Read transcript ↓↑
AI is >> [music] >> that it's most powerful when it amplifies human decision making and not [music] replace human decision making. Often times what people say and what they actually show [music] are very different. We recently did a project for a large American chain and we talked [music] to over 500 people for them in 5 days. At the end of the day, I think that people [music] don't actually care that you're building AI. What they care about are the [music] outcomes. But there's really no playbook of like, "Hey, like at this point you need to bring more people on, [music] right?" What AI is today is a ginormous data [music] set, massive compute, and then you throw a large language model [music] on top of it, right? And that's why it's very good at convincing us that it knows what it's talking about. Uber's [music] uh was it COO was like, "Yeah, you know, like we we burned through like all of our token budgets in the first quarter of our >> [music] >> the entire year." The new generation of people, younger people are also realizing like, "Hey, like this might actually cause us to stop being able to think, right?" Even they're saying like, "We don't want [music] to use AI in in our class." It disappeared at the like literally 2 days [music] before we were going to have everything signed. So, >> Hello, I'm Julio. What's your name and what do you do? >> Uh my name is Lenny and I am the founder of company called Convo. Uh that's getconvo.ai and Convo is a AI powered voice interviewing platform that allows user researchers and market researchers to scale and do different types of research that they couldn't have done before, right? So, we can now interview hundreds or thousands of people at the same time. Whereas in the past you would have to do one-on-one interviews. So, that's what we kind of enable. >> So, tell me a bit about who is your ICP or your or your client in in for your company? >> So, right now our ICP that we've been working with are, you know, large retailers that have to really understand why people buy different things for their homes and, you know, their motivations, how they decide to buy different items, and that kind of stuff. Uh and, you know, this is essentially I think people who are in the research and insights departments, right? So, we've done a variety of different industries. Right now we've kind of been focused on this, but we've also done work for real estate companies, right? That they want to build new buildings and stuff, so they want to understand like, "Hey, what kind of amenities do people want?" And, you know, what are the different types of, you know, features and things things they want inside their house. So, we've done big research projects for real estate companies as well. Um and we've also done research projects for even the Dutch pension fund, actually. So, um so yeah, we we we do a lot of different things, but like essentially that needs primary data where you interview people, that's where we can actually help a lot. >> How did you came up with this problem? How did you start the HomeBinder? >> Yeah, so it started when I was working at Miro, and I was a really early employee at Miro. We didn't even call Miro yet, actually, at the time. And I ran new user growth for them. So, I had to really understand our users to be able to kind of build the right programs and, you know, [clears throat] develop the right features and things like that that would address their needs. So, I had to do a ton of my own user interviews myself and user research myself, right? And I think [snorts] that throughout all that, I felt like we were always making product decisions on very thin data. And what I mean by that, right, is it didn't matter which method we chose, whether it's user interviews or surveys, if it was user interviews, I only was able to talk to maybe like 10 or 15 people for each product feature. If it was If it was surveys, I could send out hundreds of surveys, but then I lost the kind of the depth that I got from talking to people. So, yeah, I kind of did some research to look for solutions for that and you know, there were a lot of competitors. We have a lot of competitors out there. >> [snorts] >> But, I think what kind of sets us apart, right? And what's at the core of how we've built the company is that we believe that AI is at its most powerful when it amplifies human decision-making and not replace human decision-making, right? So, everything that we build is built around the researcher and how we can actually enhance what the researcher does, right? And we keep the researcher at the core of the most critical parts of the research process, right? So, for example, you know, setting up a discussion guide, right? Like, what are the right questions to ask and what are the things to probe for, right? That stays with the researcher. Uh doing collaborative analysis, right? The researcher is also kind of in control of all of that, right? So, those kind of very important human bits, we leave that in make sure that we keep that with the researcher. >> Are there any specific techniques that researchers use that now you amplify with Combo AI? >> I would say that one of the things that we're