Building AI Before It Was Cool: How Bad Data Quality Sparked a Revolution

Kathy Cheng started as a focus group translator in Shanghai, moved to Canada, couldn't get hired as a moderator, and reluctantly switched to quantitative research—which she hated. Her frustration with "bad data quality" led her to launch Nexxt Intelligence nine years ago, building conversational AI for surveys before ChatGPT existed.

When generative AI exploded, Kathy was perfectly positioned. But unlike most AI companies, she refuses to say Inca "scales qualitative research." Instead, she's focused on making quantitative data more engaging and reliable through conversation. Thanks again to Kathy Cheng for being our guest. You can connect with her on LinkedIn.

Transcript

Leonard Murphy: Hello everybody. It's Lenny Murphy with another in our CEO series of interviews. and today I am joined by Kathy Chung, the CEO of Nexxt Intelligence, the company behind Inca, which maybe you've heard of. Happy to have you.

Kathy Cheng: Thank you so much, Lenny. So happy to be here.

Leonard Murphy: for our audience. We were playing kind of tag for weeks trying to get this scheduled where Kathy has something come up and I have something come up so it's good that we finally made it the stars have aligned for us to have this conversation.

Kathy Cheng: Thank you so much for having me.

Leonard Murphy: I'm thrilled. For those who may not know about Next Intelligence and about Inca and about you you tell us about the company and then we'll go where the conversation takes us from there.

Kathy Cheng: Maybe I'll start with a very brief introduction of Inca maybe then I'll talk about myself but everything will come together hopefully. So Inca is conversation maybe these days conversational AI seems to become more and more popular now. a lot of people think about conversational The first thing they think about is to automate qua to skip the human moderators and all that. And interestingly we actually didn't think like that. We wanted to create conversational AI for quant. that's probably where my story might fit in. I started my career quite some time ago and when I was in Shanghai in university I had this part-time job as a simultaneous interpreter. I wor many different organizations focus groups. I did a lot of them. I really enjoy them. I would have my own little room watching this group happening. I would just talk aloud just simultaneously I'm translating the discussion to clients. At that time most of the clients were from outside of international. So pretty much every single focus group would have media translation. So I did that for a very long time. and then I got an offer from global ad agency. I was quite excited. I went to see my mentor it was at AC Nielson at that time. So I went to talked to my mentor and…

Leonard Murphy: Okay.

Kathy Cheng: I said I'm gonna join this ad agency. And then she said, "No, you are not. You're going to become a moderator." That's what she said. So really, I am just grateful for her. I think I probably didn't know myself enough at that time she said that I just thought, " okay." I just really loved them all. I really enjoy working with them. I thought why not? I'll just be trained to become a moderator. But looking back, I think I'm the I would say more nerdy kind of person. I really enjoy understanding things. advertising is really interesting. It's probably more glamorous than my personality. So yeah, I was very very very grateful that she pointed me to the right direction. So yeah, that's how I was trained to become a moderator possibly because I watched so many focus groups. So yeah, I was trained up fairly quickly and I think I did quite well. I really enjoyed my job and then I came to Canada with my family. and then as you can probably imagine for someone who was doing moderation in a different different culture to continue to do what you think you could do as straightforward, wasn't easy. Nobody would hire me as a moderator. So I learn So that's how I started to do uant. I had to learn quant. I did it okay I guess but I have to say I never really enjoyed it. I just felt I just never could write the wise in my report.  And then over time I remember one time in a meeting I just felt I really couldn't stand in front of clients talking about some quant numbers tracking study trying to find some hypothesis why didn't the changes were happening but in the back of my head I totally knew that it was just bad data quality it just didn't make any sense. So yeah in that moment I just thought I know I can't do quant anymore this just doesn't make sense to me. So that really was the motivation. so there was one time I had this very intriguing opportunity. I needed to find a solution to do better. I just thought yeah totally I would want to take on that challenge because I think that's necessary. and then it became quite obvious to me that to do quant better we probably needed to make quant a little bit more quark. That was really the intention. So that's why I said we built in to add that conversational capability to not like maybe the perception at least these days to scale qualitative research.

Leonard Murphy: And then along came AI, right? Yeah.

Kathy Cheng: Yeah. Yeah. Yeah. that's just so I started the business nine years ago. Obviously not visionary enough to predict that generative AI thing. Obviously. yeah,…

Leonard Murphy: Yeah, me either. I knew something was going to happen that would make all this easier, but I did not envision that exact form. So, go ahead.

