So I'm picturing a customer who, as you said, has this ambient noise of audio, whether they're either tuned into their news program or listening to a power ballad, or doing carpool but also having some pop song going on in the background, and then this big word cloud of all the words that they're ingesting.

Sherene Hilal

>> That's right.

Rebecca Knight

>> Talk a little bit about what you're able to unlock by getting that kind of data about the customer and the insights you're able to discern.

Sherene Hilal

>> It's discovery, right? I think personalization is so hot all the time, but people often on the technology side see personalization and discovery at odds with each other. Because to personalize, you have to fundamentally limit your option set of what you're showing to someone. And so really what this allows us to do is to introduce a new host, a new song, a new show and know you will like it. And often when people think about SiriusXM in the car, they forget we have music streaming services like Pandora. Those are systems with real time feedback loops, and so we know that you liked a song because you gave it a thumbs up. And so the ability to make things likable and relatable very quickly without a lot of load on the user, like search. Search is very outdated. The idea that you have to type something in and you're clicking buttons on a screen, it's like, "Wow, we really haven't matured much." Discovery through data is the replacement to the very manual processes a lot of us take to find things that we're going to love.

Dennis Buchheim

>> Yeah, and you beat me to saying multimodal, by the way, in our buzzword bingo here.

Sherene Hilal

>> We need a list up here, right? Power log, multimodal.

Dennis Buchheim

>> Tick them all off. But you mentioned sort of the word cloud. But the other thing to consider because it's multimodal is it's not just a word cloud. It's a word cloud and a video cloud and an audio cloud and everything you can imagine, an image cloud. So figuring out how to deduce the right context, the right insight from each of those pieces of content, and then map it to which users, which customers are going to care about it ultimately, and how do you build on it and even learn from it to better define your content programming strategy is pretty fascinating.

Dave Vellante

>> So let me pick up on that, because Sherene, what you were saying about the web browser typing into a Google search, it doesn't make any sense. When I think about Snowflake Intelligence, I think about, "Okay, that's going to be a new way in which your customers interact." I know I'm interacting, of course, with agents and LLMs. So how do you envision your system, if I can call it that, learning from the sort of traces, the exhaust of your users, and then having that feedback loop?

Sherene Hilal

>> I think the word exhaust is so perfect because it is the remnants of the data that actually complete the picture. And so for us, it's about not ignoring any way a listener might give us an input. Voice is a huge one where if you think about our product in the car, safety and accessibility is huge. You cannot be touching a screen. Some of the times the options aren't even available in the car. And so it actually forces us to be incredibly inventive about what are ways that you can interact with a service not using buttons, not using a screen, and then what are the ways that the prompts that we are now trained on with these LLMs actually teach people to interface with technology better. As search starts to advance and become more prompt based, as the models start to understand the context of your environment, those are things that we are already starting to invest in. So how do you make sure that a host when someone dials in actually knows information about the listener that makes the conversation more prepared and a bit more interesting to listen to. So those are really... It is about bringing that taxonomy back into even the exhaust, even the remnants of data to complete a full picture.

Dave Vellante

>> I mean, other than sort of Apple CarPlay trying to consume every moment of our lives, are there standards emerging across, say, automobile manufacturers so that we can in a more fascia way interact with tools like SiriusXM?

Sherene Hilal

>> Totally. Every auto manufacturer has a different set of priorities based on the market they're going after and the state of their tech. But Sirius XM has long been in the hardware systems of these vehicles, so we work with the automakers very closely and a lot of them are starting to really lean into the role that media has in giving information about the listener to make a more enjoyable ride. Think about when you buy a connectivity package from your car company and now you can listen to music and have a picnic outdoors. You want to tailgate before a sports game. You have to charge your car before you pick up your kid and you want to listen to a fun podcast. These are all small moments, but those moments are times where entertainment is actually incredibly valuable but needs to be very convenient. And so the automakers come to us and we collaborate with them on co-developing what that future of listening looks like. And their feedback is incredibly important to us, but we also work on the standards with Apple and Google and others to try to triangulate, to put the consumer back where they belong, which is kind of in control of their experience.

Dave Vellante

>> Well, you've perfected that freemium model. I mean, you get six months with your new car and then you're hooked. It's beautiful. I got to stick with you.

Sherene Hilal

>> We believe in our content.

Dave Vellante

>> Of course I'm going to re-up. You were going to say something, Dennis?

Dennis Buchheim

>> Yeah. No, I think just taking it back to Snowflake Intelligence or CoWork as well and just the notion that consumers are evolving and learning so quickly right now and how to interact with technology in ways that, sure, you typed in a search box. That's very different than having a conversation, and having that conversation with an agent that can talk with other agents and go get things done for you and transforming it from more like research into action or insight into action. Just the opportunity that we talk about just democratizing intelligence and getting the tools in the hands of thousands of business users, with hundreds of thousands and millions ultimately. But within any decent sized company, you got every business user can have access to these tools, but needs to learn how to interact, how to interrogate in that conversation to get to something that they're going to find most useful. And that's a completely novel experience.

Sherene Hilal

>> It is.

Dennis Buchheim

>> Because it used to be you'd type an email to your friendly data scientist or data analyst, and a week later you'd get an insight. Now it's a minute later or less. And you can keep learning and evolving and, "Huh, I wonder about this. I wonder about that." And to your beginning of your question, the tech is learning as you do that. And that's incredibly powerful, almost scary sometimes.