able to do is now we've turned qualitative data to we we can do so much, right? That you can actually use it in a representative way. And what I mean by that, right? Is in the past you would talk to 10 or 15 people. And I've been in so many meetings with exacts, right? Where it's like, "Oh, hey, like, what percentage of people, you know, liked the pink button?" Or something like that, right? And I'm like, "Well, I talked to 10 people, so I don't know if we can use this data this way, right?" But now, because we can talk to hundreds of thousands of people that and the the sample size is so big, >> [snorts] >> we can actually confidently say, like, oh hey like, you know, 75% of people actually like that pink button, right? So, I think that's one thing, right? And I think the other thing is we've also democratized some of the more difficult parts of research that you couldn't have done before kind of on your own, right? So, we enable what are called mobile ethnographies, and this is a really cool type of research that in the past would be extremely expensive because essentially, you know, you're you're kind of like living in the home of a person, right? Uh and and this comes from, you know, academic research. You know, we in in in business world kind of do, I would say, ethnography lights, uh but in the past like a research project like that would cost you $250,000 for 2 weeks or something like that, right? Now, because we have AI and we can interview all these people using AI and use the AI to do all this type of moderation, uh we can, you know, finish complete a project, you know, for 10 10 times cheaper, probably. >> Wow. >> Yeah. So, >> how does it work? >> Um so, for the mobile ethnography, how it works is essentially, we do we do a few different things, right? So, beyond just like a regular interview, and where this becomes very powerful, right? Is that if you do just a regular user interview, for certain things you end up priming people into thinking about different events, right? So, a good example of this is Budweiser actually in the US did an ethog- ethno- ethnography, rather, uh because if you do user interviews, people start to think about drinking events, right? Cuz so, they're like, oh like I drink at weddings or parties and things like that, and they don't think about the the non-drinking events where they also do actually grab a drink. So, they ran ran an ethnographic study, and what they realized was that Americans, in particular, really like to grab a beer when they're doing yard work. So, excuse me. So, what they did was they released smaller cans of beer, so you would finish it while it was still cold. So, that you wouldn't have any warm beer kind of left over. So, that was like something that you I don't think you would have been able to surface with a regular user interview, right? So, how we actually enable this is using AI. We will actually talk to a group of participants. And it really [snorts] kind of depends on what they're trying to understand, right? But, at the core is like we will ask them different questions each day, right? We'll have like for example, if we >> Is AI the interviewer? >> Yeah, the AI will do the interview. Um, and there's also tasks that we actually have the users complete as well. So, it's the interview. We also can actually have them record videos of their surroundings, right? So, if there's specific things we want them to record, we have them record that and also take photos. And where this becomes extremely powerful, too, right? Is because often times what people say and what they actually show are very different, right? So, how they describe themselves might be very different from, you know, the artifacts that you get like the pictures or the videos or anything like that. So, so then you can really start to surface out these different tensions that people have. And for companies who are selling different products and things like that, it's actually really good for them to start to be able to understand like, "Oh, hey, like you know, maybe, you know, you don't sell a minimalist uh uh you know, people might describe themselves as as minimalist, but maybe they have a lot of stuff. And, you know, instead of saying like, "Hey, we can give you this minimalist nice lifestyle." It's like, "Hey, we can give you storage solutions, so you can put your stuff away." Right? So, that you can seem like you're a a minimalist even though you have a lot of stuff. So, it's stuff like that. I'd say. >> Interesting. So, how does it work for my understanding is through a mobile app or or any kind of other tools? >> It's is the web app, so you can do it on desktop or on mobile, yeah. >> Well, there's essentially a lot of input devices like a audio, video are connected. >> Yeah. >> And uh Convoy AI translate all that information well, before agreeing on a survey with the researcher and translate all the information into tangible outputs >> Exactly. >> that finally you can reproduce uh >> Yeah. >> massively because before it cost a lot, right? >> Exactly. So, in the past like you you wouldn't be able to do a large ethnographic research in without paying tons of money because you need to essentially you would have