AI & The Future of Audience Understanding with David Boyle of Audience Strategies

Is very clear that people who embrace language models like ChatGPT work better, quicker and happier. That is it for sure. The question is what we do with all those benefits as organizations, as teams, as individuals, and as society.

Welcome to the Audiences Podcast, the show that helps you discover new audiences and learn how to develop your own. I'm your host, Francesco d', Orazio, founder and CEO of Pulsar, an audience intelligence platform for researchers, marketeers and PR professionals who like to put audiences at the core of everything they do.

And I'm Sal Morton, producer of the Audiences Podcast and person in charge of keeping us all on track. So what's the podcast about, Fran?

So every episode focuses on a cultural trend, an idea, a brand, or a new emerging behavior. And ask our expert guest three simple questions about it. So who's the audience of the thing? How has that audience evolved over the past few years? And where is he going next?

Adren, how are you doing?

Great. Exciting conversation today. So we're speaking to David Boyle, who has been on the pod before we.

The audience Electronic Music. Check it out everyone.

That's the one great episode on a report on the state of electronic music in the UK that has been released at the beginning of 2024. He's coming back today to discuss how AI is changing way we understand audiences. And this is the beginning of a new type of episode that we're doing in Udencies podcast where we explore ways of understanding audiences and how they're evolving, but also how do you develop an audience? So this is more like an how to part of the podcast rather than exploring new audiences that are emerging around new ideas. And it goes hand in hand with obviously that side of the podcast. But we thought there's no better person to inaugurate this, like, new format than David.

Yeah, we're zooming out, we're getting a.

Bird'S eye view of audiences and yeah.

No, better two people, in my opinion, to get changed.

So the biggest issue that we're tackling today is how AI is changing the way we understand audiences. And obviously there's a million ways that AI is changing their job from people talking about the end of market research and people talking about the value of knowledge going to zero. What we explore in this conversation with David is how that is not exactly how things are playing out. But we'll also go into the details of why AI is useful in understanding audiences and how is it changing the profession. The reason why we're asking David is because David obviously is a very well respected leader in consumer insights. He has been doing that for many, many years. He has been working in large organizations, a small organization. He has been the chairman of TMRA, the market research event, for the last 10 years. So he's got like a pretty comprehensive view of the industry. And he has been doing for a living the job of translating complex data into strategic actions for global brands. He has done one thing, though, that make him stand out in the consumer inside space. He literally, when the first ChatGPT got released, he understood that he had to drop everything and refocus on how the that fundamental change was going to shape the industry and what the opportunities and what the challenges were. He's the author of Prompt, which is a book that has been created for empowering market researchers to unlock rich insights from LLMs. And he's also the founder of Audience Strategies, which is a company that, as the name says, helps companies enhance their market understanding and achieve transformative growth using audience insights and driving strategy, using those insights.

Amazing. Should we get into the conversation?

Let's get into the conversation. As always, we try and keep it short. We failed at keeping it short. We managed to keep it within the hour. But I promise every little bit of that conversation is valuable and it ends up on a really high note. So a big note of hope and positivity that everybody needs in this day and age.

Spoiler alert, guys. Positive endings, positive ending.

So let's get into it.

All right, let's do it.

Let's do it. Hello, David. Good to have you back again.

Hey, thanks so much for having me.

So it's great to have you back on audiences. It's been almost, it's been almost a year actually since you came on the show last time. And the last episode we did was about the audience for electronic music in the uk.

This is gonna be a very different conversation, I think today it is.

But it's also this is the first episode of a new format of episodes that looks at the other side audiences. So we have one side that is about discussing new and emerging audiences about any idea or any topic or anything that is happening that you want understand. Who cares about that specific thing? But these new types of episodes is going to be about how do we understand audiences and how do you develop an audience. So I saw something that you brought recently on LinkedIn off the back of your TMRE latest sharing experience, and I thought, like, this could be the beginning of this new strain of episodes. Do you want to tell us what that post was about and then we can take it from there?

Well, First, I'm honored that you picked me to start this series because that's, that's a real treat. And this is definitely a thing I care most about, which is like, how do we get to good insights? A lot of researchers care a lot about what the insights are. I definitely am a researcher who cares about how we get there. So this is a great topic. But yeah, the biggest topic in the world and of research is AI. It's how AI is enabling us to get to insights more quickly. It's an how AI is challenging existing methods and how AI is changing how we spend our days and what our roles are. So I'm fascinated at getting to the bottom of that. What is the role of AI in market research and how do we do it well, and all these good things.

There's a lot of anxiety in any profession right now about what AI is going to do to their jobs. And obviously I'm looking at it from two angles, actually. I'm looking at it as a researcher from a market research angle, but also from a product angle. And in both cases, the narrative goes, AI is killing product management. AI is. In your case, your post was not killing. AI will eat market research. Let's start from there.

Yeah. So AI will change everything about how market research is done, full stop, no caveats whatsoever. Every single thing about how market research is done will change. Not everything will go away, by the way. We still need to talk to consumers and users. We still need qual, we still need quant, we still need all the old things, but even they will change, will change how we write scripts for those things will change how we analyze them. We'll change how we interact with users in those conversations. And we'll change how we bring to life the conclusions for strategies or for decisions in businesses. So absolutely everything about how we do market research will change. And the question then is how? And what does that mean for roles? Does anybody need to be worried about their job, our jobs going, those kind of things? Does anybody need to be worried about their job? Yes, absolutely. Everybody needs to be worried about their job. Will anybody lose their job? Yes, some will, but most people will not. Most people will not. If you learn to use AI well, most jobs for the foreseeable future at least, will just be better, more fun, more impactful and quicker. That sounds good to me.

