Reimagining the Future: How Insights Can Make AI Better (and Vice Versa)

Hello everyone and thank you so much for joining us. Uh my name is Luke and I will be hosting the last presentation that uh that we have today. [gasps] So just want to go over a few ground rules before we get started. Uh please try to be [laughter] we're kind of at the mercy of technology. I just had a little flub there myself trying to advance the slide. So please be patient with the rest of us if the tech goes down. I think we've all been in a position where that's happened. Uh be respectful. Uh I see a lot of uh good kind of interactions going on in chat. It's you know easy I think sometimes to uh to get a little harsh or critical when you're not face to face. So uh let's try to uh be the people that you would want other people to be to you. And um in the back room, the virtual back rooms, feel free to kind of engage and interact and uh use the the virtual uh you know gathering as an opportunity to to kind of network a little bit. So uh we I am pleased to introduce our last presenter for the session. Uh we have Pavey Gupta with us and uh Pavey has comes to us with uh a very what I would call storied CB of excellent companies and brands. We've got Shabbani Coca-Cola, we've got SC Johnson. So Pavey, if you're ready to go, uh please go ahead and uh take it away. >> Thank you, Luke. Can you hear me fine? >> I can hear you. Great. Thanks. >> Awesome. That's uh in the days of virtual presenting I wanted to be sure someone can hear at least. Um good afternoon everyone. Uh let me just share my screen and this is a session I did recently. Um I was fortunate to be at Harvard Business School working with one of the professors there who's put together a genai curriculum for marketing. One of the modules there is insights uh leveraging genai. So um and I call this kind of reimagining the future. I break this up into three chapters. The first kind of reflecting on where we are today. Second one talking about where do we want to go? What's our northstar? And then most importantly, how will we get there? If you look at what's happening as far as AI is concerned, it's all around us. There's a lot of new technology, new anticipation based on some of the approaches that are being shared. However, there is definitely one perspective which is around the fact that a lot of the deployment hasn't got the kind of traction that people had originally anticipated right whether it's in terms of the viability of the genai deployments whether it's in terms of the attribution of you know growth or top or bottom line growth being driven by these Genai initiatives or whether it's in terms of even just the number of initiatives that are being sunset or shut down because again they haven't kind of proven the fact that they can change the trajectory whether it's again in driving top line or bottom line. So does this mean we sitting in an hype cycle? Does this mean that this is the value isn't there to be had or does this mean that there is definitely value here? It's just that we haven't figured out or organizations are still trying to figure out how to really capture the full essence of where that value can come together. I feel this is the uncomfortable truth as far as AI specifically Gen AI is concerned. And of course you know the entire I call it the guardian kn of like investments by AI firms in in each other is is very it's tangled in a manner that seems very difficult to unfold and that again gives you belief that maybe there is more to this in terms of a bubble or a hype cycle versus it actually proving value. There's also what they call as the AI verification tax when you know AI is supposed to make our lives easier but that's on the top of the that's what is visible but beneath the surface is this furious little work happening behind the scenes where you're just trying to verify or address some of the hallucination issues that come up initially. So a lot going on there as far as AI is concerned. Anthropic ran this project went as an initiative where they actually had a shop run autonomously by AI agents. Of course, humans were helping physically move the product. There were not robots doing that, but humans were helping physically move the product. But AI was managing the entire operations of that vending machine to see what product should it stock, what pricing should it follow, what kind of promotions or discount should they be offering. And you would imagine that in that kind of a world or scenario where everything you know the data is perfect because you have every aspect of or signal of what you know what are consumers buying and what is that you have on inventory and what is the costing as far as the pricing is concerned. You would imagine that AI could manage it very very independently and prove or create value to the organization. But in this particular experiment over a period of time the value of that enterprise that one vending machine eroded the net worth of that eroded because I think some of this was also the human factor as as this vending machine was in their office environment and some of the employees were just experimenting playing with it and ordering stuff that they normally would not consume. And this led the model to believe there was more demand in those products than there actually was. And it probably ordered more than it needed to. And then it was sitting on inventory of products that no one was buying. And then it tried to liquidate that inventory and therefore ran in a situation where it was actually eroding value. So the question is when we talk about where we are today and now bringing it home closer to the insights domain are ins are AI is AI really revolutionizing insights and the promise is three-fold usually always faster cheaper better so let's think about faster is AI really making things faster but I I I talk about this example where I feel we need to ask ourselves the question are we solving the problem or are we creating a problem? There are tools that are available today which help us generate or get to insights or concept ideas at a very dramatic pace. But the question is is that solving for what the business needs or does it actually lead to creating new challenges? In this particular case, this particular example where there was work being done to optimize concepts meant that it was very easy to come up with new concepts or new ideas which could have great resonance in the marketplace. You would question how can that be a problem? Well, it can be a problem because the execution of those ideas still would take anywhere between 18 to 24 months. And if you can generate concepts in a in in a matter