 Hopefully, you're all here for starting and selling an AI company. Ramesh is in cohort eight and we're going to hear from him how he started and sold his AI company and what he looks for is an investor now when he's investing in AI startups and other types of companies. But Ramesh, before we start, just tell us a little bit about yourself. So let us know kind of prior to you starting your startups and selling them. What you've been up to. Sure. So I grew up and I was born in India. I grew up there, moved to the US for my university when I was 18, studied computer science at the University of Illinois, which is sort of in the middle of the US. And I really, really enjoyed computer science as a university student. But I also did an internship at Microsoft when I was in university. And that's when I realized I didn't want to be doing computer programming full time. I liked technology. And so after I, but I wanted, I was always more interested in the big picture. And also at that time, instead of being a full time computer programmer, meant you were working on a very small part of a bigger project. So the, so sort of the impact a single programmer could have, especially in a bigger company like Microsoft, was pretty minimal. So I wanted to have a bigger picture view of things. So I moved, so I decided, okay, let me, let me try. Some friends of mine were in technology consulting and consulting. Let me try my hand at consulting. So after graduating, I landed up working for a large management consulting company in New York, very quickly realized that was not for me. Because in that environment, if you were, if you came up with your own ideas, you were considered naive. You were considered smart and sophisticated. If you could criticize other people's ideas and come up with all the different ways, you know, something could go wrong. So that was not the right environment for me. This was the, this was the mid 90s. Some of my friends who had moved to Silicon Valley were, had become, were working on super exciting things, super exciting projects. So I sort of got bitten by the startup bug quite early on. So in the, in the late 90s, I landed up moving to the valley, landed up joining a friend startup as employee number eight. And since then for the last 25 years or so, I've just been doing startups either, either as an entrepreneur or an investor. Nice. No, that's good. Yeah. So what we're going to do team is we're going to have roughly sort of 30 minutes of presentation from Ramesh and then we can do the Q&A. But if you've got questions as we're going through, feel free to pop them in the chat as you are thinking of them so that you're a member and Ramesh may cover them during the conversation or we'll go into it at the end. Ramesh has sort of said he can go down deep in three different types of areas. So we can either go down deep into what he invests in. He can actually go quite technical in AI if there's any questions down there or actually in sort of challenges and, you know, how it's, what it's like to actually work in a startup and be a tech startup. So Ramesh over to you. You can see your screen and we'll get going. Okay. Can you, can you also, thank you, Mr. Screen. Okay, good. So I was, as Michelle said, I was thinking of going over the story of my last startup, which was called PowerQuit Labs. The different approaches we used, the challenges we faced. And then go in, go more into the general challenges. Startups, any startups building AI applications will face. After that, I was thinking of covering some of the investments I've made. I picked three AI companies that I've invested in recently. And I thought I would go over why I invested in those companies. And as an early stage investor, what I'm looking for in startups, in most specifically in AI startups. So quick overview of my last startup. It was called PowerQuit Labs. And we had a fairly, you know, grand vision in terms of extracting structure and meaning from humans, from human communication. So typically human communication is unstructured and to sort of extract something structured out of it is a challenge. A lot of community, a lot of unstructured communication is sort of just lost, unaccounted for. So we were, our goal was to bring up, bring about structure. And the first application that we were building here was to do this, to apply this for email. And the product was called Inbox Voodoo, which was basically an email assistant that would, an AI powered email assistant. So one of the challenges I had already been investing fairly actively before starting Inbox Voodoo. And one of the challenges I faced as an investor is that I would get, I would get a lot of emails, a lot of emails from entrepreneurs. And quite often I would get these long rambling emails and there would be two or three lines in those emails which were, which were relevant, which would basically tell me something like, I'm an entrepreneur, I'm closing, I'm raising, this is my startup, this is how much I'm raising and this is when my round was closing. And this is how much I'm looking for. So what we tried to do was actually try to take the, try to extract the intent or the ask of the email and summarize that and show that in a way that was easy for the recipient to digest. So that was, that was the essence of what we were doing, extracting the ask in an email. So before we were trying to understand the intent of the, of the sender. I'll get into that in a little more detail. We were a team of, we were a tech heavy team, eight people, five, five engineers. Only one of them had deep sort of NLP expertise. Duane, he had a PhD in NLP. The rest were actually a team of fairly generalist engineers who had the ability to pick up things quickly. We had