The Exchanges

Q And so talk to me, Dennis. I want to start today with how a very young Danish man came to found one of the hottest startups in the world and in AI with X.ai. So what's the brief kind of origin story for you?

A Short story is one for where our prior, prior company got acquired by Yahoo, and one part of the set of terms and conditions that comes with that acquisition was that I moved to New York, which I actually pissed very hard against because I wanted to move to the valley and do my next thing out of there, all the good things that comes with that story, but I wouldn't take it, and I moved to New York, and it's just very hard to not fall in love with the city. So I spent the last eight years here, and we built up post Having integrated into Yahoo, a predictive analytics venture out of New York as well that got acquired by Outbrain in 2013. Post integration of that company, we brought the team back together late 20 13 and thought it might just be time to For us to remove this pain of setting up a meeting once and for all by implementing this intelligent agent. And then over the last two and a half years, we've been in the basement in the dark alone, engineering away.

AI assessment note: “brought the team back together late 20 13 and thought it might just be time”

Q have to say, um, but I wanted to, we, we mentioned some of your investors there and, and obviously, uh, investments ramped up massively into the space of AI and AI startups. So I'm intrigued to hear how you approached the fundraising journey tactically then. You've probably raised more funding than most AI startups. So how was your fundraising journey approached tactically, and what were the challenges that you faced?

A So we've been quite honest from the very first days that this would be a long, very hard journey where a lot of people will sit in downtown Manhattan in a basement and not have a product in market for which we could charge. And anybody uncomfortable with anything outside of that shouldn't participate. The pitch was a little bit more slick, but that was the message. That means when we raised our seed round, we pitched nothing, but for these two million dollars, you will get an answer. Not a product, not customers, not revenue, not really much. You'll get an answer, a thumbs up or a thumbs down on whether this is even possible. Worst case, we've only invested two million dollars. And not four years and thirty five million dollars. Two million dollars to know whether is early 20 14 the point for where we should start working on intelligent agents. And remember at that point it was a whole lot less sexy than it is now in 2016. This is before the Microsoft and Facebook and Google announcements. So a little bit early. So it was all about is now the time. That I think set the tone for how we raise capital From that point on forward. The A round where we raised about ten million dollars was on, we believe now through this data set that we can prove that it can be solved, and we need a sum of capital to go engineer the product, and that is pretty much what we did for the A round. Not ge…

AI assessment note: “when we raised our seed round, we pitched nothing, but for these two million”

Q And so talk to me, Dennis. I want to start today with how a very young Danish man came to found one of the hottest startups in the world and in AI with X.ai. So what's the brief kind of origin story for you?

A Short story is one for where our prior, prior company got acquired by Yahoo, and one part of the set of terms and conditions that comes with that acquisition was that I moved to New York, which I actually pissed very hard against because I wanted to move to the valley and do my next thing out of there, all the good things that comes with that story, but I wouldn't take it, and I moved to New York, and it's just very hard to not fall in love with the city. So I spent the last eight years here, and we built up post Having integrated into Yahoo, a predictive analytics venture out of New York as well that got acquired by Outbrain in 2013. Post integration of that company, we brought the team back together late 20 13 and thought it might just be time to For us to remove this pain of setting up a meeting once and for all by implementing this intelligent agent. And then over the last two and a half years, we've been in the basement in the dark alone, engineering away.

AI assessment note: “we brought the team back together late 20 13 and thought it might just be time”

Q for that one. But I want to start with the theme of AI, and through many conversations I've had with LPs, they ask, do you focus on AI, Harry? And I always respond that AI is this enabling technology that powers the verticals we know already. So I'm looking for some validation here, my friend. I want to ask, is there really such a thing as an AI-first company, Dennis?

A I can certainly see startups take advantage of Specific technologies at certain moments in time. I'm not sure I'm a fan of companies using that as an excuse to begin the journey. I think any company should be based on some really solid foundation of pain, something or some set of things that are not working in the universe, and you're trying to fix them. You might then figure out that, hey, there's this particular technology which we can use to fix it faster, better, Cheaper, more optimal. And it happens to be that the pain we're trying to solve or trying to remove is the pain of setting up meetings. And it certainly looked like when we started that the conversational UI and the intelligent agent and thus AI would be a good way to remove it. But it's not one for where we want to be an AI first company. No, we want to be a meeting scheduling company that happens to use AI. So I'm very much in agreement with you that You shouldn't try to be mobile first, or AI first, or cloud first, or big data, or whatever the term might be of the particular era. You should just really let onto that customer pain, because that is the truest thing that you can let onto.

