The Exchanges

Q I get it, but I get it. Maybe it didn't sound like the hottest space, right? We get it. But a mid-sized market in all fairness, right? Even it's mid-sized, right? Classically.

A I think more importantly than that, you could get to Double digit AR just serving that market. And the thing that we, I think, saw that there's so many great competitors that have died around us over the last eight years was that we focused on a use case where the technology was actually good enough, not where we could dream it to be one day, but doing all sorts of cool marketing and sales content. It worked well for training because the alternative to training was a boring PDF document. And so what we've seen over the last couple of years, right, is that the better the avatar quality becomes, the bigger the tan gets. So the last 12 to 18 months, we've seen a massive shift towards product marketing with customer support, which collectively those two are now actually the biggest segments that we serve. And that's purely because the quality of the avatars have gotten better and better. We're not making like, we're not making Super Bowl ads and key brand moments, but we make a lot of tactical marketing content, right? You see the cool ad on Meta, YouTube, whatever you click to the website. And then once you're on the website, Then you learn more about the product with some easier videos, and we could just move further and further up that funnel, the better the quality of that.

AI assessment note: “you could get to Double digit AR just serving that market”

Q I get it, but I get it. Maybe it didn't sound like the hottest space, right? We get it. But a mid-sized market in all fairness, right? Even it's mid-sized, right? Classically.

A I think more importantly than that, you could get to Double digit AR just serving that market. And the thing that we, I think, saw that there's so many great competitors that have died around us over the last eight years was that we focused on a use case where the technology was actually good enough, not where we could dream it to be one day, but doing all sorts of cool marketing and sales content. It worked well for training because the alternative to training was a boring PDF document. And so what we've seen over the last couple of years, right, is that the better the avatar quality becomes, the bigger the tan gets. So the last 12 to 18 months, we've seen a massive shift towards product marketing with customer support, which collectively those two are now actually the biggest segments that we serve. And that's purely because the quality of the avatars have gotten better and better. We're not making like, we're not making Super Bowl ads and key brand moments, but we make a lot of tactical marketing content, right? You see the cool ad on Meta, YouTube, whatever you click to the website. And then once you're on the website, Then you learn more about the product with some easier videos, and we could just move further and further up that funnel, the better the quality of that.

AI assessment note: “you could get to Double digit AR just serving that market.”

Q second as a case study. Let's use it with Tomas. But what is some, with the latest models, latest everything, we can talk about DeepSeq for a moment in the next. What's something you're excited that you can do this year in an upcoming release that maybe you couldn't even do last year? What is something literally that's at the bleeding edge That you're doing for real in the enterprise.

A So we're launching our new AI video models within the next three to six months, and they're a very massive upgrade, both in terms of quality and fidelity of the avatar. So those of you who've tried the product will probably have experienced the avatars are great. They're a lot better in the last 12 to 18 months than they were previous to that, but there's still some way to go before it's completely photoreal, right? We've got to pass through that barrier, and we're also going to need you to basically prompt your avatar into a new environment. For example, if you create an avatar for yourself today, you're stuck wherever you recorded your avatar. In your office, for example, what you can do with this new tech is you can say, hey, put me in the jungle, put me in the supermarket, put me like I'm playing tennis or something like that. That opens up an entirely new creative dimension in the product. And so those two things together have always been one of the core mechanisms of Sotheby's, right? The better the quality of the avatars and voices, the more time we unlock. So a lot of people laughed at us four years ago for going out to the training market. And because it's not a very sexy market, VCs hate learning and development.

AI assessment note: “we're launching our new AI video models within the next three to six months”

Q everyone wants Snowflake as much as they can in their Snowflake spend. They want Snowflake here, right? So I don't know if this is as radical as the internet wants it to be. How do your customers want to pay, Victor? Do they want to, do they want to pay in brand new ways or do they want to pay in the same way their grandfather paid for business offer?

A We still sell seats predominantly. I think it depends just a lot of the product, right? I think the way to think about Sulthesia right now is PowerPoint two point zero, right? And like PowerPoint is one of those things where we have thousands of people using this inside big enterprises. And I think that's, that's easy to understand why we're selling seeds. I think we also want to sell outcomes. We are building kind of more specific workflows that solve, you know, we do deflecting customer support calls to the call center or increasing your sales prospecting. Which we build separately. But I think it just, I think it really just depends a lot on the product. If it's, I think there's probably a lot of companies who over indexed on seeds where it doesn't really make sense. In a creator tool, I think seeds makes sense, right? But a whole bunch of other types of tools where seeds maybe isn't like the right, isn't right now.