Kathy Cheng: That's just crazy. so yeah, so we started with the idea of making survey a little bit more conversational. but with the technology available at that time obviously we had to feel conversational pretend it to be conversational pretend there's another agent or someone on the other side listening to you. So we would program the acknowledgement the follow-up questions to make it a little bit more conversational. So we actually did spend a lot of energy training our own At that time we wanted to build AI a lot of honestly we're based in Canada we have a lot of support from the government and they really encourage R&D so that's how we were able to otherwise like a small startup how would that even be possible? So yeah we did put a lot of R&D hours into it.  I think before generative AI, we were already quite okay in the context of real AI. whatever we built at that time was kind of AIish smart enough to be able to handle some simple conversations, but we knew that it would never be truly intelligent. there was nothing truly intelligent available. but that's a blessing because it's not as good as we'd like it to be. We felt we needed to compensate and how we would compensate was that at least the ideas we had were to add more qualitative principles into quant for example projective techniques personally I just really love projective techniques like everything can be a projective really so for quad that's pretty obvious that's why we need moderators we know that sometimes people don't want to answer our questions may not be because they don't want  to it's just because our questions are really difficult. So that's why we need moderators. We have tools. We have ways to enable people to access and articulate that kind of information. In Quan, it's actually even more difficult. People are so used to check boxes and suddenly you ask them how do you feel about certain things. We can expect people to suddenly access emotions. so that's where projectives could be really helpful. So maybe we always wanted to add projectives but I would say the limited capabilities of AI was a motivation. We just felt okay maybe AI wouldn't be as good as we'd like it to be. we should just add more qualitative things. Also the engagement we wanted to build the UI as modern as possible as close to a real chat UI as possible. as close to we live our lives as possible. Right? We don't do surveys every day. we chat with people. So yeah, a lot of the other things we were building partly was…

Leonard Murphy: All right. Yes.

Kathy Cheng: Because we knew we couldn't make AI as good as possible. So we built our own conversational AI framework, built all the whole UI thing, projectives and all that. And then suddenly one day there's this thing called chat GPTP. Yeah. And then we looked into it immediately we knew this was the AI that we were looking for. And then we didn't see it as a threat at all because we knew that when the capability becomes available if we can plug it in with the com conversational framework that we had built already it will become very very strong. So yeah that's what we did. We quickly incorporated GPT technology we use multiple large language models but we incorporate large language models. We use our own conversational AI framework which we call neuro symbolic. I believe it's really important.  It's really not out of the box just large language models. It does require a lot of human expertise. So I can't thank GPT enough large language models enough. minimum we kept saying at the minimum that was the best marketing tool for us. otherwise people would be very hesitant, what do you mean by But very fortunate. Good timing. Yeah.

Leonard Murphy: Yeah. it seemed from just observationally like I was aware of you, but when Chad GPT released then, here's Inca, right? And it seems like you really grabbed on to that opportunity and even though you had been there for a while in the overall space

Kathy Cheng: Yeah. Yeah.

Leonard Murphy: But to use that momentum and excitement to engage much more you were in the right place at the right time and…

Kathy Cheng: Yeah. Yeah.

Leonard Murphy: That's very cool. let's talk about the downside though, Because here's a concern that I have had is that it is the great equalizer, right? It's decreased the barrier to entry. it is relatively easy now to create something that may not have the quality and the depth that you had built with Inca but on the surface it passes the sniff test right and…and I'm sure you've seen that as well there's hundreds of new entrance into the market and…

Kathy Cheng: Yeah.

Leonard Murphy: Everybody has some type of AI even if everybody everywhere. but without necessarily a thoughtful application that is appropriate for the needs of research. now I'm seeing more of those companies I am seeing more that are okay you did your homework right that this pretty good application.

Kathy Cheng: No. Boom.

Leonard Murphy: So you were in the right place at the right time, took advantage of that, and now there's all of these other companies that are emerging right behind you. that created a challenge for differentiation for you. how are you adapting to that different competitive set that it really is?

Kathy Cheng: Yeah, totally. Yeah, sometimes we find it's very amusing. It's really every day almost like there's something new coming up in a really.

Leonard Murphy: Yeah, it really is.