AI assessment note: “we want to be a meeting scheduling company that happens to use AI.”

Q does indeed, but as you know from, from prior interview with me, I do like to do my research, and you mentioned G Suite there with Google, and we've seen the likes of Microsoft take on Slack, and so a question from Dave Chang at DCM, he says, is it a concern for you moving forward to see such large corporates coming after startups so actively in the enterprise space?

A I think you should always be aware of the fact that come tomorrow, there'll be a competitor. And that competitor can be a weak one, or it can be a strong one, but they always arrive. If they do not arrive, I'm not so sure you're doing anything interesting. So when they do arrive, what I hope though, in my optimistic entrepreneurial setting is that I'm so far ahead that it becomes very hard to catch up. So it's not that you can't share your photos with your family on Google plus. It's just that that's too late. As in the world decided that when I need to share pictures of my kids with my mom back in Europe, that is done on Facebook. I'm not spending any clock cycles trying to persuade her to move off Facebook. That is just the standard. I think we have a chance here in turning Amy and Andrew at x.ai into a standard so that it becomes this almost single agent market. That's half a decade for me to get there, but I think there's an opening at least And I'm just trying to run towards that as fast as I can, so that if and when Google or somebody like that arrives, it's a second comer, and it doesn't matter too much. But we'll see. I feel quite confident. I'm quite chuffed, really, about where we're at today. So I don't think about it too much.

AI assessment note: “I feel quite confident. I'm quite chuffed, really, about where we're at today.”

Q have to say, um, but I wanted to, we, we mentioned some of your investors there and, and obviously, uh, investments ramped up massively into the space of AI and AI startups. So I'm intrigued to hear how you approached the fundraising journey tactically then. You've probably raised more funding than most AI startups. So how was your fundraising journey approached tactically, and what were the challenges that you faced?

A So we've been quite honest from the very first days that this would be a long, very hard journey where a lot of people will sit in downtown Manhattan in a basement and not have a product in market for which we could charge. And anybody uncomfortable with anything outside of that shouldn't participate. The pitch was a little bit more slick, but that was the message. That means when we raised our seed round, we pitched nothing, but for these two million dollars, you will get an answer. Not a product, not customers, not revenue, not really much. You'll get an answer, a thumbs up or a thumbs down on whether this is even possible. Worst case, we've only invested two million dollars. And not four years and thirty five million dollars. Two million dollars to know whether is early 20 14 the point for where we should start working on intelligent agents. And remember at that point it was a whole lot less sexy than it is now in 2016. This is before the Microsoft and Facebook and Google announcements. So a little bit early. So it was all about is now the time. That I think set the tone for how we raise capital From that point on forward. The A round where we raised about ten million dollars was on, we believe now through this data set that we can prove that it can be solved, and we need a sum of capital to go engineer the product, and that is pretty much what we did for the A round. Not ge…

AI assessment note: “That means when we raised our seed round, we pitched nothing, but for these two million dollars”

Q So what's your vision then for enterprise? Uh, Carl Louis again at DCM stated that you had a vision for enterprise and how everyone should, should have the assistant. So, so what's your, your vision for enterprise?

A I do think that these agents, including Amy and that of just managing your calendar is a setting for where you go from having a luxury that used to be provided only for the select few into a new setting for where you can completely Democratize it. As in, it is not a luxury. It is something where I don't even want the most junior account manager of ours to sit and do email ping pong to figure out when to talk to the customer. If the customer agreed to speak about this new feature come early next week, then he's on board. Then we just need an agent to figure out exactly when on Monday or Tuesday is this, and what phone number is it, or what's the address, and do we pop by? But that is not the account manager's skill set. His skill set is to know our product, have empathy for the customer, know to tease out whether he truly understands new feature. That's what we hired him to do, not to do meetings, scheduling. So I don't want, and if we can take a good example here, post acquisition of our last company, we were 450 employees, two assistants for about seven executives. That means you have 443 people left To their own devices. Amy is for them, so that they don't have to do this. This is not about making sure that any human personal assistant becomes unemployed in the future. I think all of them will continue to be employed, and I think they'll thrive and do proper work, not email p…

AI assessment note: “where you go from having a luxury that used to be provided only for the select few into a new setting for where you can completely Democratize it”

Q I'm intrigued. We spoke about monetization there. How do you think about a pricing strategy from here? So, so far it hasn't been employed and, you know, people love the product. There's obviously huge scope for monetization. So how do you start to begin thinking of a pricing model from here?