I think. Yeah, I think I'm old enough now, unfortunately, to remember another revolution that hit jobs and industries and that was about like 15 years ago, well, 20 years ago almost now, and it was Web2 and it was social media as an evolution of Web2. And there were like similar kind of headlines and narratives at the time. One was about, for example, the death of traditional media. Like TV is dead was the classic one. Obviously we had for the last 20 years the best TV that's ever been produced. So it played out differently than people thought it was going to play out. Although I think now we're starting to see social as emerging as a new broadcasting platform. And so things are actually evolving in a direction that is more similar to what was initially portrayed 20 years ago. There was also a set of narratives that played out social media against market research. And that's one where I've built a lot of my career, which was using data from the social web to try and understand audiences and figuring out what people care about. That's worked out well, but has not resulted in the death of market research. It's actually added tools to market research and added methodologies and skills. So do you think this time around, are we seeing a similar kind of impact or do you think the impact here is more severe and we can go into the consequences of that one by one. But like, do you think it's a similar kind of wave or do you think it's a more transformative wave?

I think there are similarities, absolutely. So I think one similarity is that the vast amounts of new insight will be generated that otherwise would not have been generated. That was true with web versus non web, that was true with social versus non social, and that's true with AI versus non AI. So that's new work, by the way. That's part of the reason jobs overall aren't worth worrying about, because for every job that is lost, there'll be and new jobs with new exciting use cases. Right? So for sure that similarity is the case. This is different in one sense, which is that it's more pervasive. So the web took a long time to permeate every aspect of business, and social also took a long time to permeate every aspect of business. AI, you could argue from the moment ChatGPT launched, was ready, was absolutely ready, was capable of being used by absolutely everybody in every role, literally, from kindergarten students at the very start of education to professors of physics at the top of their game. Not that it could do their jobs for them at any step on that journey, but it could help them do their jobs, help them learn, or help them to study and teach. And so instantly on day one, for free, it helps everybody do every single job, and that's every single job in market Research as well. And so that's pretty different. And it means that those people, those market research agencies, or those people writing qual scripts, or those people writing debriefs or those people writing insight reports, those people who did embrace it magically overnight got quicker, better at what they do and happier in their work because it's much more fun doing it with AI. And so you've got this superpower that a bunch of people have adopted. So that's the first way in which I think it's very different actually.

So you think there's going to be more efficiencies which at the beginning is going to result in people enjoying their jobs more rather than losing their jobs.

That's absolutely where it starts. We'll get onto the end state later. Right, but it's absolutely where it starts. Embrace AI. You do work that is better, that is quicker and that is happier. You enjoy your work more. The quicker part means you either take on more clients or you do you serve the clients you've got in the, in a shorter time period and take some time for the family or other fun things, or you do, you know, new and interesting things at that time, take on different types of project you wouldn't otherwise have taken on. So there's been tons and tons of academic studies on this topic. It's very clear that people who embrace language models like ChatGPT work better, quicker and happier. That is it for sure. The question is what we do with all those benefits as organizations, as teams, as individuals and as society. And yeah, a bunch of people end up just working harder because they do more work at the same time. Okay, a bunch of people have chosen to do better work to super serve clients they otherwise were kind of getting by with. So they've taken the benefits and thrown them into extra service. A bunch of people have taken those benefits and said, well, we'll do something new. Let's take on new lines of business or new types of role or new types of product, that kind of thing. And there are teams, there absolutely are teams that have said, well, you know What, I've got 15 people doing the same, same job. If everyone's doing their job quicker, then I can not hire some of those roles when they come around to churn. And so jobs have been lost in that sense. But in my experience so far, and in conversations with teams trying to engineer processes around AI, it really is places where you've got a large number of people doing a very similar, slightly mechanical job. That's where you have the opportunity to not Hire a role when it, when it becomes vacant, or to not grow the team when you otherwise might have done, or make people redundant when costs are getting tight. So there is some of that as a choice versus growth, versus diversification and versus super serving clients. But there is some of that and I think it really is only places where there's a very large number of people doing pretty much the same job.

When I look at this though, what I worry about is unintended consequences of the change. Not on the job itself, but on the perception of the job. And one of the things that pretty much everything we do is based on scarcity in terms of the value that we create. And we're entering a space where the value of knowledge is plummeting because the scarcity of knowledge is also plummeting. Like all of a sudden now you've got something that you can ask questions on every possible single study that has ever been done that public and accessible, obviously, and that's not a small caveat, but you've got access to all the knowledge out there and it's at the fingertips of anyone that can type in any language in natural language. And so what I see as a potential big shift is in the perception of the value of market research and the perception of value of knowledge. And therefore jobs will probably be lost. Not because of how optimized the workflow in research are, but because of how cheap research is going to become in the eyes of the client. And how much resource are we going to have available to pay for roles in research because of that perception of value? What do you think of that?

Yeah, I think that will happen and I think it'll be mostly bad decisions being made. So we've seen some of this before when self service research came along, suddenly stakeholders, decision makers could log into a platform, run their own market research studies. They didn't need a research team anymore and teams were either not hired when they otherwise would have been or were let go. Now I've been employed in at least one large company which was relying entirely on that method. And I came in basically to fix the mess that they'd made because it turns out if you're not very good at market research, it's very hard to run self service market research, which is absolutely good.

Yeah, sounds counterintuitive, right?

Some terrible decisions were being made because people thought they knew what they were doing, doing. And we're making writing like bad questionnaires and analyzing them poorly, you know, and so, okay, we're in a world of AI where the same Thing is happening. A decision maker, a stakeholder can open up. ChatGPT can ask about unmet category related needs in their industry or themes or trends. They'll get an answer. It might not be the right question. It might not be the right answer. They'll get one and they'll say, well, I don't need to call the market research team because this is instant and they take a long time or you know, things like that. It's not a good answer though. It's not a good decision. Every single time. I've worked in great detail with somebody on using AI on absolutely every single topic, from market research through to Grammy winning songwriters and Hollywood screenwriters, through to CEOs of major corporations, through to personal assistants helping executives. Like every single time. It's better if the person with the skills does the thing with AI than if somebody else without the skills does the thing with AI. You're better able to ask the right question, you're better able to judge and response, you're better able to refine the response and make it actually good, actually useful, actually truthful. And so for example, I would not dare use AI in a meaningful way on a topic I'm not expert at unless I was willing to put vast amounts of time into fact checking the work. But you know what? Want better to have an expert run it for me, but I want an expert to run AI, not an expert who doesn't run AI. But no, I think losing teams because you're doing it yourself, using AI is a road to mistakes. Yeah.