to have like moderators talk to people, right? One-on-one, right? So, at most you might be able to get like 25 people or something like that, right? We recently did a project for a large American chain and we talked to over 500 people for them in 5 days. >> Well. >> Yeah. >> Well. And what challenges have you found during the whole process of building this company? >> I think some of the challenges I face are more you know, around I would say that here's this is following, I think. One of the big challenges I think when you're building in AI is that something that seems very novel today could be very archaic and old tomorrow, right? And a great example of this is you know, 2 and 1/2 years ago when I started building this I spent a lot of time and effort making the AI conversation cuz we use voice AI as part of it, right? And that's that was like at the time that was a differentiator from our competitors. Our competitors use chat or they use kind of like a walkie-talkie like you push to record. So, I made it like full duplex and we try and we reduce that latency so it could feel like a natural conversation like you and I are having right now. So, I spent tons of effort doing this, right? And, you know, I think initially it was pretty cool. Like people saw this and they're like, "Oh yeah, that's pretty neat. I can see ourselves using this." And then OpenAI, like 6 months later, came out with their multimodal model, right? Which then like this just became a natural thing. Like I'm I'm we're just like, "Okay, well, now we just we spent all that money of like time and effort on this." Uh but So, I think things like that uh are challenges, obviously, but the the the tech is just changing so fast, right? That you sometimes are like, "Oh well, I spent all this time on this." So, you know, I think the the lesson there is, you know, And and I think it's good because we're still a pretty small team of like six people, that we can adapt to these changes relatively quickly. I think the other thing that is a challenge, right? I would say is that I think that the when it comes to AI, right? I I I a challenge and and I think that it's important, right? Is the at the end of the day, I think that people don't actually care that you're building AI. What they care about are the outcomes, right? So, what is it that you can actually deliver for me, right? And, you know, that's the insights and that kind of stuff, right? That we can deliver to them. They don't care that it's an AI doing the interviews. They don't care, right? It's like at the end of the day, it's like, "Hey, this is the output and this is what I can present to my stakeholders. This is how you're going to save me time or allow me to do different research that I couldn't do before." And I think that's really important, right? Is that like we don't have to keep, I think, telling people like we're building this AI company, right? And and I think that dovetails into the next piece, which is the trust, right? Because I think at this point, everyone knows, right? These things just lie all the time, right? They They hallucinate. They make stuff up, right? So, you know, for us something that is, again also at our core, right? Is this trust piece, right? So, everything that we do is as transparent as possible and as explainable as possible, right? Like those are kind of core tenants for us, right? So, every insight that we deliver to our customers has evidence attached to it, right? So, whether it's an image, whether it's a quote or like a still from the video, right? We show them like, "Hey, like if we tell people like, 'Oh, hey, like uh 90% of the people have beige rooms or whatever, right?' Well, we can show them like every single picture, right? Like, "Hey, like look at this. You can see all the photos here of all the rooms, right?" And I think that helps kind of build that trust to so so that they can just kind of know that like, "Hey, like you this thing is not just making something up." And And I think that's also really important. So, I think those are some challenges from like a building an AI perspective. I think the other part of building a startup, I think that's challenging, is it's hard to Excuse me. It's hard to know when to bring additional help on, I think, right? Cuz there's no there isn't like there's there's a lot of playbooks for like, "Oh, hey, like this is how you build your pitch deck." And, you know, this kind of stuff, right? But, there's really no playbook of like, "Hey, like at this point you need to bring more people on, right?" So, I think kind of figuring that out is is is is very challenging. It was very challenging for me, at least, right? Cuz I felt like you know, "Oh, I can do everything myself, right?" Especially, you know, using AI and stuff like that. Make things a lot simpler for myself. But, then at the same time, it's like you actually can't do everything yourself, right? Like there's only so much time in a day, even if you have AI like doing different thing different tasks for you. You can only do so much. And then there's also the context switching, as well, right? So, if I'm spending all my day like mornings like doing product work or or engineering work, it's