A So we've always had this idea in mind of this being for the masses, not for the few. So we wanted a price that was completely disconnected to the idea of humans being involved in any way, shape, or form. This is a software SaaS product, and if it's a software SaaS product, there's certainly a number of good examples that we can lean up against, whether that be a Dropbox, or a Salesforce, or something similar. I'm very confident you'll see us be in market with three editions. A free edition where you can schedule a few meetings in the name of Amy or Andrew at x.ai. A professional edition for where you can schedule unlimited meetings, tune the signature, and do some of those pro things you would expect, and pay me 29, 39 dollars, whatever that might be a month. I think that'll be the very High price. And that would probably decrease over time. And then you'll see us have a business edition as well for where you can move the whole thing to your domain, buy 89 seats or 450 seats for the whole organization, get one invoice, get an admin and security setting, so you can manage it. And there's nothing new or innovative about what I just explained here, and that's not where we want to be innovative. We think that is a tried and tested setting, and we think our product Fits really well into that setting.

AI assessment note: “you'll see us be in market with three editions”

Q It is indeed, but I want to kick off today, and for those that made the cardinal sin of not listening to our first episode, I'd love to start by how did a native Dane make his way to New York and come to found one of the hottest AI startups of the day in x.ai?

A The very short story in my very short life, given I'm only 46, is one for where I did a few ventures back home in Denmark, then moved to England for a while, then moved to Budapest for a while, where we spent four years doing a analytics venture, and that one got acquired by Yahoo, and part of the acquisition terms were that I moved to New York, and I thought, I'll just spend a couple of years in New York, and that was 10 years ago. Now the kids are in school in New York, and really anything I'll do in the foreseeable future will be from Wall Street, where I live, and I've spent the last 10 years. Now, how did I get started on x.ai? I think it's one of those things, and it's gonna sound like that eBay story, which is kind of not really true, whether they started the whole thing based on that sort of pest dispensers, but you know what? I actually did start this whole venture by going back into my calendar and seeing exactly how many meetings that I did. And as I looked at that, I just thought, a thousand meetings a year, which you sit and schedule yourself in your underwear at 11 p.m. Is that your life, Dennis? And I thought, that can't be my life. There needs to be some sort of escape, and we're working on that escape right now.

AI assessment note: “part of the acquisition terms were that I moved to New York”

Q no longer my life, thanks to Amy, but I do want to ask Dennis, you mentioned there are a couple of your ventures beforehand. Three companies before X.ai is founder and CEO, and COO at another. Success ranges from acquisitions to Yahoo to, as you said, going bust. Which you describe as an expensive MBA. I did love that. But tell me, what did you learn from the success first?

A I like a little bit of extra money in my pocket, but that's not the good part of it. I think the good part is that when you do see success, then you dare to do something slightly more brave next time. As in anything for where you do a 10 mile run in your head. I could do a 15 mile run. Or you do a three K swim. I can do a five K swim. And before you know it, you're swimming the English Channel. So some success breeds bravery to do something even more grand next time. That's certainly what I'm fond of, because if you do five things and none of them are working out, you could end up being just slightly timid in your ideas and your approach and your attacks, and I'm certainly fortunate in having started out this whole journey with a win.

AI assessment note: “some success breeds bravery to do something even more grand next time”

Q Well, I'm so pleased to hear that, and I'll be sending this episode to all our LPs. But I do want to ask then, given that there's a lot of talk about the standardization of tools, libraries, what TensorFlow allows, Et cetera, et cetera, but will this largely remove much of the barriers and defensibility, and how do you think about this as a company that is powered by AI?

A I'm certainly a fan of good, strong libraries, and I'm a fan of things being standardized, and I'm not sure that by those things being commoditized that there's no room for startups. If anything, I actually don't even think that is the biggest obstacle. For me, and if I look at all of the time and energy and money that we spent over the last good four plus years, that went into defining the universe that we exist in. As in, go pick up any self-driving car project right now. It is not their ability to build a model that makes a prediction. It's trying to figure out, what am I even trying to predict here? Am I trying to predict whether there's a pedestrian, or whether there's people, or whether there's a sign, or what am I even trying to predict? And if I'm trying to predict something On what data am I trying to predict it? How do I get that data? If I get it, do I need to label it? If I need to label it, how do I label it? If I do label it, in what application? Because it's obviously not in Microsoft Word, but if not that, where? So there's a ton of challenges before you're trying to kind of craft your first model, and I think many of the models that we've done here, if that's based on some output of a TensorFlow installation, fine, because that is the least of challenges. So no, I don't think that is an obstacle for a good entrepreneur to kind of jump into the pool and play alo…