Sherene Hilal

>> I mean, trust is huge.

Dennis Buchheim

>> Trust is huge.

Sherene Hilal

>> The trust component comes in.

Dennis Buchheim

>> Governance is critical.

Dave Vellante

>> I guess there's a whole vector we could go down with trust. Throw a few more buzzwords in there. But I want to ask you about the state of AI at your respective companies. There was sort of this top down mandate that was thou shalt become AI native, and then it became, well, how do I do that? And now you're seeing this bottoms up activity with things like NemoClaw and OpenClaw and Claude CoWork and Perplexity Computer, all individuals are now. It's like the original PC, you guys are all too young to remember, but it was a personal productivity boom.

Rebecca Knight

>> Well, because you were using it in your personal life, and then it really made sense to use it at work.

Dave Vellante

>> Of course, and it was kind of self-funding. How would you describe the state of AI inside of SiriusXM?

Sherene Hilal

>> For us, it's speed. It's about speed. Oftentimes when we talk about the content that we have, so podcasts and news and music, we think about it as one shape and one size. But how we take a Satellite to distribute it in the car versus stream over a network for music versus make it available on platforms like YouTube, each supply type is a snowflake. That was not intentional, but it is unique. And so part of what AI allows us to do is to start to standardize our supply types in a much more regular way. So I can take our Satellite news, talk, and sports and the channels like comedy and I can make them available digital and I can learn in the environments much faster to make it easier to access and listen to. So speed has been the number one application for us. And when you have our volume and variety of content and talent and distribution, speed is everything. It's what creates the advantage though.

Dave Vellante

>> And would you say that it's, it's kind of a trite question, but top down? Bottom up? Or is it kind of middle out because you got your developers now driving that speed?

Sherene Hilal

>> It's coming from all directions. I would say the top down concept is really about permission, like permission to work with these tools, to try new things, to build KPIs and measures that maybe are much more ambitious than what you would report out publicly, but you want to hold your teams accountable. The stuff that actually is translating to that speed is what's coming in from the middle, which is our tech teams being like, "We never want to build this ever again, so let's build it through AI so it is automatically able to adjust to new inputs." So that's where I've seen the most velocity to the company.

Dave Vellante

>> And obviously at Snowflake it's very much product driven, but you're also customer zero and there's an internal activities going on. How would you describe it inside Snowflake?

Dennis Buchheim

>> Yeah. Well, this is my first anniversary today at Snowflake.

Rebecca Knight

>> Congrats.

Dennis Buchheim

>> And a year ago we didn't have Snowflake ?Intelligence or CoWork. We didn't have CoCo. We didn't have these tools. We were using Gemini a bit. It was starting to just kind of take shape organically. But the push, and I think it came from top down and bottom up in our case, from Sridhar on down saying, "Okay, you have to take advantage of our own tools." And of course we allow the use of all these different wonderful models that are evolving so quickly. So using the tools on a day-to-day basis for everything from insights to development happened ridiculously quickly. And I mean, I've never seen anything like it.

Sherene Hilal

>> I can attest to that. We had a meeting with our Snowflake team, and everyone was using the tools, the prompts, the search. It was part of the demo. It's part of what our teams are learning about. So it really was as daily as what you're saying.

Dennis Buchheim

>> Yeah. It used to take us weeks to develop a demo, and now it happens real time sometimes, or certainly within a day. It's pretty incredible, and that all happened in the last few months. We have our hands on so much more data at the business level than we did a year ago for sure, or even six months ago. I think perhaps the most interesting thing actually came up in a customer meeting yesterday. A customer was asking, "What's on Snowflake's roadmap? Where's this going in five years?" First of all, we all laughed like-

Rebecca Knight

>> Who knows?...

Dennis Buchheim

>> "Years?"

Dave Vellante

>> Or five months even.

Dennis Buchheim

>> Exactly. But kind of the point that I put at the end of it was I thought it was interesting that he even asked about a roadmap because yes, we have a roadmap that's about enabling a foundation and infrastructure for apps and agents and all the governance and everything that we talk about and the control plane. But really any company's roadmap now can be, again, kind of democratized. Anybody can add an app, a skill, a capability, and that becomes part of your tool set, your solution set, internally, with customers, whatever the case may be. I think that's just fundamentally game changing.

Sherene Hilal

>> Yeah. I mean, we've evolved from this idea of roadmaps and how we govern to KPIs. So we have key performance indicators. We share them with our partners and we say, "Here's where we believe we're behind and here's where we have to win always." And we work together to say, "Here's how we're driving the result." And the roadmap bends to the result. It needs to drive that result with whatever is needed. And so the roadmap is becoming, I think, as outdated as search, I guess.

Dennis Buchheim

>> Yeah, very good. Yeah. .

Rebecca Knight

>> Well, Sherene and Dennis, this has been so helpful and so much fun talking to you.

Dave Vellante

>> Thank you guys.

Sherene Hilal

>> Yeah, it was great.

Rebecca Knight

>> We could talk to you all day.

Dennis Buchheim

>> Have a good one.

Rebecca Knight

>> Thank you so much for coming on the show.

Sherene Hilal

>> Yeah, thank you.

Dennis Buchheim

>> Appreciate it. Thank you both.

Rebecca Knight

>> I'm Rebecca Knight for Dave Vellante. Stay tuned for more of theCUBE's live coverage of the Snowflake Summit. You're watching theCUBE, the leader in enterprise tech news and analysis.