The 80% of Marketing that AI Should Take Off Your Plate with Rafa Flores

Welcome to Ad Speak, Adweek's podcast about the business of marketing, media, and creativity. In this episode, we're sharing a conversation recorded live at one of Adweek's events featuring leaders from across the industry discussing the ideas shaping brands right now. Let's get into it. >> Good morning, everybody. So, you know, as you all probably know, personalization used to be mean, you know, like greeting somebody by name or, you know, showing them an ad based off of what's in their shopping cart, which we all all love. But, you know, now with AI, you can kind of anticipate what people are going to want and what they're going to do before they actually do it a lot better. And so, for this fireside, we're going to kind of dig into how AI can really help supercharge your marketing and what are the obstacles that you're going to encounter as you try to implement. Uh with me is Rafa and I'll let you introduce yourself. >> Yeah, good morning. I am first of all very grateful for those of you who are here. Name is Rafa Flores, chief product officer, uh company called Treasure Data. We're leading with AI agent foundry. We have an AI marketing cloud and we have a marketing super agent. A lot of buzzwords in there, but there's a lot of good behind that and so I'm excited to share a little bit more. Hopefully, you take a couple of new things from this. >> And uh there's going to be 5 minutes for Q&A at the end. So, you know, as you're listening, if you have a question, you know, keep that in mind. Uh you'll have a chance to ask it to Rafa. >> Just be Just be nice. >> So, uh Rafa, so um let's start at the highest level. Like, what is the biggest problem that marketers typically have that AI can potentially solve? >> I think what's important to think about is and I'm going to take you back in a little bit of a journey here. I'm going to go back 250 years ago. Does anyone know what happened 250 years ago? Y'all need to wake ups. 250 years ago, the Industrial Revolution. That is correct. That is what happened 250 years ago. Now, the reason I bring that up is it gave something new to the world, right? A lot of people talk about machinery, right? It brought coal and use of machinery. I think it gave a little bit more. It gave momentum, right? It allows somebody who had to build something with their hands to use a tool that could actually now build for scale. I believe AI can help marketers in the same way. It can give a marketer momentum, right? There's that 20% of things that you still need a human element that if you put 100% of your time to that, it can just change the game. Obviously, >> everyone wants to implement AI in in some capacity. Um what's the what's the biggest hold up? >> I think what's important and look, I I lead a company that's leading AI in many ways. I'm not here to tell you that AI is going to replace your job. Okay, so you can breathe. I'm not going to say, "Hey, you're going to lose your job to AI." I don't believe that, right? I used to be at Arm Holdings back in the day. I led the Internet of Things. We thought we were going to change the world. Nobody was going to actually have to even use a washer dryer anymore. It's going to be fully autonomous. How many of you in the room actually go use AI mode in your washer dryer? >> I don't. I have it, paid a lot of money for it, >> and I don't use it. I like the time dry, and I like to manually change it, right? And so, I think you have to take some of that, and when you think about, "Okay, what is the the biggest hold up for marketers to adopt AI?" Eliminate from your brain that it's going to eliminate your job, number one. Number two, everyone's teaching and talking about agents, right? "Oh, you can use my agent. You can use this agent." It's all siloed. There's a cost of fragmentation, not just with data, but also with the agents themselves. And so, the biggest hold up is, "How can I bring it all together so I can use it all in one place to then augment my muscle of doing things on a day-to-day of of a lifestyle marketer?" >> Is there any way to kind of really move past fragmentation because that's, you know, I mean I mean, that's usually the thing that just kind of stops people in their tracks. >> Yeah, I mean, fragmentation there's also collaboration, which is a big element of that, right? So, how do you move forward in the era of, "Okay, data's all broken. It's everywhere. Agents everywhere. I'm hearing all this noise." I think you have to sit down right via CML CXO with your team and have a frank conversation around, "Hey, how can we actually solve a business outcome?" Right? Cuz when you think about AI, we have a saying at Treasure Data that, "Look, your AI is as good as what you feed into it." Now, part of that is the data that goes into it, right? If you go on Chat GPT and you put a bad prompt, it's going to give you a bad answer. Now, it doesn't mean the AI is bad. It just means the prompt was not good. Right? But, there's also an element of collaboration, which is have you sat in the room and said, "Hey, here's what we can get out of using this AI product." A lot of times that conversation hasn't happened. So, then it's fragmented. Right? Cuz some folks in the room may be using Chat GPT, others may be using Gemini. Yeah, it's it's fragmented. >> Yeah, I mean, one of the things I've kind of talked about if on