hard to actually switch gears and then move into to do sales or marketing and that kind of stuff, right? So, um yeah, I think it's kind of like that that part is tricky and and you also have to kind of balance too. It's just like, "Hey, like I don't have a lot of money, right? Cuz like I'm bootstrapped. So, like I think that also makes it tricky of like, "Oh, okay, like do I really want to spend I don't know, 2,000 euros a month or you know, if I'm if I'm lucky, right? Like if I can get someone to believe in the dream and I'm only paying 2,000 euros a month then I'm lucky, right? Uh and I I think that's that's also part of the trickiness of like, "Hey, like when do I bring that person on uh to I guess accelerate and amplify the work that we do." >> Interesting. Yeah. How how do you see the world in the upcoming years with these AI and advancements that we are having in the coming years? >> Yeah. Um I think if I'm being honest, I don't think that I don't think that we are going to reach AGI with AI in its current form. >> [snorts] >> Right? Cuz at the end of the day, what we have today, I think the intelligence piece is a bit of a misnomer, right? Cuz what AI is today is a ginormous data set, massive compute, and then you throw a large language model on top of it, right? So, it's this predictive model that kind of like understands human language, and that's why it's very good at convincing us that it knows what it's talking about, right? >> [snorts] >> But, I [clears throat] think we're also starting to see that like just throwing more compute at this is not actually solving this problem, right? Like when we went from the first public release of ChatGPT to um uh was it three? The three >> 3.5 >> Yeah, 3.5 or whatever. what >> so. >> Like that was a huge jump, right? From the first one and then you're like, "Wow, this is incredible." But then when we went to four, it was like >> Wow. >> Yeah, but then now we're seeing like it's like that that we're not seeing these giant leaps anymore, right? Like sure like on some benchmarks or whatever, it can code better, it can do this or whatever, but I think we're also starting to see too even at companies, right? Like, you know, like last week Uber's uh was it COO was like, "Yeah, you know, like we we burned through like all of our token budgets in the first quarter like of our the entire year." Right? So, now they've restricted you know, each engineer they can only spend like $1,500 I think a month on on tokens. But the problem, the other problem is they're like, "Yeah, we burned through all these and we're not actually shipping any faster, right? Like, we're not actually producing better outcomes for our customers, right?" And I think I think the issue there, right? Is and I and I'm noticing this myself too, right? Is that I'll use like if I'm not writing the code myself, if I'm doing prompts or something like that to to to generate the code a lot of times I'll have it generate like four different versions and I'll be like, "Oh, I like this little piece from here. I like this piece here." So, you're actually like you know, in a way you're sure you're you're you're you're you're you're able to produce code faster, but then like your output time is still probably the same because now you're like picking and choosing from like all these different versions, but then that's also why your token costs are like so high, right? >> And also everybody uses four anything. So, they I think we are we're using it, right? >> Yeah, [clears throat] yeah. I think I think that's the other that's the other part too, right? Is that I think that they're very good at convincing us that it can think and you know, it really it really kind of doesn't think, right? Um and I think there are a lot of I would say there's a lot of danger there, right? Like I think that with for example, Google's announcement like was last week, right? Where they're like, "Hey, we're going to do just basically all AI results." I think that's very dangerous, right? Because I think that part of you know, humans being able to think critically, right? Was like you know, if you Googled something, you could look at the sources and be like, "Okay, like this is actually trustworthy and this makes sense. So, I'll like I'll read this or whatever." Or I won't read it cuz this doesn't make sense. This is silly, right? But now, like if you're kind of replacing all that and it's just giving like an AI result, then you're kind of like, "Well, I don't know where this is coming from, right? Like it could be it could have been like on the 10th page of the serp and somehow like the AI is like picking this up and surfacing this." I think that kind of stuff on a social level is is very dangerous. Um so, hopefully we can figure out some guardrails on how we can you know, protect ourselves from that and you know, we're even seeing different things in in education as well, right? So, my partner is a is a is a professor actually at at a university. And one of the things that she's noticed is what what's actually interesting is she said the new um the new class this year actually was like, "Yeah, we don't want to use AI on on any of our stuff." Um cuz like the >> What