While AI technology is the shiny new object, is it time to shift focus to where humans, the consumers, shine through? Artificial intelligence is undoubtedly beneficial, revolutionizing industries with its ability to analyze vast amounts of data and identify patterns at unprecedented speeds.
However, does AI truly grasp nuanced human behaviors and the critical concept of being consumer-centric? The panel drilled down on the intricacies of understanding consumer demands, discussed why being consumer-centric is so important, and unearthed how AI can be incorporated to illuminate bringing the consumer to the forefront of your business.
Moderated by Aarti Bhaskaran, Global Head, Research & Insights, Snap Inc., the panel included Michael Nevski, Director, Global Insights, Visa; Suann Griffin, Former Senior Director, Insights, Serta Simmons Bedding; Jolanta Oliver, Director, Digital, Menu, Foundational Insights, McDonald’s; and Jennifer Ng, GM, Advertiser Insights, The Trade Desk.
Bhaskaran: We are hoping to have a super interesting discussion today on when technology and humanity intersect. How do you find the balance between humans, consumers, and tech?
I would love the panelists to introduce themselves as well to the audience. Give us a one liner of what your day to day looks like. And it’s a fun question. What is the most overused word when it comes to AI? I’ll go first. For me, anything that says AI powered.
Ng: I would say automation. Probably a bit overused, sometimes confused. I think AI and automation can work hand in hand, but they are not the same thing, and not to be feared as well.
Griffin: Excited to be talking with you all today. My word is deep fake. I think of this as we think about synthetic panels and the emergence of deep fake imagery and deep fake audio. Are qualitative sizzle reels inevitable? Something to think about.
Nevski: Probably AI enabled. Like many of my colleagues, I feel that right now, there’s a lot of talk about enabling and incorporating AI everywhere it’s possible and even impossible. We need to really have a good strategy and a thoughtful approach in how we incorporate AI in all our processes and what AI really means.
Oliver: For me, the overused word is fully automated. When you think fully automated, there’s so much that goes into GenAI that you need to be careful about that. You can’t assume that what is generated is correct. You always need to question, is this correct? And what kind of tests do we need to do to prove the efficacy of this new tool?
The AI Journey
Bhaskaran: How have you incorporated AI? Maybe we can focus on generative AI into your insights toolkit, and would you be able to share specific examples with the audience?
Oliver: We have been on a journey with AI tools for the last three to five years. The one big area where we wanted to get to somewhat full automation was the drive-through. How do we get to a place where a robot could take the order from a customer and deliver an experience that’s far superior than what we have today in terms of order accuracy, speed, and friendliness?
It’s been a two-year journey. When you think about how to implement an automation of that size, it’s not something that you take lightly. You need to test it and retest it with different regions of the country where customers speak differently. They may behave differently. Taking all that into account, we realized that a fully robotic agent could be faster in some ways and slower in others. If you want it to be super friendly and talk with the customer, well, guess what? That sounds all good, but it doesn’t help our operational metrics. It slows things down. And when you want to do suggestive sell as well for a new product that the customer didn’t order, that also slows things down.
When thinking about what you want to achieve with a robot, we find that setting our standard of what the KPIs are going to be and whether we’re going to meet those is something you need to do at the offset of the project, and you need to keep checking whether those KPIs are being met. We’ve actually canceled that project because we just could not achieve the level of customer experience, crew experience, and operations automatically with the robot. It just means we’re not there yet. We’re trying, and failing fast is something that companies need to embrace.
When you get to a point where you’ve learned but the current technology can’t get you there, just take what you’ve learned and try to do something a little bit different for the future to get to the customer experience you want.
Bhaskaran: I love the point you made about it’s a balance between operational efficiency versus cultural nuance and a customer centric approach. Michael, how have you incorporated AI or GenAI in your world of insights?
Nevski: About nine months ago, my company acquired OpenAI’s product for us on premises. So we have a ChatGPT now for that. We also have a Copilot, and the company created a whole structure of ambassadors and training in how to incorporate that in the daily and weekly activities. But to me, generative AI, that’s the big talk on the street right now, and especially in our industry because, overall, AI has been around for a long time. Let’s say, in my business, it would be security, risk, transactions, and safety.
When you use generative AI versus the human being when you conduct the research, whether it’s a quant or qual, when it comes to actually putting it all together and presenting to your key stakeholders, it takes human oversight to really decide on what you put together, what needs to be changed or tweaked. That’s why while technology enables us to be more efficient, we still need to provide guidance, leadership, and input as human beings and especially research professionals.
Bhaskaran: It’s so true. You touched upon in terms of use the tech that you have and learn, but you also need expertise. Suann, what have you observed in terms of GenAI being used in insights?