And I think we're just going to see this dynamic play out, I think with social media when it became a tool for research, what helped the more traditional side of the industry was that the people in charge of budgets were not exactly skilled at the social media and the big data side of things. So they kind of protected in many ways correctly traditional streams of research instead of replacing them with big data streams because that's what they knew. I think in this wave we have a slightly different position though, where the AI morphs in ways that make it look familiar to people that are not familiar with it. And it makes it look like accessible and easy for people to use and put to, you know, in a production workflow, in whatever workflow you're trying to manage from audience understanding to strategy and planning. And so I think that's more of a risk here, like bigger systemic mistakes in terms of like insights function. And I'm seeing that a lot in large organizations where they collapse, the insights function into other functions like the marketing function. And that means that the specific skill of the INSIGHTS function is now hierarchically like beneath other skills, and therefore is seen as more expendable than it was before. Are you seeing some of that, especially from an industry point of view, now that you've been at the helm of TMRE for how many years now?

10 years now.

10 years, exactly. So where do you think we are in terms of Insights budgets in large organizations?

Yeah, that's a good question. I mean, I think one big realization for me is how slowly the world will change. You know, I said earlier that we've instantly got access, everyone on day one to this free technology which changes everything. But we're still, we're two years into that journey now. We're shortly after the two year birthday of ChatGpt as we record this, and still a lot of people have actually not logged onto it or used it in any meaningful way. So change is slow even when you have this magical gift which is well proven to be transformative to your work, but it's still a slow process. So there are a number of organizations who have adopted it in a very clever way. First of all, for me, that is making everything they do enhanced slightly by AI, as I've described earlier. So script writing, analysis, ideation, thinking, strategizing, every single part of the process enhanced slightly by AI and they're seeing transformative results. Two different consumer goods companies I've worked with are releasing more new products, ideating around more new products that meet the consumer testing threshold. They score higher when they do hit the consumer testing threshold and they get there a lot quicker than they did before, thanks to AI being a little bit of every step of the process mostly. So there are a bunch of organizations doing that. They're not necessarily talking about it, by the way. They're a little bit embarrassed about it, sometimes a little bit shy, a little bit unconfident about it and nervous. So that's not being talked about, but that's absolutely happening here and there. There are also organizations who have not touched this at all yet, where they really aren't, haven't got corporate policies in place that enable the team. In fact, they've probably got corporate policies in place that hold the team back from doing innovation. And the team are kind of happy and safe where they are. So there are absolutely a bunch of those. And then there's certainly teams that are being shaken up massively. Let's give Procter and Gamble as an example, not of a team who's being shaken up massively, but A team who's spoken on the record about the transformative change. So I had the pleasure of talking to Kurti Singh, who's the chief analytics, insight and media officer at Procter and Gamble at tmre. And he talked about how studies that used to take three or four weeks in the old world can now be done in less than a week. And studies that used to cost $30,000 in the Old World can now be done for $1,000. Now, he's managed that change elegantly with Impropter and Gamble, and not in a transformative way, because he cares deeply about quality and process and things like that. But that transformation that many people aren't talking about, that should change everything about how you put pressure on your insight team to deliver the taste in which you expect your insight team to deliver. But this should change everything. So I think there's more transformation, there's more challenge that's coming. It's just the organizations are slower than you might expect to adapt.

Yeah, it's always way slower than you think it is. But then all of a sudden it happens, and then you realize that change was always there and it was just progressing as an undercurrent. Let's talk about before we go into the things that are changing because of AI. I think for people that are not necessarily familiar with how to use AI in audience research, can we just list a few of the ways where AI helps within the workflow Audience research.

Great. Yeah. I think there's four ways. There's four broad categories, and then each has lots of derivatives. Right. So the first is synthesis, and that's taking a bunch of text of any kind. Maybe it's reviews, open ends, maybe it's process documents, technical documents, and doing something with it. Finding themes, finding patterns, synthesizing it down and turning it into something else. You know, AI is absolutely incredible at that. So anytime you have a large block of text, you absolutely should be using AI for synthesis. Second, learning, learning from the data the AI was trained on, or if the AI has search capabilities, learning from the Internet or from documents as well. But if you just want to learn, like, tell me about this industry, tell me about this client, tell me about this audience need, tell me about this product or the category. Trend learning is a great thing you can do using AI. Third is thinking. So that's okay, here's the audience insights. Help me think about which might be the most important. Or here's the audience insight. Help me think about the implications for my product or my brand. Or here's my thoughts on the key conclusions from this piece of research. Help me think through the big themes here, because I'm not seeing them. Help me think through the tagline of the slide. Help me think in some way. What strategy should I do as a result of this work, but thinking. And the last one, which ironically is the one Most people think ChatGPT and AI is for, is writing. That's the one I use least because I'm much more interested in synthesis, learning and thinking than I am about the written word at the end of the day. But if you care about the written word, like elegant prose, catchy titles, translating documents into different languages or different industries, or for different stakeholders, then getting really good at using AI to help you just do the writing part, maybe taking your raw ideas, reprocessing them, reformatting, and then communicating them more elegantly. Writing is the fourth of the four, but they're all very, very different. Right? And they all have massive implications. And we do all four of those things frequently, every single day. Yeah.