AI assessment note: “no, I don't think that is an obstacle for a good entrepreneur”

Q no longer my life, thanks to Amy, but I do want to ask Dennis, you mentioned there are a couple of your ventures beforehand. Three companies before X.ai is founder and CEO, and COO at another. Success ranges from acquisitions to Yahoo to, as you said, going bust. Which you describe as an expensive MBA. I did love that. But tell me, what did you learn from the success first?

A I like a little bit of extra money in my pocket, but that's not the good part of it. I think the good part is that when you do see success, then you dare to do something slightly more brave next time. As in anything for where you do a 10 mile run in your head. I could do a 15 mile run. Or you do a three K swim. I can do a five K swim. And before you know it, you're swimming the English Channel. So some success breeds bravery to do something even more grand next time. That's certainly what I'm fond of, because if you do five things and none of them are working out, you could end up being just slightly timid in your ideas and your approach and your attacks, and I'm certainly fortunate in having started out this whole journey with a win.

AI assessment note: “some success breeds bravery to do something even more grand next time.”

Q Well, I'm so pleased to hear that, and I'll be sending this episode to all our LPs. But I do want to ask then, given that there's a lot of talk about the standardization of tools, libraries, what TensorFlow allows, Et cetera, et cetera, but will this largely remove much of the barriers and defensibility, and how do you think about this as a company that is powered by AI?

A I'm certainly a fan of good, strong libraries, and I'm a fan of things being standardized, and I'm not sure that by those things being commoditized that there's no room for startups. If anything, I actually don't even think that is the biggest obstacle. For me, and if I look at all of the time and energy and money that we spent over the last good four plus years, that went into defining the universe that we exist in. As in, go pick up any self-driving car project right now. It is not their ability to build a model that makes a prediction. It's trying to figure out, what am I even trying to predict here? Am I trying to predict whether there's a pedestrian, or whether there's people, or whether there's a sign, or what am I even trying to predict? And if I'm trying to predict something On what data am I trying to predict it? How do I get that data? If I get it, do I need to label it? If I need to label it, how do I label it? If I do label it, in what application? Because it's obviously not in Microsoft Word, but if not that, where? So there's a ton of challenges before you're trying to kind of craft your first model, and I think many of the models that we've done here, if that's based on some output of a TensorFlow installation, fine, because that is the least of challenges. So no, I don't think that is an obstacle for a good entrepreneur to kind of jump into the pool and play alo…

AI assessment note: “I'm not sure that by those things being commoditized that there's no room”

Q DCM to Brad at IA. I wanted to ask because the commonality amongst my conversations with them was your ability to retain your Hi, our founding team, which has been with you since 2013. Obviously, I don't need to tell you how rare that is, but I do want to know, Dennis, what is the secret to not only hiring these managers, but also scaling and retaining them over time?

A So you, of all people, perhaps even more than me, know the few reasons for why many startups end up not succeeding to the extent you had hoped for, and certainly one of them is that the founding team either couldn't figure it out together, Kind of stay together, that we're super happy on day one, a whole lot less happy on day 800. It certainly makes it a little bit easier if you find a team of people that you know you work well with. So of the four founders, where I'm one of them in x.ai, three of us already worked together in our prior venture. So we worked together for, yeah, about 10 years. So we're not surprised by all those kind of little kinks that comes along with being A human being. And we're together because we have specific qualities where there's a little overlap, but not much. As in, I know exactly what Alex is good at. He knows exactly what I'm good at. We know exactly where, Dennis, don't fiddle with that. Just leave it to Alex. And he knows exactly, nah, that's not my responsibility. Dennis will fix it. I don't even, don't even need to look at it. But it doesn't mean it's easy. So any Any startup is just hard. As in, sometimes, surely, you win the lottery, and you stumble into an Airbnb, and if you're Brian, good for you. The rest of us, we just have to plow through a ton of shit years on end, and see if we can get out on the other side. And that is hard. I do c…

AI assessment note: “three of us already worked together in our prior venture. So we worked together for, yeah, about 10 years.”

Q speak, with the schedule, but I didn't expect to go off piece quite this early. But I have to ask, combining both the success and the failure, how do you determine when something's not working, when to follow the vision and the mission, so to speak, versus when to really just realize that you're drowning and fighting a losing game? How do you make that decision and fight the balance?

