What are the practical ways AI is going to change research - not just in the future, but in the here and now? Arguably, there’s no one better to talk to on that front than Kathryn Topp, the CEO and co-founder of Yabble. As Kathryn describes, she and and her co-founder, Rachel O’Shea, introduced their first generative AI tool in 2020 with Open AI since ADD and I was lucky enough to talk through:
All episodes available at https://www.insightplatforms.com/podcasts/
Follow FutureView on Twitter at https://twitter.com/FutureView7
What are the practical ways AI is going to change research - not just in the future, but in the here and now? Arguably, there’s no one better to talk to on that front than Kathryn Topp, the CEO and co-founder of Yabble. As Kathryn describes, she and and her co-founder, Rachel O’Shea, introduced their first generative AI tool in 2020 with Open AI since ADD and I was lucky enough to talk through:
All episodes available at https://www.insightplatforms.com/podcasts/
Follow FutureView on Twitter at https://twitter.com/FutureView7
the future's different because you it's not about the data telling the story. It's actually about you as an individual, kind of bringing people along that journey and having because you're going to have infinite knowledge. So ai will provide you with infinite knowledge, and then it's about how you use that knowledge to create sway in an organization. So if I was a young woman entering the industry now, I'd spend a lot of time ensuring that my communication skills were really strong. I was really confident in speaking, and particularly speaking with big groups of kind of corporate businesses or big conferences. Particularly speaking with big groups of kind of corporate businesses or big conferences, I'd make sure I had a really good understanding not just of the tools that are available, but how AI works and where you can then apply different models and different techniques to different types of data.
Speaker 1:I don't know about you, but I'm intrigued and quite excited about the idea of synthetic or augmented data. So I'm delighted to have on Catherine Topp, the founder and CEO of Yabel, one of the leading companies in the space. Cath's a brilliant guest, not just because she's at the forefront of understanding and creating practical applications for new technology, but because she incorporates a deep understanding of conventional consumer insights into that process. In this interview, she gives a fascinating insight of conventional consumer insights into that process. In this interview, she gives a fascinating insight into a vision to build a new type of research company, servicing both brands and agencies, by integrating AI tools that enable you to perform a variety of critical tasks in the workflow. It's not just about creation of virtual segments. It's also about analysis and visualization of unstructured data to integration of surveys. Listen in to find out more, as well as something you probably don't know about, catherine. So on to the episode. So, catherine, firstly, thanks very much for joining today, also very early in the New Zealand morning for you.
Speaker 2:Pleasure to be here, and I love an early start, so you're lucky.
Speaker 1:Well, it's very kind of you to do it and so to wake you up. You probably don't need to be here, and I love an early start, so you're lucky. Well, it's very kind of you to do it and so to wake you up. You probably don't need to be woken up, but I'm just going to get going and just find out something about you that most people wouldn't know. So it doesn't have to be a deepest, darkest secret, but it could just be something that I don't know, you wouldn't be able to find on the web or through a linkedin search or anything along those lines so I can share two things with you.
Speaker 2:So one is I, I can fly an airplane, so I have most of my pilot's license. So, um, if you ever need a little holiday break no commercial flights, but just small, short, regional ones I can help you out there, um. And the other one is I have a complete obsession with labradors, and me and my 35 kilo Labrador go scent trialing. So we go and trial, we do a whole lot of scenting, both outdoors and indoors, and he can find all sorts of wonderful things that I train him to find. So it's lots of fun and it's a good way to get an hour and hour away from the business.
Speaker 1:Well, yeah, that sounds awesome.
Speaker 2:When you're over in the UK, will you come and do some training of our dogs we really need, I'd love to my mother was a New Zealand champion, obedience trialist and a gun dog trialist actually, and we had boarding kennels growing up, so my entire childhood was all about dogs and going around the dog shows and dog competitions. So dogs have always played a big part of my life and I absolutely love them, so more than happy to help you out.
Speaker 1:Or even just some online tips would be helpful as well. And the pilot's license is interesting as well. So what does part of a pilot's license mean? Does that mean you can only fly up to a certain type of plane?
Speaker 2:So definitely so you're always rated to your plane, but for me I can fly solo, but I can't actually take passengers at the moment. But yeah, so there is restrictions on where I can fly. I had to put it on hold a little bit because I've been very busy with building the business, but hopefully get back to it soon and start getting my hours back up so I can have a little bit more freedom.