Griffin: It’s actually been a long time coming, both AI and GenAI. Aperio Insights introduced me to the Remesh team in 2017, and I was an early adopter of that and remain a huge fan of their research approach. I’ve seen HR teams use it. I’ve seen consumer insights use it. It’s just an amazing tool to get qual at scale, very quickly.
But I think it also presents a problem of how do you incorporate all of these tools? I think that our bandwidth becomes an issue. How do you look at all of the tasks-based AI solutions that are available and it takes you time to vet each one. Do they have ability to address specific gaps? Can they talk to each other? How do they nest with your other solutions? How do you use one versus the other and have the team bandwidth to train each of the tools and ensure that you’re still keeping the consumer centric mindset. A lot of times, legal gets involved to look into the provider data and privacy practices.
With Memory’s recent GenAI supported solution, they accelerate the collection of real consumer videos, and that AI gives you insights within minutes. I think it’s a great quality scale for people who want to test early, test often. That’s one of the things I’m looking for when I look at a GenAI solution or any kind of consumer research solution. Can we go in quickly, affordably, rather than planning out a large project? For an insights application, can we test a little thing, test something else tomorrow, and really use it as we go so that the consumer remains at the focus through the entire stretch rather than coming in one time. That’s one of the biggest potentials in our category for Gen AI solutions.
Bhaskaran: I love the point you made about stitching different tools or sources together. That’s not going to change regardless of whether we use AI or not. That’s always been a challenge.
Ng: In its essence, the Trade Desk is an ad tech programmatic platform, so you can argue that the basis of the entirety of the company is based off AI. However, I think from an insights perspective, we’re still very new at it. I run a very small insights team, and so the two main ways that we use AI from an insights and research perspective is just simply data aggregation, data analysis, to make our lives a bit easier. If we’re taking thousands of rows of consumer attributes, we use AI to kind of help summarize, aggregate, and create better profiles of those consumers using that data.
Another form of that is we can see content across the open Internet, and we have a language model that helps us understand the sentiment of the content. It would be nearly impossible to use humans to analyze all of that content and all of that sentiment across thousands and thousands of pages. That’s another form that we’ve been experimenting with, understanding the sentiment of the language and of the content in our work.
A Balanced Approach
Bhaskaran: It’s great to see so many different examples from customer experience to efficient qual perhaps all the way through to training and large language models. At Snap, one of the ways in which we use AI is to test creatives, because we have trained the model on over ten thousand plus video ads, just to identify what creative assets are present or not. A lot of benefits we have spoken about in terms of speed, efficiency, but, obviously, our panel is a showdown between AI and consumer intelligence. How do you find that balance between AI driven insights and consumer-centric approaches?
Griffin: It’s about training your model. It is uploading your segmentation so that you’re getting the right personas. It’s scraping the Internet to find what is the topic, what are the conversations, what are the conversations on social. I think it is using our decades of expertise and keeping the consumer at the forefront of the decisions that a business makes and understanding the strategic direction that your business units need to make—and being the nexus, the connector of the dots there to ensure that we’re kind of maintaining that balance. It is a balance right now. We are walking on a balance beam and struggling not to fall off.
I heard an analogy recently that we’re on a highway. We are all going 200 miles an hour. Everyone’s going to see different things, but you must go the speed limit. And the speed limit is 200 miles per hour, and we’re going to see great beautiful scenery, and we’re going to be in moments of darkness. But you can’t just pull over and you can’t just slow down and so I think that it is up to us to keep that human element in there. And by human element I mean insights professionals as well as the consumers that we are here to represent and we’ve always been here to represent. It’s making sure the findings that we’re applying are in the best interest of those we are applying them for. It’s really trying to ensure that you’re keeping the right mindset, and it’s not just faster, cheaper, more agile. There’s a purpose, and most of us are very purpose built for that consumer-centric mindset.
Nevski: Suann is right on the money, in terms of what we need to consider—consumers. We’re building these insights or tools or processes. Your last example is a great illustration that AI is not a panacea. In my research, I see that there is accelerated adoption by consumers in the U.S., for example, and some other developed countries. Accelerated adoption of generative AI, not only for everyday tasks or for different purposes. And they use the generative web for different purposes, from creating artistic work or evaluating videos. Or they are shopping, and they would like to use it for product reviews, ingredients, created lists. And it’s about one-fifth of the population here. Not only do we need to be human centric while we’re driving at 200 miles an hour, but we need to think about the consumers, how they’re adapting to this technology in their everyday lives.
Is AI going to replace me? It’s not. It’s still the technology which is going to help us be professional insights leaders, who will rely partially on or incorporate the technology, but have that judgment, have that testing, and incorporate that into the process while we’re overseeing it.