One of the things that I find amazing about using AI in audience research is how many steps it compresses into one execution that you don't realize you're doing manually when, as a research, you go through the data, but you're actually doing it. So when you try and then build like we do it on full service about workflows for agents that do different types of research using the data, you just realize how many steps you need to break down the workflow into for it to be performed by an AI, and then it's quite effective. But actually, if you don't have the expertise of understanding what you do as a researcher, the steps are going to be wrong, and then you're going to end up with an output that looks actually valid on the surface, but when you start digging underneath, you just realize that it's not. And one of the best examples of that is in the synthesis phase, the one thing that people do is always assume that whatever data you chuck into an LLM, you're going to get a representation of whatever that data set is like that is complete, and then you're going to get something that the LLM is going to sell you on as an analysis of that completeness. But actually, unless you have ways of checking what is being represented, most of the times you're just going to get a specific subset of what you tracked into an LLM in terms of data, and you're never going to know which one it was. And so the analysis could be like, inherently flawed. Like we've Done this, for example, with a product that we're launching in February called Narratives. And with Narratives, the problem was like, how can we account for every single stream of public opinion that exists without losing granularity and quality of the signal and then getting AI to actually produce an interface for people to understand what is in each of those streams? But I don't want to lose streams or I don't want to collapse streams together that shouldn't be collapsed together. So you have to put in, into the system a very traditional, if you like, clustering, tiered clustering system so that the LLM only gets fed the blocks of information that you want it to synthesize in a way that you know is going to represent completeness. Do you think, like, what do you do in your, for example, in your workflows when it comes to synthesizing large blocks of text or like studies that you're analyzing, how do you ensure that the outputs you're getting from LLM are actually representative of the whole piece?

Yeah, well, that's a very good question and you've hit on a number of really good points there. And this is something that most people really struggle with. So we do a lot of training for market research teams as well as product teams or strategy teams in loads of big companies. And one of the fundamental points is exactly your point, which is that you need to break a task down into steps to make and get the AI to help you with each step rather than the task. So you could absolutely throw a bunch of open ended comments in and say, what should my strategy be as a result of these comments? You can do it, but you might not get a great answer. And the answer is completely, completely unsteerable if you do that in a single step. So you're much better off taking the open ends, synthesizing them in at least one and possibly multiple ways. First, prioritizing using your expertise and probably other data sources, which of those insights is most important, is most powerful, is most relevant given your current situation, then ideating a little bit for those key insights. What might be the options in terms of product or in terms of marketing that would meet those insights? Then thinking through how to act on them, how might I act on each of those? And then maybe you can choose which one you care about. And yeah, so you can absolutely do that in one step, but good luck steering that, good luck judging it. Whereas if you do each step by step, each step can be checked relatively easily against the previous step and against your expertise, against your knowledge and against other data sources. So that is not the promise of AI that most people want to hear. They want to hear. I could just jump straight to the answer. And you can. And AI is more persuasive than a human. And so it's persuasively wrong when it's wrong. So no, don't do that, break it step by step.

That persuasion element I think goes against one of the key tenets of research, which is not about persuading an audience that your answer is correct, but it's about providing proof that your answer is methodologically correct. And therefore people can make up their own minds. And I think that's one of the clashes between genai and market research, which is you have on one end a sales guy trying or a saleswoman trying to tell you what you should be believing and in a very convincing way that you want to believe because it's convenient. And on the other end you have a science of people that are relatively shy and relatively not trying to sell you onto anything. And they're actually trained not to sell you onto anything. And they're just trying to go like, look, this is the steps I've taken. This is from the steps I've taken, what results and this is what, you know, you have to review and then you can make up your own mind. I think it's a very different attitude.

Yeah. The good news for decision makers and stakeholders out there worrying about this is that there's a profession of people who are trained very well to navigate these kind of problems. They're called researchers. And so having them run AI to do research is going to be a lot better than having anyone not trained in research running AI to do research.

Yeah. Honestly, I've got the same problem with product management where people, as we were saying before, they claim AI can do the whole job of writing product requirement documents and managing different steps of the process. And then you end up with people building stuff that just goes counter the fundamental needs of human beings or doesn't take into account risks. Like what happened with the character AI episode where the character that the teenager was talking to effectively agreed that, you know, if you commit suicide at some point and the guy actually did. You know, there are ways of understanding how these products play out in people's lives that are quite complex and quite nuanced and is not, however good that model is going to become is there's no way that you can just automate that and you shouldn't. That's like humans are gonna want to hear from humans, but also humans should have Input in what. What humans interact with, because that's what we're good at.

One of the funniest moments, as I'm exploring the world of AI and learning about applications in different industries, was working with a Hollywood screenwriter. And so she had sent me a kind of a setup for a scene. And I said I would use AI to kind of sketch out the flow of the scene. I did that. I thought, this is great. I sent it back saying, look how great AI is. This is a really good. I took your stimulus and I created this great scene. Look at this. And she wrote back, just basically laughing at me, saying, that is the worst scene I've ever read in my life. So I was completely unable to judge the quality of this scene. As an expert, she was, by the way. Then she was able to tell me what was wrong with it. We were able to fix the prompt, and we're able to get it to generate a very good scene. So it is possible, but it requires an expert to judge and refine and make sure you're asking the right questions. And the same is true in research.

Yeah, totally. And I think so, as we were saying that this, like, four steps where AI really helps. Synthesis, learning, thinking, writing. I think we can break down those steps further. For example, when it comes to the synthesis side, obviously you have automated analysis of unstructured data, which is pretty, pretty amazing. Then you have a very nuance, for example, in our space, sentiment analysis, that brings into, you know, play the context of the analysis that you do in the audience, the cultural issues, the news cycle. You know, you can bring a lot more into the analysis that than traditional machine learning used to bring into the analysis of an expression of opinion. Obviously the analysis of emotions, pretty incredible. We have tested it recently for the analysis of credibility as well. AI is pretty amazing at analyzing the credibility of users or publications. And it does so by looking at the context around them and how people are reacting to them and the opinions that people have on them. So that's a pretty interesting way of using that. What I really like is this crosses the synthesis and the learning side of things, and it's the ability of identifying patterns across multiple touch points. So you bring together customer service transcripts, social media, data forums, product reviews, or anything you have from qualitative interviews. And then you can build a cohesive view of that data that will take a researcher a lot longer to synthesize and turn into Personas. But actually, what I find amazing is that if I feed an agent that we're building for audience profiling, A bunch of touch points around demographics or interest or opinions on certain things. Then I can get really fully formed audience Personas that before would have taken a lot of qualitative work to do and a lot of time and expertise. But now I can get a draft as a starting point, and that sets me on the path to develop thinking that before would take me no longer.