Speaker 1:Good stuff, but anyway and I am aware you're very busy and I should probably stop asking you irrelevant questions and jump onto Yable, which is really what I wanted to talk about. And so, just to get directly into that, could you just let me know a little bit about Yable, just a headline summary as to what the company does and then, secondly, why you founded it?
Speaker 2:Absolutely so. Yable is a generative AI company and we build tools specifically for the insights industry. And we build tools across the entire insights process, from data creation through to data analytics. And, for me, why I founded the business was I saw a real gap in the space of technology for insights where, in order to kind of complete tasks, you had to buy lots of different platforms. And you know you had to buy this to run a panel, or you had to buy this to kind of use market sample, or you might need to buy this to run some qualitative. So what we initially wanted to do with Yabba was to create a platform where you could bring most of your insights work together into a single place. You could house your own communities, you could buy from Market Sample, you could run advanced analytics and then, more recently, you had the ability to create your own custom synthetic data as well. So the business has been going. The platform's been live since 2017.
Speaker 2:So we started in a more traditional, I guess, research space and then, in 2019, we started building with generative AI. So we had our base platform up and running and we looked and we said what are the two biggest problems that we see for the insights industry, and so the first was related to being able to easily generate insight from unstructured data. So we said there's so many open-ended questions and sources of open-ended or unstructured data available and it's really not being utilised to its best potential. So that's when we created our analytics tools. So we brought our first generative AI tool to market.
Speaker 2:I think it was in late 2020, early 21, when no one actually knew what generative AI was, and it was quite funny actually. You'd go into these businesses and you'd go, hey, look, we've got this really cool question and answer tool. You just load up your data, you ask it a question and then it will give you an answer. And we'd have the text kind of coming up on the screen and I'd have people kind of looking under the desk and looking behind the computer and they're just like how is that possible? That's not possible.
Speaker 2:So those were the days before ChatGPT, when no one kind of knew what generative AI was. So it was quite a fun journey. And then, once we had a really good suite of analytics tools, we actually moved on to our next problem that we wanted to solve, and this is a problem that kind of goes right back to why we created the Able platform in the first place and it was a problem around sample. So how do we create really reliable, cost-effective quality sample in the insights industry? So how do we open up access to that? And, with our knowledge and use of generative AI and the advancements that we saw happening in that space, we started to develop our augmented data authoring or the industry often calls it synthetic data and in December last year we brought to market the first of those tools, which is our virtual audience tool.
Speaker 1:When you started in 2017, was the AI piece always part of the vision.
Speaker 2:Yeah, so the two problems we wanted to solve are always there, so we were always kind of wanted to solve those problems. How we solved them, though, changed. So when we first started out looking at unstructured data, we were looking at more traditional kind of the more traditional machine learning methods of solving those, but we had just one of my new my head of product innovation actually had just started with the business as a product manager at the time, and he said look, the traditional method is a really valid method, but actually there's a whole new type of AI that's coming, and I really think that we should investigate whether it's possible to use it in terms of thinking differently about how we process and access this unstructured data. And so we started that journey with OpenAI, and very early, like, openai was a very small, much smaller company at that point in time and worked really closely with them for a 12 to 18 month period and developing out the first of the analytics tools.
Speaker 1:Got it, and so I was like, well, we might veer off schedule and I'd like a good geek out as well. So what were the original solutions for unstructured data, like the original machine learning approach that you've had, as opposed to the open AI development?
Speaker 2:Yeah, we didn't actually build any cause. We decided that the genFAI approach was better, but we looked at the more traditional machine learning, where you kind of and all your researchers who are listening will know what this is where you essentially have to train the machine. So you come, so they'll get data come in. There might be a code frame against it. You'll then kind of train, you'll go okay, so this comment goes with that code, and you'll start mapping it back and the machine obviously learns over time and it can slowly start to automate that process. But it is very much a. You know, you need to do all of that training and it's not dynamic. So if you kind of have kept the same code frame and you haven't updated the training, then any new data that comes in that's different obviously won't fit into a lot of those buckets. So, very human intensive and very reliant on the human, continuously kind of updating the training for the machine. So that's your more kind of traditional text analytics.