a data analyst and, and I had a product as well. Looking back, if I could do things with a team, we had a phenomenal technical team. Looking back, if I could do things differently, we would, we would get, we would sort of invest more in people to help us with user acquisition and growth. So I was the, I was the only, I was the only person handling that. And there was a whole lot of things that we did not do fast enough with, with growth and with experimenting with user acquisition growth that I would do differently today. If I were to start this, if I could sort of move the clock back. Small team, but I, I love this quote from by, by the CEO of Twilio, which, which talks about how, you know, a small cohesive team can really do a lot of damage. So, and in terms of ParaQuit, I spoke about, I briefly covered Inbox Voodoo, which was our email assistant. Before that, we tried a bunch of other products on top of our platform. So we had built, we had built this platform, a natural language processing platform, which was capable of extracting relevant text from, from a large corpus of text. So the, in fact, the first product that we built on this platform was called Intent Pulse, which was, which was a product to get your status update highlights in your Facebook and Twitter feeds. I was, I was getting a lot of, as a, as a, as a, as a Twitter user and a Facebook user, I was beginning to get a lot of useful information in my Twitter feeds and in my Facebook feeds, but it was quite often just drowning in all that noise in that information. And that they'd be like one or two, out of a hundred updates, they'd be like one or two updates that were relevant and time sensitive that they didn't want to miss. And those would be just lost in the noise. So the, our first product was geared at trying to find the status updates that mattered. One, one, one of the sort of the catalyst to start this was I'm a big fan of tennis and a couple of friends of mine were at a tennis tournament. And they, I saw that later on Facebook after the event, after the fact. And I asked them, why didn't you tell me about it? And they were like, dude, we posted it around Facebook. So it seemed like, you know, that the growing trend was as, was just to broadcast all this information. So I thought it would be very, very useful to build something that would save me time, not have to go through a lot of, you know, a lot of noisy updates to find the ones that mattered. So this was, this was one product we experimented with. And basically what we found out was these updates that were about entities that were about products. People were using events that were going to destinations they were traveling to books they were reading. We found categories and those at least as founders, we found categories in those updates and those categories more interesting. And my philosophy has always been if given the size of the internet today, if there's something that's useful for you, there's a very good chance it can be useful for 100,000 other users. And worst case, I mean, you have, you have, you have, you know, product market fit for at least one person. So that has been my sort of philosophy in general in terms of building products. If I can build something that's useful for me and my 20 closest friends, that's a good start. But here it did the, this was, this was, did not work in this instance with intent pulse. It turned out that the overlap of users who cared enough about their Facebook and Twitter status updates, but not enough to check it, you know, all the time was very, very small. So the intersection was just not big enough for a full-fledged product. And then we experimented with a product for book recommendations. It's the, the status updates we found that were about books, people discussing books, especially on Twitter, were very, were a very, very high quality. We experimented with the book recommendations product that grew to a few 100,000 users, got a lot of great sort of initial, got a great initial press coverage, et cetera. But still that was also not enough for us to build a standalone company in the space. So one takeaway from our experience with Book Vibe was that, you know, just because you get very, you know, super positive feedback from an early set of users, from the early adopters, does not mean you can create a standalone successful consumer product. Especially in a category like books. So the third product we tried, so we were, we were quite experimental in trying to figure out what worked on our NLP platform, the application that worked. The next application we built was Inbox Voodoo. And that was the application we eventually landed up getting acquired for by Google. So what was Inbox Voodoo? I think I covered that as a personal email assistant. We would prioritize emails. We would summarize emails. And we also had features that would sort of make dealing with email a lot more efficient by providing responses, smart replies that could sort of quicken, quicken responses. Yes, no, see you there. Confirm that type of stuff. We would, this is a different types of emails. Happy to jump into this in more detail. If any of you have questions. And our longer term vision was to be a full-fledged, intelligent assistant. So we're starting with email to go into, to go into being a much smarter, all-encompassing assistant. Had lots of very, very difficult tedious problems to deal with dealing with language because people express intent in a lot of different ways. So dealing with that was a huge sort of, was a very, very ugly, hairy problem at times. Brief overview of our tech. We try, we experimented with a bunch of approaches. We started off with an old school rules-based technique which was easier to tweak, easier