A That's an awesome question, right? And I'm not sure any one of us have a really good answer. I do think, though, and this is pure opinion, that most people give up too quickly. I certainly believe that there's a real dramatic cost when you give up too late, and you try to kind of find that equilibrium, and I'd certainly much rather err on the side of giving up a little bit too late than giving up a little bit too early, because the whole most likely outcome of any startup is that We will die. But the only way you're going to figure out that you will not die is to take it so far into the future that you find something that was not visible in the beginning, and you have to kind of keep pushing forward. So I would certainly suggest, not based on my prior commentary, that people kind of keep plowing forward. Now the signal, I think, is having at least determined up front what you would agree is some sort of end to this, which could be, uh, No more money in bank, but not necessarily. No more money in bank might just mean that you're not paying yourself any salary anymore. Then it could be not paying myself salary for eight months, or not paying me and my co-founders any salary for eight months, or not paying me and my co-founders any salary for eight months, and not paying taxes for four months, or however aggressive you want to be. But I think the best way to do it is to do it up f…

AI assessment note: “the best way to do it is to do it up front while you're still sober”

Q the application of AI and really the delivery mechanism behind it. For you firsthand, you've experienced That kind of boom of the conversational UI, and we've seen that from Alexa to Google Home. The big question for me here is, is this conversational UI, is it truly a paradigm shift in the way we as humans interact with our devices, or is it really an iterative advancement in our communications?

A So I'm a believer, so anybody who listens to this should expect me to have some bias, but no more bias than tomorrow, And then kind of being mixed in UIs, but let me just take a couple of steps back, so I'm certainly old enough to have some memory of having taken my CS degree on the command line, and that being pretty much the primary UI for any type of access to compute back in the late eighties, and certainly years prior to that, and I certainly also remember the introduction of the graphic user interface, and my mom getting access to word processors, and a whole lot of other people who didn't have kind of A specialized education, understanding how to use computers, and I certainly also remember the mobile UI and my kids getting access to touch interfaces, and that probably being a distinct UI in its own right, and having said that, I do believe that the conversational UI is a distinct UI in its own right, and I think it's very important to kind of make a distinction between a UI and what happens on that UI. The conversational UI does put different demands on the user. If anything, I actually think it's Might finally fully democratize access to compute so even the least tech-savvy person can now figure out what the time is in Singapore without knowing how to boot up a browser and go to Google and search for it. But I can actually just ask some assistance. Though I am very bul…

AI assessment note: “I do believe that the conversational UI is a distinct UI in its own right”

Q Absolutely. No, I do agree with you, but, but staying on the product foundation that we started on there, I'm intrigued now. How do you package a business edition of the product when the core utility really is quite user focused?

A I think that's a good, really good question, and we've thought a lot about that, and I think really the industry go no further than chat to the guys over at G Suite at Google or Office at Microsoft, where what they have is a single user focus, as in you need Harry to put a financial model in place for a potential investment that you're making, and you will open up Excel, and you will sit on your laptop, and that'll be that. The future Seems though, to be one for where we'll collaborate more and more. And that means that we probably need to figure out that these individual tools, how do we make them team oriented? And if you look at what Amy is doing, which is that Harry and Dennis needs to meet up on Skype on this Friday morning, at least in my end. And that is something for where she helps me. And in that regard, how do we then move into a team setting? And that is not always easy. And for some, they go no further than just saying, hey, oh, Our team edition is just you pay more, as in you've got more seats, and then if it's 17 seats, I multiply the price by 17. If it's a thousand, I multiply by a thousand. I'm not so sure that you've exhausted your imagination if that's what you get to. So, I think we've done or thought about this on three dimensions, really. One, there is the absolute necessary requirement to put some admin in place to manage a team. This is non-sexy stuff, b…

AI assessment note: “thought about this on three dimensions, really. One, there is the absolute necessary requirement”

Q And then the one from Brad at IA, how did your upbringing in Denmark impact you as a CEO?

A I think, or at least this is what the economists tell us, there might be some truth to it, that we are the happiest country in the world, and I think we've been that four out of The last five years, and there's some level of enthusiasm back home. I'm not sure where it's coming from. It's certainly not the high taxation, but there is something in the water, and I think that unbridled enthusiasm, which I have, serves me well. So even when it gets dark, I tend to be quite optimistic to that last moment where, but Dennis, there's only 17 dollars in bank, and we've got payroll of 850 K Come 10 days. Yeah, but I'm sure this is going to work out. Here's what I think we should do. And I'm just generally just very optimistic. So I think that would be the one thing I would take away from that 28 years there.

