AI Agents Are Starting To Buy. Stripe Is Building How They Pay.

Oh, we didn't change anything in the Oh, we didn't change anything in the product and all of a sudden just like product and all of a sudden just like product and all of a sudden just like went vertical. Yeah. And it was cuz the went vertical. Yeah. And it was cuz the went vertical. Yeah. And it was cuz the agents had found it. agents had found it. agents had found it. >> In that world, how do we scale trust? >> In that world, how do we scale trust? >> In that world, how do we scale trust? How do we scale trust when the agents How do we scale trust when the agents How do we scale trust when the agents are supposed to be transacting are supposed to be transacting are supposed to be transacting >> and most of the sort of existential >> and most of the sort of existential >> and most of the sort of existential abuse that AI companies are facing is abuse that AI companies are facing is abuse that AI companies are facing is actually pre-transaction outside of the actually pre-transaction outside of the actually pre-transaction outside of the transaction. And that is, you know, um transaction. And that is, you know, um transaction. And that is, you know, um basically stealing tokens instead of basically stealing tokens instead of basically stealing tokens instead of stealing money. Well, guess who doesn't stealing money. Well, guess who doesn't stealing money. Well, guess who doesn't talk to sellers? talk to sellers? talk to sellers? Who doesn't talk to sellers, Emily? Who doesn't talk to sellers, Emily? Who doesn't talk to sellers, Emily? >> Agents don't talk to sellers. Are you >> Agents don't talk to sellers. Are you >> Agents don't talk to sellers. Are you with me? Yeah. Agents are just with me? Yeah. Agents are just with me? Yeah. Agents are just relentless. Like, they're willing to relentless. Like, they're willing to relentless. Like, they're willing to negotiate as long as they need to negotiate as long as they need to negotiate as long as they need to negotiate. They'll come back again and negotiate. They'll come back again and negotiate. They'll come back again and again. And you could imagine that like again. And you could imagine that like again. And you could imagine that like um you know, consumer surplus, for um you know, consumer surplus, for um you know, consumer surplus, for example, like example, like example, like >> yeah, >> yeah, >> yeah, >> could could be eroded relatively >> could could be eroded relatively >> could could be eroded relatively quickly. quickly. quickly. >> I'm here at Stripe because Stripe is >> I'm here at Stripe because Stripe is >> I'm here at Stripe because Stripe is what is making the future of buying and what is making the future of buying and what is making the future of buying and selling on the internet possible. So selling on the internet possible. So selling on the internet possible. So right now most of the AI that we have is right now most of the AI that we have is right now most of the AI that we have is in our pockets and we think of it as it in our pockets and we think of it as it in our pockets and we think of it as it gets the email done. We think of it as gets the email done. We think of it as gets the email done. We think of it as maybe it helps us put a dock together.

maybe it helps us put a dock together. maybe it helps us put a dock together. For a lot of us we're not thinking of it For a lot of us we're not thinking of it For a lot of us we're not thinking of it as something that buys and sells. And if as something that buys and sells. And if as something that buys and sells. And if we do that comes up in like the sci-fi we do that comes up in like the sci-fi we do that comes up in like the sci-fi movies we don't want to watch. movies we don't want to watch. movies we don't want to watch. What would happen if we actually made What would happen if we actually made What would happen if we actually made that safe? What would happen if we that safe? What would happen if we that safe? What would happen if we actually made that something that we actually made that something that we actually made that something that we could trust? That's the question I could trust? That's the question I could trust? That's the question I wanted to get at chatting with Emily at wanted to get at chatting with Emily at wanted to get at chatting with Emily at Stripe today because for me, if I think Stripe today because for me, if I think Stripe today because for me, if I think about all the stuff that I have to deal about all the stuff that I have to deal about all the stuff that I have to deal with and and I'm a busy dad and I've got with and and I'm a busy dad and I've got with and and I'm a busy dad and I've got a lot to do, so much of the time I am a lot to do, so much of the time I am a lot to do, so much of the time I am making transactions, purchase decisions making transactions, purchase decisions making transactions, purchase decisions that I have to do that I don't want to that I have to do that I don't want to that I have to do that I don't want to spend a lot of time on, but I'd better spend a lot of time on, but I'd better spend a lot of time on, but I'd better approve them because I don't want approve them because I don't want approve them because I don't want someone else doing it. And then, you someone else doing it. And then, you someone else doing it. And then, you know, who knows what will happen. So the know, who knows what will happen. So the know, who knows what will happen. So the whole story of agents is the story of whole story of agents is the story of whole story of agents is the story of getting a computer to do something that getting a computer to do something that getting a computer to do something that you can 100% of the time trust. Now I'm you can 100% of the time trust. Now I'm you can 100% of the time trust. Now I'm not here to tell you that we are at 100% not here to tell you that we are at 100% not here to tell you that we are at 100% yet. I wouldn't want to give you that yet. I wouldn't want to give you that yet. I wouldn't want to give you that false confidence, but that percentage false confidence, but that percentage false confidence, but that percentage has been rising really, really rapidly.

has been rising really, really rapidly. has been rising really, really rapidly. And that is one of the reasons I keep And that is one of the reasons I keep And that is one of the reasons I keep talking about AI on this channel all the talking about AI on this channel all the talking about AI on this channel all the time. And one of the things that that time. And one of the things that that time. And one of the things that that has made possible is this conversation has made possible is this conversation has made possible is this conversation with Stripe because we are with Stripe because we are with Stripe because we are [clears throat] now at a point in our [clears throat] now at a point in our [clears throat] now at a point in our trust story where I can actually give AI trust story where I can actually give AI trust story where I can actually give AI a wallet and I can trust it to go do a wallet and I can trust it to go do a wallet and I can trust it to go do something like buy me coffee. Would I something like buy me coffee. Would I something like buy me coffee. Would I trust AI to buy me a couch yet? No, not trust AI to buy me a couch yet? No, not trust AI to buy me a couch yet? No, not yet. Uh would I trust AI to sign up for yet. Uh would I trust AI to sign up for yet. Uh would I trust AI to sign up for a subscription yet? Probably not yet. a subscription yet? Probably not yet. a subscription yet? Probably not yet. Would I trust AI to do something Would I trust AI to do something Would I trust AI to do something specific for me within a given budget specific for me within a given budget specific for me within a given budget like buy coffee? Yeah, I would. And so I like buy coffee? Yeah, I would. And so I like buy coffee? Yeah, I would. And so I sat down with Emily and I got to chat sat down with Emily and I got to chat sat down with Emily and I got to chat with her about what Stripe is doing in with her about what Stripe is doing in with her about what Stripe is doing in that trust space and why it matters. And that trust space and why it matters. And that trust space and why it matters. And ultimately, if we get this right, we ultimately, if we get this right, we ultimately, if we get this right, we actually get a lot of our time back, we actually get a lot of our time back, we actually get a lot of our time back, we get money back in our pockets. We can get money back in our pockets. We can get money back in our pockets. We can actually spend time on the things we actually spend time on the things we actually spend time on the things we care about and not the things that feel care about and not the things that feel care about and not the things that feel like drudgery to us. So, I'm really like drudgery to us. So, I'm really like drudgery to us. So, I'm really excited for this one. Uh, I hope you excited for this one. Uh, I hope you excited for this one. Uh, I hope you enjoy this conversation. I had fun enjoy this conversation. I had fun enjoy this conversation. I had fun chatting with Emily. chatting with Emily. chatting with Emily. >> Emily, tell me about you. How did you >> Emily, tell me about you. How did you >> Emily, tell me about you. How did you get to Stripe? I've been at get to Stripe? I've been at get to Stripe? I've been at [clears throat] Stripe about five years. [clears throat] Stripe about five years. [clears throat] Stripe about five years. Okay.

Okay. Okay. >> Boy, a lot has uh changed over that >> Boy, a lot has uh changed over that >> Boy, a lot has uh changed over that period. Hasn't been a quiet one. period. Hasn't been a quiet one. period. Hasn't been a quiet one. >> Not been a quiet one. >> Not been a quiet one. >> Not been a quiet one. >> Um I'm an economist by training. So was >> Um I'm an economist by training. So was >> Um I'm an economist by training. So was pretty obsessed with how we could create pretty obsessed with how we could create pretty obsessed with how we could create more equality of opportunity in the more equality of opportunity in the more equality of opportunity in the labor market uh back as I was in grad labor market uh back as I was in grad labor market uh back as I was in grad school. And so uh defected from academia school. And so uh defected from academia school. And so uh defected from academia after the PhD and went to Corsera. I after the PhD and went to Corsera. I after the PhD and went to Corsera. I don't know if you've ever been a don't know if you've ever been a don't know if you've ever been a learner. learner. learner. >> I I I have been a learner. I've used >> I I I have been a learner. I've used >> I I I have been a learner. I've used Corsera. Corsera. Corsera. >> Yeah. I thought uh hey what better way >> Yeah. I thought uh hey what better way >> Yeah. I thought uh hey what better way to to create more opportunity in labor to to create more opportunity in labor to to create more opportunity in labor market than to kind of level the playing market than to kind of level the playing market than to kind of level the playing field in education. I was there about field in education. I was there about field in education. I was there about eight years and then joined Stripe eight years and then joined Stripe eight years and then joined Stripe originally to look after data science. originally to look after data science. originally to look after data science. Uh and as we started sort of working Uh and as we started sort of working Uh and as we started sort of working through like what's possible with stripe through like what's possible with stripe through like what's possible with stripe data um sort of built out the org to data um sort of built out the org to data um sort of built out the org to cover everything from ML infrastructure cover everything from ML infrastructure cover everything from ML infrastructure and agent infrastructure to our and agent infrastructure to our and agent infrastructure to our userfacing data products and a bunch of userfacing data products and a bunch of userfacing data products and a bunch of our experimental bats. So it's been it's our experimental bats. So it's been it's our experimental bats. So it's been it's been a fun run. It's been a fun one and been a fun run. It's been a fun one and been a fun run. It's been a fun one and and of course like part of why we're and of course like part of why we're and of course like part of why we're talking is that the last six months have talking is that the last six months have talking is that the last six months have been completely insane from an agent and been completely insane from an agent and been completely insane from an agent and even from a Stripe perspective and so even from a Stripe perspective and so even from a Stripe perspective and so I'd be really curious maybe from an I'd be really curious maybe from an I'd be really curious maybe from an insiders perspective. Tell me a little insiders perspective. Tell me a little insiders perspective. Tell me a little bit about what that's been like working bit about what that's been like working bit about what that's been like working at a company that you know isn't an AI at a company that you know isn't an AI at a company that you know isn't an AI lab but is deeply involved in the lab but is deeply involved in the lab but is deeply involved in the transformation we're all living through.

