Bonus: Got AI Anxiety? Learn How 875 Brands Are Using AI: DTC x Triple Whale’s State of AI Report

0:00Most people don't know the majority of what Triple Oil is doing. They started as really a data platform to centralize your data. That was all in service of being able to automate tasks across your business. We recently, about a month and a half ago, released MMM in our Mobi chat feature. A brand of ours was running a giveaway. They asked Mobi to look at, on the last day of our giveaway, how should we adjust our spend? Mobi then gave them a plan to do so that had channel by channel specific allocations to look at, and then an action plan of what to do strategically to do that. And the brand saw their highest revenue day in history.

0:45Anthony DelPizzo:Welcome to the D2C podcast. Today we have a very special bonus episode where we are diving deep into the state of AI in D2C marketing with one of our long-term partners, Triple Whale, Anthony Del Pizzo. Welcome to the podcast. What's your role as director of product marketing? That's correct, Eric. Yeah. What is your day-to-day consist of at Triple Whale? Yeah. So I own and my team owns all of our product positioning, our messaging, bringing products to market. So when we're developing, say, new AI features or new features across our measurement tooling, basically understanding what are the problems that we're looking to solve for our customers?

1:24And then how can we ensure that we bring those to market in a way that effectively solves those problems?

1:28Anthony DelPizzo:So we're going to dive into the results of our survey where we surveyed 875, over 875 DTC operators to get a clear picture of how they're using AI. I feel like there's a bit of anxiety in the audience about whether you're doing enough with AI, what you should be focusing on with AI. So that's what we're going to dive into. But I wanted to start with just with a little because a little catch up with Triple Whale. It's been a long time since I've kind of chatted with someone from your side, and I know your suite of tools has evolved quite a bit. And I think it makes for a good sort of intro to what we're talking about today.

2:02Anthony DelPizzo:So maybe give me a little overview on how Triple Whale has evolved over the years. Yeah, totally. And so I joined Triple Whale about nine months ago. And when I joined Triple Whale, I had a very singular view of what the company did, that they did attribution. I'd previously come from spending four years at Klaviyo and admired Triple Whale at how quickly they were growing, I joined Triple Whale and quickly realized that most people don't know the majority of what Triple Whale was doing. So Triple Whale, while they started as really a data platform to centralize your data, we had a mobile app, we still do, that's widely used to centralize your data across every single tool in your stack to give you real-time insights so you can make real-time decisions.

2:41That was all in service of being able to automate tasks across your business. because then once we had your data in one place, we then built measurement models on top of that so that data was trustworthy. So best in class of multi-touch attribution is what the team started with. We've since expanded to a marketing mix model and MMM and adding incrementality testing, really so that brands can just triangulate, hey, what of my marketing is actually driving revenue? And then how can I lean into that more? And then we've built over the years ways to send that really robust data that we've built outbound.

3:16So like a lot of data enrichment capabilities. And then all of our AI capabilities sit on top of that. So I have access to all of this data layer that then now we have MobiChat, which is our generative AI chat. So you can chat with Mobi, ask any questions about your business. It's like chat GPT, but with real-time context into every single layer of your business. And then we have MobiAgents, which is what we launched this year, which are really our AI agents that are essentially superhuman teammates that we're giving to all of our brands across different functions in their brand, really trying to be the one-stop shop where they can actually leverage AI and feel confident that they can expand with Triple Oil and AI in their business concurrently.

3:58Anthony DelPizzo:Agentic AI. Agentic is such a cool word. I still use AI every day, but just as sort of like an individual, like each task is unique kind of thing. And I'm building, you know, I built a few GPTs and stuff like that, but I haven't got to that agentic level. So I'm interested to dive into how that's working with brands. So we surveyed 875 operators, 93 and a half percent of them are using AI. So what are those other six and a half percent of people doing? And I think the main learning is right now the entry point for AI is creative. What are you guys seeing in the data on that side? Yeah, it's really similar in that we, I mean, we surveyed all of our customers just to get an AI perspective too around like everyone is using some notion of, hey, can I use a copy generation, creative generation?

4:47They're feeding AI. Okay, hey, we have this creative that works. Can you generate something similar? What we're seeing is that brands find that copy is a really quick thing they can iterate from and trust it. And then they've tapped into creative, mainly on like the creative elements perspective as AI has evolved to view images, even generate images. And that's why we at Triple Elb actually invested a lot in our creative analysis and generation agents, where we actually can look at all the brand's data, say, hey, these are the elements. We use AI vision to actually view the creative across every platform and say, hey, across all your channels, this is what's working and this is what isn't.