to read, but also required manual intervention. And we also experimented with deep learning. It was the early days of deep learning in 2014, 2015, which is all the sort of rage back then. For us, we realized that our old school rules-based technique worked a lot better than a more sort of, glamorous deep learning approach. The sheer amount of data that we needed to get deep learning working was just not practical for us at that time. The key lesson from that was we actually landed up spending more time than we should have trying deep learning. In the end, we settled for a hybrid approach with taking some aspects from deep learning, but the rules-based approach was what predominantly worked the best for us. The other sort of key lessons learned from this was, sounds obvious, but have very, very clear product focus in terms of who you're sort of building this product for. We were, especially with something like email, which has a ton of, you know, different, a ton of nuances. Looking back, we should have probably picked a specific category, as opposed to coming up with a general-purpose email productivity tool. One company that had a lot of success in this space was a company called RelateIQ, which built an email, which built a product. An email was a part of that, sort of just for salespeople, and they were acquired by a sales post for about 400 million. And technically, it wasn't a very sophisticated product, but they sort of really nailed the use case and also had an amazing UX as well for salespeople, really understood that domain really well and built something very, so we were sort of trying to be more of an all-purpose, general-purpose productivity tool, which was much harder in terms of just the sort of the use cases we had to deal with for an email. And then measuring improvement, especially when you're dealing with something like language, it's very, very hard to get it perfect. And it can be frustrating when you sometimes, when we would see that sometimes it was getting what seemed like very obvious detections, flagging emails that were very obviously not actionable as an actionable email, getting false positives. So it can be demoralizing for the team after months and months and months of working if you're not sort of, if you come back with what customers perceive to be basic mistakes. But just look at Siri. Siri still after hundreds and hundreds of millions of investment still can get some, sometimes can get basic questions right. So for a startup, it's critical to sort of measure how you're improving and to see that so that it does not continue to become the sort of massive, unwieldy problem that just keeps growing and growing. So we did, we incorporated that, but we should have really focused on measuring improvement in the sectors that mattered for us. We tended to use a more sort of academic way to measure improvement, something called the F1 score, which looks at in machine learning, which looks at precision and recall. Whereas what we should have looked at was really what sort of constitutes improving in the sector that we're targeting. How accurate is it for our user group in terms of picking up actionable emails? Some of the challenges we faced was accessing to build any machine learning system. You need training data, and accessing training data is very easy if you're building, say, a classifier for web pages, where you have billions of publicly available web pages. We were building something for email, and email data tends to be very private, quite sensitive. So it is very, very hard to acquire email data. We landed up initially training our corpus, based on Enron data. I don't know how many of you are familiar here with Enron, but Enron was the massive company, the energy company in the US, was this massive scandal, and then their emails were deemed to be public domain. So a dream for machine learning scientists to be able to look at those emails, and we were actually, we went through those emails, had a team, and we were labelling those emails. This is an urgent email, this is a non-urgent email, which was great for a start. But one challenge was it was not the most relevant dataset for our users. Enron users were these people at Enron. This was mostly the emails that were sent by the senior management team, Harvard MBAs using correct, very sophisticated English. And there is our initial user set was not that technical in their use of the English language. It was more colloquial, more real, so it was not the most relevant user set to train our corpus on. So we landed up later on training it on internal company emails, on our own emails, and that worked a lot better for our target market. And the whole, and this is something that most startups underestimate, getting clean label data is extremely tedious. This was, nowadays we use mechanical Turk, which provides people who can also label data once we started getting a larger email dataset to be labeled. Once all the sensitive information was stripped off, we would use some of that to label the label data. But it was still, we had to spend a lot of cycles getting the process in place. Nowadays it's much easier. There are many companies that provide labeling as a service. So that lowers, this is a challenge. But for us, a few years ago, this was a massive challenge. Coming back to the general challenges, even getting access. So email, getting access to email data, in our case, was a challenge initially until we started training it on our own emails. But it also can be very, it can also set up, so it's a double-edged sword. If getting access to data is very, very easy, that means the barriers are also a lot lower. Any startup can do it. There's no sort of data mode. Whereas for us, once we got access to