Speaker 2:When we started looking at generative AI, though, we went well, first of all, we can do this amazing thing where it can generate text and give us an answer. So, um, we kind of started with that and we launched a product called query, which we're actually just about to retire because we've launched a um, an updated version of it called gen. But, um, that was where you could just type a question, you could ask a question of your data and then it would give you back an answer which was, um, quite revolutionary at the time. And then the market because they didn't know what generative ai was were kind of like oh well, that's really interesting, I love it, but how do I know it's true? And so, um, from there we went on to develop our yabble count product, which is a theming encoding product which takes um traditional text analytics but actually processes it with generative AI, which allows us to do 100% automated theming encoding of unstructured data.
Speaker 1:I really like it on the website, the way you've split out well these ideas like data creation and data analysis and then the flow between them, and I thought that actually may be the easiest way for everybody to understand it. Could you talk that through in terms of you know, you've explicitly called it data creation rather than data sourcing or anything like that.
Speaker 2:Sure. So data creation is literally that we're using generative AI to generate data, but it doesn't come from nothing. So the data creation tools have and the one that's really the commercial tool that people can kind of buy and run at scale is our virtual audience tool. So that tool is being designed with this concept of I've got a business question, I can, in less than 30 minutes, I can actually get a really robust, solid understanding and answer to that business question. So that's kind of the idea behind it our ability to generate data from existing data. It means that the whole need for doing traditional field work so going out kind of specifying a sample, going out, putting all these questions in that kind of are there just to manage human behaviour, collecting the data and then bring it back in running tables, all that kind of stuff you actually don't need to do that anymore. So that will slowly disappear because with the ability to bring multiple sources of data together, to augment them, to synthesize them, to create new learnings from them, we can then answer most of the business questions purely using our data models.
Speaker 2:So for me, this is the future of the insights industry purely using our data models. So for me, this is the future of the insights industry. I'm probably some people would disagree with me, but I truly believe that this is the future of where the industry is going, and it's also how we're going to democratize the industry as well, because anyone can come into the Apple platform. You don't need to be a, you know, a really classically trained researcher. You can put your business question in and you can immediately kind of get an answer to that question, guided through by AI, and it's a really reliable and robust answer. So how that? I'm not sure. Do you want me to go into how the data is created in that?
Speaker 1:process? Yeah, if you could. I guess that's the obvious kind of question. So how does it work? So I go in, or maybe if we could talk through, you know, what's a kind of a case study.
Speaker 2:I go in, I've got a certain question, yeah, so. So I've got a business question. So we're kind of an example. We had a customer um recently who um had a three-stage research process. So it was kind of desk research, qualitative research and quantitative research. But their timeline had um kind of condensed and so they wanted to see if they could find a faster way to kind of do that desk research and qualitative research phase. So with our virtual audiences we were able to kind of complete that phase and the actual data runs took about 30 minutes, but by the time we kind of pulled the insights from that it took about a day and a half. So really accelerated process. But how this works is you have your topic or your business question. You come into the platform, you put your topic in. As soon as you put your topic in, it activates a whole lot, an array of kind of search algorithms.
Speaker 1:So would stage one would be sorry to butt in on it but say, if it was the kind of the desk research you might be going, who are the major competitors in this given space? What's their market share? Something like that. Would that be an example of a starting point?
Speaker 2:yeah, you might um do like a lot of big brands will do desk research where they kind of go what's going? You know what are the trends in the market um, what are some consumer habits, what are some evolving um evolving behaviors that we need to be aware of, all of those types of things. So the virtual audience product allows you to do that a lot more effectively by just literally putting a topic in. It's not just a desk research tool, though I think that's really important. So it can absolutely um kind of replace or complement traditional desk research. But it can go a lot further than that, um with what it does I see and so sorry I butted in.
Speaker 1:So you're activating search algorithms are now going out and pulling, I assume, on a variety of different information sources yeah, and we've spent a lot of time developing this product.
Speaker 2:So, um, we've had extensive r&d and validation work in terms of its development. So you put your topic in. We've got some quite sophisticated search algorithms that go out and we search for data across two data frameworks. So our first framework we call a leading indicators framework and this is a macro framework that looks at your topic and it looks at macro influences of behaviours around that topic. So it might look at the economic conditions, it may look at a regulatory environment, but it will look at a framework that indicates macro influences of your topic and consumer behaviour around that topic.