AI assessment note: “I think that unbridled enthusiasm, which I have, serves me well.”

Q Absolutely. No, I do agree with you, but, but staying on the product foundation that we started on there, I'm intrigued now. How do you package a business edition of the product when the core utility really is quite user focused?

A I think that's a good, really good question, and we've thought a lot about that, and I think really the industry go no further than chat to the guys over at G Suite at Google or Office at Microsoft, where what they have is a single user focus, as in you need Harry to put a financial model in place for a potential investment that you're making, and you will open up Excel, and you will sit on your laptop, and that'll be that. The future Seems though, to be one for where we'll collaborate more and more. And that means that we probably need to figure out that these individual tools, how do we make them team oriented? And if you look at what Amy is doing, which is that Harry and Dennis needs to meet up on Skype on this Friday morning, at least in my end. And that is something for where she helps me. And in that regard, how do we then move into a team setting? And that is not always easy. And for some, they go no further than just saying, hey, oh, Our team edition is just you pay more, as in you've got more seats, and then if it's 17 seats, I multiply the price by 17. If it's a thousand, I multiply by a thousand. I'm not so sure that you've exhausted your imagination if that's what you get to. So, I think we've done or thought about this on three dimensions, really. One, there is the absolute necessary requirement to put some admin in place to manage a team. This is non-sexy stuff, b…

AI assessment note: “we've done or thought about this on three dimensions, really. One, there is”

Q very much like my recording patterns. Um, but, but I mean, I want to start then by discussing that kind of the dark basement and the training by discussing the core technology behind Amy and Andrew. We can't forget Andrew. So what did the training look like? How, how long did it take? Uh, and how much data have you kind of painstakingly gone through to, to build accurate patterns?

A That's a fantastic question. And I think a lot of people who today, June, 20 16 decide that the solution to their pain is some intelligent agents will be surprised about the amount of effort they have to put into both their definition of this Conceptual model of this universe in which they exist, and that of assembling the training data set. Those two things seems like, sure, it's a whiteboard session to define the universe in which we exist, and it's a couple of weeks on getting just some raw data, and then we start kind of engineering and train the models we want to put in place, but that is just not the case. I think of all the things that I've been surprised about, and any startup is just one big surprise, is the magnitude of the work which you have to apply to those two. Something as Simple as defining time, as in Amy's understanding of you saying Wednesday, one o'clock, June something, EST, daylight saving. The sad thing is that that's just not how you speak, so however you say that, I need to be able to translate that into a true time. That alone, we probably spent a year and a half coming up with the right model, because humans will say stuff like, we'll meet up later, this afternoon. Upon my return. Before the demo. They're all items where you're trying to say something about when you and I should meet. So that is massive. Just defining it. You haven't even collected t…

AI assessment note: “That alone, we probably spent a year and a half coming up with the right model”

Q the application of AI and really the delivery mechanism behind it. For you firsthand, you've experienced That kind of boom of the conversational UI, and we've seen that from Alexa to Google Home. The big question for me here is, is this conversational UI, is it truly a paradigm shift in the way we as humans interact with our devices, or is it really an iterative advancement in our communications?

A So I'm a believer, so anybody who listens to this should expect me to have some bias, but no more bias than tomorrow, And then kind of being mixed in UIs, but let me just take a couple of steps back, so I'm certainly old enough to have some memory of having taken my CS degree on the command line, and that being pretty much the primary UI for any type of access to compute back in the late eighties, and certainly years prior to that, and I certainly also remember the introduction of the graphic user interface, and my mom getting access to word processors, and a whole lot of other people who didn't have kind of A specialized education, understanding how to use computers, and I certainly also remember the mobile UI and my kids getting access to touch interfaces, and that probably being a distinct UI in its own right, and having said that, I do believe that the conversational UI is a distinct UI in its own right, and I think it's very important to kind of make a distinction between a UI and what happens on that UI. The conversational UI does put different demands on the user. If anything, I actually think it's Might finally fully democratize access to compute so even the least tech-savvy person can now figure out what the time is in Singapore without knowing how to boot up a browser and go to Google and search for it. But I can actually just ask some assistance. Though I am very bul…

AI assessment note: “I do believe that the conversational UI is a distinct UI in its own right”

Q DCM to Brad at IA. I wanted to ask because the commonality amongst my conversations with them was your ability to retain your Hi, our founding team, which has been with you since 2013. Obviously, I don't need to tell you how rare that is, but I do want to know, Dennis, what is the secret to not only hiring these managers, but also scaling and retaining them over time?