transformation we're all living through. transformation we're all living through. >> Yeah, so you know Stripe is really >> Yeah, so you know Stripe is really >> Yeah, so you know Stripe is really building the economic infrastructure for building the economic infrastructure for building the economic infrastructure for AI and that has a lot of different AI and that has a lot of different AI and that has a lot of different flavors to it, right? It's like how do flavors to it, right? It's like how do flavors to it, right? It's like how do AI companies uh monetize and bill? How AI companies uh monetize and bill? How AI companies uh monetize and bill? How do they fight fraud? How do platforms do they fight fraud? How do platforms do they fight fraud? How do platforms like, you know, Verscell and Replet like, you know, Verscell and Replet like, you know, Verscell and Replet serve this sort of whole new host of serve this sort of whole new host of serve this sort of whole new host of builders that are building new builders that are building new builders that are building new businesses with AI? Um all the way to, businesses with AI? Um all the way to, businesses with AI? Um all the way to, you know, how do we enable agents to be you know, how do we enable agents to be you know, how do we enable agents to be independent economic actors, right? Make independent economic actors, right? Make independent economic actors, right? Make sure businesses can sell to them, make sure businesses can sell to them, make sure businesses can sell to them, make sure they can um spend on behalf of sure they can um spend on behalf of sure they can um spend on behalf of buyers. Um and you know, Stripe is very buyers. Um and you know, Stripe is very buyers. Um and you know, Stripe is very data driven. We love to look at metrics. data driven. We love to look at metrics. data driven. We love to look at metrics. Uh most mornings we look at metrics, Uh most mornings we look at metrics, Uh most mornings we look at metrics, most afternoons as well. Um and you most afternoons as well. Um and you most afternoons as well. Um and you know, you may have seen when we know, you may have seen when we know, you may have seen when we announced um the acquisition of Open announced um the acquisition of Open announced um the acquisition of Open Router that Patrick in his in his letter Router that Patrick in his in his letter Router that Patrick in his in his letter to investors uh wrote a bit about the to investors uh wrote a bit about the to investors uh wrote a bit about the singularity and mentioned that sort of singularity and mentioned that sort of singularity and mentioned that sort of January 1 was the first day of the January 1 was the first day of the January 1 was the first day of the Singularity. We could debate what does Singularity. We could debate what does Singularity. We could debate what does it mean and what was the date, but it mean and what was the date, but it mean and what was the date, but really like we were looking at the really like we were looking at the really like we were looking at the metrics as we always do and we saw just metrics as we always do and we saw just metrics as we always do and we saw just a sharp up and to the right of a sharp up and to the right of a sharp up and to the right of >> such a broad swath of metrics that was >> such a broad swath of metrics that was >> such a broad swath of metrics that was since sustained um that we thought hey since sustained um that we thought hey since sustained um that we thought hey we're just fundamentally living in a new we're just fundamentally living in a new we're just fundamentally living in a new era and for us you know there there were era and for us you know there there were era and for us you know there there were again many dimensions but the three that again many dimensions but the three that again many dimensions but the three that really popped were one just more new really popped were one just more new really popped were one just more new businesses being created so this businesses being created so this businesses being created so this economic dynamism right the um number of economic dynamism right the um number of economic dynamism right the um number of uh businesses being incorporated through uh businesses being incorporated through uh businesses being incorporated through Atlas [clears throat] Atlas [clears throat] Atlas [clears throat] was doubling year-over-year. Um the was doubling year-over-year. Um the was doubling year-over-year. Um the number of new businesses joining Stripe

number of new businesses joining Stripe number of new businesses joining Stripe at all in any form was up 50% at all in any form was up 50% at all in any form was up 50% year-on-year. Um the second trend was year-on-year. Um the second trend was year-on-year. Um the second trend was like, hey, these aren't just like like, hey, these aren't just like like, hey, these aren't just like signups or new businesses being created signups or new businesses being created signups or new businesses being created and then spun down. These are businesses and then spun down. These are businesses and then spun down. These are businesses that are monetizing and monetizing that are monetizing and monetizing that are monetizing and monetizing faster and more than ever. So, um if you faster and more than ever. So, um if you faster and more than ever. So, um if you look at our 2026 cohort, they're look at our 2026 cohort, they're look at our 2026 cohort, they're monetizing about 50% more on average. So monetizing about 50% more on average. So monetizing about 50% more on average. So the median um than was the prior cohort. the median um than was the prior cohort. the median um than was the prior cohort. So more businesses monetizing faster and So more businesses monetizing faster and So more businesses monetizing faster and and more and a lot of that is driven by and more and a lot of that is driven by and more and a lot of that is driven by AI. Um and then we saw the agents AI. Um and then we saw the agents AI. Um and then we saw the agents directly, right? Like um I think the CLI directly, right? Like um I think the CLI directly, right? Like um I think the CLI is a good example of this. You know, is a good example of this. You know, is a good example of this. You know, we've had our command line interface for we've had our command line interface for we've had our command line interface for maybe seven years now and it was these maybe seven years now and it was these maybe seven years now and it was these sort of very nichy developers that would sort of very nichy developers that would sort of very nichy developers that would use it in small numbers. We weren't use it in small numbers. We weren't use it in small numbers. We weren't investing at all. We didn't change investing at all. We didn't change investing at all. We didn't change anything in the product and all of a anything in the product and all of a anything in the product and all of a sudden just like went vertical. Yeah. sudden just like went vertical. Yeah. sudden just like went vertical. Yeah. And it was because the agents had found And it was because the agents had found And it was because the agents had found it. And we were like, well, agents are it. And we were like, well, agents are it. And we were like, well, agents are helping form these businesses. These helping form these businesses. These helping form these businesses. These businesses are monetizing and reaching businesses are monetizing and reaching businesses are monetizing and reaching real users, right? We see a very real users, right? We see a very real users, right? We see a very different slice of the AI economy than different slice of the AI economy than different slice of the AI economy than than GPUs or models. We see actually than GPUs or models. We see actually than GPUs or models. We see actually where like supply and demand intersect where like supply and demand intersect where like supply and demand intersect and where real customers are buying um and where real customers are buying um and where real customers are buying um these AI solutions. Um and so seeing these AI solutions. Um and so seeing these AI solutions. Um and so seeing that inflection was really powerful. and that inflection was really powerful. and that inflection was really powerful. and then and then agents as actors first as then and then agents as actors first as then and then agents as actors first as uh sort of users of Stripe and now uh sort of users of Stripe and now uh sort of users of Stripe and now increasingly as buyers on the internet increasingly as buyers on the internet increasingly as buyers on the internet uh has been has been a big part of the uh has been has been a big part of the uh has been has been a big part of the shift. So that's where we are and I shift. So that's where we are and I shift. So that's where we are and I would say we're just really focused on would say we're just really focused on would say we're just really focused on you know we've built handinhand with um you know we've built handinhand with um you know we've built handinhand with um basically every successive wave of basically every successive wave of basically every successive wave of startups. Um the SAS platforms were one startups. Um the SAS platforms were one startups. Um the SAS platforms were one and the marketplaces and really the and the marketplaces and really the and the marketplaces and really the current wave is is AI and they're more