the data and built something, it was actually quite a barrier to entry as well. So it's a double-edged sword. Getting access to a unique data set can be hard, but if you do get access to that, it can be quite powerful. And it's a barrier to entry for other startups. AI learning techniques are now getting commoditized very, very quickly. So something that can differentiate your startup or a startup you invest in from the rest is access to proprietary or unique data. And most important, the product still, no matter how bleeding edge or how advanced your AI is, most end users could care less about what's behind the scenes AI. You still need to really focus on making the product intuitive and easy to use. As an investor, I get a lot of pitches from companies, from founders, that really hammer in the point that their AI is so advanced. But if they don't have a product end of the day, that's useful and easy to use. It's not going to succeed no matter how advanced or sophisticated their AI is. So that was sort of a quick overview of my lessons learned from Parakwit. We can go into question-question, now, and then I can jump into some portfolio companies or I can jump into some of that now. Yeah, do you want to have any questions on sort of the learnings of an AI startup at this point then? OK, so a few of my recent investments in AI startups. So before that, a quick sort of overview of the types of AI startups. So there's core AI startups, core AI startups would be something that actually helps the deployment of the AI itself, building the AI itself. So for example, a startup that may provide tools that help you with labeling, that help you build the AI in the first place. Then there are application AI startups. So we started off a bit like an application AI startup. We had tech, a platform that could analyze text and it could be applied in different ways. Applied in different sort of industries, we build different sort of products on top of that. We built a product for book recommendations, as I mentioned, we built a product for prioritizing emails. And then there's industry AI startups which are applied to AI for a specific business use case. So I would say 90% of the startups that are out there are typically industry AI startups. AI for this, sorry, AI for a specific use case. Yeah, I'll quickly go over some of my recent focus as an entrepreneur as well, before getting into some of my portfolio companies. So I'm also the founder of a company that does, that's building sports AI solutions. And we have built, using computer vision, we have built technology that can do event detection broadly speaking. So we have a product for tennis which quickly condenses a two-hour tennis match into 20 minutes the actual time that's spent playing points. So essentially offering tools that typically pro athletes have access to amateur athletes, but pro athletes can watch the highlights of a match because it's shown on TV or it's recorded by someone. So we can offer similar tools to amateurs. We've also built products for ruin and we're experimenting with the product for golf as well, which essentially from a one hour session on a driving range, we'll sort of condense that into 30 swings with just the swings without all the wasted time that goes on. And also we wanna just use your smartphone to do that as opposed to using sensors, hardware or expensive launch monitors. So that's what I'm working on now as an entrepreneur. We do have a question from Facebook that was taken actually. So can you tell us how you went about the financing and pricing strategy for consumers? Right, so our strategy was we had a free version of the product and then we also had a paid version, which was for small businesses and teams. So it was basically a freemium model, yeah. Thank you. So yeah, so this was, which was quite tricky for us because in terms of getting, coming up with a model here for the as a productivity tool, it was quite difficult, especially we found it a challenge to get people to pay 10 bucks for a tool that would boost their email productivity. Companies have now, companies like Superhuman seem to be pulling it off now, which is amazing, but we had a hard time convincing people to pay 10 bucks a month for an email productivity tool, which is why looking back, I would have targeted a more specific verticals such as email for sales professionals, where it's easier to justify charging businesses or sales professionals 30 to 50 bucks a month if it can help them, if it can help, it's a product that can help salespeople close more deals. Okay, some companies, some of my recent investments in AI startups, one of them is a company called Kamua, which does AI based video editing for market, for marketers. So one really sort of simple but interesting use case is all these brands that have built, that have created videos on YouTube, and the formatting just does not work, that does not work on Instagram stories, for example, these new channels. So using the very simple, so editors have to spend a lot of time porting it, getting it to work on a different channel. So using that tool, it's very, very, it automates a lot of the painstaking effort and as a video editor, more of your time is freed up to work on the fun creative stuff as opposed to the nitty-gritty stuff. So this is, I was quite impressed with their, with the founder, a deep experience working with brands, knew the domain well, teamed with great technical chops and they had really, really good early traction amongst freelancers. Had they run a company before? Like what other characteristics had they done that made you invest in them? In them. So the founder had not run, was the first time founder, but he had a ton of sort of management experience