This episode is sponsored by Pulsar. What is Pulsar? Pulsar uses AI to analyze live data from the web and the media to help you understand people at scale and with nuance. We're talking about social media like X Instagram, Reddit, Pinterest, YouTube, as well as search data and any media from TV and radio to print, news and podcasts. And of course, you can bring your own audience data like NPR or CRM to analyze alongside everything else. Brands like Amazon, agencies like Macann, media outlets like the Guardian, and organizations like the UN use Pulsar to understand their audiences and create products and messages that matter to them. If you'd like to get a live signal from your audience, get in touch@pulsarplatform.com.

So I want to go back to your 10 Commandments for how AI is changing research audience understanding. So the first point that we touched on was the speed. So you have point number one in your. In your. In your 10 points is from big reports to rapid answers. Is that how you see people commissioning research now as well?

Yeah, absolutely. Which goes back to Kurti Singh's point, about three or four weeks. Goes down to one week. And so you can get great answers really quickly now. So 100% definitely. There's no reason to wait for a very long report. You should have the quick version done asap. Using AI help then if you require a higher level of granularity, a higher level of proof, evidence, a higher level of specificity and all sorts of other things, then absolutely commission the big project you would have done before. Absolutely still do that, that make sure the team who are doing that big project are using AI to make it better and quicker, possibly cheaper or better as your choice, but absolutely still do that project. So I don't think it will stop big research projects from being necessary. It just will change the role of them and they should change in speed and quality as well.

Great. Point number two, three days to strategy.

Yeah, I mean, it's really simple because it used to take a very long time and it used to be very hard work to think through the issues and trends that you're facing. The possible implications of that for Your strategy and to prioritize which strategies make sense, which ones are most useful. And all of those things can now happen much more quickly. AI is very good at synthesizing and thinking in all those ways and ideating around options and prioritizing options. So you can get to a so what gets an answer, get to a recommendation much more quickly. If we're honest in a lot of market research, it presents data or facts or maybe even insights, but without so what's attached to them now, there's no excuse not to add loads of great, so what's to your market research, because you can do that piece of work much easier. So we're doing that a lot more than we used to do because there's no longer a big extra phase of work we need to charge for and the client needs to invest in. It's now something that we just kind of may as well do with every single piece of insight. It's great.

Does that change though, the composition of the team that needs to work on those projects? As I, I found a similarity in one of the issues that we discussed actually with Julian's Electronic Music, where because small clubs are dying, you don't have training grounds for new emerging DJs. So if you're saying that someone needs to be. Someone working in research needs to be strategic and looking at the so what's on the implication on the insight? A lot more than generating the insight itself, does it mean that we have less space and opportunities for training younger people that are coming into the business?

Yeah, I think that is definitely an issue. It's an issue if the senior people decide to do it all themselves. And so I have five people or so in my organization and I have a choice in the world of AI, do I do everything myself? Because you know what, I'm perfectly capable now of doing absolutely everything myself without a team. I can do the data science bits myself now. I couldn't before AI. Do I do the qual bits now? I never would have done that before AI. You know, I can do it all myself. But as we've discussed before, it's not the best way of doing it. It's much better to have somebody who really does understand qual, who does the qual bits with AI. And it's much better to have somebody who does the data science bits with AI. And so if anything, we've grown our team over the last year and people are doing more and more diverse work, more interesting work, more creative work. So I think it's a decision for leaders. Do you Want to build a team, grow a team, do more, do bigger projects and invest in that next generation, or do you want to do it yourself instead of hire a team, which I think is slightly lazy and risk quality because you're skimming everything rather than investing the time required in every task. So it's a decision, I think, but.

Also is a higher risk strategy because if you lose the people that are going to replace the more senior people that, that you have in your team at some point then you might end up with losing everything the moment that senior person leaves because you don't have anybody else coming into that role. And you need to hire someone as senior and as expensive to do the job completely because you will, you've bet all your chips onto that specific resource rather than having a range of options that you can bring into the mix when you need them.

Yeah. And each task is something that you can get very good at. So while I can do data Science now with AI and I couldn't before ChatGPT came along, I can do quick and simple data science. Data science is not that important. It's much better to have my data scientist do that work with AI because he actually knows the questions to ask and the quality control to apply to it. So you're better off with the team. You really are.

Let's talk about point number three, and this I think needs a bit more explaining because it's not obviously clear why this helps, but point number three is deep needs. Tramp surface data.

Yeah. Well, I've been on a personal crusade for many years now that demographics are wonderfully useful and interesting in many ways, but not the key thing that we should use to describe audiences. Not the most useful way to think about audiences, not the most powerful way to think about audiences. There's other things that are much, much, much more useful and their needs, they're category related needs and some people call them pains or unmet needs if you. But these are the underlying things that drive people to want something different from your product or your category or service. They're the underlying things that drive people to interact. And these are relatively hard to think about. It requires a bit of brain work and a bit of nuance that we used to spend a lot of time really wrestling with what the underlying needs might be in a category. And then we do survey research to validate which underlying needs were most important. And then we'd use those needs to cluster and segment audiences in really powerful ways. That was great. With AI we can find the likely list of those needs much more Easily language models know a lot about audiences and their needs in all sorts of different categories. And so you can kind of trivially now come up with a thoughtful, inspiring list of needs. So yeah, stakeholders who previously maybe didn't quite understand the difference between demographics and needs now can. They can now trivially explore idea around needs in different categories and think about the implications of those needs. So I think this is going to help the research industry take that final step away from demographic based decision making and towards this deeper consideration, more widespread consideration of underlying needs, I hope. Yeah.