current wave is is AI and they're more current wave is is AI and they're more demanding than any of our previous demanding than any of our previous demanding than any of our previous users. So, it's been a it's been a busy users. So, it's been a it's been a busy users. So, it's been a it's been a busy and fun time. Um, but very much just and fun time. Um, but very much just and fun time. Um, but very much just like building handin-hand with them like building handin-hand with them like building handin-hand with them because their needs are so different because their needs are so different because their needs are so different than traditional SAS. The business than traditional SAS. The business than traditional SAS. The business models are so different. models are so different. models are so different. >> We all are using agents more and more. >> We all are using agents more and more. >> We all are using agents more and more. Some of us don't realize it, but the Some of us don't realize it, but the Some of us don't realize it, but the things in our pockets are agents now. things in our pockets are agents now. things in our pockets are agents now. Basically, um, Basically, um, Basically, um, >> in that world, how do we scale trust? >> in that world, how do we scale trust? >> in that world, how do we scale trust? How do we scale trust when the agents How do we scale trust when the agents How do we scale trust when the agents are supposed to be transacting? And I are supposed to be transacting? And I are supposed to be transacting? And I think it's become very salient even in think it's become very salient even in think it's become very salient even in the last couple of weeks as we've talked the last couple of weeks as we've talked the last couple of weeks as we've talked a lot about agents breaking trust. What a lot about agents breaking trust. What a lot about agents breaking trust. What does it look like for Stripe to play a does it look like for Stripe to play a does it look like for Stripe to play a part in the agent trust economy? part in the agent trust economy? part in the agent trust economy? >> Yeah. So I mean there's there's so many >> Yeah. So I mean there's there's so many >> Yeah. So I mean there's there's so many dimensions of trust, right? One of the dimensions of trust, right? One of the dimensions of trust, right? One of the dimensions of trust we obviously care dimensions of trust we obviously care dimensions of trust we obviously care deeply about is when an agent's buying deeply about is when an agent's buying deeply about is when an agent's buying something for you, when an agent's something for you, when an agent's something for you, when an agent's spending money on your behalf, how do spending money on your behalf, how do spending money on your behalf, how do you trust them to spend that money? And you trust them to spend that money? And you trust them to spend that money? And then on the other side of the equation, then on the other side of the equation, then on the other side of the equation, if you're a business where an agent is if you're a business where an agent is if you're a business where an agent is buying from you, how do you trust that buying from you, how do you trust that buying from you, how do you trust that that agent is a good buyer? Um, and so, that agent is a good buyer? Um, and so, that agent is a good buyer? Um, and so, you know, we have we have Link. Are you you know, we have we have Link. Are you you know, we have we have Link. Are you a link user? Okay. So, Link started out a link user? Okay. So, Link started out a link user? Okay. So, Link started out as our consumer wallet. We have about as our consumer wallet. We have about as our consumer wallet. We have about 300 million consumers who use Link, but 300 million consumers who use Link, but 300 million consumers who use Link, but Link has very quickly transformed to be Link has very quickly transformed to be Link has very quickly transformed to be the wallet for agents, the wallet for agents, the wallet for agents, >> right? So, if you're using Muse, for >> right? So, if you're using Muse, for >> right? So, if you're using Muse, for example, um, your agent can be example, um, your agent can be example, um, your agent can be outspending for you through Link. Um and outspending for you through Link. Um and outspending for you through Link. Um and with that, you know, your identity is with that, you know, your identity is with that, you know, your identity is connected to link. So to some extent, connected to link. So to some extent, connected to link. So to some extent, the end seller knows who you are. Um but the end seller knows who you are. Um but the end seller knows who you are. Um but we also have a whole suite of trust and we also have a whole suite of trust and we also have a whole suite of trust and abuse controls for um the customer and

abuse controls for um the customer and abuse controls for um the customer and for the buyer and for the business are for the buyer and for the business are for the buyer and for the business are the controls and protections in the the controls and protections in the the controls and protections in the hands of that individual. And then as hands of that individual. And then as hands of that individual. And then as the business is making a sale to that the business is making a sale to that the business is making a sale to that agent, um we're very rigorous about agent, um we're very rigorous about agent, um we're very rigorous about passing along many of the same fraud, passing along many of the same fraud, passing along many of the same fraud, abuse, trust, safety scores that we abuse, trust, safety scores that we abuse, trust, safety scores that we historically had on individuals um about historically had on individuals um about historically had on individuals um about the agents so that businesses can can the agents so that businesses can can the agents so that businesses can can decision. But I think we're in a whole decision. But I think we're in a whole decision. But I think we're in a whole new world of of trust and fraud and new world of of trust and fraud and new world of of trust and fraud and abuse. And maybe um you know, one one abuse. And maybe um you know, one one abuse. And maybe um you know, one one slice that uh brings it to life for me slice that uh brings it to life for me slice that uh brings it to life for me is actually the work we did with Cursor. is actually the work we did with Cursor. is actually the work we did with Cursor. Like we're at Stripe, we're we're 11 Like we're at Stripe, we're we're 11 Like we're at Stripe, we're we're 11 years into um fraud prevention for our years into um fraud prevention for our years into um fraud prevention for our users, right? Millions and millions of users, right? Millions and millions of users, right? Millions and millions of businesses use Radar, which is our fraud businesses use Radar, which is our fraud businesses use Radar, which is our fraud protection product. And it wasn't that protection product. And it wasn't that protection product. And it wasn't that many months ago that cursor came to us many months ago that cursor came to us many months ago that cursor came to us and they were like, "Radar is world and they were like, "Radar is world and they were like, "Radar is world class, but it is not solving my class, but it is not solving my class, but it is not solving my problems." And you know, when you when problems." And you know, when you when problems." And you know, when you when you pull the thread on why it very you pull the thread on why it very you pull the thread on why it very quickly becomes obvious, which is radar quickly becomes obvious, which is radar quickly becomes obvious, which is radar historically was built at the historically was built at the historically was built at the transaction level. And most of the sort transaction level. And most of the sort transaction level. And most of the sort of existential abuse that AI companies of existential abuse that AI companies of existential abuse that AI companies are facing is actually pre-transaction are facing is actually pre-transaction are facing is actually pre-transaction outside of the transaction. And that is, outside of the transaction. And that is, outside of the transaction. And that is, you know, um basically stealing tokens you know, um basically stealing tokens you know, um basically stealing tokens instead of stealing money. And you know, instead of stealing money. And you know, instead of stealing money. And you know, tokens are not yet a currency. Maybe tokens are not yet a currency. Maybe tokens are not yet a currency. Maybe they will be, we'll see. Uh but they are they will be, we'll see. Uh but they are they will be, we'll see. Uh but they are certainly a very very valuable asset.

certainly a very very valuable asset. certainly a very very valuable asset. And so, as always happens, fraudsters And so, as always happens, fraudsters And so, as always happens, fraudsters figure out how to steal valuable assets. figure out how to steal valuable assets. figure out how to steal valuable assets. And the same way they used to steal, you And the same way they used to steal, you And the same way they used to steal, you know, credentials or money, they're now know, credentials or money, they're now know, credentials or money, they're now trying to steal tokens. Um but again trying to steal tokens. Um but again trying to steal tokens. Um but again it's often disconnected from the moment it's often disconnected from the moment it's often disconnected from the moment of transaction. So in the case of cursor of transaction. So in the case of cursor of transaction. So in the case of cursor um you know it was it was uh users um you know it was it was uh users um you know it was it was uh users fraudsters coming in and spinning up fraudsters coming in and spinning up fraudsters coming in and spinning up accounts and then stealing the free accounts and then stealing the free accounts and then stealing the free credits. credits. credits. >> Yeah. The spinning up free trials or uh >> Yeah. The spinning up free trials or uh >> Yeah. The spinning up free trials or uh creating a a payo um cart and then creating a a payo um cart and then creating a a payo um cart and then actually not paying when the bill comes actually not paying when the bill comes actually not paying when the bill comes due. And you know, traditional SAS due. And you know, traditional SAS due. And you know, traditional SAS probably wouldn't have cared about that probably wouldn't have cared about that probably wouldn't have cared about that because, you know, the marginal cost of because, you know, the marginal cost of because, you know, the marginal cost of selling your traditional SAS product is selling your traditional SAS product is selling your traditional SAS product is near zero. But AI certainly cares about near zero. But AI certainly cares about near zero. But AI certainly cares about that because inference costs are that because inference costs are that because inference costs are extremely high. And these, you know, extremely high. And these, you know, extremely high. And these, you know, these these free trials, these premium these these free trials, these premium these these free trials, these premium offerings are um they're loss leaders. offerings are um they're loss leaders. offerings are um they're loss leaders. >> Um well, they're loss leaders if it's a >> Um well, they're loss leaders if it's a >> Um well, they're loss leaders if it's a good user. If it's a bad user, they're good user. If it's a bad user, they're good user. If it's a bad user, they're just lossers, [laughter] just lossers, [laughter] just lossers, [laughter] losers, one might say. Um and so you losers, one might say. Um and so you losers, one might say. Um and so you know we uh embedded with with cursor and know we uh embedded with with cursor and know we uh embedded with with cursor and this is actually a good example of like this is actually a good example of like this is actually a good example of like how we work handinhand with our users.