and experience building teams from scratch at some of the bigger companies it worked for. So, and the, and his co-founder had experience working at startups before. Yeah, does that answer the question? Okay. Yeah, that's fine. Okay, okay, okay. Headliner is another company that I invested in. This is in the podcasting space and they make it really easy to select relevant clips from a long podcast with the tool targeted at publishers. So one reason I've not used podcasts as much as I probably should is because it takes a long time to find that one clip you're interested in. There's no sort of, still there's no real good preview for a podcast and that's what these guys are doing. And it's also what they do that's really cool is you're on a webpage reading an article about say investing in startups. They'll pick a relevant clip from a podcast in that article. So really enhancing the user experience and a tool and a way for both the podcasters and publishers to monetize better. Here, this was one of the, why did I invest in this company? I was an early investor. In fact, the first investor and the founder is previous company which had a successful exit. He sold the company. It was a content recommendation company. So he's been, he's been immersed in the space for a while and he knows what it takes to sort of take a company from inception to a successful exit. This was a company sold to IAC for two years after he started it. So he's had that first successful exit and now he's actually swinging for the fences looking for something really big. So AMP control is a recent investment in a very, very exciting space. And so I think an overlooked space, some simple but valuable using AI powered EV charging software to optimize EV charging. Once the founders really worked for an automobile company before understands the domain really well as a really, really good team, really has put together a really good tech team as well. So in terms of my investment criteria, it's very hard to predict what's going to work and before once you launch something, once you get, you can do all the analysis before but once you actually get something out there, you'll always learn something new. So I think it's very, very important to be experimental in figuring out what's going to work. But you also, in the first place, you also need to be product thoughtful. So you're not just taking a lot of random bets all the time. The other factor is, I think founders always underestimate how difficult and tough user acquisition is today. I think I touched on that in terms of you may have all the AI in the world but if you don't have a credible plan for user acquisition, that makes it very, that's a big red flag. It's very hard to predict that as well but at least you should have a plan. And if you have done some sort of experiments that tell you this is how much you're spending to acquire one user and this is how much we expect to make off of one user, that is a lot more compelling to me than market sizing and sort of the overall sort of macro look. So I tend to prefer more of a bottoms up approach as opposed to something that's like a GDP forecast that tells you we'll get 1% of this mark of a $10 billion market. How are you going to get your first thousand users and how are you and how much are you going to spend to get those users? And then I think the most important factor is founders. You really want to go for something meaningful in terms of the outcome that they want to create and to make sure it sort of makes sense as an investor because startups are risky and if in the best case, if let's say I'm coming in at a $5 million valuation entry valuation and if in the best case the founder is going to take the first $10 million offer that comes their way, it's a 2x return for me in the best case as an investor and that quite often does not make sense. So you want founders who are sort of going to make a real attempt at going for something meaningful. I think that's most of it. I have a bunch of other slides, more technical slides in the appendix and I'm happy to jump into questions now. No, that sounds great. Thank you very much. That was fabulous. If you unshear your screen then, and then people can see you and we can start unmuting and feel free to ask any questions about whether it's the investment criteria a bit more. I have a few there myself. The actual AI in depth stuff or actually startups in general in the challenges faced. I have a question if I can jump in in front of everybody. Thank you, Ramesh. A really great presentation. My name is Chris Rodley. I'm from Snap IT. I noticed that you've designed, in the case of 40 live, you've designed an AI to solve a specific, really clear targeted outcome and then you also have the rowing startup. So I imagine you're taking a similar tech stack and just reapplying that into another entity. Is that correct? Yes, yes. So good. So it's a strategy that we're also using in my company. And so you have an IP sharing. Can you do you mind just talking briefly about how that works between those entities? Because it looks like they're separate entities. So it's not, no, they're not. They're actually all part of the same. So the name of the company Slice Backend, we started off with a very heavy sort of tennis focus. But they're all owned by Slice Backend, all these. So they're three products within the same entity. Yeah, I think it would be very complicated and messy. In my opinion, just sort of spin them off into different companies. So good. Thank you. Yeah, it's really helpful. I have a question as well. Yep. So thanks for sharing your experiences. It's really helpful, I think, for those of us at the outset to try and learn the easy way. So my question is around the key sort