of hours, then how many more times or how many more concepts can come up between now when you generated a concept and started working on it to the time you actually launched the concept. And that's when you realize that having a tool which can generate concepts really fast is great, but maybe it's answering or addressing an issue of like generating concepts but not addressing the issue of speeding up the entire process of execution. And when that happens now your bottleneck is execution and not kind of concept generation. And if your engine keeps generating even more concepts, that bottleneck continues to become harder. So I argue saying you're actually creating a problem in this case. Of course, with AI and some of these DIY tools, you can field research at a dramatically lower cost than what we used to in the past. But here again, I argue that having access to cheaper resources, is that actually the right thing to do or is that actually creating what I'll call as inside slop? Because just having access to consumers on call to field surveys over and over again could lead to lazy market research. We might already have those answers but we're still trying to find those answers from consumers and [clears throat] now generating so much input and information which can start creating confusion in an organization. So just having this cheap doesn't work for us because having it cheap means we we're producing more than what we need creating confusion. And then the last one is more in terms of this example is is it better or worse. And here I [clears throat] have a use case from synthetic data. And synthetic data this use case is oh if you have low sample sizes we can use synthetic data to boost the sample size. Awesome. Great. Um but would you have been better off being ignorant or are you better off creating this illusion because now you have a data structured in a manner that let's say it was you know less than five or 10 respondents but you've mimicked the statistical properties of those 10 respondents to make them into synthetic respondents of like 200. Um and if an organization start beginning to use this data at scale, are you misleading the organization to believe that there is statistical rigor? Even though there isn't, it's still those 10 people, right? So these are examples where you could argue that AI isn't really revolutionizing insights because some of these use cases or applications are making things worse for us rather than making it better. And the reason for that is typically with AI we are always starting with like we've got the solution we've got a technology we've got AI oh this is this is a clever tool and then we try to see how can we fit that technology into our world of insights right so if we think about where do we want to go do we always want to use this or what if we flip the order around you know what if we actually started with insights and then started thinking about you know where is where is insights how can insights be used for AI or actually start thinking about both of these inter interdependently right so I call this that infinity growth loop so what if this was moved from a linear to an interdependent process where insights could help us create better AI and AI itself helps us get to better insights because it is able to see patterns and it's able to help us scale different approaches at a dramatically fast pace or lower cost. So then comes the third chapter of how will we get there right so where are we today I think it's a mixed bag there are some successful use cases some not that successful that means the real value of AI still needs to be proven where do we want to go well we want to take it to a level where actually we recognize the fact that having insights can make AI better even as AI is making insights better right then the question is how will we get there and To get there, I call it the MOA algorithm, which is really insights can help us understand motivations or the why. Insights can help us get to the right focus through that opportunity identification. Where should we focus? because AI can be used pretty much across every every aspect of the business but it requires some more calibration to the b to that business use case. So that means we're going to put a lot of effort into what we're doing. Then it makes sense for us to have some prioritization framework to say where should we start from and that way we can tap in that opportunity and optimize that. And the last piece is you know insights can help us with the how you know how do we leverage that human understanding and that's the ability that we have and that if we can unlock that ability in a manner that taps and optimizes what humans can bring then the AI solutions that we're building will be that much more impactful. So today the motivation really starts from tech solution backwards um because we start with a tech or platform we have data and then we are trying to understand what is the business problem right however where I believe we should start from is make sure that we have the business problem outlined then look at what is the data and model and then drive adoption because now you are solving for a problem that you know needs solving. So having the insights will help us clearly articulate and identify why are we leveraging AI to solve what are we solving for and therefore the tools or the deployment of AI will actually become smarter because we know what we are solving for in the first place. It's not like we're not trying to force fit it. As far as that opportunity identification is concerned, I I have this kind of 2x two framework for all insights. And it goes on the x-axis from the known to the unknown. The y-axis goes from intuitive to counterintuitive. And this is where I believe AI can or should be prioritized on the bottom left quadrant. The bottom left quadrant is what we already know and what is already intuitive. And this is what I will call as validatory insights. in the organizations um I have worked with and other adjacent segments and categories that I have been keeping an eye on a lot of effort today both in terms of time as well as resources whether it's spend or you know u people manpower is going after validation and this is where AI can play a dramatic role because as you if You can look at ingesting all the data that you already have or whether it's structured or unstructured data. You could actually then interrogate that model to start addressing what you know and what is already intuitive. And organizations need this. There is a need for validation because the validation might be needed just to quantify the rate of change. the validation might be needed just