stage a little bit is just is the change management aspect of implementing AI and I know there's a bunch of different ways to kind of push through that. Do you guys Do you have any advice on that? >> Talk about who is doing what when it comes to AI, right? So, collaboration is also very strong when you define roles ahead of time. Right? And so, we're firm believers that for change management to take place and the agent take by the sign era, everyone needs to know what role they play and what role the AI plays within their role. Right? That's the only way you can accelerate collaborations. I'll give you a great example. We work with many global brands. One of them is Samsung Motors, right? Everyone knows them. They use a lot of our agents, right? These are audience agents to build segments just by interacting with the chatbot. They don't have to go and say, "Okay, what is our ICP?" They can just chat and it creates a segment for them. And so, that is great, right? Because they know that the role that the agent is playing for them is is just actually building the segment, but they're playing the role of still giving it the ICP, right? And defining that for the agent, so the agent can just fine-tune it and make it even better. But, you have to have that conversation. >> Cool. So, once you move past the issue of fragmentation, you know, how should marketers think about evolving your marketing strategies? >> It's a million-dollar question, right? We always talked about it from the lenses of the predictive versus proactive, right? Predictive is you're trying to predict the next behavior, right? As a marketer or an agency supporting a marketer, you're trying to figure out, okay, what is the next best action, right? What is the next best offer? What is the next best offer product, for example? Proactive is a little bit differently. It's actually engaging with that potential consumer in real time. And what I mean by that is you could actually go log in, use a persona agent that says, "Hey, this is Rafa Flores, right? He likes to dress down on Wednesdays, as you could tell, right? He loves sports. He lives in Orange County, California, blah, blah, blah." And it will actually pull my profile and engage with me, right? And so, you're having this back and forth, and as a brand, you literally tested everything. It wasn't predictive, it was proactive. You're not trying to predict my behavior, you're seeing my behavior in real time. But it's all virtually done, right? So, then when you actually get in front of me, you are more likely to convert me, right? It's all about a point point of conversion. >> So, when you're thinking about like predictive versus proactive, the predictive it feels like those technologies are available pretty readily available today. Proactive is where you have the AI power where the AI implementation can really, you know, enable that sort of capability. Is that >> Totally. I mean, predictive, right? It's that's been out for a long time. And it seems like machine learning people forget about machine learning nowadays, and I believe AI is agentic plus machine learning, by the way, right? If I can give you a brand sentiment, which is a model, but I can also give you a persona agent to test that brand affinity, it all comes full circle. And so, there's a lot of predictive technology. I think proactive is when you start bringing some of those two together. >> What do you need to do in order to start building out a proactive campaign strategy, and then how do you know if you actually even need it? >> Good question. I'm going to walk you guys through an actual scenario. So, at Trust Your Data, one of the things that I love doing right as head of product is, can we put ourselves in the shoes of the brands we serve and try to target their consumers, right? Like if we were to launch a company, what does that look like? So, we actually did that exercise a couple weeks ago. And we created a persona called Emma. Now, Emma is not a real person, just so you'll know. Okay? It's kind of like the whole Tilly Norwood uh debate. We created Emma, but Emma is not real. But we want people to feel like Emma is real, right? Cuz when we test this company that we launch, we want this to be as real as a person as possible, right? And so, we walk through the journey with Emma, right? And this is a typical journey that all of you do many, many times. Which is, okay, if we launch a new company and we're trying to go and pitch it to an Emma who is our target consumer, what are we trying to do? And the first thing is, well, what does Emma do first thing in the morning? She wakes up, she grabs her coffee, and she shops. Right? And so, how do we get that in front of her? How can we get that VIP lookbook, right? To her top of inbox, so she can be on our website and actually click it. The problem is Emma may not buy, right? Like an ideal scenario is every morning everyone goes and purchases and that's it. There's a drop-off. You may be on a rush or you just you're not sold yet, right? So, then you leave. What happens next? Well, we know with all the data in the background that Emma likes to go on a walk, right? When she's at work. We want to target Emma at that point in time. It's all about the cadence, right? It's being on autopilot, being autonomous above the brand. And so, Emma then gets a push notification. And in that push notification, you have a signal now, which is, hey, you