5:27And here's an example of what would work best that you can actually hand off to your creative team. So what we're seeing is people are excited about creative. AI isn't really there from a creative generation standpoint, but it can cut the creative briefing process, the creative ideation process down 80 plus percent. Because your AI is able to basically analyze all your data across platforms and then give you those ideas to then work with your creative team on how to execute on. And so it's been a really good way to at least start from and that you can feel confident in that, hey, this is a use case that is something that our creative team just spends so much time ideating that we can cut.

6:08Anthony DelPizzo:And there are some things, I guess, like some product shots or product orientation, like some simple things that AI, I think, can do well now. But there also is that really uncanny valley thing that you can, whenever an image is glowing in a certain way, you can just tell that it's AI. And there's probably a little bit of a reaction to that at this point. Totally. And that's why we're seeing it's like our best brands who are leveraging our creative agents are using it to get 80, 90 % of the way there and then make the final touches to put it in their brand ethos and make sure it's on brand. Because that's kind of the ways to leverage AI to its most effectiveness right now without making it seem inauthentic.

6:50Anthony DelPizzo:I think copy is something that everyone's using it for. Any AI tells out there? I have two big AI tells for when it's AI copy. One is the use of the double dash. I love the M dash, the double dash to like break up two ideas in text. But I have to, I've had to go back entirely to semicolons because the M dash is a total AI tell. And the other thing I find is when it, when it always says the thing you're thinking about isn't this, it's actually this. And it does that sort of constantly. It's not this, but this, it does that so much. So those are, those are my two big AI tells and any, any on your side?

7:25Yeah, I will say the classic one that comes up is the M-Dash. And as a marketer, I have loved the M-Dash for years and I still like to use it. But to your point, it's sad it's getting all of the AI force right now. We haven't really seen anything too much on the generation side. I think when you think about headlines, sometimes the overuse of emojis can be a really big giveaway. We have agents that produce email copy, email subject lines, previews. And sometimes when we look at some of the emails they produce, they're all back to data. But we really have been looking to see, hey, if we take this emoji out, if we cut that, will it actually have an impact?

8:04Anthony DelPizzo:Do like a thorough A-B test there. I think one of the things that I loved in the report was the data that showed that as brands get larger, so brands that are like over 10 million, their shift, I'm sure they're still using it on the creative side, but their shift focuses to include a lot more like higher level operations when it comes to budgeting for the business, sales targets, things like that. I assume that's your bread and butter. What are you guys seeing? Yeah, it's exactly that. It was super validating because we're seeing a lot of our bigger brands basically have two perspectives, where they'll have the day-to-day execution, where they'll be looking at things like their creative because their creative, just the requirements now from platforms to constantly create new creative is so vicious at this rate.

8:51But we're seeing so many of our larger brands also think about, okay, how can I use forecasting agents? We have agents that look at our marketing mix model, which are just things that are more strategic. So things that you're going to be doing monthly, things that you're going to be doing maybe quarterly to think about, okay, hey, how can we have AI start planning ahead for us and then use the real-time data and real-time tools like your anomaly detection agents, your agents that focus on your daily insights of your marketing performance or even your weekly insights to then kind of converge those ideas of like, hey, how is our daily performance and weekly performance kind of mapping up to like these create the longer term strategic agents that are looking at models that just are thinking more ahead or are very much used for strategic thinking like an MMM.

9:43And so it's been a really interesting way of, I think it's just also like a bandwidth perspective when you have the team that can kind of handle both the strategic and the day-to-day, it's a good way to kind of bring those two together.

9:55Anthony DelPizzo:Can you walk me through kind of an example where, just so I understand how, like using Triple Whale to understand my media mix model, what does that look like? And what would be a really, or do you have any examples of like pot really, you know, great cases where people looked at that and determined, because everyone's looking for halo effects, right? Everyone is looking for their marketing to create flywheels. And quite often what happens is the reverse happens where you end up spending way too much on the bottom of your funnel. You're not bringing enough new customers in. Can you give me an example of how, yeah, like Triple Whales unlocked some MMM insights for brands?

10:27Yeah, I think there's two examples that I think there's one specific to MMM that comes to mind that I'll give, but there's another that I think is really relevant to your question of just like, how am I able to understand like where to invest in my spend and how things correlate. So first is that we recently, about a month and a half ago, released MMM in our Mobi chat feature. So basically you can ask Mobi in chat, just like ChatGPT, hey, using MMM, can you look at my previous quarter's performance, look across all of my channels and identify where I should invest more in, what tests I should run, and where, how I should basically readjust my budget allocation.