to justify what we're already working on because there are different stakeholders and different entities that we need to be convincing whether it is internal or external stakeholders that we're working with. So if if we can leverage this opportunity identification based on where to focus on is on validatory insights then we can free up our teams to do more of the exploratory and incremental and eventually transformational work. And the last piece is around our own superpowers, our human superpowers because that is what is going to help us drive action because anything that we're doing in the space of insights or data or analytics it's we have to remember that the data analytics and insights are a means to an end. At the end of the day, we are looking to drive action, to drive impact in the marketplace. And to do that, insights are very important. The first piece in terms of our abilities is curiosity. And I love to bring my son into my presentations whenever I can. And you know, to me, what his faces and pictures bring is a reminder for all of us that the most curious people are humans. And of course, kids amongst humans, kids are even more curious than adults and grown-ups. And that is what actually makes us connect the dots and understand what's going on even better and continue that path of discovery to therefore bring value to an organization. So if we have the right curiosity to ask the right questions, then the solutions we are building are going to be that much more impactful. Of course, the other superpower for humans is empathy. It's to understand and if we have empathy, then it will actually address the entire adoption challenges that AI is facing because we will then understand what are the benefits or what are the barriers towards adopting any particular solution that is being built leveraging AI. And the moment we are able to understand that like the grief cycle of how humans are reacting to a new solution then we can make sure that those emotions are being addressed proactively rather than reactively to actually drive adoption that happens at scale. And then the last piece in this will be around the work around the integration and you know with everything going on as far as information is concerned there's so much more information there's so much democratization and this is where again the human ability to look for patterns and look for kind of this machine and human connection to drive better value comes together as well. So if you think about it, the role of insights is to drive curiosity. It's to drive that empathy and to drive connection across different sources. The role of AI will be to drive that discovery, to scale that understanding and to accelerate the connections. So these are kind of you know working together towards getting to the same goal and that's why I call like you know driving that curiosity building that empathy and driving that IQ when that comes together that is what will lead to better AIQ for any organization and that is that infinity growth loop. So the you know the the aspect of making sure that the technology that is being developed we understand who it is being developed for why it is being developed and what can make it what can drive the adoption of that technology will make things better for everybody. At the same [clears throat] time the technology itself AI itself will actually continue to make our processes more efficient and more effective. So you know that kind of continues to drive the growth at our end as well and eventually all of these coming together is what leads to that infinity growth loop. So with that I can now open the line for maybe I don't know how we are going to field questions is it u look if you can help me with that is it uh through the chat or whatever because you know I I just wanted to talk about that conceptual framework and then leave it open to see how people are reflecting on it and u what what are you know if there questions or if they're reflections on if this is sparking any thoughts then happy to have that dialogue or conversation. Yeah, absolutely. We'll be taking questions through the chat from this. We have two different sources. Uh one is the chat panel from folks in the uh back room right now. If you have a question, please uh please post it in the chat and I will draw a attention to it. Uh we also are taking questions from our live YouTube stream and if any questions come through there, uh I will pass them along to you. I am going to start with a with a with a question I had while you were going through it. Um, one of the examples you showed was sort of this idea that if a base size was low, uh, data would be synthesized to in order to increase the base size. Um, as a market researcher, you earned a little gasp on that one from me. Is that a straw man example, an example of an extreme, or is there a practical application for that uh, that someone would would do or choose to do that in that situation? It's uh it just really stood out to me. >> Yeah. And and again, this is where some of those solutions are already being touted as the greatest thing to happen to the insights industry after slice of bread. And that's where I feel like you know there are while there are dramatic benefits that any technology can bring there likewise there could be certain watch outs in the way we use these and this particular use case of saying I will boost the sample size synthetic synthetic synthetically or synthetically um to um to make it more robust. I feel that is a very weak application of leveraging AI. Now don't get me wrong, synthetic data definitely has its uses. I I feel one of the strongest uses of synthetic data is to give us the ability to simulate before the work actually has been done. And that sometimes becomes very very critical because it helps us almost visualize what potentially could be coming out of the research itself even before we actually feel the research. Um without internalizing the data but actually seeing how the flows or structure of the survey methodology that we're using. If we could synthesize like if we could synthetically create an output um then at least we can start answering the question are we asking the right questions in the first place or not. Are we able to get all those kind of crossstabulations or analysis frameworks that we believe we will use at the back end uh after the work has been finished. we could probably try it out sooner and see if we needed to refine our questioning approach or methodology. That's one use case. There could be multiple other use cases of synthetic data. >> Okay. Thank you. And and another question for me. So, you've got a lot of great material here