Speaker 2:The next thing it will do is it will look and search at a micro level and pull really specific information related to your topic. So this might be market research reports, it might be specific consumer conversations that it can surface, but at this point in time it's surfacing publicly available sources of information, so it's not going behind paywalls or other things. Once we have that data, we bring it together into what we call and we process it and bring it into a knowledge lake. Now, if you're a brand and or an agency or a panel company and you sit on big sinks of proprietary data, so that data could be years of survey research, it could be academic articles, it could be reports that you've paid specifically for. You can add all of that data into this knowledge lake and it will continue to build up and enrich that knowledge lake. And from that knowledge lake we then use it to essentially segment the market relative to your topic, so to create the different groups of consumers that exist in that marketplace.
Speaker 1:Groups of consumers that exist in that marketplace. So I guess two questions, because I'm just trying to check that I'm understanding it fully. So you've created the segments out of the knowledge lake and within each segment obviously you've got this whole body of historic information around them. So you're then interrogating that segment to answer questions based on all the information that's attached to the segment is. Is that a fair summary of the of the first part of it?
Speaker 2:it is. It's a little bit more sophisticated than that in terms of how the data comes together and how it so there's a whole lot of predictive modeling that sits in this particular product. So, based on the knowledge lake and the information we have in the knowledge lake, we then use predictive models to generate data based on the question you're asking. So we are using a whole lot of different data, we're augmenting it together and using it to generate new data. So, that's ultimately how it works.
Speaker 1:Yeah, I see it. So that's what I was wondering is so it's not just a stagnant leak that you're pulling on, you're actually projecting off it.
Speaker 2:Yes, yeah, it is a big predictive model that sits within the virtual audience tool and it uses multiple sources of data and there's a lot of steps in the process to get to the output, but we do it as speed, efficiently as we can without impacting inequality.
Speaker 1:Do you always see a role for humans in it? So for every project, do you always have a human component of the study, and do you think that will continue to be the case?
Speaker 2:if so, right now a human is in the process, because a human's kind of putting the, the question in um and it's looking at the output and kind of taking the output back into the business from a strategic decision making point of view. So, um, so there's definitely a human in the process in terms of activating and using the tools, and I think it's an interesting question. I'm speaking at a conference locally on Monday actually, and my colleague was pulling up some slides and one of the slides had on it like AI won't take your job, but a human using AI will. And I think that's a really important point here, that AI is not going away. Synthetic data is not going away.
Speaker 2:The ability to do insight generation a lot faster and a lot more efficiently and, I would say, actually a lot more accurately, is also not going away.
Speaker 2:So, as human researchers, we actually need to adapt, be really open-minded and bring these tools in and use them day to day.
Speaker 2:So the Yabble analytics tools, for example, allow you to process data a thousand times faster than a human. So if you've got a competitor kind of processing data a thousand times faster, how do you compete if you're not adopting these tools? So I don't think the role of the human is going to disappear, but I do think it's going to change. I think from a data collection point of view we will move to a lot more intimate conversations with consumers. So probably those one-on-one, those much deeper contextual inquiries, the really kind of deep emotive understanding of humans, is kind of where that fieldwork component will go as more and more synthetic data becomes, more and more reliable at a kind of scale, quantitative level. From a researcher point of view, I think the ability to talk at a really strategic level, understand how to bring multiple sources of data together and use the right tools to analyze those and make sense of them, is really where the benefit of the researcher will sit in the future.
Speaker 1:Yeah, I think it's a really good point. I wasn't actually thinking about the stakeholder on the agency or the brand side of it, but, as you say, it's a very important component. Everybody should be using these tools or they're just going to get left behind, realistically, aren't they. Also interesting that point around keeping a human element within your data sources, but almost kind of like zoning in, I guess, on what's the most relevant or the most kind of fertile source of information. You know, rather than just feeling like you've got to do a sample of 1,000 people because it's statistically robust. But all of us who've worked in the insights space, once you actually look at the data, particularly when you look at the open ends, you can actually see, say, out of that 1,000, you might have 200 who really really contributed in terms of they paid more attention to the survey and they've given detailed, open-ended responses. So do you think it might actually improve the quality as well of outputs?