how we work handinhand with our users. how we work handinhand with our users. They come to us with an existential They come to us with an existential They come to us with an existential need. We embedded folks um in there was need. We embedded folks um in there was need. We embedded folks um in there was no like PRD or proper product no like PRD or proper product no like PRD or proper product development process like no practices development process like no practices development process like no practices just get in there go [laughter] just get in there go [laughter] just get in there go [laughter] >> and in this case we could get in there >> and in this case we could get in there >> and in this case we could get in there and move super quickly because of the and move super quickly because of the and move super quickly because of the stripe network right so we process like stripe network right so we process like stripe network right so we process like two trillion a year is about uh 2% of two trillion a year is about uh 2% of two trillion a year is about uh 2% of global GDP but it's a much much larger global GDP but it's a much much larger global GDP but it's a much much larger share of the AI economy right so if you share of the AI economy right so if you share of the AI economy right so if you think about you can look at all think about you can look at all think about you can look at all different indices but take something different indices but take something different indices but take something like the like the Forbes AI50 like 88% like the like the Forbes AI50 like 88% like the like the Forbes AI50 like 88% of them run on Stripe. So we see that of them run on Stripe. So we see that of them run on Stripe. So we see that side of the economy and then all of the side of the economy and then all of the side of the economy and then all of the new businesses, many of which are AI new businesses, many of which are AI new businesses, many of which are AI businesses getting spun up on Verscell businesses getting spun up on Verscell businesses getting spun up on Verscell and Replet and Manis and you name it, and Replet and Manis and you name it, and Replet and Manis and you name it, right? Those guys are also right? Those guys are also right? Those guys are also >> building on Stripe because we are deeply >> building on Stripe because we are deeply >> building on Stripe because we are deeply integrated into those AI dev tools. So integrated into those AI dev tools. So integrated into those AI dev tools. So we see sort of the the full swath of the we see sort of the the full swath of the we see sort of the the full swath of the AI economy, which means in our cross AI economy, which means in our cross AI economy, which means in our cross network data, we actually did know who network data, we actually did know who network data, we actually did know who was an abusive customer, right? Right. was an abusive customer, right? Right. was an abusive customer, right? Right. So, so the name of the game, what cursor So, so the name of the game, what cursor So, so the name of the game, what cursor needed and what it turned out every AI needed and what it turned out every AI needed and what it turned out every AI company needed was for us to move from company needed was for us to move from company needed was for us to move from scoring transactions scoring transactions scoring transactions uh to really identifying abusive or uh to really identifying abusive or uh to really identifying abusive or risky customers. Not about the risky customers. Not about the risky customers. Not about the transactions by the time they create an transactions by the time they create an transactions by the time they create an account or start a free trial or start account or start a free trial or start account or start a free trial or start to rack up overages like tell me if to rack up overages like tell me if to rack up overages like tell me if they're likely abusive. And because we they're likely abusive. And because we they're likely abusive. And because we could see it across the network, you could see it across the network, you could see it across the network, you know, a a fraudster is rarely only know, a a fraudster is rarely only know, a a fraudster is rarely only operating on one service, right?

operating on one service, right? operating on one service, right? >> They're playing a [laughter] big game. >> They're playing a [laughter] big game. >> They're playing a [laughter] big game. They're playing a big game. Um and so They're playing a big game. Um and so They're playing a big game. Um and so that allowed us to move really really that allowed us to move really really that allowed us to move really really quickly um for cursor and and literally quickly um for cursor and and literally quickly um for cursor and and literally have again not some fancy product but have again not some fancy product but have again not some fancy product but just pipelines live to that like an API just pipelines live to that like an API just pipelines live to that like an API they could hit in days right and then we they could hit in days right and then we they could hit in days right and then we scaled that up and um you know basically scaled that up and um you know basically scaled that up and um you know basically every AI company doesn't just use Stripe every AI company doesn't just use Stripe every AI company doesn't just use Stripe they use Stripe for their full financial they use Stripe for their full financial they use Stripe for their full financial infrastructure they use us for payments infrastructure they use us for payments infrastructure they use us for payments yes but also billing and tax and revenue yes but also billing and tax and revenue yes but also billing and tax and revenue recognition and fraud solutions like recognition and fraud solutions like recognition and fraud solutions like Radar and our data pipeline because Radar and our data pipeline because Radar and our data pipeline because guess what they want real time access to guess what they want real time access to guess what they want real time access to all of that data to build their all of that data to build their all of that data to build their applications on top. They need to know applications on top. They need to know applications on top. They need to know who's buying what for how much, who's who's buying what for how much, who's who's buying what for how much, who's retaining insurance and shurning their retaining insurance and shurning their retaining insurance and shurning their subscriptions, who has credit, who subscriptions, who has credit, who subscriptions, who has credit, who doesn't, who's going fraud, who's not. doesn't, who's going fraud, who's not. doesn't, who's going fraud, who's not. Um, and so they're they're using really Um, and so they're they're using really Um, and so they're they're using really really all of Stripe, and that includes really all of Stripe, and that includes really all of Stripe, and that includes um Stripe Radar. um Stripe Radar. um Stripe Radar. >> So that that this is part of why you and >> So that that this is part of why you and >> So that that this is part of why you and I had so much fun talking. I think one I had so much fun talking. I think one I had so much fun talking. I think one of the things that got my wheels turning of the things that got my wheels turning of the things that got my wheels turning then and that got my wheels turning in then and that got my wheels turning in then and that got my wheels turning in your answer now is that you talked about your answer now is that you talked about your answer now is that you talked about this idea that Stripe is moving beyond this idea that Stripe is moving beyond this idea that Stripe is moving beyond the transaction. the transaction. the transaction. >> Yes. And so you're talking about in the >> Yes. And so you're talking about in the >> Yes. And so you're talking about in the token fraud space, but to me when I token fraud space, but to me when I token fraud space, but to me when I think about what I hear from folks in think about what I hear from folks in think about what I hear from folks in the audience, um I hear about this the audience, um I hear about this the audience, um I hear about this larger crisis of trust in the way uh not larger crisis of trust in the way uh not larger crisis of trust in the way uh not just economic transactions but the way just economic transactions but the way just economic transactions but the way software is delivered. Mhm.

software is delivered. Mhm. software is delivered. Mhm. >> And so I think the question that I would >> And so I think the question that I would >> And so I think the question that I would have, it's a very open-ended question so have, it's a very open-ended question so have, it's a very open-ended question so we can kind of kick it around is if we we can kind of kick it around is if we we can kind of kick it around is if we look at token fraud as the first foray look at token fraud as the first foray look at token fraud as the first foray into a larger conversation about the into a larger conversation about the into a larger conversation about the reliability rails that the new AI reliability rails that the new AI reliability rails that the new AI economy needs. economy needs. economy needs. >> Yes. >> Yes. >> Yes. >> Where else do you see that going? >> Where else do you see that going? >> Where else do you see that going? >> Yeah. So, so what does the AI economy >> Yeah. So, so what does the AI economy >> Yeah. So, so what does the AI economy need? The AI economy needs, well, let's need? The AI economy needs, well, let's need? The AI economy needs, well, let's just take the models as given because just take the models as given because just take the models as given because there's plenty of people working there there's plenty of people working there there's plenty of people working there for the AI economy to work. So all of for the AI economy to work. So all of for the AI economy to work. So all of that supply of stuff and the stuff that that supply of stuff and the stuff that that supply of stuff and the stuff that can be built on that stuff to connect can be built on that stuff to connect can be built on that stuff to connect with the demand which I think you with the demand which I think you with the demand which I think you yourself noted is not chat. It is like yourself noted is not chat. It is like yourself noted is not chat. It is like agents operating on the internet. Um agents operating on the internet. Um agents operating on the internet. Um there's a few things that need to exist. there's a few things that need to exist. there's a few things that need to exist. First First First >> you need to be able to monetize like the >> you need to be able to monetize like the >> you need to be able to monetize like the good functioning product that you've good functioning product that you've good functioning product that you've built and you need to monetize in a way built and you need to monetize in a way built and you need to monetize in a way that doesn't put you in the red and some that doesn't put you in the red and some that doesn't put you in the red and some of that is usage based billing and why of that is usage based billing and why of that is usage based billing and why we acquired metronome. I was actually uh we acquired metronome. I was actually uh we acquired metronome. I was actually uh last week I was in Soma with about 300 last week I was in Soma with about 300 last week I was in Soma with about 300 metronome users and they were all going metronome users and they were all going metronome users and they were all going through the journey of like let me move through the journey of like let me move through the journey of like let me move from from subscriptions to usage based from from subscriptions to usage based from from subscriptions to usage based so I cover my cost but the best among so I cover my cost but the best among so I cover my cost but the best among them were also moving to outcomesbased them were also moving to outcomesbased them were also moving to outcomesbased because part of trust is knowing that because part of trust is knowing that because part of trust is knowing that when you pay for a thing you're actually when you pay for a thing you're actually when you pay for a thing you're actually getting the value and you don't really getting the value and you don't really getting the value and you don't really want to be paying for like the infrared want to be paying for like the infrared want to be paying for like the infrared cost or the GPU you want to be paying cost or the GPU you want to be paying cost or the GPU you want to be paying for like the value to your business the for like the value to your business the for like the value to your business the outcome delivered so I think there's outcome delivered so I think there's outcome delivered so I think there's this whole important world of of sort of this whole important world of of sort of this whole important world of of sort of the the billing and monetization stack the the billing and monetization stack the the billing and monetization stack Then there's the there's the important Then there's the there's the important Then there's the there's the important world we talked about around fraud and world we talked about around fraud and world we talked about around fraud and abuse. And obviously why why does that abuse. And obviously why why does that abuse. And obviously why why does that matter? It's not just for the businesses matter? It's not just for the businesses matter? It's not just for the businesses to be healthy. It's like you can't have to be healthy. It's like you can't have to be healthy. It's like you can't have a functioning market if you need some