and um you know, it's it's sort of an approach or an intellectual framework to start considering some really deep questions. Can you break it down into some practical advice? Um, imagine a market researcher or insights organization who hasn't yet put their toe in the water with AI. What what advice would you give uh practical things that they should be thinking about right now as they're starting to to to sort of think about or experiment with AI? Yeah, I think you know when I talk about the MOA algorithm that's that would be what I would highly encourage people to be asking the question and even if we cover the M at least solidly then we would have made dramatic progress from where we are. See what happens with AI I believe and I've seen enough of it in organizations that I've been exposed to [snorts] is there is a a lot of the AI is being driven by what I will loosely call as a FOMOled objective fear of missing out. So a lot of the technology people or organizations that are selling these tools are able to appeal to some of the seauite business leaders. they're able to appeal to their perspective of like hey do you want to be the seuite leader on whose watch your organization got left behind because it wasn't adopting AI and hey look at this another organization this is how they're using AI and so on and so forth right and before you know lo and behold those seuite leaders are coming to us um you know our teams saying why are you not doing this how how is it this other organization is able to do this and why are we not doing this and then it's like we start chasing our tail because instead of knowing what we needed to solve for our own business we now have a solution that someone else has deployed cleverly at times and maybe hyped up at other times but we don't know whether that's fit for purpose for us but just because that has been touted as a amazing case study. We then are chasing our tails trying to see how we can apply that or bring that to life in our own reality. Through the process, we missed an opportunity to actually go and address our biggest pain points or our biggest areas where AI could have been even more powerful for us. So that would be my advice to people to say when anyone's coming your way and asking you to do work in the space on AI ask the question back saying what is it that we need to solve for what are the business challenges that we need to solve for then do the assessment of what type of AI tools or technologies are available which can actually be leveraged to solve for that and in all likelihood it's Not it you know it might not be a plugandplay solution that is that already exists. In all likelihood you might need to adapt or adopt some of the solutions that are available to make them come to life for your context or your organization. But at least the least we can do is to start with the problem first. >> Okay. Thank you. Another question from the back room. This one is from Sesh. I like the idea of infinity loop. Can you share any project where that approach or framework has been applied? What were the learnings or challenges in adopting it successfully? >> So, it's still a conceptual framework. Um, it's something that I'm planning to bring to life through the new startup that I'm working on right now. The idea here is any solution that you're building as far as AI is concerned don't just think of it as a AI solution to solving for you know the insights needs that you have but also think of it as you know kind of understanding who the audience is understanding what the problem is and understanding what makes them tick or not tick and therefore start also thinking about you know the insights of the audience for whom the problem is being generated and therefore solve for that adoption upfront rather than start thinking about adoption after you've deployed a problem uh a tool set. >> Thank you. Anything you can share about uh the startup that you're working on? >> Uh still early days u as I said this is the conceptual framework. [laughter] Um the idea is to think about you know question the paradigm of how we are already working and what through my 30 years of experience what I've seen as some of these assumptions that we make where you know we hold ourselves kind of hostage to situations or data sets or um solutions that you know we just we just continue to refine what we're doing and try to make it faster. faster or cheaper or better but instead of that how do we reimagine you know what are we doing in the first place I believe there is a there is an opportunity to improve the linkage between business decision making strategic business decision making and the initiatives that insights and analytics teams are working on we're all doing a lot of busy work and a lot of our busy work does get used pretty much on ongoing basis for tactical decisions which is excellent it's great don't get me wrong I don't want us to lose that ability but the space of strategic business business decision making and connecting that to the initiatives that we are working on today I feel that is where there is a big opportunity to improve >> okay thank you uh I don't see any other questions coming through is there a question you would have liked to have been asked that you have a good answer for and take an additional opportunity to educate Yeah. So I need to first think of the question or maybe think of the answer that I know and the question I could have asked and therefore the questions. Okay. Now I I think again I feel like you know I I feel like with these kind of um remote or um hybrid models is always difficult because you don't get a pulse of how is the audience reacting to the conversation that you're doing. Um, no, I don't have any like whatever preset questions that I wish would have been asked, but yeah, I would I I'm I'm happy to connect and hear more from people and get perspective on, you know, if this if this kind of triggered some provocation or inspiration for them and what are they thinking and would be more than happy to like connect even offline later if if there are questions that come my way. Yeah. >> Okay. Excellent. Well, thank you so much for the presentation. Uh, very educational, a lot of really good food for thought in there. And I don't see any more questions coming through, so we will wrap the conference today. Thank you so much, Pavey. And just want to let everyone who's listening know that uh we will be posting archival presentations. Uh, if there's anything you missed that you want to catch up on or something that was particularly good that uh you you want to watch again, uh those will be posted, I believe, to our YouTube channel. Thank you so much.