of points where you see the company really increasing in value, both from an investor side and from your experience and as a founder. So it seems maybe like 1,000 users is one of those. What other sort of things should entrepreneurs be shooting for to really increase the value of their companies? To increase value. So depending on the on the market, you're targeting this sort of different sort of benchmarks for enterprise SaaS companies, for example, versus a consumer company. So that depends on the market you're targeting. I think if you're looking, which sort of market are you considering, are you looking at? Yeah, we make search engines and data visualization for scientists. That's what my company does. Okay, so my guess is there's probably a higher willingness to pay in that market of the very sort of targeted segment. For consumer apps, the bar is a lot higher. So I think I briefly touched on book vibe, which is a book recommendations product. For that to have been really interesting to investors. We had a few hundred thousand users, just not interesting enough if your sort of revenues are going to come from primarily, we had two revenue streams, affiliate revenues from Barnes & Noble or Amazon are a tool for self-published authors to promote themselves. So the bar was very, very high for to even raise a series around millions of users. So it depends in different categories. So for me personally as an investor, since I do very early state stuff, a few thousand users, if the valuation is right, is a good start. And I think what's really more important than anything else is the trajectory, the shape of the graph. Cool, and yeah, one other area that we've looked at in this sort of targeting, where that's going to increase value is the technology. But as you've sort of said, I want to be fair to say that technology without some sort of validation in a use case is really not worth it that much. Yes, yes. So yeah, that's something I learned after my last experience, my last startup. It is, yeah, your value is, because so many startups, so many companies nowadays that make tall claims as well, it's very hard. If you don't have users, it's very hard to validate those claims. As I mentioned, a lot of tech is getting commoditized. It's easier and easier for people to build products on top of hard tech. As an investor, I used to, so my strategy has changed. In the early days of my investing, when I first started about 15 years ago, they were in that many hard core, hard tech companies, or even companies that I worked with. You know, hard core, hard companies, or even companies that made claims that they had, that they had revolutionary, yeah. And that number has just gone up. So I always used to think that if the entry valuation was right, if there was a really good core team, it could be sold to a larger company for the tech alone. But nowadays, those types of acquisitions are getting much, much harder to get to sell a startup for just for the core tech. There are exceptions, of course, but in general, it's just much, much harder today. Right. Okay. Cool. Thank you. Yeah. Sure. On the investing side, Ramesh, how are you going to help the New Zealand companies? What is it that you can offer the New Zealand companies when it comes to sort of connections to other investors, or actually even if they're trying to go into that growth phase, because you're starting at the startup end, what is it that, you know, how are you going to help the companies along the way on the journey? Yeah. So I think looking at companies in the UK, for example, when I compare them with companies in the valley, I think one factor that separates valley companies, the top companies in the valley, is how much sort of more sophisticated they are when it comes to, compared to the European companies in general, when it comes to growth. So they are just far more, they've just have people on their teams quite often that have, many of them might have worked for a Facebook, might have worked for an Uber, they just have much more sort of familiarity in terms of tactics that can drive growth. I'm sort of generalizing in here, but so just, so I think connecting people in New Zealand with people in the valley who understand growth is something I can do. Connecting them with investors who have actually, who have run companies before. So I was actually quite very fortunate in my case that one of our seed investors, David Jesky, was someone who had sold a company to Google before. So I was able to get it, so we were able to present to Google and have a chance to pitch Google in the first place because of that. So had we not had that sort of access to people at Google, it would have been much, much harder, may not have been possible. So that's something I can try to do, have a good network in the valley. This is sort of, I think it tends to be, if a startup has tens of millions of users, then these things are less, less valuable. But if a startup is sort of in that range where they're not really quite sort of, they're not quite a rocket ship yet, I think in those situations, these sort of connections and introductions are much more valuable. So that's something I can help with having gone through the whole process, starting a company, raising rounds from, most of them were investors, who had also been operators before. So I've got a very good network there when it comes to people that can add value as investors. Nice. Yeah, that was actually going to be one of my next questions was how do you get in front of your potential acquirer? So you think one of the key things is actually having someone on your team or in that close network who has actually been part of that organization or sold to that company, had some sort of