AI assessment note: “We still sell seats predominantly. I think it depends just a lot of the product”

Q second as a case study. Let's use it with Tomas. But what is some, with the latest models, latest everything, we can talk about DeepSeq for a moment in the next. What's something you're excited that you can do this year in an upcoming release that maybe you couldn't even do last year? What is something literally that's at the bleeding edge That you're doing for real in the enterprise.

A So we're launching our new AI video models within the next three to six months, and they're a very massive upgrade, both in terms of quality and fidelity of the avatar. So those of you who've tried the product will probably have experienced the avatars are great. They're a lot better in the last 12 to 18 months than they were previous to that, but there's still some way to go before it's completely photoreal, right? We've got to pass through that barrier, and we're also going to need you to basically prompt your avatar into a new environment. For example, if you create an avatar for yourself today, you're stuck wherever you recorded your avatar. In your office, for example, what you can do with this new tech is you can say, hey, put me in the jungle, put me in the supermarket, put me like I'm playing tennis or something like that. That opens up an entirely new creative dimension in the product. And so those two things together have always been one of the core mechanisms of Sotheby's, right? The better the quality of the avatars and voices, the more time we unlock. So a lot of people laughed at us four years ago for going out to the training market. And because it's not a very sexy market, VCs hate learning and development.

AI assessment note: “we're launching our new AI video models within the next three to six months”

Q Do you think that's true from what Victor think we can do more with fewer engineers with AI and the new world? Or is that a bit of a bit of a myth?

A No, I think we can, but I think, I think there'll be many different shapes of companies. I don't think there'll be one type of company where you really want to go for it. I still think if you want to build a hundred billion dollar company, You've got to need a lot of salespeople. You've got to need a lot of engineers. But I think there's no doubt that productivity is going through the roof. It feels like we are seeing a lot of these types of companies, as Thomas was saying, where you can, and also I feel like the mindset has changed a little bit, where there's a lot of reluctance to like hiring more people and like just being more efficient, which I think is super awesome. But yeah, I think it's never black and white, right? But I think that's a new type of company, which I think is super cool. But if you want to tackle the enterprise like we are, That's still very difficult to do without adding significant account.

AI assessment note: “No, I think we can, but I think, I think there'll be many different shapes”

Q everyone wants Snowflake as much as they can in their Snowflake spend. They want Snowflake here, right? So I don't know if this is as radical as the internet wants it to be. How do your customers want to pay, Victor? Do they want to, do they want to pay in brand new ways or do they want to pay in the same way their grandfather paid for business offer?

A We still sell seats predominantly. I think it depends just a lot of the product, right? I think the way to think about Sulthesia right now is PowerPoint two point zero, right? And like PowerPoint is one of those things where we have thousands of people using this inside big enterprises. And I think that's, that's easy to understand why we're selling seeds. I think we also want to sell outcomes. We are building kind of more specific workflows that solve, you know, we do deflecting customer support calls to the call center or increasing your sales prospecting. Which we build separately. But I think it just, I think it really just depends a lot on the product. If it's, I think there's probably a lot of companies who over indexed on seeds where it doesn't really make sense. In a creator tool, I think seeds makes sense, right? But a whole bunch of other types of tools where seeds maybe isn't like the right, isn't right now.

AI assessment note: “We still sell seats predominantly. I think it depends just a lot of the product”

Q Do you think that's true from what Victor think we can do more with fewer engineers with AI and the new world? Or is that a bit of a bit of a myth?

A No, I think we can, but I think, I think there'll be many different shapes of companies. I don't think there'll be one type of company where you really want to go for it. I still think if you want to build a hundred billion dollar company, You've got to need a lot of salespeople. You've got to need a lot of engineers. But I think there's no doubt that productivity is going through the roof. It feels like we are seeing a lot of these types of companies, as Thomas was saying, where you can, and also I feel like the mindset has changed a little bit, where there's a lot of reluctance to like hiring more people and like just being more efficient, which I think is super awesome. But yeah, I think it's never black and white, right? But I think that's a new type of company, which I think is super cool. But if you want to tackle the enterprise like we are, That's still very difficult to do without adding significant account.

AI assessment note: “No, I think we can, but I think, I think there'll be many different shapes”

Q And there are certain ways. And if I'm, I think of everyone that's like Stripe-esque, price is Stripe-esque that I've seen, right? And we may, if you're, if, and I think we may be playing, spending too many mental calories and playing too many games on pricing when we should focus more on closing customers in a low friction manner, right? I think this is over overdone this issue, right?