Kathy Cheng: So, really people say this is just This a new wave of internet like everything just like it's one day we probably won't be able to see anything without AI in it. but I think overall It's good in a sense that it creates awareness. hopefully it makes AI less scary. because if everything is already AI it's just like the internet everyone's already on the internet. Why are you so worried about it? So in a way it is good but like you said yes we need to do more to differentiate ourselves maybe not too long ago we were at least one of the few if not the only conversational AI so now we need to do more again in a way this is probably good we realize we really need to think through what we stand for what's our purpose how we really try to continue to stand out in the crowd. and we think It has become quite clear now. We just really need to double down. We just need to be very very clear in our messaging. Maybe our messaging hasn't been very clear because the competition wasn't as much at least. but now we feel like we need to be clearer. I still don't feel very comfortable talking about scale replacing qua. I don't know. I find it personally I can't say that. I'm just deeply qu I can't say that. I can't make myself say scale mass like all that we do have a lot of clients using in kind of qua at certain scale that's totally fine but I think It's a different kind.

Leonard Murphy: Yeah. Yep.

Kathy Cheng: So to me to use scaling qu as the core message it's my personal problem. I just find it very very difficult to say that. But going back to our original origin story, we wanted to fix uant. That sounds probably a little bit too big, but at least we try to contribute to a lot of the innate problems that Quant has.  And then we do believe we have seen evidence that if we add engagement, if we add that probing like I call that we do see richer information we do see more explainable data we do see more coherent data. So it's beyond just the deeper insight at scale. It is about better quality of quant overall because when people are engaged you expect them. Yeah. So we feel like we need to really focus on that. Because fundamentally I do believe core even though yeah we now have the ability to scale but the real in-depth core I think it's a very different mindset it's a whole different thing it's like I keep thinking back of my mentor she said to make a focus group successful three things have to happen first is the moderator has to be  happy. The morator has to feel like they have asked all the questions, they received all the answers and feel very satisfied. And then the participants have to be happy. They feel they must feel they have shared everything they wanted to share. and then the clients have to be happy because they're sitting in the back room. They have to feel their questions are answered that they can move forward and they have more discussions beyond what they heard. So it's simultaneous. So there's a lot more going on than just getting those verbatims. It's the connection. It's the instant evolvement of ideas and findings and all that. So yeah, that's probably just me. I think yes,…

Leonard Murphy: What's up?

Kathy Cheng: We can scale qual but it's very complicated. It's so human. plus now there are just so many companies technologies trying to disrupt that. We should probably just stay away from it. We should just stick to what we wanted to do. We wanted to make quant.

Leonard Murphy: That's great. And I get exactly what you mean, And so it's actually really interesting that maybe we need to be more precise in my lang in our language because I'm just as guilty as anybody saying qual at scale or quants quant be more qual but the reality is as I'm listening to your thing when I say what I mean is that the form factor of a quantitative exercise is going to be more conversational. Right with more probing, more of a back and forth process. the while also including kind of traditional question types because example I always use a conjooint exercise. I can't imagine a conversational conjoint…

Kathy Cheng: I think it has to be part of it.

Leonard Murphy: Right? What a pain in the butt that would be. so maybe we'll see…

Kathy Cheng: Eventually when quant really is 100% conversational, it is just part of it. Yeah, it is.

Leonard Murphy: How we get there, that I know it was done back in, the Katy days as well, so I'm glad you pointed that out because I think it's the form factor of the engagement mechanism that changes…

Kathy Cheng: Yeah, I think it's the

Leonard Murphy: But qualitative is still a very in-depth process to try and get to know a person and a topic at a much deeper deep.

Kathy Cheng: Yeah, one at a time. So, yeah…

Leonard Murphy: Yes. 

Kathy Cheng: I think that's the difference. It's like quant we probably do care about the aggregated information more but I see a friend

Leonard Murphy: My family just left on an errand, so the dogs now, they're just used to coming into my office. So, that's Iggy Pop.

Kathy Cheng: So yeah…

Leonard Murphy: Go ahead.