get 15% off if you actually go back and buy that loungewear that you were thinking about. So, now this is hot in Emma's brain. Emma's thinking, okay, this is cool, right? I can picture myself, I'm tired, I hate my boss, I want to go home. And then she does, and we want Emma to go right back to where she was and use that discount code. But again, it's not perfect. We thought about, okay, what can go wrong? A lot of time this gun code doesn't work. Right? You plug it in and it doesn't work. And that sucks. You got them engaged, you got them excited, they're ready to purchase, and it doesn't work. That's when you can then also use AI for client telling. Right? Have a chat right in place that says, "Hey, yeah, here's no problem. We'll give you an additional discount on top of that." And then they buy, right? But, they buy the wrong size. Here's another doomsday scenario, and so they want to return it or exchange it. That's a blessing in disguise, cuz guess what? When they go to the store and they say, "Hey, I need a different size." There's a signal that hits the person at the point of sale that says that Emma is actually a very good ICP for their store credit card. And so now you went all the way down funnel past just a purchase to true loyalty, right? And so this is a real scenario that many of you do. There's a lot that happens in the background, of course, but you have to think about every single engagement and be autonomous. >> Is there an extent to which all of these communications though are too much? Because, you know, we are kind of in this world where there's so much digital digital noise. Everyone is trying to get everyone's attention every at all times. You know, so like AI is doing a very good job, you know, kind of pushing out these messages, sequencing these messages. Um how about using it to, you know, pull back on those messages? >> Yeah, yeah, yeah. I mean, omni channel orchestration, right? That's that's been a hot one for some time. And uh it's also you got to it's contextual, right? I mean, I can sit up here and say whatever, but it everyone's in a different context. Every brand also sits in a different context and how they want to run their marketing. Some of them want to bombard the potential consumers, right? The shoot and spray approach. Others want to be more tailor and specific. You could use AI if you want to be more specific, so it's not noise. And what I mean by that is when I think of personalization, I think about something that is very unique, that's one-to-one. Right? I think about one-to-one personalization. I think if I was to lead a brand, I would want that every single person who may get an email, right? So the channel is email, who has similar interests, they get the same context, but the images are different. The offer may be different because it's personalized specific to you. No one email should be the same. And so, to me, if you want to be very tailored and specific, you can reduce the noise, but make sure that the noise you do bring forward has to be one-to-one personalization. I think that's the difference. And you could do that with AI. >> And the other thing I was kind of curious about are is uh the synthetic persona because I was talking to a few other marketers who, you know, said that they use them, not just for marketing, but also for like product development. Um obviously, you you talked about Emma, and I know we're not talking necessarily about she's real. Real real synthetic. Um what's the value of a synthetic persona though? Like, do you get like accurate data from them? >> It's as good as what you feed the model, right? And so, uh when it comes to synthetic personas, I think everyone wants to use them. It's a little creepy, by the way. Like, I've actually chatted with myself. And sometimes I'm like, man, that guy is tough, >> Right? So, did you create a synthetic persona of yourself and chat with >> I know. >> Yeah, and and I haven't even told my wife that there is a synthetic persona of me cuz that'll be kind of fun, right? Like, how can I get my way all the time and she may beat me at every conversation. But uh I think I think when it comes to synthetic personas, everyone wants to use them. I recommend you try them, but I think it's only as good as what you feed the model in the sense that again, if you're given a bad data, right? Where it lacks context, it's no use to you, right? Because and then the AI may hallucinate. Now you have a synthetic persona that's hallucinating and it's completely it's completely wrong, right? And so, you have to think about what you are feeding AI. >> How do you actually start figuring that out though also because there's you have a lot of data. I'm sure you have a lot of Everyone has a lot of garbage data. So, how do they kind of figure out which signals are necessary? >> Yeah, so AI decisioning, right? That's like the My clicker stopped working, by the way, so I will Oh. We're back. All right. Uh AI decisioning, right? How do you figure out the signals? I think it's not just about figuring out the signals, it's about figuring out what you want to do with those signals, right? And that's where the whole concept of AI decisioning comes in, which is, "Hey, as all these signals and triggers are coming in, what pathway should I take ex-consumer down?" Right? And on in an autonomous way. You all sleep, right? I mean, probably not last night, but you all sleep on a