A I think the deeper insight here, which I have had confirmed so many times over the last years is just to not overindex on social media, I think, right? Or what VCs want or like what I want as a founder. I was like, just talk to your customers, listen to what they want. Help them deliver awesome value, and the rest will take care of itself, right? Yeah, it's a bit like, you know, I think with all this like model stuff, it's like most customers don't really care about models, right? They buy products. And I think in the tech world, we love talking about technologies and whatever, it's like agents, latest like hype cycle. But when the rubber hits the road, eventually, you can't just live on like bus and grand ideas. Yeah, that's always, that's a big lesson for me the last eight years, for sure.

AI assessment note: “just talk to your customers, listen to what they want”

Q I want to share a story, but Victor, will this, is your COGS gonna, is your COGS a huge issue and is it gonna plummet? Or just from the perspective of Synthasia, is this not as big a deal as it sounds from COGS and the rest? What's your learnings in 48 hours?

A We have great software managers already, so I wouldn't say it's not something we, like, put a lot of those in emphasis towards, but obviously as we scale up, it's great. I would say, though, I think I totally agree with Thomas. It's a massive technical breakthrough. I think it's interesting when you look at the market, I think if anything, to me, this is just their confirmation that humans and markets are driven by stories and emotions, not fundamentals, right? It's absolutely hilarious to see all the world's tech analysts who know nothing about AI just panicking because there's like a new model out. I think people have a tendency to overestimate the impact of this in the short term and maybe underestimate it in the long term. I think the long term tail effect, as Thomas just said for the application there, is huge. I think that's really massive. But I think in two months, I think, 99.9% of the people who are using Anthropic and OpenAI today is going to go back to using their products, not DeepSeek's model. And there's a big differentiation between those two, right? I tweeted the other day that smart people, Talk about models and benchmarks, and customers talk about app flows, apps and workflows, right? And there's such a big difference of a model that performs better, and if they technically, right, they'll probably have a big, again, long-term effect in the market, but, but, …

AI assessment note: “We have great software managers already, so I wouldn't say it's not something”

Q I want to share a story, but Victor, will this, is your COGS gonna, is your COGS a huge issue and is it gonna plummet? Or just from the perspective of Synthasia, is this not as big a deal as it sounds from COGS and the rest? What's your learnings in 48 hours?

A We have great software managers already, so I wouldn't say it's not something we, like, put a lot of those in emphasis towards, but obviously as we scale up, it's great. I would say, though, I think I totally agree with Thomas. It's a massive technical breakthrough. I think it's interesting when you look at the market, I think if anything, to me, this is just their confirmation that humans and markets are driven by stories and emotions, not fundamentals, right? It's absolutely hilarious to see all the world's tech analysts who know nothing about AI just panicking because there's like a new model out. I think people have a tendency to overestimate the impact of this in the short term and maybe underestimate it in the long term. I think the long term tail effect, as Thomas just said for the application there, is huge. I think that's really massive. But I think in two months, I think, 99.9% of the people who are using Anthropic and OpenAI today is going to go back to using their products, not DeepSeek's model. And there's a big differentiation between those two, right? I tweeted the other day that smart people, Talk about models and benchmarks, and customers talk about app flows, apps and workflows, right? And there's such a big difference of a model that performs better, and if they technically, right, they'll probably have a big, again, long-term effect in the market, but, but, …

AI assessment note: “to me, this is just their confirmation that humans and markets are driven by stories”

Q And there are certain ways. And if I'm, I think of everyone that's like Stripe-esque, price is Stripe-esque that I've seen, right? And we may, if you're, if, and I think we may be playing, spending too many mental calories and playing too many games on pricing when we should focus more on closing customers in a low friction manner, right? I think this is over overdone this issue, right?

A I think the deeper insight here, which I have had confirmed so many times over the last years is just to not overindex on social media, I think, right? Or what VCs want or like what I want as a founder. I was like, just talk to your customers, listen to what they want. Help them deliver awesome value, and the rest will take care of itself, right? Yeah, it's a bit like, you know, I think with all this like model stuff, it's like most customers don't really care about models, right? They buy products. And I think in the tech world, we love talking about technologies and whatever, it's like agents, latest like hype cycle. But when the rubber hits the road, eventually, you can't just live on like bus and grand ideas. Yeah, that's always, that's a big lesson for me the last eight years, for sure.

AI assessment note: “I think the deeper insight here... is just to not overindex on social media”