Speaker 2:I absolutely think the quality of output improves through the use of AI.
Speaker 2:So in particularly, you know and there's some things that you know the synthetic data products in the market are very early, so they will continue to improve. But one thing that they provide that a human doesn't provide is they provide a lot of context and understanding at a broader market level to answer a question. So there's a lot more enrichment in that answer than there is with a human, who can often be very lazy, particularly in a quantitative survey and answering that question. So I do think we get richer data, provided that it's been created in the right way. And that's where you know this is a new part of the industry. There will be a lot of tools out there that are kind of quick and fast and dirty, that kind of don't provide that real kind of richness and quality or reliability. But then there'll be other tools that have been really well researched, validated, kind of have a deep understanding of the insight requirement and can provide a real richness, and certainly for us, yable is on that side of the development of these products.
Speaker 1:Yeah. So, catherine, the million dollar question, maybe literally, how do you sell this in? Because there must be a lot of skepticism. I know the GPT series will have helped, but do you find brands and agencies are pretty receptive or are they quite skeptical and a little bit befuddled by this new approach?
Speaker 2:you know, quite skeptical and a little bit befuddled by this new approach. It certainly changed over the last 18 to 24 months. So, obviously, when we first started selling into generative ai and nobody knew what it was and there was a lot of fence sitting and confusion, the launch of chat gpt absolutely helped to kind of drive people's open-mindedness towards these tools and we went from kind of in late 2022 to people kind of not really understanding to late 2023, you know where the brands and agencies are kind of coming to us going. We need to adopt, kind of help us adopt tools into our businesses. So certainly from an adoption point of view, I think 2024 is going to be a really significant growth year of AI adoption, particularly in analytics For data generation.
Speaker 2:It's a new area and I think there's a lot of. So we're really confident in our data and we've got Fortune 100 customers working with us testing it it kind of using it to ultimately replace traditional research. And I think over the next 12 to 18 months, as more of those case studies become available, as more of those brands are kind of willing to stand up on stage and talk about the adoption of this new way of conducting research, then the marketplace will gain more and more confidence as well. But it's absolutely the future and, as a researcher, once you start working with it and you kind of appreciate the freedom that it delivers you and your ability to kind of move at the speed of thought, absolutely you don't want to go back to traditional methods. I like that phrase, by the way move at the speed of thought. You don't want to go back to traditional methods.
Speaker 1:I like that phrase, by the way move at the speed of thought rather than the speed of my. Well, it's probably limited by the speed of my incredibly slow typing, still, unless you've got some form of audio interface. So what's the balance at the moment, then, in terms of brands as opposed to agencies who are using you, and where do you see that balance moving in the future?
Speaker 2:So for us, brands are still a predominant customer base, and that's by design. So we've always been, even from day one, a platform that was designed firstly for the brand customer. But particularly as the generative AI market knowledge and awareness is growing, we've had more and more agencies come to us and ask if they can kind of start adopting our tools, which has been really nice because I do think there's a real opportunity within agencies to drive a lot more efficiency using generative AI within their organizations. So I'm seeing adoption across the board. One thing I hypothesized and I was actually wrong about One thing I hypothesized and I was actually wrong about was I hypothesized that our virtual audience product or our synthetic data product would be kind of only a product that really appealed to brands, and I've been wrong in that space. So I would say interest levels equally as high among agencies as it is among brands, which is really nice to see because often agencies are later adopters in this industry.
Speaker 1:How does the business model, the charging structure work? I mean, I see it's mainly based on a subscription-based model, but is there an ad hoc?
Speaker 2:component as well. So, yes, there is. So we've. Obviously we're a volumetric-based business, so the more volume you put through our engines, the more affordable we become. But having, I guess, worked in the industry for a very long time, I very much understand the project-based nature of how the industry works and, particularly for you know, there's a lot of independence in this industry and we wanted to create a product that kind of gave access to the small players as well as the big players. So we have a we do have a pay-as-you-go feature that's designed specifically for the smaller agency, the independent, that allows them to adopt technology and kind of keep pace. And then for our bigger brand customers, our bigger agency customers who are doing volume through the engines, we have subscription models that ultimately allow them to get better pricing based on those volumes.