a functioning market if you need some a functioning market if you need some huge markup. huge markup. huge markup. >> That's right. >> That's right. >> That's right. >> On um on the product to cover what the >> On um on the product to cover what the >> On um on the product to cover what the fraudsters are are costing you. And then fraudsters are are costing you. And then fraudsters are are costing you. And then there's this third piece around making there's this third piece around making there's this third piece around making agents independent actors. And that's agents independent actors. And that's agents independent actors. And that's actually not unrelated to one and two, actually not unrelated to one and two, actually not unrelated to one and two, right? So maybe take the cursor example. right? So maybe take the cursor example. right? So maybe take the cursor example. If cursor was so existentially hurt by If cursor was so existentially hurt by If cursor was so existentially hurt by free trials or um uh the premium tier free trials or um uh the premium tier free trials or um uh the premium tier and we couldn't help them, what were and we couldn't help them, what were and we couldn't help them, what were they going to do? They're going to turn they going to do? They're going to turn they going to do? They're going to turn off retrials. They're going to shut off retrials. They're going to shut off retrials. They're going to shut down. They're going to move to purely down. They're going to move to purely down. They're going to move to purely salesled. They're going to have some salesled. They're going to have some salesled. They're going to have some sort of like big go to market team out sort of like big go to market team out sort of like big go to market team out selling everything and productled growth selling everything and productled growth selling everything and productled growth is going to be thrown to the wayside is going to be thrown to the wayside is going to be thrown to the wayside because it's too risky. because it's too risky. because it's too risky. >> Yeah. >> Yeah. >> Yeah. >> Well, guess who doesn't talk to sellers? >> Well, guess who doesn't talk to sellers? >> Well, guess who doesn't talk to sellers? >> Who doesn't talk to sellers, Emily? >> Who doesn't talk to sellers, Emily? >> Who doesn't talk to sellers, Emily? >> Agents don't talk to sellers. Are you >> Agents don't talk to sellers. Are you >> Agents don't talk to sellers. Are you with me? Yeah. So, so like why do you with me? Yeah. So, so like why do you with me? Yeah. So, so like why do you need PLG? You need PLG for the need PLG? You need PLG for the need PLG? You need PLG for the individual developers who by the way are individual developers who by the way are individual developers who by the way are like in many ways more exciting than the like in many ways more exciting than the like in many ways more exciting than the very large traditional enterprises and very large traditional enterprises and very large traditional enterprises and you absolutely want them on your product you absolutely want them on your product you absolutely want them on your product but you also need the agents on your but you also need the agents on your but you also need the agents on your product which means you need to have the product which means you need to have the product which means you need to have the monetization mechanism generally usage monetization mechanism generally usage monetization mechanism generally usage based billing because you know I'm not based billing because you know I'm not based billing because you know I'm not going to tell an agent they can go and going to tell an agent they can go and going to tell an agent they can go and sign up for an annual contract. I'm sign up for an annual contract. I'm sign up for an annual contract. I'm going to tell them I'm happy if they pay going to tell them I'm happy if they pay going to tell them I'm happy if they pay Stripe a cent per query or whatever Stripe a cent per query or whatever Stripe a cent per query or whatever >> microtransaction >> microtransaction >> microtransaction >> a microtransaction. Uh well and then you >> a microtransaction. Uh well and then you >> a microtransaction. Uh well and then you need the financial infrastructure for need the financial infrastructure for need the financial infrastructure for that which we haven't talked about.

that which we haven't talked about. that which we haven't talked about. Yeah, that's a whole piece. That's a Yeah, that's a whole piece. That's a Yeah, that's a whole piece. That's a whole piece like the whole the whole uh whole piece like the whole the whole uh whole piece like the whole the whole uh I mean stable coins brings us a long I mean stable coins brings us a long I mean stable coins brings us a long way. It's not the only way you can do way. It's not the only way you can do way. It's not the only way you can do stored balances and treasury and a whole stored balances and treasury and a whole stored balances and treasury and a whole bunch of other stuff but but I think um bunch of other stuff but but I think um bunch of other stuff but but I think um so far stable coins seems to be an so far stable coins seems to be an so far stable coins seems to be an important part of u micro important part of u micro important part of u micro microtransactions. Um then you need the microtransactions. Um then you need the microtransactions. Um then you need the fraud and abuse and then you need to fraud and abuse and then you need to fraud and abuse and then you need to make it actually um easy for agents to make it actually um easy for agents to make it actually um easy for agents to discover, know what they need to pay and discover, know what they need to pay and discover, know what they need to pay and pay. And there's the business side of pay. And there's the business side of pay. And there's the business side of the house which is our machine payments the house which is our machine payments the house which is our machine payments protocol, right? Like as a business, as protocol, right? Like as a business, as protocol, right? Like as a business, as a software business for example, I need a software business for example, I need a software business for example, I need to be able to say this is what I sell. to be able to say this is what I sell. to be able to say this is what I sell. This is how much it costs and this is This is how much it costs and this is This is how much it costs and this is how you pay. I need to be able to say how you pay. I need to be able to say how you pay. I need to be able to say that in a machine readable way um that in a machine readable way um that in a machine readable way um through the machine payments protocol. through the machine payments protocol. through the machine payments protocol. And then the agent needs to be able to And then the agent needs to be able to And then the agent needs to be able to actually send me those funds. Um, and as actually send me those funds. Um, and as actually send me those funds. Um, and as a consumer or a business, I need to task a consumer or a business, I need to task a consumer or a business, I need to task an agent, give an agent the ability to an agent, give an agent the ability to an agent, give an agent the ability to pay on my behalf, which is again going pay on my behalf, which is again going pay on my behalf, which is again going back to things like the link link as the back to things like the link link as the back to things like the link link as the agent for the wallet for agents. agent for the wallet for agents. agent for the wallet for agents. >> Let's touch on the link piece a little >> Let's touch on the link piece a little >> Let's touch on the link piece a little bit because this gets at like current bit because this gets at like current bit because this gets at like current events and all the news about the events and all the news about the events and all the news about the hugging face hack and this and that. hugging face hack and this and that. hugging face hack and this and that. >> Yes. When I think about where the >> Yes. When I think about where the >> Yes. When I think about where the popular narrative around agents is and popular narrative around agents is and popular narrative around agents is and where it's sort of moving into this where it's sort of moving into this where it's sort of moving into this space where yes, agents are goal space where yes, agents are goal space where yes, agents are goal motivated, agents are driven, agents motivated, agents are driven, agents motivated, agents are driven, agents will hack their own rewards environments will hack their own rewards environments will hack their own rewards environments to get what they need to do.

to get what they need to do. to get what they need to do. >> I think a lot about the difference >> I think a lot about the difference >> I think a lot about the difference between a goal and the shape of the goal between a goal and the shape of the goal between a goal and the shape of the goal along the way, the way the agent gets along the way, the way the agent gets along the way, the way the agent gets there, the side quest the agent goes on. there, the side quest the agent goes on. there, the side quest the agent goes on. How do you think like I I feel like a How do you think like I I feel like a How do you think like I I feel like a lot of the link product space right now lot of the link product space right now lot of the link product space right now is around that goal guaranteeing that is around that goal guaranteeing that is around that goal guaranteeing that goal is accomplished correctly. goal is accomplished correctly. goal is accomplished correctly. >> But if we live in a world where agents >> But if we live in a world where agents >> But if we live in a world where agents also have the [snorts] ability to do also have the [snorts] ability to do also have the [snorts] ability to do inappropriate side quests, inappropriate side quests, inappropriate side quests, >> how do you think about where that >> how do you think about where that >> how do you think about where that product space needs to evolve so that we product space needs to evolve so that we product space needs to evolve so that we have the full rails of the transaction have the full rails of the transaction have the full rails of the transaction guarantee? guarantee? guarantee? >> Yeah. Yeah. Yeah. So I think there's >> Yeah. Yeah. Yeah. So I think there's >> Yeah. Yeah. Yeah. So I think there's maybe the maybe the most basic piece is maybe the maybe the most basic piece is maybe the maybe the most basic piece is go on all the side quests you want, but go on all the side quests you want, but go on all the side quests you want, but before you spend my money, I need to before you spend my money, I need to before you spend my money, I need to approve. approve. approve. >> And I can say I don't care about $5. I >> And I can say I don't care about $5. I >> And I can say I don't care about $5. I don't want to approve every $5 or I don't want to approve every $5 or I don't want to approve every $5 or I don't want to approve the first $100, don't want to approve the first $100, don't want to approve the first $100, but like at some point I need to approve but like at some point I need to approve but like at some point I need to approve and I think those rails are important. and I think those rails are important. and I think those rails are important. Um, Um, Um, but then I think there's this piece. So, but then I think there's this piece. So, but then I think there's this piece. So, we talked a lot about the revenue and we talked a lot about the revenue and we talked a lot about the revenue and the monetization and sort of the trust the monetization and sort of the trust the monetization and sort of the trust layer within that. Then there's sort of layer within that. Then there's sort of layer within that. Then there's sort of like what are the rails for like what are the rails for like what are the rails for intelligence? Not what are the rails for intelligence? Not what are the rails for intelligence? Not what are the rails for money for intelligence, but what are the money for intelligence, but what are the money for intelligence, but what are the rails for intelligence? And actually um rails for intelligence? And actually um rails for intelligence? And actually um you know the the recent announcement you know the the recent announcement you know the the recent announcement around acquiring open router is an around acquiring open router is an around acquiring open router is an important part of this because actually important part of this because actually important part of this because actually open router is core intelligence open router is core intelligence open router is core intelligence infrastructure right that's what they do infrastructure right that's what they do infrastructure right that's what they do it's it's it's you know what is the um it's it's it's you know what is the um it's it's it's you know what is the um it's a very actually complex combination it's a very actually complex combination it's a very actually complex combination of things that the user is generally of things that the user is generally of things that the user is generally optimizing for on open router but what optimizing for on open router but what optimizing for on open router but what model do I need given my objective model do I need given my objective model do I need given my objective function which includes quality and cost function which includes quality and cost function which includes quality and cost and latency and reliability and and and and latency and reliability and and and and latency and reliability and and and and right and I think you know the agent