And I think this extends to beyond needs. This extends to the more qualitative views of audiences that we thought was inherently something that humans was going to get better at grasping. But if you look in the 90s, there was this French sociologist called Maffasoli that worked on the idea of tribes and how the idea of people coming together around their passions and interest was a lot more powerful as a predictor of behavior than grouping people together by demographics, for example, or by how much money they earn. So what that happened with social media was like, finally we can have a view of these tribes because we can see how they interact and we can build a model of how they get into certain brands, brands or certain activities or sports and some passions that they have beyond who they are as a person that earns a certain amount of money and as kids and as a certain age. But what is interesting with AI is how good AI is at understanding the correlation of those affinities to build a connection between a system of affinities and give you that Persona. And I think that's been the most amazing discovery for me in terms of working with audiences, audience research and AI was like how the get capability as leapfrog our ability to segment audiences.

Yeah, I think AI is really a pattern finding machine and maybe that's the reason I'm personally so excited about it, because as a mathematician by training, the joy I find in the world is finding these hidden patterns that aren't at all obvious to people but turn out to be very useful in predicting things. And AI is very good at finding hidden patterns, whether it's needs or affinities or you talked about credibility or others. It will find those, those hidden patterns and articulate them wonderfully. It's so helpful.

And that I think plays into point number four that you have, which is moving from signals to symphony.

Yeah, we've got so much data these days, so many potential signals of things that are important and useful in the world. Whether it's open ended comments or Digital interactions online or likes or comments or followers on Facebook. There's so many signals out there pouring into data and insight teams, you know, it's very hard to make sense of them. And suddenly we've got this gift of a tool that can AI which can look across multiple data sets or within multiple data sets and help to find these hidden patterns. And so from data overwhelm or. Yeah, or data that's not being used because we haven't got the human time to process it, suddenly now we can expect to process vastly more of that data in a way that actually is useful. Finding those hidden useful patterns.

And point number five that you have is moving from being human readers to AI listeners.

Yeah, well this is the simple point really, that text used to be something that was only readable by humans. I mean really, numbers have been machine readable for a very long time, but words had to be read by humans. That's really hard work actually. It's time consuming, it's mentally taxing, requires a lot of memory and all sorts of other things. So even things like processing open ended responses. I was working with one of the biggest companies in the UK the other day and they said that they set on a bunch of important research and nobody has read the open ended responses. Literally nobody read them because they didn't have time. But now that's trivial to read them. Machines can read them. I think, by the way, this is a bit like the transition from spreadsheets that used to be rooms full of and it was was mostly women, rooms full of women on paper, writing ones and zeros and passing them to the next person. That was how we processed numbers at scale in the old days. That went away as machines learned how to process numbers. Well now we're doing the same thing with words. There aren't many places where there's rooms full of people reading words. So the analogy doesn't quite work. But we're going from having humans reading words at scale to machines reading words at scale and other guiding and leading the machines and doing that work. So suddenly all these words that weren't reading or you weren't reading properly or you're delegating someone else to read, suddenly we can read them or read the insight from them, find those hidden patterns from them at scale and that's a complete transformation. Analytics that used to be reserved for numbers can now be done on words. That is amazing.

That's this phrase in a TV series that I watched recently called the Diplomat. And in that series at some point it's very stuck with me, the ambassador talks one of his aides and says, nobody likes to be briefed. And I was like, yeah, that's true, actually, nobody likes to be briefed. Like, nobody wants to sit through a briefing session and everybody wants to get to the key points. And I think AI is doing just that, like cutting the brief down to seconds and giving you the key nuggets that you need to get from that brief. And for you to then be able to act off the back of the insight, rather than spending most of the time getting to the insight to then being able to dedicate some time to the thinking of the actions that you need to be taking. And obviously, if we take this step now further, and we move one level up from the specific tasks that we do at every single step of the process, we're kind of moving away from the individual tools that you use. And I think we need to rethink how the research system works with this new capabilities. So you have one point, which is point number six, which is from tool collection, tool capability architecture.

Yeah. Insight teams have always had this challenge, and it's just got 10 times as difficult as it was before. But the challenge is there are an incredible number of important data sets out there. Mostly each data set has a different tool attached to it. You know, you end up with two or three market research tools or dashboards or systems, two or three social media analytics dashboards or systems, and internal dashboards and systems. So you end up with a lovely suite of tools. And every single day, people are pitching new tools. The cost of writing software has gone to close to zero, as thanks to AI. And so you've got all these other companies spinning up now with new dashboards, new systems, new tools that market researchers could and should be experimenting with at the very least. And so how the heck do you integrate all of those into a real world workflow? That's very hard, actually. And so that's where we need what I call insight operating systems. So a system, a way of operating which correctly causes people to use the right tools for the right things in the right process, and that smoothly and elegantly integrates them all together. When I was at the BBC, we struggled with TV shows like Doctor who. You know, there were so many different dashboards, different pieces of analytics, different pieces of research. We had a team especially dedicated to just taking all of those different insights for Doctor who and turning it into one single presentation for stakeholders that integrated messages across all those things. You know, that's the kind of insight integration that we need. Much, much, much More of. And it's a brand new skill.

It is, and that's a really interesting point because what I see in RFP and research briefs that we get involved in at Pulsar is that the onus of the integration of different signals to provide a coherent vision usually is put on the supplier of a specific part of their workflow that is supposed to advise the client on how to integrate what they do with everything else they do. But we haven't seen yet a centralized function in enterprise organization that owns the Insights operating system. And I think that's a problem because a lot of the people that run consumer intelligence operations, they are not technologists. And so you have the issue now that you need to be a researcher, but you also need to be a technologist to be able to design that system. And most enterprise organizations got traditionally trained research people that lack that technology skill. So is that what you're seeing or are you starting to see some organizations where the technologist researcher is in charge of the operating system? And if you have examples, which. Where are they?