Oliver: If you want to create powerful stories for your leadership, the AI could get you part of the way there, maybe 60% of the way there. But at the end of the day, a live video with consumers that match your segmentation, that tells their truth about this is what it’s like to be living in my body, to be living in this world. This is my day to day. That’s what moves hearts and minds. AI is great at perhaps getting to the story, but I believe the insight is generated by humans. The data collection and aggregation are created by the AI, and you need both to tell that clear story along with videos and just being strong in the boardroom to have your leadership develop empathy for the customer or the crew member or a store employee. That empathy is what’s going to move your board to make decisions that’ll make the customer experience. AI can take you only so far.
Ng: It can only take you part of the way. You can analyze a thousand different attributes and behaviors and understand the consumer, but are you going to really truly understand the why? Nothing can replace surveys or direct feedback from a consumer to really understand the ins and outs of the whys and how they are thinking, and so that will never be replaceable.
Navigating the Highway
Bhaskaran: We also spoke about how to use those tools. Whatever tools you’re building, how do you align that with customer centric values? And most importantly, what are the guardrails that you keep in mind apart from becoming besties with legal and privacy?
Ng: In terms of guardrails, you must think through where, when, and how you use AI versus your more consumer centric other methods of insights and research. Having a clear path and decision-making process on where and when you use it and for what is really important. It’s not going to solve all of our problems. I have a really small team, and I think a lot of companies maybe make the mistake of saying, I can actually replace an analyst on my team with AI. But an AI isn’t going to do everything that you need an analyst to do. There’s the human component. But one of the most important guardrails remains the same. It’s around privacy, and consumer privacy and understanding that we are talking about humans. We are talking about actual consumers. What components of their data should we be using and for what purpose? Are we being transparent? Are we giving and getting the right permissions? Privacy doesn’t change, whether it’s AI or not, and that’s the probably most important guardrail that we need to consider.
Nevski: I agree that data privacy, how we store it, how we process it, but also the ethics of how the study and how the data is being used. While you can set up probably some kind of training model, whether using ChatGPT or Gemini or anything else. But still, that kind of a nuanced human aspect needs to be controlled and overseen.
But the second one, just to give an example, we’re switching, let’s say, to a synthetic panel from quant to qual, considering who you’re researching or what segment, what audience, older generations, Xers like myself or boomers, they prefer human interaction. And if you have that moderator, they’re not going to connect as fast and as efficient as younger Generation Z and Millennials. They would require human interaction when you conduct the research. If I do any kind of a qual and it involves older generations, I know right off the bat, if it’s about moderation, conversations, it needs to be human on the other side. Utilize the tools, to record, to analyze, to process, but interaction is a human.
From my study, I know that early adopters of generative AI, they prefer chat bots. They prefer technology when they need to interact with us as brands. Much higher percentage of those saying, yes. I want to come into my business very quickly and get out. They would be very open to generative AI moderating focus groups or in-depth interviews. But all the generations, they will not. Those aspects of human-centric needs to be considered when we build those processes and we go about the research.
Griffin: We’re keeping the human involved and staying focused on privacy. If we think about what artificial general intelligence is going to evolve to become, and I know that that’s so new and it’s a very black box for us. But thinking about how it should behave, what goals it should pursue, which problems it should solve, we now have the opportunity to train AI. How do we think about freeing up humans even more, to solve the bigger problems, to bring our ethics, our morals, our values, and then how do we look globally? Because it’s not my ethics, morals, values that I want AI to imbue as it solves problems for humanity. It’s kind of the collective. Can we use GenAI to gather that collective? It’s the ultimate survey research. How do we do that and think it through with our futures hats on. Thinking ten years ahead of time where the human element we’re talking about is so important right now might be missing or might have been replaced, and we’re now putting humans in a different position.
Oliver: As long as GenAI is helping us do our work, we’re going to use it, and we’re going to continue experimenting with it. A big challenge: In big multi-national organizations, the data sits in many places. And to the extent that we could get a GenAI to touch all the different databases and bring data together, that’ll be a win. But right now, all that data is still sitting siloed in different places. One person can’t necessarily make one decision. It’s still about collaborating with people within your company and saying, here’s what I’m learning with the data I have access to. What are you learning? And coming up with a solution, a story, to tell the business? There’s a long way to go as far as I’m concerned with GenAI being the king. But I’m just really excited about the next few years and what they bring in terms of how to make our jobs easier, better, how do we get the time back so that we can focus on real insights and shape strategy.
Bhaskaran: What an excellent way to wrap up. I’m sure we could talk about it for ages. So maybe it’s not a showdown between consumer intelligence and AI. It’s how we navigate that highway at 200 miles per hour. AI can get us halfway there, but it does need human intervention, values, and expertise to reach the destination.
Check out the video for more of the Consumer Intelligence vs. Artificial Intelligence panel at the Road to TMRE 2024, plus a question-and-answer session afterwards. Click here for more of the content during Road to TMRE 2024.