and right and I think you know the agent and right and I think you know the agent is in many cases actually consuming is in many cases actually consuming is in many cases actually consuming additional inference and so the tighter additional inference and so the tighter additional inference and so the tighter we can get to understanding the co the we can get to understanding the co the we can get to understanding the co the task understanding the cost of the task task understanding the cost of the task task understanding the cost of the task understanding the quality of the output understanding the quality of the output understanding the quality of the output and then understanding the downstream and then understanding the downstream and then understanding the downstream revenue that's going to come from that revenue that's going to come from that revenue that's going to come from that output if you think about stripe today a output if you think about stripe today a output if you think about stripe today a lot of what we're doing is you know lot of what we're doing is you know lot of what we're doing is you know maximize your profits as a business by maximize your profits as a business by maximize your profits as a business by having the right monetization and the having the right monetization and the having the right monetization and the right pricing and the right fraud and right pricing and the right fraud and right pricing and the right fraud and abuse controls and the right customers, abuse controls and the right customers, abuse controls and the right customers, right? But really, you want to maximize right? But really, you want to maximize right? But really, you want to maximize your profit on both the revenue side and your profit on both the revenue side and your profit on both the revenue side and the true cost side, which requires the true cost side, which requires the true cost side, which requires getting deep into the guts of what's the getting deep into the guts of what's the getting deep into the guts of what's the underlying intelligence. You let your underlying intelligence. You let your underlying intelligence. You let your product, your agent, you know, product, your agent, you know, product, your agent, you know, >> and how are you smart about it? >> and how are you smart about it? >> and how are you smart about it? >> And how are you smart about it? Yeah. >> And how are you smart about it? Yeah. >> And how are you smart about it? Yeah. >> And how do you let your agents be smart >> And how do you let your agents be smart >> And how do you let your agents be smart about it? Because the agents will also about it? Because the agents will also about it? Because the agents will also be consumers of open router. be consumers of open router. be consumers of open router. >> So, yes. And here I think it's going to >> So, yes. And here I think it's going to >> So, yes. And here I think it's going to be like people are going to start with be like people are going to start with be like people are going to start with very very constrained parameters when it very very constrained parameters when it very very constrained parameters when it comes to you know what they're willing comes to you know what they're willing comes to you know what they're willing to let the agents do freely if that to let the agents do freely if that to let the agents do freely if that thing comes with cost or if that thing thing comes with cost or if that thing thing comes with cost or if that thing is directly impacting their end customer is directly impacting their end customer is directly impacting their end customer right and I think that's wise. It's the right and I think that's wise. It's the right and I think that's wise. It's the same way like when you and I started same way like when you and I started same way like when you and I started shopping online like we didn't go buy a shopping online like we didn't go buy a shopping online like we didn't go buy a leather couch because we had no idea leather couch because we had no idea leather couch because we had no idea what we were going to get what we were going to get what we were going to get >> and then gradually there is >> and then gradually there is >> and then gradually there is >> infrastructure data tooling built up >> infrastructure data tooling built up >> infrastructure data tooling built up around that experience that gives you around that experience that gives you around that experience that gives you confidence and conviction right so oh confidence and conviction right so oh confidence and conviction right so oh well now we're billing for outcomes and well now we're billing for outcomes and well now we're billing for outcomes and now we have these great evals that now we have these great evals that now we have these great evals that measure the outcomes and now we have

measure the outcomes and now we have measure the outcomes and now we have this great router that's actually this great router that's actually this great router that's actually optimizing for our profits based on the optimizing for our profits based on the optimizing for our profits based on the end need of the customer the thing we're end need of the customer the thing we're end need of the customer the thing we're billing for and what it's going cost to billing for and what it's going cost to billing for and what it's going cost to provide that thing and all of a sudden provide that thing and all of a sudden provide that thing and all of a sudden people are sort of gradually gonna people are sort of gradually gonna people are sort of gradually gonna >> It's like when you have like a puppy, >> It's like when you have like a puppy, >> It's like when you have like a puppy, right? Those retractable leashes like right? Those retractable leashes like right? Those retractable leashes like you let out like a little more little you let out like a little more little you let out like a little more little more. Yeah. And they're definitely I more. Yeah. And they're definitely I more. Yeah. And they're definitely I guess guess guess >> I'm not there with a corgi yet, but >> I'm not there with a corgi yet, but >> I'm not there with a corgi yet, but we're getting there. we're getting there. we're getting there. >> They're so cute. >> They're so cute. >> They're so cute. >> They're so cute. >> They're so cute. >> They're so cute. >> Deep love of corgis. Just one >> Deep love of corgis. Just one >> Deep love of corgis. Just one >> so far. There's there's a campaign for >> so far. There's there's a campaign for >> so far. There's there's a campaign for more in the household, so we will see. more in the household, so we will see. more in the household, so we will see. >> I I do think there's sort of a lot of >> I I do think there's sort of a lot of >> I I do think there's sort of a lot of sort of talk narrative around these sort sort of talk narrative around these sort sort of talk narrative around these sort of agents doing the the hugging face, of agents doing the the hugging face, of agents doing the the hugging face, you know, whole situation. But I think you know, whole situation. But I think you know, whole situation. But I think when it comes to money and building when it comes to money and building when it comes to money and building businesses with agents, we're actually businesses with agents, we're actually businesses with agents, we're actually seeing still like a relatively seeing still like a relatively seeing still like a relatively constrained rational approach. Um, and a constrained rational approach. Um, and a constrained rational approach. Um, and a lot of the demand for example in our in lot of the demand for example in our in lot of the demand for example in our in our link wallet um has been around build our link wallet um has been around build our link wallet um has been around build those controls. Um, and for businesses those controls. Um, and for businesses those controls. Um, and for businesses selling to agents like you know tell me selling to agents like you know tell me selling to agents like you know tell me exactly the quality of the agent that's exactly the quality of the agent that's exactly the quality of the agent that's coming to me and whether or not I can I coming to me and whether or not I can I coming to me and whether or not I can I can trust them to make good use of my can trust them to make good use of my can trust them to make good use of my product. And I'm curious like you're in product. And I'm curious like you're in product. And I'm curious like you're in this position of trust on the internet this position of trust on the internet this position of trust on the internet economy. How are you thinking about economy. How are you thinking about economy. How are you thinking about facilitating facilitating facilitating confidence from these traditional card confidence from these traditional card confidence from these traditional card networks that even if the agent is networks that even if the agent is networks that even if the agent is handling payments and it flows on your handling payments and it flows on your handling payments and it flows on your rails, there is not going to be an rails, there is not going to be an rails, there is not going to be an increase in chargebacks. You're not increase in chargebacks. You're not increase in chargebacks. You're not going to have a bunch of bad debt that going to have a bunch of bad debt that going to have a bunch of bad debt that piles up. Like there's going to be all piles up. Like there's going to be all piles up. Like there's going to be all those traditional indicators of loss of those traditional indicators of loss of those traditional indicators of loss of quality in the network are going to be

quality in the network are going to be quality in the network are going to be kept at reasonable levels. kept at reasonable levels. kept at reasonable levels. >> Yeah. So >> Yeah. So >> Yeah. So we actually have a lot of data and a lot we actually have a lot of data and a lot we actually have a lot of data and a lot of models doing that today for of models doing that today for of models doing that today for traditional businesses and traditional traditional businesses and traditional traditional businesses and traditional consumers. And so while the signals are consumers. And so while the signals are consumers. And so while the signals are quite different when you're talking quite different when you're talking quite different when you're talking about an agent, right, the way an agent about an agent, right, the way an agent about an agent, right, the way an agent navigates navigates navigates >> uh a purchase is very different than a >> uh a purchase is very different than a >> uh a purchase is very different than a human clicking through. And so a human human clicking through. And so a human human clicking through. And so a human clicking through, you would see certain clicking through, you would see certain clicking through, you would see certain behaviors that would indicate that behaviors that would indicate that behaviors that would indicate that they're potentially fraudulent while an they're potentially fraudulent while an they're potentially fraudulent while an agent is doing something different. So agent is doing something different. So agent is doing something different. So it's different signals, different it's different signals, different it's different signals, different inputs, but a lot of the outcomes, a lot inputs, but a lot of the outcomes, a lot inputs, but a lot of the outcomes, a lot of the sort of precision and recall of the sort of precision and recall of the sort of precision and recall you're looking for in any given model you're looking for in any given model you're looking for in any given model that you're talking to the regulators that you're talking to the regulators that you're talking to the regulators about and the banks about or the issuers about and the banks about or the issuers about and the banks about or the issuers about, it's quite similar. So we're kind about, it's quite similar. So we're kind about, it's quite similar. So we're kind of starting there and then when we learn of starting there and then when we learn of starting there and then when we learn that there's a different regime or sort that there's a different regime or sort that there's a different regime or sort of um a fundamentally different attack of um a fundamentally different attack of um a fundamentally different attack vector as we saw with sort of the the vector as we saw with sort of the the vector as we saw with sort of the the the customer abuse upfunnel of the customer abuse upfunnel of the customer abuse upfunnel of transaction, then we build new transaction, then we build new transaction, then we build new solutions. But but our base case is hey solutions. But but our base case is hey solutions. But but our base case is hey the whole industry understands the sort the whole industry understands the sort the whole industry understands the sort of objective functions as we've defined of objective functions as we've defined of objective functions as we've defined them today. And so let's understand how them today. And so let's understand how them today. And so let's understand how those perform with agents and make them those perform with agents and make them those perform with agents and make them really work for agents. Uh and then really work for agents. Uh and then really work for agents. Uh and then evolve from there.