usual basis. Your brain shouldn't sleep, right? The consumer may be awake at 1:00 in the morning. I wake up at 3:30 in the morning every single day. I'm I'm at the gym by 4:00 a.m. Between 3:30 and 4:00 a.m., that's when I shop, that's when I do stuff, that's when I use my credit card to do whatever I need to. None of you are going to want to target me at that time cuz you're sleeping, right? And so it should probably be autonomous. There should be a an AI decisioning in place that says, "Hey, you know what? Based on the brand sentiment between 3:00 and 4:00 a.m. for Raphael, this is perfect for him, and we are going to shoot a specific push notification that he can then go and make a purchase." Right? But all those signals need to come at play, too. >> How long do you have to train her for before you can actually deploy it? >> Depends how much data you give it a chance, right? But um I think it's okay to let your AI fail, right? I think you just have to contain it. And what I mean by that is we've all heard the automotive um story that those in They They implemented AI, they put a chatbot on the website, and people got free cars, right? That's doomsday scenario for AI. Not good, right? And so But it failed. Now, that automotive company, number one, did a major campaign after that, right? That spoke to, "Hey, yeah, we effed up." And it gave them a lot of PR, but number two, they realized that, "Hey, you know what? It failed. We didn't give it enough of the data that it needed, right? We also didn't think about different regions, different markets, different territories, global, right? Language. Those are things that you have to look into, but let it fail. It's okay for it to fail. >> I mean, right now, you know, your failures, I mean, you know, if you're a big brand, chances are your failures are going to be very, very public. >> Yeah. So, is there appetite to really kind of experiment >> in the sort of live setting and then, you know, if it fails, suddenly, you know, Bloomberg's starts writing it up or, you know, we might >> Yeah. You guys make um There's a risk to it, right? I I think Look, how many folks here in the room 100% trust AI? I can't see it super bright, but from what I can see here, nobody raised their hands. Not many. Right? And so trust is something that's built, right? And so, but failures do happen. I'm not going to sit here and say AI never fails. Like, sometimes it does. You have to just contain it, right? So, when you look at a vendor, per se, not just us at Treasure Data, but you can look at any vendor that pitches you AI, which I'm sure happens every day. When that's happening, think about use cases you can trial and error that are very contained, right? To build Again, going back to 150 years ago, to build momentum, right? Build credibility. If you're a CMO and you just take a big bet and you bet all in AI and it fails, yeah, you'd probably lose your job. But, if you start proving value with something small that then scales, that's your victory. But, you have to build trust. >> All right. So, then, how about ensuring that what you do generates results? >> Yeah. Business outcomes. >> Right? And so, I think one of the things that I want you guys to take away from this, too, is and I have it on the big screen, there's a lot of asset factories going on right now, right? Everyone builds digital assets, creative agencies. I love what somebody said the other day on the stage of, "Hey, you know what? I'm a CMO and we get five agencies in the room and I like to pick this and pick that and I kind of put it all together, right? Great. But, what is the outcome? Why are you pick kind of picking and choosing what you need, right? It's Think about the business outcome first and then think about, okay, how can I design this all together so that this program makes sense end-to-end versus thinking in silos of, okay, I'm generating a digital print for advertisement. Right? Like that is predictive versus a proactive strategic approach. >> Right, so in other words, start with what you want to happen and then work your way backwards from there. >> So start with the problem, right? It's I think with AI everyone is shoving it down your throat, right? Board of investors, even to me, right? You got to do more with AI. I must see it all the time. You got to do more with AI. But for what? Right? Like what is the problem that that piece of AI can do for me? >> And then how about, you know, ensuring that you can kind of scale this and yeah, in a way that's sort of reliable and and accurate? >> Scale is extremely important, right? I think again, it you don't just get to scale, right? You build up to scale. It doesn't happen overnight. You don't go from a Well, now I guess you can go from zero in revenue to 100 million overnight really quickly. But it's rare, right? And so scale is something you build and you work towards. It's the same concept with AI, right? Build the confidence towards trusting your AI fully across your organization. Use AI over time to scale your reach, right? Scale your cadence, scale the number of programs you can run if you're an agency, right? Concurrently. I'm a firm believer and I say this to my team, some of them are here. It's not that AI is going to replace your job. I don't believe in that at all. There's a human element, right? But it can take away 80% of the junk that you do all the time. For example, for me, I'm in a thousand