A So you, of all people, perhaps even more than me, know the few reasons for why many startups end up not succeeding to the extent you had hoped for, and certainly one of them is that the founding team either couldn't figure it out together, Kind of stay together, that we're super happy on day one, a whole lot less happy on day 800. It certainly makes it a little bit easier if you find a team of people that you know you work well with. So of the four founders, where I'm one of them in x.ai, three of us already worked together in our prior venture. So we worked together for, yeah, about 10 years. So we're not surprised by all those kind of little kinks that comes along with being A human being. And we're together because we have specific qualities where there's a little overlap, but not much. As in, I know exactly what Alex is good at. He knows exactly what I'm good at. We know exactly where, Dennis, don't fiddle with that. Just leave it to Alex. And he knows exactly, nah, that's not my responsibility. Dennis will fix it. I don't even, don't even need to look at it. But it doesn't mean it's easy. So any Any startup is just hard. As in, sometimes, surely, you win the lottery, and you stumble into an Airbnb, and if you're Brian, good for you. The rest of us, we just have to plow through a ton of shit years on end, and see if we can get out on the other side. And that is hard. I do c…

AI assessment note: “three of us already worked together in our prior venture”

Q speak, with the schedule, but I didn't expect to go off piece quite this early. But I have to ask, combining both the success and the failure, how do you determine when something's not working, when to follow the vision and the mission, so to speak, versus when to really just realize that you're drowning and fighting a losing game? How do you make that decision and fight the balance?

A That's an awesome question, right? And I'm not sure any one of us have a really good answer. I do think, though, and this is pure opinion, that most people give up too quickly. I certainly believe that there's a real dramatic cost when you give up too late, and you try to kind of find that equilibrium, and I'd certainly much rather err on the side of giving up a little bit too late than giving up a little bit too early, because the whole most likely outcome of any startup is that We will die. But the only way you're going to figure out that you will not die is to take it so far into the future that you find something that was not visible in the beginning, and you have to kind of keep pushing forward. So I would certainly suggest, not based on my prior commentary, that people kind of keep plowing forward. Now the signal, I think, is having at least determined up front what you would agree is some sort of end to this, which could be, uh, No more money in bank, but not necessarily. No more money in bank might just mean that you're not paying yourself any salary anymore. Then it could be not paying myself salary for eight months, or not paying me and my co-founders any salary for eight months, or not paying me and my co-founders any salary for eight months, and not paying taxes for four months, or however aggressive you want to be. But I think the best way to do it is to do it up f…

AI assessment note: “do it up front while you're still sober and sane”

Q And then what about the next five years for you and x.ai? Obviously, we're going to dominate the enterprise, but what is it specifically in the roadmap for you?

A Anybody listening to this still with us should pull out their phone, jump into their calendar app, look at that 17 events they have on the calendar for next week. If I didn't put them there, then I haven't done my job yet, so I am hunting that moment for where, when you look at your calendar, it was Amy or Andrew who added those events, and there are so many things that I need to do to make sure that that turns out to be true, whether that be That's the title of the event is so solid that even if you could add it yourself, you don't want to, that the address is so succinct that it is always mappable. That's the people you meet with, the bio which I provide in the event, it's just solid enough where you're never confused about who you meet with. That her ability to reschedule a meeting, even though that is super fucking scary, is so good because she knows the network that you should probably just reschedule that one-on-one With a team member, because you've got a customer meeting instead. So everything which I can do around the event, that is what we are chasing. Not 15 other things. We just want to be the guys to schedule meetings, but do it really well.

AI assessment note: “We just want to be the guys to schedule meetings, but do it really well.”

Q Absolutely, and so you should be. But, but moving back to the product and kind of the entry into the enterprise space, uh, naturally as a VC now, uh, I have a slightly, the Uh, more strong inclination towards monetization. So how do you evaluate the thought process then behind the, the move into enterprise and the pricing of a business product as a result?