Speaker 1:I'm conscious of time and you've probably got to get on with your day, but I just had a couple of other quick questions with your day. But I just had a couple of other quick questions, One of which was around an idea of if you were to characterize usage or synthetic or augmented data, or however we want to kind of describe as almost like good uses and bad uses. What would a good use be as opposed to a bad use?
Speaker 2:So I think and I can only talk for the Yable data products so I think it's less about good and bad and more about the capability of the product right now. So, like any form of generative AI product or product that's being developed with kind of cutting edge technology, the capability of the product will change over time and with AI based products it's changing really rapidly. So if you were going to come to the Yable platform right now today after this podcast, because you were convinced you needed to, which is fantastic so where the synthetic data product that Yable provides would work really well for you is where you've got traditional desk research. So absolutely we can kind of replace that. We have got kind of a more kind of exploratory piece of work that might be about understanding market behaviours, might be about understanding brand attitudes or experiences, might be about testing kind of basic concepts and ideas and across lots of different audience groups. So we can work B2B, we can work consumer sample all of those things and so those more kind of exploratory, initial stage qualitative or short quantitative projects. So the current virtual audience product is fantastic for that.
Speaker 2:Over the next kind of six months we'll be rolling out more sophisticated offers within the product. So the ability to run full quantitative questionnaires will become available shortly, the ability to kind of develop tracking, brand tracking studies using synthetic data so all of that will become available over the next kind of six to 12 months. So it's less about good and bad and more about kind of looking at the capabilities of the product and deciding is it right here today, does it have the capability to do that? And, as I said earlier, there'll always be some use cases where actually it makes more sense to maybe do kind of really deep one-on-one interviews with a consumer as opposed to doing a piece of synthetic sample. But that's no different to what it is today when you're choosing whether you do a kind of qualitative piece of work or a quantitative piece of synthetic sample. But that's no different to what it is today when you're choosing whether you do a qualitative piece of work or a quantitative piece of work.
Speaker 1:So if we were to just move on to a quick far round, I was interested in what advice you'd give to young people, but particularly young women looking to make their way in the insights world. So it's a subject that's been discussed on quite a lot of the interviews we've done as part of this podcast, and the consensus seems to be that we're certainly making progress as a sector, but there's still a way to go. So what's your take on how organizations can more effectively encourage and promote young women within this sector?
Speaker 2:you know, encourage and promote young women within the within this sector. So I mean being a um, an advocate and a champion for for women in business, is definitely something I'm really passionate about, and yabel is a, a woman-led business, so myself and my co-founder are both women and we have our head of product and technology is also also a woman. So I think think for me it's a really important part of any industry and I think diversity in any business is very important. So I think if you're coming into the industry right now, it's a very different industry to the one that I started in. So my advice would be start reading, learn, experiment with all of the different types of AI tools that are becoming available and think about how you can use them to become more efficient. But equally, I think the role of the future researcher will be different in terms of your ability to be bold, and what I mean by that is you need to be able to talk strategically at the executive table and to have the weight and confidence to kind of create strategic strategies and have businesses buy into those strategies.
Speaker 2:So often, as an industry, people within the insights industry tend to be quite introverted, but I think the future research needs to be less about the data and more about the storyteller and the strategic kind of voice in an organization. And I know the industry talks a lot about storytelling and how important it is and all of those things. But the future is different, because it's not about the data telling the story. It's actually about you as an individual kind of bringing people along that journey and having because you're going to have infinite knowledge. So AI will provide you with infinite knowledge. And then it's about how you use that knowledge to create sway in an organization.
Speaker 2:So if I was a young woman entering the industry now, I'd spend a lot of time ensuring that my communication skills were really strong. I was really confident in speaking, and particularly speaking with big groups of kind of corporate businesses or big conferences. I'd make sure I had a really good understanding not just of the tools that are available but how AI works and where you can then apply different models and different techniques to different types of data. Because once you understand the ins and outs of how it works, you can then really start to innovate and think about the best way to use it. So that would be my advice. I think it's about kind of being bold, being more extroverted and having a lot of confidence in and around AI and how you can utilise it.