connection. So you think that's an excellent mask? Yeah, that's a huge help. You can do it in other ways as well. They're more sort of organic. If someone on there, let's say you have a deep, you have a technical company and someone on your tech team knows, has worked, someone on the tech side, let's say you're speaking to Twitter, someone on the tech side of Twitter, has used your product, is familiar with them. So when someone on the business side speaks to someone on the tech side for validation, they're like, yeah, we know those guys. It makes a big difference. So it can happen. But these sort of, it's just a little easier if you have someone who can give you that credibility when it comes to an introduction, just because there's so many more startups today. It's just harder to stand out. Yeah, yeah. Hi, Michelle. Michelle, can I ask a question? Of course. Can you make your own? Go for it. Hi there. I'm so happy to be here today. My name's Lane, and we're Chatterize. And what we do at Chatterize is we build conversational chatbots that live in a virtual world and help young Chinese kids speak English with confidence. So the question, yeah, so the question I have is the way that we've gotten started is that we've just plugged in a speech-to-text engine. And we've plugged in a speech-to-text engine instead of building it ourselves. After quite a bit of debate in the beginning, and then we've focused on the user experience of it all, to prove out the concept. And I'm wondering, what is your opinion about kind of building out the technology before you prove the product itself and the concept itself? How much effort did you guys put into doing that? Would you do it differently if you could pull something off the shelf? Right? And I'm just always kind of thinking about at what point is the... We're collecting data, right? So the good thing is we have a way to collect data. And I was always thinking about at what point is the right point for us to start to build our own engines? We... Our target market is four to 12-year-old Chinese kids speaking English, right? So it's always a difficult ask for an off-the-shelf engine. Yeah, so if it's not really core, especially in the early days, to the overall product, if the tech is not really that core, I wouldn't spend too much time building, tweaking the model, building a really spending too much effort building an in-house model there. I would sort of be very... I would use whatever you can be very... Do something quick, more quick and dirty on the tech side. And then if it's... But if you were to tell me that the tech is a core part of the product, then I'd spend more time upfront in the model there, investing in the tech. So how core is that? Well, I mean, our product innovation is that we provide over 80% more spoken opportunity than any other language learning product on the market, right? So it is core in that we are a conversational world where people can practice their oral English, right? But it is that also... I'm just wondering if... Because we have a lot to do. We have a lot to do. We've got the Chinese market to deal with. We're a business to consumer company. We also have like a user interface for children, you know, who can't read, right? And so I just sometimes wonder if building our artificial intelligence engines anytime soon will create a situation where we'll never be better than a company who would just solely focus on that. Like how do you split resources? It's a tricky one. So looking back, we would definitely not spend so much time. We spend way too much time trying to get the system to be very precise. So it was all about surfacing urgent emails. So I felt that as a user, if the system sent me a notification and the email was not urgent, I'd just never use it again. So my personal bar was very high in terms of how accurate it needed to be. But it didn't need to be... It didn't looking back. It didn't need to be... We could have sort of gotten something as long as we had messaged it correctly, especially for our initial users because our early users were actually providing us good feedback, wanted to help us. So we tried to sort of... We set the bar too high on accuracy as opposed to have gotten something out there quicker and tweaked it based on that. Okay. Yeah, so that's what we did in terms of but spending too much time initially trying to get the technology right. Okay, thank you. Thank you for that. I really appreciate it. Sure, sure. Any other questions today? So going forward then, Ravish, looking at the New Zealand market, if you were going to invest in some New Zealand companies, how is it best that people contact you? Do you want to put your email address in the chat? Yeah, email them. Yeah, email them. If you put that in the chat and can cut and paste it. Yeah. And what other ways can folks help you in the New Zealand ecosystem? Oops, so we end it. To see more deal flow, what would you like to see? Yeah, so one area I'm obviously, I'm very interested is the sports AI I'm working in that area. So startups in that space, I'd be very interested in that area. AI, application AI companies in general is interesting. Even core AI companies building stuff in that category. What else? I think for me, in terms of, yeah, as I said, in terms of criteria, the most important factor is founders who have a credible plan for user acquisition. That is the most important factor in terms of criteria. Let's see. What other sectors am I interested in? Health tech. I'm looking at that area. As I get older, I want to start, you know, use products that will sort of improve the quality of my life. Stay healthy. That's an area. Productivity, as I mentioned, is an area that I was immersed in. I still like that area, even though