meetings, right? Instead of chief product officer, I become chief meeting officer. It's frustrating and annoying. I do my job in the late into the night cuz I have stuff to do. I have road maps and reports and whatnot. I put meeting notes through AI because 80% of my day is in meetings. And I just copy and paste it. Right? And what was interesting is I've been in the room where somebody is using AI had take notes and somebody else calls out in the room, "Oh, why did he use AI?" In my head, I'm like, "Why Well, why wouldn't you?" Right? It's annoying to put together meeting notes. So, that's the 80%, but the 20% still needs a human element. Apply it there. >> Um we have about 5 minutes left. Do we have any any questions from the audience? I cannot see you guys. >> Can't see a thing, so yeah. >> We've got one free Wait. Oh, uh that right there. Do we have mic runners? >> Good morning. Uh >> Someone had caffeine. >> I was hoping you could speak to how workflows are changing from AI. So, like, that only is it changing the way that we're delivering experiences to consumers, but how's it changing the way that marketers their day-to-day looks and actually how they work with agencies and all of the partners that go into the campaigns. >> Yeah, it's a good question, right? And it kind of goes through collaboration, right? And and being a system designer. Being a system designer doesn't mean you're going to go and build the product. It means you have to build the the next program altogether, the next campaign, coordinate across different agencies. And so, when it comes to using AI for that collaboration, it's you need to sit down with all those different folks in the room and say, "Okay, here's the 80% of stuff that none of us here want to do. How can we put that on autopilot so that the rest of the time we can focus on that 20%?" Right? And it's also about giving you a running start. I I think one of the biggest things if if you use AI to give you a PowerPoint template, right? It gives you a running start. You can take that and make it pretty and better and on brand, right? Above the brand. But you don't have to start from scratch. Right? It starts with a little bit of momentum. And so, again, I think you can use it in a wise way to enforce collaboration, but also to get a running start on your day-to-day basis. You guys are half asleep. Come on now. Ask me a tough question. Make me uncomfortable. I always tell folks, make me uncomfortable on stage. >> Hello. Hi. My name is Johnny. I have a question. I'm really curious to know that when you're talking about to free ourselves up from this 80% of junk work, do you think people need direction on what to do with that newfound 80%? Very good question. It's not about what you So when I say 80/20, right? It's not that suddenly you freed up 80% of your time. It's that 100% of your time can actually go to that 20%, right? So all your time should be going to the things that actually do matter. So I'll give you an example. Again, going back to my example, I spend a lot of time in meetings, right? I use AI to write meeting notes. Now it doesn't mean that suddenly I'm not doing anything. It just means that I have more time to actually focus on again, on the road map, on catching up with my team, having one-on-ones, right? I'm more hyper-focused on the the things that matter. So you could put the 80% on autopilot, put your best effort on that 20%, right? It's not about finding what's new, it's about actually having the time to focus on the things that that matter, that make a difference in in your day-to-day. You spend most of your time at work. Make it worthwhile so you can get promoted, right? That's the biggest thing. Reach your own scale. Don't just aspire for scale as a business or as a brand or an agency. Make yourself very scalable. Good question, by the way. Any from you? >> Yeah, what's the biggest mistake marketers typically make when they start out, you know, kind of when they first start out kind of trying to deploy some sort of AI solution? I think the biggest problem, not just marketers, but anyone with using AI is they just do it to do it, right? Like either somebody's trying to force you to use it or you think it's cool and you're going to try it. If there's no purpose to it, you are going to fail, right? I mean, there's certain things that are just logical things of life. They apply to AI, too. Just cuz it's hyper-intelligent, it doesn't mean that it's going to solve every single problem you have. And so, my biggest advice is have a purpose to why you want to use AI. Going back to the Internet of Things, right? The washer dryer. There's no purpose to that other than a company throwing billions of dollars into P&L to make something cool that's just there's no need for it. Actually, did you try it yet, your AI washer dryer? I have not, to be quite honest with you. I know. I know. I like time dry. Give it a shot and then report back. I'll be here next year, we'll talk about AI mode on the washer dryers or the fridge. You have it in the fridge? All right, Rafa, thank you so much. Yes, thank you, by the way. Appreciate it, Fletcher. That was a conversation recorded live at an Adweek event. You can find more coverage, analysis, and interviews from across the marketing and media world at adweek.com. Thanks for listening to Adweek. If you like the show, be sure to follow or subscribe wherever you get your podcasts.