Yeah, I mean I think the big consumer goods companies tackled this early and so Unilever had their version of this and Procter and Gamble and all the others all have their versions as well. Well, as I understand it. So that's the. But that's the very tech version of it. I think that's pretty rare. And, and you gotta have. Well, you used to have a very big budget to do that. But as I say, the cost of writing software has collapsed thanks to AI. And so maybe that's tackleable now by people that wouldn't have been able to tackle it with budgets in the old world. But I think there's a very human version of this as well if you just think in a systems way. So with Doctor who it was very human. We took each of the individual aspects of aspects, dashboards, reports, and a human said what's the most useful part of each of these? And how do they come together? And so there is a very human way of doing this as well, which is sometimes done in the Insight teams as we tried in Doctor who. It's often outsourced. So quite a number of companies come to us and say, hey, I've got this big decision to make. Here's a Google Drive folder full of market research reports on this topic from various different trends, agencies and qual and quant agencies can. You just can bring it all together for us. And then. But the worst of all is when the, the Insight team lets the stakeholder make that decision. There's a big global media company we're working with and one of the key decision makers in the business, the key creative decision makers in the business came to us and said, well, look, I, from the Insight team, I have 10 different reports on how this one piece of content is doing. You know, I've read them but like I don't feel qualified to, to hold all of that in my head and, and triangulate all these issues and challenges and, and chall know, can I use AI to help me to understand the breadth of the insight on this topic and get to an answer? Well, you know, yes, you can. So stakeholders or technology or insight teams or outsourcing it, but there's a problem here and you need an architect of some kind to design the system and process technologically or in a human way in order to make this much better. Because the number of data sources is only increasing and they need integrating at some point.

So we're saying we need more insights. Architects and some of the most advanced organizations that have been deploying market research at scale like Procter and Gamble and Unilever have been doing this for a while. I mean, I remember the people Data lab at Unilever for the last maybe more than 10 years already. But I think most organizations do need an insight architect and that will change the way AI is deployed. Point number seven from finding to forging. So what do we mean by that?

Yeah, well, in the old world it used to be exciting to find insights. You know, we didn't have a lot of data or it was hard to get hold of. And so coming up with facts, coming up with data points, coming up with evidence was exciting in and of itself. That's finding data, but that's not good enough anymore. Now we have an abundance of data, abundance of dashboards. We've got an abundance of insights, insight, especially with AI helping us to find and uncover even more. And so now it's about forging, it's about taking the those numbers, taking those data and turning them into strategies, turning them into useful, actionable recommendations and carrying them across the line. Not just taking them into recommendations, but persuading people, bringing your leadership team with you on that decision. And so it's more like a producer maybe who's trying to gather all the elements, integrate them together and fundamentally tell a story that's compelling to an audience about that. So a big change in skills, big change in capabilities needed.

Well, and that actually connects to the next point, which is from the return of creatives and strategists, which Reminds me of this photographer that commenting on AI said photographers are always going to be around because humans want to hear stories from humans, not from machines. Is that why you think creatives and strategies are coming back as a core skill?

Yeah, I mean, I think that's part of it. Because at the end of the day, you're absolutely right. The leadership team making a decision on a topic want a human being to confidently tell them they believe it's the right thing to do. A human being who can be accountable and responsible for that recommendation. So, yes, partly through delivery, partly through responsibility, accountability. Absolutely. But also partly, as you say back to this previous point, that it's not good enough to just have facts and figures that we need to bring them to life. And a fact can be brought to life in so many different ways. A fact can be brought to life in a very kind of cold, logical way, a fact which. Which often misses the mark in terms of implications. A fact can be brought to life in a way that doesn't fully take into account the nuance of the company and the culture and the goals and aspirations of the brand in question. And that misses the mark sometimes, but also it can be brought to life in a way that fails to pick up the cultural nuances of the time and the macro tre where society and culture and consumers is going to. And if you miss that, then you miss the mark as well. So all of that requires quite a lot of skill, quite a lot of strategic thinking and quite a lot of creativity. And I guess over the last few years, as I've worked with more and more like creative strategists, I've learned myself the importance of creativity alongside strategy. And not just cold hard recommendations, but cold hard recommendations that are brought to life or ideated around and with implications that are just, just much bigger and much bolder than before. And most insight teams, most research teams don't have what I think of as creative strategists and I think they're missing out, they're missing out on convincing their audience, but they're more importantly missing out on those implications, which would require a bit of, you know, joining the dots in order to get there. A bit of foresight, a bit of creative vision.

Yeah. And I think that joins the dots on another problem, which is that a lot of the traditional training in research doesn't involved any of the execution side of research. So it's not easy, but it's easier to make recommendations in abstracts. But at the moment you're trying to turn that recommendation into a creative strategy or A product innovation, that's when you see the value, that recommendation at play. And if you fail to produce that example in a tangible way, that it's not just the abstract recommendation, then actually the insights falls apart.

And like everything we've talked about so far, AI can absolutely help you to do that. And so I would say that broadly is not a skill of mine. I perhaps thought it was until I worked with people who are much more creative than me. But I recognize now that's not a skill of mine, that really creative strategy. But you know what, with AI I can get quite a lot of the way there, which is an amazing transformation of my abilities to be a creative strategist. But it's no duplicate, no replacement for having an actual creative strategy. Charlie on our team, I usually have a crack at it and come up with something that's quite creative. I send it to Charlie and he says, no, David, let me show you what I should do with that. He uses AI to do a better job than I would have done on it. So we can all get closer to this goal now, which is great, but we still need those people more than ever.

Yeah, I don't think I've ever used a creative output from stable diffusion or an LLM as it was. But what it did provide me was cutting the time massive between zero and a draft. And that initial draft has sparked ideas, has generated the hypothesis for what to do next. And it definitely helps you explore a lot more angles before you go and double down on a specific one that you think and you know is right with the experience you have. And so point number nine is beyond the quarterly polls. And I think it speaks to the long term strategy and the idea that what's really valuable is having someone with a strategic view on the inside. But why is AI better at providing that long term vision versus the traditional short term analysis of what an audience needs right now?