A Again, I'm just using Dropbox as an example. I'm a huge admirer of what they built over the years. Even with the last few years where people kind of fell out of love, I kind of rooted for them. And then what happened, say over the last couple of quarters where people figured out that they're doing a billion dollars annual recurring revenue, uh, Everybody's fallen in love again. I think they built a splendid business. What I do think though, if you really ask them, hand on heart, should you not have launched that business edition a little bit earlier, they would say yes. I'm sure there'll be no interview for where they will say that, but hand on heart, alone, in the dark, I'm sure Drew would agree to the fact that, hey, we should not have waited that long or given away that part of the market to Box or somebody else. We knew from the get-go that Or at least assume from the get-go that the vast majority of our revenue will be derived from the enterprise. So for us, it's been a drive towards launching the business edition, and it's been a focus on how do we best package this business edition, not the single utility. That was a stepping stone to where we will be on Tuesday at 11 o'clock for where we're going to launch that, uh, business edition. So for us, it was not one for where, hey, this is going to be An interesting change. It's been a set of design thoughts, product design th…

AI assessment note: “So for us, it's been a drive towards launching the business edition”

Q Absolutely, and so you should be. But, but moving back to the product and kind of the entry into the enterprise space, uh, naturally as a VC now, uh, I have a slightly, the Uh, more strong inclination towards monetization. So how do you evaluate the thought process then behind the, the move into enterprise and the pricing of a business product as a result?

A Again, I'm just using Dropbox as an example. I'm a huge admirer of what they built over the years. Even with the last few years where people kind of fell out of love, I kind of rooted for them. And then what happened, say over the last couple of quarters where people figured out that they're doing a billion dollars annual recurring revenue, uh, Everybody's fallen in love again. I think they built a splendid business. What I do think though, if you really ask them, hand on heart, should you not have launched that business edition a little bit earlier, they would say yes. I'm sure there'll be no interview for where they will say that, but hand on heart, alone, in the dark, I'm sure Drew would agree to the fact that, hey, we should not have waited that long or given away that part of the market to Box or somebody else. We knew from the get-go that Or at least assume from the get-go that the vast majority of our revenue will be derived from the enterprise. So for us, it's been a drive towards launching the business edition, and it's been a focus on how do we best package this business edition, not the single utility. That was a stepping stone to where we will be on Tuesday at 11 o'clock for where we're going to launch that, uh, business edition. So for us, it was not one for where, hey, this is going to be An interesting change. It's been a set of design thoughts, product design th…

AI assessment note: “We knew from the get-go that... the vast majority of our revenue will be derived from the enterprise.”

Q That's fantastic. And then, and then I want to finish on the next five years for you and for Amy and x.ai. Where's the roadmap going?

A Here's the thing. There is a little bit above ten billion formal meetings being set up in the U.S. alone Every year. Not date nights. Not walk the dog. Formal meetings. You and I talking on Skype this morning. Oh, Chasing those. No more, no less. I want you to get to a point later in life where you look back and say, hey, ha, fuck, we did email ping pong for 30 years. Then these dudes came along and solved it, and it disappeared, and I can't even remember why would we do this? Why would I suffer through email in and out to manage that calendar? We should have done that agent decades ago But here we are. So that's what I'm chasing. That's us.

AI assessment note: “So that's what I'm chasing. That's us.”

Q those, uh, big corporates, I've got a question now from, from one of your investors in, in Amy and X.ai, and that's David Cheng from DCM. Uh, and he says, oh, still on the strain of this corporate, David asks, what is your take on how AI startups will interact with the giant corporates, also investing heavily into AI research, meaning your Googles, Facebooks, Microsofts. So, so what's your take?

A I think there is so much work to be done in traditional productivity. So the way you and I have worked pretty much since we got out of college is in a setting for where software assisted you and I in doing a job a little bit faster A little bit more accurate, whether that is playing around with a data set in Excel, and you create a pivot table, or whether that is submitting your receipts, or creating an invoice. All of that is somewhat done on software, but you, Harry, are doing the job, not the machine, and this new setting is a setting where you'll hand that job over to an agent. You won't create the invoice. You'll tell Stephen, The invoicing agent. Can you go create an invoice for Nike on 19,000 dollars for these three services? And then it's up to the agent to kind of figure out who's the customer, what are these services, how do we create an invoice, how do I send it to them, and so on and so forth. And this new setting will have a massive positive impact on productivity. And I don't see the Enterprise, take that on. That is, we've seen that play out already. So it used to be that any massive enterprise would go create their own ERP system. Nobody does that today. You somehow decide whether you want to use SAP or Oracle or Microsoft, and you use one of those platforms. Anybody in the past will be setting up and engineering their own CRM system. Nobody would do that today.…

AI assessment note: “I don't see the Enterprise, take that on.”