does she teach? >> She teaches busi- uh international business and yeah, yeah. So, she's in corporate governance. So, well, that was really interesting, right? Is like I think even the new generation of people, younger people are also realizing like, "Hey, like this might actually cause us to stop being able to think, right?" Even they're saying like, "We don't want to use AI in in our class. Um whereas, you know, like I think the last 3 years like a lot of people use AI. Um and what she noticed is that the people who rely really heavily on AI have a really hard time actually like they might be able to produce work that looks really nice. But then when you ask them you know, very specific questions or whatever, right? They have a very hard time actually explaining those concepts, right? Because it's convinced you like, "Oh, this is the right answer." right? So, yeah, I think these these these are the things that I think are interesting on a social level. Like definitely like, you know, outside of uh building a startup. >> Interesting. >> Yeah. >> And you worked in US, right? >> And I worked in the US. I worked in Canada as well, yeah. Yeah. >> And how do you see the startup ecosystem in the US and Europe? There are a lot of critiques. Europeans ourselves, we critique a lot of our ecosystem when we compare to to US. >> I think I think there's there's there's a few I think differences that I've I've noticed, right? So, I think the first one that I've noticed here, at least in the Netherlands, right? Is it's very expensive to set up a company, right? Like it cost me over 2,000 euros to set up my BV, right? So, I imagine, you know, if I was just out of university like that's super expensive. Like I And like even if I asked my parents, that's like a lot of money, right? Like it's like, you know, 2 grand is not not nothing, right? In the US and Canada you can start a company for like 200 bucks, right? Like $200 is a significantly easier ask of your parents. Like, "Hey, can you lend me $200 to start a company?" Of course, right? Whereas like 2,000 is like, "Oh, wait, hang on a second." right? Like 2,000 is, you know, like that's a month's worth of rent, right? That that you're you're kind of throwing away. So, I think that part already reduces, you know, the number of of people who want to take the plunge to to to start a company, right? And I wouldn't say I I wouldn't say that Europeans are less innovative than the US, right? Cuz like I think that's like one of the big big things like Americans are like, "Yeah, like well, we have way more innovation here, like all this stuff happened here, right?" And you know, like all this AI stuff happened there, right? And that's partially true, I think, right? Because DeepMind was actually originally a European company, right? So, you know, the original kind of AI company was a European company, right? It's just that I think that Europe in certain ways has a lot of different rules and regulations, right? To to protect consumers, right? And to protect civilians, right? And and I think it's I I it's it's tough, right? Because I think as someone who lives here, I'm like I I appreciate that, right? In certain ways, right? That like, "Hey, like I know that at least the government will step in and make sure that my data privacy is taken care of and I'm protected, right?" Whereas in the US, it was just like it's just like, "Okay, like we're going to take your data and do whatever we want with it, and you know, you don't get to say anything about it until something really bad happens, right?" So, I think it's a balance, right? But I think that when it comes to, you know, the next, I think, 10 years or whatever, right? Like I think that we are we are seeing a lot of people actually shift away from the US in certain ways, and perhaps come to Europe or whatever, uh to to build companies, right? And I hope that there are going to be more innovative companies that come out of Europe, because I think that, you know, Europeans have a really good education system, right? Like it's it's it's incredible. You don't have to you don't have to go into massive debt, right? To to to get a good education here, right? Um and I think, you know, there's very interesting problems that, you know, Europeans are solving, right? Like in Amsterdam, for example, in the in the Netherlands, for example, right? Like all the agriculture and these types of very different problems, um you know, they're solving in very novel ways, right? That you know, I kind of hadn't heard about some of these things, like using using fungi to to to to get rid of um >> Pollutants in water. >> Pollutants and things like that in water, right? Like I think that stuff's really cool and I think that's really innovative, right? And I think that what we need is probably like we we need the government to essentially help us from a almost like, you know, if we think about it, right? Like we live in this kind of like more socialist society, right? But like but when it comes to starting your com- your own company, you're kind of on your own, right? There isn't a lot of government help, right? That will you know, give you a bit of like, "Hey, in case this fails, you