11:04And so we saw a really cool example of that being used with a brand of ours who was running a giveaway. It was on the last day of their giveaway. They asked Moby to look at, hey, you know what, on the last day of our giveaway, how should we adjust our spend using a MMM model? Moby then gave them a plan to do so that had channel by channel specific allocations to look at, and then an action plan of what to do strategically to do that. and the brand saw their highest revenue day in history by following the exact step-by-step plan. We've also seen brands take the MMM plan and say, hey, let me plan my next quarter.

11:43And you basically use that to influence their quarterly planning. So there's like both the near-term and long-term components. An interesting use case that I think has been also really interesting was we had a user who asked Mobi to understand the correlation between MetaSpend and Amazon purchases. And Moby was able to basically give them a really strong understanding of, hey, when I put a dollar into Meta, how does that translate downstream to Amazon purchases? To understand like, hey, actually our Meta is impacting, our MetaSpend and Meta ads are impacting Amazon. And so they invested more in Meta in specific ad sets that actually were having an impact downstream on Amazon Correlation.

12:31And that was really only possible because we're bringing in all the data together.

12:34Anthony DelPizzo:Strategy is the key word across the entire pilot house organization these days. Because tactics without strategy is generally going to waste your money. And so one of the things we asked the 875 marketers was how they're using AI, whether they're focusing on efficiency, whether they're focusing on ROAS. And it seemed a lot of people cite, 83.5 % people cite efficiency as their top goal with AI. That's sort of like the lowest hanging fruit in a way where you feel like you've got to take advantage of these tools to get more work done. What are you guys seeing on your side? Yeah, that's consistent.

13:11I think saving time is truly step one. Like that is most tangible for brands. It's like, hey, I used to run an example that we have here is Dixon Flannel, an early adopter of Mobi Agents. They used to run creative analysis across all of their channels. It would take about 10 hours a week. I think we cut it down to 10 minutes. And immediately, you can see the output within a couple weeks. So it was really profound. I think what we're seeing is the long-term value of being able to have that time back to be more strategic and have insights that you perhaps couldn't uncover because you didn't have the resources on your team.

13:47Your analysts were just too backed up. We're seeing then agents have impact on things like ROAS, where LSKD, a global retailer in Australia, they even saw a significant impact on their ROAS by leveraging our marketing channel performance agent because they were just able to find things like whether it was recommendations, channels that were performing better, ad sets that they could switch out earlier a lot faster because they had an agent doing that versus having to ask someone on their team to run a query. And so it's like the time savings is probably most immediate, but then pretty soon after, it's like that time gives you a lot of more time to be strategic that then you can actually make smarter decisions that have an impact on your bottom line.

14:32Anthony DelPizzo:Who are your cast of agents at this point? Which discrete agents have you guys built out? Yeah, so we actually have a library of about 70 pre-built agents, which maybe sounds overwhelming, But the way that we've categorized our agents are all around the jobs that we know our customers they're looking to get done. Whether that's creative analysis, that's retention marketing, that's acquisition marketing, that's operations, conversion rate optimization. So we have basically a cohort of 10 or so agents in each of those that are very supportive on like, hey, this specific job. So maybe it's analyzing my meta performance week over week.

15:11Understanding what creative is working in Google. We have this library that lives inside the Triplewell app. We've also externalized a lot of it on our website just to provide some color into like the things that we're thinking about. And we're constantly getting feedback from our customers of like, what are the jobs that we're not solving well? And like, what are the things that you would love to automate? Because then let us automate that for you.

15:33Anthony DelPizzo:I think everyone has a different approach to the way they use AI on their team. And I think a lot of people, you know, I use ChatGPT and I've built a few GPTs that help me, you know, do things with the podcast. I have whenever we're working on a new project, I have a new folder. And I'm finding it's getting better and better at remembering all of the data that I put into it, which must be just a huge amount of memory that they're storing for us. But I think one of the cool things about what Triple Whale represents is it's that you're building off of a library that doesn't – the whole company gets based on the data from your Shopify store, the data from your meta accounts.

16:10Anthony DelPizzo:and it becomes an ongoing learning library where the learnings are persistent and kind of always building. Is that how you guys see it as well, that you're just sort of building this incredible database of actionable data? Absolutely, Eric. I mean, I think that's the biggest thing with like why Triple Oil and not ChatGPT. It's like when you compare the same prompt, you're going to get two different answers because we ultimately have access to all of your real-time data. And that means like updated live. And so like ChatGPT doesn't have access to the campaigns that you just ran an hour ago or your real-time ROAS.

16:44When you think about what makes Triple Whales AI like so powerful, it all goes to the data that sits on top, like below it. Because we're actually, we're working with ChatGPT, we're working with Cloud, we're working with all of the models because they're really building the best in class models and intelligence. It's how can we make sure to build the fastest and most effective way to get your data into that model. And that requires building a unified schema. So understanding how things are mapped and what data points are mapping effectively. So that purchases, website browsers all get mapped to the correct place and are fed to the AI models with the correct context.