Kathy Cheng: Qual, yes a lot of times I do also believe in that scaling qua is necessary because that perspective is important and sometimes we want to compare different segments and all that scale is necessary but the real power of qua I believe it will stay exist is that single insight right sometimes it's about that one line that really changes the world so it is about that single attention the single person interaction. and then on top of that some certain skills can provide that perspective but that's all good but at the end of the day it's that one single insight that can really make a difference. yeah so I would think the language we're still working on it because it's really really difficult. It's really hard not to say co but then it's also really hard to say co and quan. It's really difficult. We're still working on that. But I think the key difference is the dynamic for quant has always been just single dimensional. People just do their own exercises but it's a information exchange. it's a dialogue. it's two It's not one way. So I think that's the most important thing that we should introduce we should bring to quant to change the dynamic so that people feel that they're not just blindly doing their own things. there's someone on the other side that truly truly appreciates what they say and truly interested in what else they have to say. So yeah, we believe that's the fundamental shift we will see happening more and more in the quant world.

Leonard Murphy: I think it's happening now. the well underway so you've been in the thick of things right talking to clients building a product and…

Kathy Cheng: Yeah. Yeah.

Leonard Murphy: Selling that product I have this kind of macro observational view my sense is that the advent of chat GPT and…

Kathy Cheng: Mhm. Whoa.

Leonard Murphy: Everything that has accelerated a change state in a way that I had not experienced in my years in research. and that continues to accelerate kind of exponentially. so the way I think about is fundamentally the need to understand people and to utilize data to drive business decision-m that's fundamental. That's not going away. That's probably even growing. But the process and the tools that are used to accomplish that and even the constituent stakeholders involved in leveraging that changing rapidly and within literally just a few years I give a kind of a three-year horizon from here we are in early 2025 that the norm for the…

Kathy Cheng: Mhm.

Leonard Murphy: What we think of as a research process will be utilizing tools like Inca that don't look like our parents ain't our parents what we grew up with right in research qual overlaid with a service capability which I assume that's more what next does as well to kind of all right here's what it means and here's the implications of that and that's just going to be the norm across the board but everything happens in the middle is going to be driven by some application of artificial intelligence across the board.

Leonard Murphy: Are you seeing it now? am I being overly optimistic or pessimistic or however you want to view that something new. Yes.

Kathy Cheng: I see that as optimistic. Yeah, I think the speed of change is really astonishing. maybe not even in terms of years. I feel like every conference we try to go to as many conferences as possible. it's almost like every conference you feel there's some change. what I've been wanting to do? know how in the past we always have syndicated study to check the pulse of the consumers that kind of thing. I think it would be really interesting. What if there's a way for us to get a transcript of all the conversations going on in a conference and then we really check the polls one conference from another. I feel like this probably will be really interesting because this time I'll say the recent conference we were at Corks Los Angeles I really felt things were really changing. I felt more voices about rapid changing changes are needed.

Leonard Murphy: Yeah, that is interesting.

Kathy Cheng: Rapid changes are happening like all that. So yeah, I think that would be really interesting, wouldn't it? If yeah,

Leonard Murphy: It used to be I haven't done it in a long time. I stopped even going to some of my own events, but I used to do a write up. So, going into the event, I would kind of say here's what's happening going into it. and then afterwards, I would summarize a lot of those themes. but you're right,…

Kathy Cheng: Yeah. It's really

Leonard Murphy: We actually could do that as long as we record the sessions and at least the sessions we probably can't go around ease dropping on everybody but I use events as a data collection tool. So to inform my own perspectives right for that but you're right I think that it was a gradual thing. I think you're right. It six months is a long time now. So I want to be we could go on for a long time chatting. it's really been a delight having this conversation but I want to be conscious of your time as well as the audience. so you build this company. It's very innovative. You're in the right place at the right time. That has its pros and cons because also the technology that accelerated you is also potentially a threat in some form or fashion, right? at least competitively it's more crowded marketplace. What's next? no pun intended.

Kathy Cheng: Oops. I love that. Yes. I think the tools will become more sophisticated.