Speaker 1:Yeah, thank you. I think those are all great areas of advice. So, on a personal level, I was also wondering just about how you manage everything. I mean, as you said, you're based in New Zealand, you're working across multiple time zones, you're heading up this really innovative company, so how do you manage to maintain work-life balance? Would you have any recommendations there in terms of how people can achieve work-life balance?
Speaker 2:So yes and no is the answer to that one. So it depends what you want to do. So I mean, the YAML business is a highly ambitious business in terms of what it wanted to achieve and, especially coming from New Zealand, in order to kind of be known on the global stage, it does require a lot of energy and effort and pure resilience, to be perfectly honest. So I would be lying if I said I didn't work a lot of hours, because I absolutely do, but I love what I do and I'm incredibly passionate about it. So often it doesn't feel like I'm working, but it is important to take time out and we have always myself and Rachel always kind of worked and ensured that we have permission for ourselves and also our team to take time when they need it to do things that they like to do. So for me, you know, I absolutely every morning an hour and a half, I will always take my dog out for a big walk and that's a really important part of my kind of mental health and well-being. It means I get moving, it means I get fresh air, it means I get some thinking time and reflection time, and then for Rachel, she does other things that are really important to her. So as her daughter was growing up, it was, you know, I really want to be able to coach the netball team, so making sure that she had time available for that.
Speaker 2:So I think, on a personal level, giving yourself permission to take time is really important. But then as an organization especially as an organization grows I mean the Yable business. I started this business literally from my kitchen table and it's now grown into a global business. That's kind of leading the industry from an AI point of view, and you have to let, go and let and trust your team, and so I'm really big on succession planning in any business and also in letting I hire incredibly talented people and I need to let those people do their jobs, because if I'm trying to do all of their jobs for them, then I'm gonna obviously burn out. So for me, it's about trusting a team, you know, letting them do the job that you've asked them to do, and then giving yourself permission to take the time that you need to ensure that you're maintaining your health and wellness.
Speaker 1:Yeah, again, great advice. And, by the way, I think it's very difficult for founders to do that in some cases to start hiring people who are actually smarter than you in certain areas and not doing everything. Final couple of questions. Sorry, a slightly cheeky one, but I do like to ask it. So what would your partner say? Your best and worst characteristics are?
Speaker 2:It's a really hard question. So I probably I mean best characteristics. I mean I'm probably quite highly empathetic, very generous, and I always try and kind of see the best in people and kind of view new things from from their viewpoint. My worst characteristics are that I probably work too much, to be honest, in terms of hours of the day, and this is a good and a bad quality. I'm incredibly determined, so I just won't give up, and so if I really believe something can happen or we can do something, then I will absolutely just be so stubborn and determined to make sure that we get there. And that can be both a good and a bad quality.
Speaker 1:Final question was around your favorite or most impactful book or recent book. It doesn't have to be a book. Actually, it could be a piece of media, like you know, a film or a podcast or TV series or anything like that.
Speaker 2:Do you know? This is a slightly embarrassing question for me because I mean my every day when I'm working, I'm constantly, I guess, learning because of the industry we're in and kind of trying new things. So when I'm consuming literature or TV or what have you or, I tend to go in the escapism route. So probably my most enjoyable book that I've listened to recently because I do a lot of Audible is the Thursday Murder Club series by Richard Osman, a good British comedian. Just super great escapism and ability to turn off, and I guess that fits with that whole kind of work-life balance thing. But I am constantly like for me, a lot of the things that we're learning and doing in the business is learning about ways to obviously work with models, with AI models, in order to create really reliable synthesized data. So that's kind of what I'm reading and doing a lot of at the moment.
Speaker 1:Well, I'm glad you also have a chance for some escapism. Thank you so much. It's been really really fascinating. I've learned a huge amount and incredibly appreciative.
Speaker 2:No problem, thanks, henry, it's been great.
Speaker 1:It was great talking to Kath and I hope you enjoyed our conversation. Lots of thought-provoking stuff in there and some very hot topics. It's really interesting to hear from someone who's very immersed in the technicalities of the subject. From that perspective, I also thought her point around the importance of humans to interpret in a world where knowledge becomes infinite was really well put. Thanks, as always, to Insight Platforms for their support. Thank you to Kath again for the interview and thanks to you for listening. See you next time.