17:21And then on secondarily, it's like we have all of these, we're now 45 ,000 brands that use TripleL. And so that feeds in and trains our TripleL models and our TripleL data and the AI models that we're working with on the fact that, hey, we know what good looks like for your industry. We have real-time benchmarks. So when something's underperforming, we know why it's underperforming and what is working best. And so being that e-commerce and retail expert really has significant output on our AI output, not to use the output buzzword too much, but really because it's like, Like, hey, Mobi speaks ROAS, CAC, LTV out of the box.

18:02And you can even customize it for what ROAS means for your business, how you want to focus on NC ROAS versus regular customer ROAS. It's really, really, really just like the e-commerce expert that can sit working closely with all of the top AI tools.

18:17Anthony DelPizzo:Do you use Mobi to set up ongoing reports that are made all the time? Or is it more of just like, as you think of things, you fire off questions? Yeah, great question. So really it's two ways. So Mobi consists of MobiChat, which is exactly that latter use case you mentioned. Like I have a question. I need an answer. MobiChat can also build you reports. It can build you dashboards, build you plans. You can also then turn those chats into an agent. But we also have what's called MobiAgents, which are those recurring basically reports and questions that MobiAgents can also now take actions for you.

18:53So if you tell an agent, say, hey, if my ROAS dips below three on this channel, like I need you to shut off this ad set. Moby agents can now do that for you. And so they're autonomous, they're recurring, and they're leveraging the same data that sits out on your triple L.

19:11Anthony DelPizzo:One of the cool findings that I like hearing as a marketer was that there doesn't appear to be a lot of displacement of marketers yet. People are not replacing full team members with AI. It's not taking our – the clankers aren't taking our jobs yet. That's my new favorite thing that I'm hearing on social media is people calling AI clankers. Yeah, that's great. And that was reassuring to me where people are just using this to become better marketers, to have more time. But I'm wondering, is the next stage of AI agents is like, are we going to have AI employees in the next three years? where instead of using the agents in these discrete individual ways, we'll just have an ongoing person that knows how to use an ongoing AI employee that knows how to use all the different kinds of agents and just becomes one layer removed from having to manage it even.

19:58Yeah, it's a great question. And we're seeing a couple of different things. We're seeing kind of both where like, Hey, we view AI as giving your team superpowers, allowing you to do more with the resources that you have and really allowing you to uncover things that just transparently aren't humanly possible. And we also have seen, and what we're seeing more of is exactly that, Eric, where we actually had a brand under Alfit. They're a global woman's apparel brand based in Israel that they actually were looking to hire five data analysts. They onboarded with Mobi agents. They then, instead of hiring those five analysts, they hired one more senior analyst who could manage all of the Mobi agents.

20:39And so I do think we're evolving in a world where AI and working with AI agents will not become a nice to have. It's a need to have. And it will be more strategic to be able to, Hey, how can we think about team structure in my org chart as both human and agents?

20:53Anthony DelPizzo:Yeah. Human, human in the loop. Yeah. Love that expression. Cause I think, cause even just when it comes to writing, like we recently wrote a landing page for a new project, our agency community that we're building out and it takes so much working with it still to get it the way you want. And it's still, by the end of it, was a full AI landing page, and it's done extremely well. But the amount of guidance that it still needs to get it the way you want it is critical. Oh, totally. And understanding how to prompt effectively is so critical now. I think we've seen even with GPT-5, we've gone away from just like, oh, yeah, use any prompt.

21:28It's like the way you prompt has direct impacts on the output. And so understanding that and something that we're trying to become as a company, AI fluent, which takes a lot of education and re-education and rethinking how workflows exist, is definitely critical as we continue to evolve.

21:46Anthony DelPizzo:One of the questions that Pilot House was trying to tackle recently was like, how do we make sure that AI is like an imperative across the entire organization? How do we, because one of the key results here was that 40 % of people say that nobody like owns AI in the company. Nobody owns AI development as an initiative. It's like entirely decentralized where everyone across, because you never know where these insights are going to come from. If that's the case where you have all of these people using AI in different ways, what's your advice to organizations that are looking for ways to either wrangle it or just to get the best results out of it?

22:25Yeah, so I mean, we're seeing our best users start small, like start with a use case, but that use case spans across maybe a few different people. because what we don't want to happen and what we found from our brands who have tried 10 different AI tools is that those AI tools aren't speaking to each other and they're running off of different data sets. And so they're giving you different insights that you perhaps can't trust. So it's like establishing that data foundation first of, hey, we're going to be looking at the real-time same thing. And the AI is going to be looking at that too is like rule number one.