evolve from there. evolve from there. >> I have a brain teaser for you. >> I have a brain teaser for you. >> I have a brain teaser for you. >> Oh goodness. >> Oh goodness. >> Oh goodness. >> Um >> Um >> Um >> it's a Monday morning. I don't have >> it's a Monday morning. I don't have >> it's a Monday morning. I don't have enough coffee. enough coffee. enough coffee. >> I know. We can pre bring you coffee if >> I know. We can pre bring you coffee if >> I know. We can pre bring you coffee if we need to. Uh we need to. Uh we need to. Uh >> I'll phone a corgi. >> I'll phone a corgi. >> I'll phone a corgi. >> That's right. Uh so it's a pricing >> That's right. Uh so it's a pricing >> That's right. Uh so it's a pricing question. Oh y question. Oh y question. Oh y >> and it's one that I don't have a good >> and it's one that I don't have a good >> and it's one that I don't have a good answer to and no one knows answer to and no one knows answer to and no one knows >> y >> y >> y >> that that I know of has a good answer. >> that that I know of has a good answer. >> that that I know of has a good answer. >> By the way, can I just say the whole >> By the way, can I just say the whole >> By the way, can I just say the whole industry is figuring out pricing in real industry is figuring out pricing in real industry is figuring out pricing in real time. I kid you not. It was like it was time. I kid you not. It was like it was time. I kid you not. It was like it was like hundreds of last week Soma, like hundreds of last week Soma, like hundreds of last week Soma, beautiful Soma. You think you'd like be beautiful Soma. You think you'd like be beautiful Soma. You think you'd like be outside drinking wine. No, we're like outside drinking wine. No, we're like outside drinking wine. No, we're like hundreds of like monetization, payments, hundreds of like monetization, payments, hundreds of like monetization, payments, pricing nerds sitting in a room being pricing nerds sitting in a room being pricing nerds sitting in a room being like my company is selling products like my company is selling products like my company is selling products >> that look nothing like the products >> that look nothing like the products >> that look nothing like the products we've sold in the past. we've sold in the past. we've sold in the past. differentally differentally differentally different different different properties we've sold before and yet properties we've sold before and yet properties we've sold before and yet consumers are still used to subscription consumers are still used to subscription consumers are still used to subscription based pricing. So how the heck do I get based pricing. So how the heck do I get based pricing. So how the heck do I get to something that makes sense for my to something that makes sense for my to something that makes sense for my business? Okay. So I'm just saying the business? Okay. So I'm just saying the business? Okay. So I'm just saying the whole the whole industry is stumped but whole the whole industry is stumped but whole the whole industry is stumped but I'm ready for your brain cheater. I'm ready for your brain cheater. I'm ready for your brain cheater. >> Great. So then you can solve it. >> Great. So then you can solve it. >> Great. So then you can solve it. >> Um so the question is this. When you >> Um so the question is this. When you >> Um so the question is this. When you talk about outcomebased pricing, talk about outcomebased pricing, talk about outcomebased pricing, >> yes, >> yes, >> yes, >> you are characterizing >> you are characterizing >> you are characterizing value that will be different for value that will be different for value that will be different for different customers.

different customers. different customers. >> Um, that they may not fully articulate >> Um, that they may not fully articulate >> Um, that they may not fully articulate back to you. And in fact, from a market back to you. And in fact, from a market back to you. And in fact, from a market incentives perspective, they probably incentives perspective, they probably incentives perspective, they probably don't have the incentive to fully don't have the incentive to fully don't have the incentive to fully articulate the value back to you. articulate the value back to you. articulate the value back to you. >> And you can fully know your cost and >> And you can fully know your cost and >> And you can fully know your cost and you're looking at the open router on the you're looking at the open router on the you're looking at the open router on the cost side. cost side. cost side. >> And you also know the product will be >> And you also know the product will be >> And you also know the product will be dynamic. in that world, how forget dynamic. in that world, how forget dynamic. in that world, how forget individually pricing correctly, how do individually pricing correctly, how do individually pricing correctly, how do you even get to a pricing schema or you even get to a pricing schema or you even get to a pricing schema or market structure around outcomes? Yes. market structure around outcomes? Yes. market structure around outcomes? Yes. That is logical. Okay, again I don't That is logical. Okay, again I don't That is logical. Okay, again I don't have the answer and actually I was I was have the answer and actually I was I was have the answer and actually I was I was um talking on Wednesday to Michael who's um talking on Wednesday to Michael who's um talking on Wednesday to Michael who's the the president and head of AI at the the president and head of AI at the the president and head of AI at Replet and they have sort of like long Replet and they have sort of like long Replet and they have sort of like long tossled with this question of can we tossled with this question of can we tossled with this question of can we move to you know they had a lot of move to you know they had a lot of move to you know they had a lot of backlash when they move from backlash when they move from backlash when they move from subscription to usage based um and subscription to usage based um and subscription to usage based um and they're thinking like can we move to they're thinking like can we move to they're thinking like can we move to outcomebased because wouldn't that be outcomebased because wouldn't that be outcomebased because wouldn't that be lovely for aligning incentives and lovely for aligning incentives and lovely for aligning incentives and they're looking at their like 50 million they're looking at their like 50 million they're looking at their like 50 million plus users who range from you know plus users who range from you know plus users who range from you know students to solopreneurs to Fortune 500s students to solopreneurs to Fortune 500s students to solopreneurs to Fortune 500s and they're like I don't know what what and they're like I don't know what what and they're like I don't know what what the standardized outcome is. There's one the standardized outcome is. There's one the standardized outcome is. There's one case that I think is solvable which is case that I think is solvable which is case that I think is solvable which is when the outcome can be directly tied to when the outcome can be directly tied to when the outcome can be directly tied to token consumption.

token consumption. token consumption. >> Mhm. that combined with eval right you >> Mhm. that combined with eval right you >> Mhm. that combined with eval right you basically have token billing which we basically have token billing which we basically have token billing which we offer for example on metronome um where offer for example on metronome um where offer for example on metronome um where you track in real time and price to the you track in real time and price to the you track in real time and price to the cost of the underlying cost of the underlying cost of the underlying >> LLM >> LLM >> LLM >> and then you have eval operating under >> and then you have eval operating under >> and then you have eval operating under the hood that say well given this this the hood that say well given this this the hood that say well given this this function that the firms defined for me function that the firms defined for me function that the firms defined for me about quality cost constraints what about quality cost constraints what about quality cost constraints what should be the most efficient model to should be the most efficient model to should be the most efficient model to choose so like the router chooses the choose so like the router chooses the choose so like the router chooses the model bas based on eval bunch of model bas based on eval bunch of model bas based on eval bunch of performance criteria and then the price performance criteria and then the price performance criteria and then the price that the business has for selling that that the business has for selling that that the business has for selling that good is just the cost of the tokens plus good is just the cost of the tokens plus good is just the cost of the tokens plus some markup. So that's like that's like some markup. So that's like that's like some markup. So that's like that's like clean but assumes you have good evals clean but assumes you have good evals clean but assumes you have good evals which then gets you to the question of which then gets you to the question of which then gets you to the question of what's the outcome people care about and what's the outcome people care about and what's the outcome people care about and what do you do in a world where the what do you do in a world where the what do you do in a world where the outcome is very heterogeneous and that I outcome is very heterogeneous and that I outcome is very heterogeneous and that I don't know yet. I do think, you know, don't know yet. I do think, you know, don't know yet. I do think, you know, we're probably beyond I love that you we're probably beyond I love that you we're probably beyond I love that you can create like, you know, infinite SKUs can create like, you know, infinite SKUs can create like, you know, infinite SKUs with any combination of numbers, but I with any combination of numbers, but I with any combination of numbers, but I think in practice, we're probably beyond think in practice, we're probably beyond think in practice, we're probably beyond the world of SKUs. Like there's no the world of SKUs. Like there's no the world of SKUs. Like there's no longer such a thing as a skew. There may longer such a thing as a skew. There may longer such a thing as a skew. There may be for some like very sort of proumer, be for some like very sort of proumer, be for some like very sort of proumer, you know, monthly subscription thing.