it's a hard area to monetize. What else? Are you willing to follow on rounds? Like how far through the investment journey will you go with one of your companies or? Yeah, I'm open to it. I typically like coming in really early stage. So if it's, so when a company gets to a series A stage, etc. I typically don't follow on, but I'm open to it. I'm open to it. Yeah, typical between 50,000 and 100,000 U.S. dollars. It's not making it to New Zealand. Yeah, so it's typically don't come on as a lead investor. It is in the U.S. that would be a $100,000 investment would typically be about, the typical seed round is about a million to $2 million. So it's about 10% off the seed round of the overall round. In the U.K., the typical seed round tends to be closer, about 400,000 pounds, about half a million dollars. So $50,000 investment is about 10% of that. I used to be put in right larger checks, negotiate terms, be the take both seats, but I don't do that anymore. I just don't have the time for that since I'm also running my own startup. Nice, that's actually really good intel to know that you'd prefer to syndicate with others and have someone else do the lead and do the DD. Yeah, good intel. Yeah, that's typical, but having said that, I'm also, I'm not afraid to be the first investor and I'm happy to consider coming on as, even at a pre-seed stage as the first investor or the only investor in a smaller round, if there's a pre-seed round they want to do. Some of my best investments have actually been deals where I've been the only investor and without any other investors. So I'm open to that as well. Thanks a lot for the session, really appreciated because you're so open, also sharing what worked but didn't work, I really think that's very helpful. And in fact, my question was, so it's related a little bit how you could support the New Zealand ecosystem from, let's say, mentoring perspective because we have many, we have many, many ventures in EHF who are interested in the area and who have started to work, like Kyle is on the call, he was part of the Huddles as well. And I was wondering, are you open to do some mentor sessions on areas which would fall into your area without necessarily investing? It could lead to investment if you think it's interesting but just help some of the ventures to get some of your experience. Yes, absolutely. I'd like to get plugged in to the New Zealand startup ecosystem. I was actually hoping to have moved to New Zealand in August to be, I'd been planning on moving to Queenstown. So I haven't been able to get as involved remotely as I would as I would like to. So definitely something I'd love to do. Well, we've run out of time unless there's one more question that someone would like to do. I've got one little sort of misquestion that popped up earlier. So you described a situation where you had an engineering heavy team and looking back, you wish you had like a little bit more focus on growth. Yeah. In terms of, say, hires or team, what would that ideal person have been? The ideal person for growth or the ideal composition? I guess the ideal person that you would have added or the way you'd change your team to sort of focus it more on growth. Focus, focus it potentially one of the, one more. Yeah, that's a very good question. I wouldn't replace any of the people on the engineering team that was also Yeah, yeah, yeah, it's a very good question. I would get a team of 10. I would have at least two people on growth. Cool. And you sort of mentioned with Silicon Valley that it was much more common to have, I don't know, more focus on growth. Is it materializes people whose core skill set is that user acquisition and growth? Yeah, there are people there who are, I think most startups there as well fail. They can't crack growth. So growth is difficult even there. I think the difference is there are more people with experience in growth thanks to all these companies that have grown enormously over the years. Someone who has worked, who has worked at Instagram or someone who has been at Twitter in the early days at LinkedIn, all these. So it's easier to have, to get access to people who understand, who understand growth. Yeah. Cool, great. Thank you. Rick, I've got one question that's come in from Facebook that says they're looking at starting an AI company, but they were surprised you didn't list patented idea or patent pending as a criteria. Do you feel safe investing if that's not, if they're not patented? Yeah, I feel, I know some investors love it. Many acquirers like it too. Personally, to me, it's not a big factor. It is a nice to have. But personally, I find it quite a bit annoying actually when startups sort of overemphasize that early on when they talk about IP it generally takes years to get these for patents to be granted. So it's a nice to have, but not a criteria for me. Thank you. Well, thank you very much for your time. Thank you everyone for joining. Our next live session we have in February is investor fellow Mark Brigham. He's been interviewed by investor fellow Chris Wake from atypical. And we will also have another session that was postponed from our earlier in the month because of speakers having to get called away to governments. And that's about innovation. And how does New Zealand look at innovation to create jobs and that in the regions, which we will have a few fellows on a panel being interviewed by Peter Crabtree from MB. So thank you for joining us. And thank you very much for staying up for us and you can now go home. Yeah, okay. Thank you very much, Michelle. Enjoy your evenings. Enjoy your day. Thank you. Thanks all. Thanks a lot. Really, really enjoyed it. Great questions. Looking forward to meeting in all of you in person soon.