know, you'll still be you'll still be okay or whatever, right?" Like cuz, you know, for instance, I did like how I did this, like I basically just I quit my job and and have my I and I just kind of funded this myself, right? So, like having some government programs, um that are a little >> [clears throat] >> like more helpful, I think, on the financial side uh would be really powerful for for for European companies, I think. Um but I yeah, I I I don't agree necessarily that like innovation innovative stuff doesn't happen here. I think there are a lot of rules, uh like I said before, and that definitely does slow things down, but I understand the reason behind them. Um and yeah, so it's like a it's it's tricky, right? Cuz do you want to have crazy innovation and no rules to the point where, you know, everyone around is getting hurt by this or, you know, do we like slow down innovation a little bit and you can still do things that are really innovative? Like I think that's the That's the difference that I see um in the I guess the question that we have to ask ourselves. >> Okay. Interesting. Um any lesson that you would like to give to the viewers? >> Um I think the biggest lesson that comes to mind is the the the building a startup is is is a marathon. It's not a sprint. Right? Resilience is such an important part of, you know, building your own startup, right? Like there's Like it's definitely a roller Like emotionally, it's definitely a roller coaster cuz it's just like "Oh my god, we got like a massive contract." And then it's like, "Oh no, actually they decided they didn't want it with us anymore." And it's it's So it's like crazy kind of up and down, right? And and if you cannot kind of handle that, right? Like Like if you cannot kind of try to regulate it so that you don't get too high and don't get too low, then that'll be very difficult for you, right? So I think like I've Excuse me. I know people who, you know, every little thing that happens will have a massive impact on on on them, right? On their life, right? And I think or on their personal kind of well-being. And I've seen that cause people to burn out because you're just like so happy and then so sad and so happy and then so sad. And I don't think that that's a good I don't think you can you can last like that, right? Like I think that's just way too much on our kind of nervous systems and and and that kind of stuff, right? So I think, you know, uh that's all I think part of the resilience of it, right? Like knowing that there's going to be stuff that just doesn't work, right? That just fails really spectacularly. Um there's going to be deals that fall through that are, you know, kind of of no fault of your own, right? Like for us, this this happened like 2 months ago. We had this deal that we were about to sign on the dotted line, and then that company had layoffs, and then our champion was one of the people that got laid off. So, we just like this deal that we worked on for 3 months that would have been, you know, like a a pretty big deal for us. It disappeared at the like literally 2 days before we were going to have everything signed. So, you know, if I were one of my friends who get really kind of caught up with this, I would be extremely depressed. Uh but then I kind of looked at this as like, "Oh, hey, like this is actually This is a good sign like cuz this person really believed in us. Now, we just have to find We've learned so much about this industry. So, I'm going to take those learnings and try and find other people in the same industry and basically explain to them how we can solve their problems for them because now I know, hey, like these are the big pain points that this guy actually wanted us to solve for him, and he was willing to pay us big money for this, right? So, now I just have to find another person in a different company. Like there's so many companies out there, right? So, I think again that's a part of the resilience. I think resilience is like super important. And And then I think actually also just outside of the startup having like a really good support system is really important, right? Like my partner, she is She's incredible, right? Like she is so encouraging all the time. And I think that that helps me actually like on those days where like, you know, where I lost that contract, right? Like it's like it's it's, you know, having dinner with her that makes life like feel okay, right? Um cuz like sometimes you do feel like pretty crap. Um but yeah, having that strong support system I think is also really important. >> Amazing. >> Yeah. >> Well, for the last question, it's not a question anymore. I leave you the camera, and you can ask for a connection or or anything to to the network and the viewers are watching this. >> Cool. So, yeah, if anybody needs help with market research or user research, um we would love to actually show you what we can do. We'd love to, you know, um get you access to the platform and show you the type of analysis and, you know, how you can uh really accelerate and do way more research and different kinds of research that you never thought you could have done before. Sounds good? Thank you so much, guys. Thank you so much.