you know, monthly subscription thing. you know, monthly subscription thing. But in practice, the product you buy is But in practice, the product you buy is But in practice, the product you buy is going to be this complex combination of going to be this complex combination of going to be this complex combination of not just the product, but like the not just the product, but like the not just the product, but like the underlying models that power it, which underlying models that power it, which underlying models that power it, which themselves are constantly changing. And themselves are constantly changing. And themselves are constantly changing. And so I think that's going to have a whole so I think that's going to have a whole so I think that's going to have a whole bunch of downstream implications for, bunch of downstream implications for, bunch of downstream implications for, >> you know, what's your pricing table? >> you know, what's your pricing table? >> you know, what's your pricing table? What's your billing model? How do you What's your billing model? How do you What's your billing model? How do you talk about your product to the customer? talk about your product to the customer? talk about your product to the customer? Uh yeah, price, yeah, price Uh yeah, price, yeah, price Uh yeah, price, yeah, price discrimination is like a whole other discrimination is like a whole other discrimination is like a whole other world there. And you know, you said world there. And you know, you said world there. And you know, you said something really interesting, which is, something really interesting, which is, something really interesting, which is, uh, customers may not have full uh, customers may not have full uh, customers may not have full incentive to reveal the outcomes they're incentive to reveal the outcomes they're incentive to reveal the outcomes they're optimizing for. optimizing for. optimizing for. >> Yeah. >> Yeah. >> Yeah. >> Um, and there are two, you know, sort of >> Um, and there are two, you know, sort of >> Um, and there are two, you know, sort of competing factors there. Uh, they will competing factors there. Uh, they will competing factors there. Uh, they will want to say the outcomes they're want to say the outcomes they're want to say the outcomes they're optimizing for because they want those optimizing for because they want those optimizing for because they want those outcomes to actually happen, outcomes to actually happen, outcomes to actually happen, >> but they don't want to fully reveal how >> but they don't want to fully reveal how >> but they don't want to fully reveal how much they value those outcomes. That's much they value those outcomes. That's much they value those outcomes. That's right. right. right. >> And so, I think what's going to happen >> And so, I think what's going to happen >> And so, I think what's going to happen is they are going to say what the is they are going to say what the is they are going to say what the outcomes are because they're going to outcomes are because they're going to outcomes are because they're going to want those things to happen. Yeah, want those things to happen. Yeah, want those things to happen. Yeah, >> but then they're going to, you know, but >> but then they're going to, you know, but >> but then they're going to, you know, but then but then the market's going to be then but then the market's going to be then but then the market's going to be testing like, well, how much are testing like, well, how much are testing like, well, how much are [snorts] you willing to pay for that?

[snorts] you willing to pay for that? [snorts] you willing to pay for that? And in a world where it's not sort of And in a world where it's not sort of And in a world where it's not sort of human to human procurement, but agentto human to human procurement, but agentto human to human procurement, but agentto agent procurement, you could imagine agent procurement, you could imagine agent procurement, you could imagine markets getting like very efficient, markets getting like very efficient, markets getting like very efficient, right? Because um agents are just right? Because um agents are just right? Because um agents are just relentless like they're willing to relentless like they're willing to relentless like they're willing to negotiate as long as they need to negotiate as long as they need to negotiate as long as they need to negotiate. They'll come back again and negotiate. They'll come back again and negotiate. They'll come back again and again. And you could imagine that like again. And you could imagine that like again. And you could imagine that like um you know consumer surplus for example um you know consumer surplus for example um you know consumer surplus for example like like like >> could could be eroded relatively >> could could be eroded relatively >> could could be eroded relatively quickly. quickly. quickly. >> That's really interesting. Okay, one >> That's really interesting. Okay, one >> That's really interesting. Okay, one last question for you. last question for you. last question for you. >> When was the last time an agentic >> When was the last time an agentic >> When was the last time an agentic interaction on stripe? Someone using interaction on stripe? Someone using interaction on stripe? Someone using wallet or whatever it is surprised you wallet or whatever it is surprised you wallet or whatever it is surprised you like this is a new use case. This is like this is a new use case. This is like this is a new use case. This is something I haven't thought about. something I haven't thought about. something I haven't thought about. I mean, I was surprised when people I mean, I was surprised when people I mean, I was surprised when people started using agents to do physical started using agents to do physical started using agents to do physical world things like I people using agents world things like I people using agents world things like I people using agents to provision software like didn't really to provision software like didn't really to provision software like didn't really surprise me. That feels like what agents surprise me. That feels like what agents surprise me. That feels like what agents are. Okay, so now you can have agents are. Okay, so now you can have agents are. Okay, so now you can have agents out buying your physically sending cards out buying your physically sending cards out buying your physically sending cards so that your mother receives it on so that your mother receives it on so that your mother receives it on Mother's Day and like the agent is Mother's Day and like the agent is Mother's Day and like the agent is making the purchase. I guess I guess I making the purchase. I guess I guess I making the purchase. I guess I guess I actually should have expected this actually should have expected this actually should have expected this because humans are intrinsically lazy.

because humans are intrinsically lazy. because humans are intrinsically lazy. But I don't know. I had this mental But I don't know. I had this mental But I don't know. I had this mental model in my head like in person will be model in my head like in person will be model in my head like in person will be in person will be in person and then in person will be in person and then in person will be in person and then there will be this there will be this there will be this >> brave new world where the internet you >> brave new world where the internet you >> brave new world where the internet you know which traditionally was an internet know which traditionally was an internet know which traditionally was an internet built for humans gradually becomes an built for humans gradually becomes an built for humans gradually becomes an internet built for agents or there's a internet built for agents or there's a internet built for agents or there's a separate agent internet and all of the separate agent internet and all of the separate agent internet and all of the digital stuff will be happening there digital stuff will be happening there digital stuff will be happening there and sort of humans will be less and less and sort of humans will be less and less and sort of humans will be less and less involved in the digital and then I saw involved in the digital and then I saw involved in the digital and then I saw these examples of like people using the these examples of like people using the these examples of like people using the agents to like literally reproduce the agents to like literally reproduce the agents to like literally reproduce the physical and my mind was a little physical and my mind was a little physical and my mind was a little >> I don't know if I was disheartened I >> I don't know if I was disheartened I >> I don't know if I was disheartened I think it's a cool company and like you think it's a cool company and like you think it's a cool company and like you know maybe the mother otherwise would know maybe the mother otherwise would know maybe the mother otherwise would not have gotten the Mother's Day card not have gotten the Mother's Day card not have gotten the Mother's Day card but that was just surprising for me like but that was just surprising for me like but that was just surprising for me like leaping into like the real physical leaping into like the real physical leaping into like the real physical world in a way that's world in a way that's world in a way that's >> quite different than sort of like the >> quite different than sort of like the >> quite different than sort of like the B2B SAS AI commerce land I generally B2B SAS AI commerce land I generally B2B SAS AI commerce land I generally think of for for MPP think of for for MPP think of for for MPP >> and I think that's one of the things >> and I think that's one of the things >> and I think that's one of the things that has been really interesting to me that has been really interesting to me that has been really interesting to me is like I get to like talk with the is like I get to like talk with the is like I get to like talk with the audience and I get to hear all these use audience and I get to hear all these use audience and I get to hear all these use cases and the ones that surprise me the cases and the ones that surprise me the cases and the ones that surprise me the most are pretty consistently most are pretty consistently most are pretty consistently >> consumer use cases >> consumer use cases >> consumer use cases >> that come from a different part of the >> that come from a different part of the >> that come from a different part of the adoption bell curve. It's not the like I adoption bell curve. It's not the like I adoption bell curve. It's not the like I live in the valley and I have all of my live in the valley and I have all of my live in the valley and I have all of my max plans and this and that. No, it's max plans and this and that. No, it's max plans and this and that. No, it's it's the people who it's the people who it's the people who >> Silicon Valley can't surprise you.

>> Silicon Valley can't surprise you. >> Silicon Valley can't surprise you. >> It doesn't >> It doesn't >> It doesn't >> We're too old for that. It can't >> We're too old for that. It can't >> We're too old for that. It can't surprise us anymore. surprise us anymore. surprise us anymore. >> Look at the gray hairs. Um yeah, it >> Look at the gray hairs. Um yeah, it >> Look at the gray hairs. Um yeah, it doesn't surprise. And I think that doesn't surprise. And I think that doesn't surprise. And I think that that's what I'm really excited about that's what I'm really excited about that's what I'm really excited about when we look at the next year, the next when we look at the next year, the next when we look at the next year, the next two years, where we're going with the two years, where we're going with the two years, where we're going with the economy and AI. economy and AI. economy and AI. >> There is this tremendous um potential >> There is this tremendous um potential >> There is this tremendous um potential surprise factor in how people are going surprise factor in how people are going surprise factor in how people are going to interact with agents. Yep. to interact with agents. Yep. to interact with agents. Yep. >> And I don't know, like I'm excited to >> And I don't know, like I'm excited to >> And I don't know, like I'm excited to chat again in 6 months and see where chat again in 6 months and see where chat again in 6 months and see where we're at because I think there's going we're at because I think there's going we're at because I think there's going to be a lot more surprises. to be a lot more surprises. to be a lot more surprises. >> Likewise, I think it's going to be >> Likewise, I think it's going to be >> Likewise, I think it's going to be actually six months going to be a very actually six months going to be a very actually six months going to be a very different world. Maybe three. Maybe we different world. Maybe three. Maybe we different world. Maybe three. Maybe we should do three. We'll do three. We'll should do three. We'll do three. We'll should do three. We'll do three. We'll do three. Very good. Thank you. I do three. Very good. Thank you. I do three. Very good. Thank you. I appreciate you coming on. Thank you. It appreciate you coming on. Thank you. It appreciate you coming on. Thank you. It was great to see you. It was great to was great to see you. It was great to was great to see you. It was great to see you.