Leonard Murphy: But yeah, here we are beginning of 2025. Where do you see yourself and the business being in the next say two years that's a foreseeable horizon I think

Kathy Cheng: Client's requests will become more sophisticated. now we're saying there one conversational AI can be different from another Certain conversational AI is built with more thought through framework, more opportunity for researchers to be part of it to brief the AI like all that. But at the end of the day, it's conversational AI. People think about conversational AI. the conversations or the work that we're doing with clients now clearly show that it's not going to be just one thing called conversational AI. It's going to be more domain expertise related. we have been training custom built chatbots for different applications, different types of research.  We have been helping clients train AI coding for very specific areas like social values for example such a complex topic. and then people say one thing it could mean this social value it could also mean another social value. traditionally we do need a lot of human experts to really kind of decode what people say and to understand what kind of social values are reflected behind that. What if we could turn this into a process where AI can help us a little bit because only when we do that this type of insight social that's the deep down we can finally understand people what drive them to make certain decisions what if we can get a little bit of help through AI so that kind of information can be more accessible for more companies organizations so yeah I think that's definitely trend just a small large language models that's happening

Leonard Murphy: Okay, that's

Kathy Cheng: I feel like a small conversational AI. I really love the idea. I think deep down I'm just so cool. I really like that small conversational very purposeful, very specialized. I feel like that's probably what's next.

Leonard Murphy: That's very cool. I'm thinking about my own interactions. I turned Grock 3 loose on all of my content and said analyze it. Tell me about me, and amazingly well done. But what was really interesting was the nature of the interactions with the interface, right, of the sessions changed. it adapted in real time to…

Kathy Cheng: Hello. But ultimately more human,…

Leonard Murphy: What now knew about me. and the nature of the recommendations were different, It kind of got a sense of, you're a researcher. this is your context and how you think about things. Point being, I could see that process of it being a very adtive, very focused, almost scary personal. I'm not totally comfortable with the experience yet. And…

Kathy Cheng: Isn't it? It's like you really treat you as an individual human at the mo.

Leonard Murphy: There yes but also it's scary human right to an extent right I'm like no you're not yeah yeah it interesting times so said we wanted to kind of also touch on any lessons learned obviously you have had a really interested entrepreneur reneurial journey.

Kathy Cheng: Yeah. Yeah. Yeah. A little nosy. Yeah. Yeah. build a good team.

Leonard Murphy: Anything that you would pass on to other folks that are emerging into your leadership roles or entrepreneurial roles that here's my wisdom. Take this to heart.

Kathy Cheng: I think that's really the most important. It's probably sounds pretty cliche. but really it's so important especially for people like me coming from the research world. Building a technology company that's just something that I still don't know. so we need a good team to work together on things. Sometimes it's really luck seriously I mean things already have changed when I started to convince an engineer to work on something market research with the perception of surveys it wasn't as easy…

Leonard Murphy: Yeah.

Kathy Cheng: But now we are cool seriously market research is becoming more and more interesting now so hopefully this is already becoming easier for newer entrepreneurs in our space I think the second thing I would say financial literacy I guess again again right we are researchers we're so nerdy we're ve very good at looking at numbers but we don't necessarily know how to read financial reports I think yeah that's one of the things I'm still learning and I think as the company grows that's definitely critical I felt like if I had a better financial literacy I would probably have seen opportunities and…

Leonard Murphy: I hear you that I often describe myself as a recovering CEO from kind of the first half of my career of building and…

Kathy Cheng: Challenges at least very differently probably would have been very helpful. but still it's a learning process.

Leonard Murphy: Running research companies. And that particular piece of things I was worse at than anything else and also didn't like doing it probably because I wasn't very good at it. And gleefully enjoy not having to deal with those things from a management standpoint. So I get that entirely. Kathy, is there anything you wanted to touch on that we did not?

Kathy Cheng: I really enjoyed the conversation. I think if we had an opportunity to talk again in two years like what we just said, it would be really interesting to see where our industry will be at that time.

Leonard Murphy: It's a date. So, there we go. we're good. We will definitely do that. where can people find you?

Kathy Cheng: Let's do it. It's LinkedIn/Cathy Chang. I believe it's fairly straightforward. our website is Nexxt Intelligence is the company Inca is the product name. Yeah, these are my email Kathy with a K. k a t h y at nexxt.in.

Leonard Murphy: That's great. Thank you so much. This was really I'm glad finally the stars align for us to have this conversation. It was great. Appreciate you sharing your story. best of luck in continuing to innovate and drive change in the industry.

Kathy Cheng: Thank you so much, Lenny.

Leonard Murphy: No, Thank you to our listeners. without you, Kathy, I probably would not have kept really pushing to make sure this happened. thank you for giving us a reason to connect and have this conversation. Hope that you enjoyed it. thank you to our sponsors. Thank you to producers, and we'll see you on the next edition of the CEO series. That's it for now. Bye-bye.