Well, I think the first point is that there are, in my experience of working with quite a lot of different companies and agencies on this. There are a lot of people who are just very stuck in the moment. They're stuck in the quarterly routine, the quarterly reporting. You know, the immediate priority is and just don't have the time or attention or the budget to step back and think about the longer term implications, the longer term trends, the longer term evolution of audience needs, the evolution of adjacent categories and their impact on your category. You know, all these long term things that are happening. It's just something that so many people struggle to prioritize, struggle to budget for struggle to allocate time for. So I think there's a really big need here. And again, like everything we're talking about, it's a need that AI can unlock. And we talked earlier like AI is very good at finding these themes, finding these long term patterns, finding these adjacencies, translating knowledge or insight from one industry to another. You know, what if my category evolved the same way this other category has evolved? Or what insights can we learn from this other category that will lace up to come to mind, this category that leads mine? Maybe. So in the same way as everything else, AI just unlocks that power should you choose to do it. And so I just think many more people will do it because it's such a big gap in the world right now. It's not done nearly often enough. Yeah.

And I think it also speaks to the statistical nature of the machine, which is something that is designed for scenario modeling and predictive insights. So once you deploy that on human behavior, you're capable of playing out the consequences of what you see as an insight right now into a long term trajectory a lot more easily than you would if you were a human, effectively. Last one, insights everywhere.

Yeah, this has been something that I wanted to see realized for a very, very long time. You know, working inside large organizations and even organizations where I was proud to have transformed the Insight team into something that was super, super, super impactful. You still have to be honest that the vast majority of decisions that get made in the smaller brands or the smaller countries or the smaller partnerships, the smaller media, decisions like the vast majority of them are made without useful insight right there in the moment. Insight teams help with the big stuff, the plan stuff, the stuff that you know is going to happen. But so much of decision making is spontaneous, is in the moment or is not worth bothering the insight team with or, or have you bothered the insight team to ask for help with with it? You might not make their priority list. And so there's this vast untapped set of decisions that insight's practically not helping with these day as is right now, today. And so this is another way that finally AI can unlock this long standing problem in the world, which is that those people in that moment on their phone on their way to the meeting that they didn't think to call the Insight team to get prepped for can now ask of insight by that's trained on or that's got access to a vast swathe of proprietary insight data. Let's say they can now ask it a quick question and go in with at least some useful insight into that meeting and maybe even great insight. So I'm excited about the trickle down effect that not just big brands and big countries get decisions and insight, but that every decision might get insight that's needed to help make it better.

So it sounds like there's a lot of work to be done on a technology that has hit the market at really broadly and quickly in the last couple of years, but there is still a ton of work to do to metabolize what the technology can do and how we can implement it. Do you think this is the focus of the next decade of the evolution of research, or do you think there's going to be some other major technological leaps that we're going to make that are going to speed this trajectory even faster than it's been right now?

No, I think this is by far the biggest, biggest change in technology or business or life that has happened and will happen in our lifetimes. It's more far reaching than any other technology. It's instantly available to everyone on the planet. It's capable of helping with absolutely everything we do all day, every day. Not doing it for us, but helping us with it. It's an incredible accelerant to your skills, capabilities and curiosities. There's no part of the market research world that this does, doesn't touch, change, improve, make better, quicker and happier in doing it. And this is not promise or vague LinkedIn leadership nonsense. This is hard. Scientific tests on real world people have shown this. And the classic case of the future is there. It's just not evenly distributed. If people like Kurti Singh are saying, well, four weeks goes to one and $30,000 goes to $1,000 dollars, then you know the future is there. It's your choice. It's our choice whether we want to go leap into that world and be better and quicker and happier in our work or whether we don't want to. So no, this is the most transformational thing we'll ever see in our lifetime. And it was big enough for me to decide to stop everything I was doing and spend my time on this. I really enjoyed my job. Until late November 2022, something like much more exciting came along. I feel so lucky to be alive at this time, helping other people to realize it. And yeah, I still do a lot of my old job, which I deeply love as well, but so I'm very grounded in real world market research. But this is the biggest deal in the world.

It is, it is. So if your kid came to you and said, dad I want to study market research and become a researcher. What is your advice going to be? Go and do it or get out of the house?

No, go and do it. You know why, though? Because following your passions is the thing to do in this world. And as long as your passion is something that's vaguely productive to society, you should follow your passions. That's my number one piece of advice to anybody. But yeah, absolutely. But do it in an AI first way. Don't do it the way your lecturers are telling you because they're one one generation behind. Don't do it the way your boss is telling you in the new company because you're younger and smarter and better able to adopt this stuff than they are. Help to lead the lecturers in using AI in their courses. Help to lead your boss by challenging the mentality them. Actually AI can help me to do this job. So do it in an AI first way, which is radically different than what you'll be encouraged to do by people who are older than you in life. But absolutely. I think market research teaches such incredible foundational skills that are useful. Maybe the most. Actually, you know what? Maybe the most useful set of skills for using this technology. So more than one large company, Clorox in the US is one have said publicly that the market research team are leading their work on AI across the organization because the market research team is the team that's best suited to work out what's good and bad and what good use cases are for this technology. So market research is a very helpful skill in this modern world.

There you go. That's what I wanted to get to. Good wrapping. David, thank you so much for joining us again. Insightful as always and looking forward to the next 10 years now.

Can't wait. Let's see where it goes. Thank you.

Thanks David.

You've been listening to the Audiences Podcast, the podcast that helps you discover new audiences and learn how to develop your own. Me and the team would love to hear your feedback on the episode and on the podcast in general. Let us know which audiences we should explore next or anyone we should get on the show. Do reach out on our social media or email us@highaudiencespodcast.com as always, please rate, review and subscribe till next next.