Ragnarokast Episode 18 Transcript

Steven (00:00):

I’m Steven.

Spencer (00:01):

And I’m Spencer. Welcome to Ragnarokast, your podcast for all things marketing and MarTech.

Steven (00:06):

Hello everyone. 

Spencer and Steven (00:08):

We’re the co-CEOs of Ragnarok!

Spencer (00:11):

Welcome back everyone. By everyone I mean Steven, because you’re the only one that’s been here before other than me. Welcome back Steven and myself.

Steven (00:19):

Oh, I feel so welcome. Thank you Spencer

Spencer (00:24):

How’s it going Josh? Thanks for joining us.

Josh (00:26):

Oh yeah, it’s going great. Thanks for having me.

Spencer (00:29):

So Josh, would you mind doing a little intro into yourself and your meaning in life and while we’re all here

Josh (00:38):

Awesome. Yeah, absolutely. I run Solutions consulting for our partnership team at Tealium, a software company, CDP Tag Management, API hub. Lots of experience in MarTech and digital advertising. Worked at some digital agencies in the past and originally worked at DoubleClick way back in the day. So in and out of Ad Tech and MarTech, and been at Tealium for about 10 years now and I’ve been working on the partner team for the last four or five.

Spencer (01:09):

Wow, 10 years. Did they give you a medal or a jacket or anything at that?

Josh (01:15):

Well, I think I cleared it out from the background before we started. I had a really nice glass statue that’s on the cabinet behind me most of the time. I got a two month sabbatical this past summer and eight weeks off, got to travel southeast Asia with my teenage son and go to a bunch of fun places and eat some awesome food, and get a little bit of history and culture in between. And then just came back to New York City where I’m born and raised and still live and enjoyed a beautiful New York City summer. It took me about two days back at work to forget about all that relaxation and just sort of get back into the swing of things.

Steven (01:55):

Well, that’s awesome. I mean for a teenage boy to go back to high school and be like, you guys don’t even know what I’ve been doing.

Spencer (02:03):

That’s true. That’s a flex as they say, as the kids say these days. Alright, so guys, I want to start out before we get into the topic, I didn’t ask Steven this before. So Josh and Steven, I’d like for each of you to think of without thinking too long on it, give us a hot take on MarTech at the moment. Your MarTech hot take.

Steven (02:23):

Ooh, I could start. I could start. I got this. Alright, hot take MarTech buzzword ai. Let’s talk about a little AI MarTech. Alright, so I have been seeing, I will say a lot more call it grounded examples, not just like experimentational use cases, but grounded examples and people using tools like Canva and some of these other gen AI type features like Dall-e from Movable Inc. And I’m starting to see it more and more where they’re using it to build a, I think of it as an image repository, a content repository at scale. I mean, I want to say it’s not just like it’s starting to work, it’s always been working, but I’m just starting to see it more and more and I’m having a hard time now actually recognizing if I am being targeted by it or not, which I think is the more impressive thing. So I think hot take wise, I think we’re getting closer to the singularity, so to speak, in the gen AI content generation, AI for marketing type messages when it’s personalized but you don’t realize it. So I think that’s great,

Spencer (03:32):

But you don’t mean the singularity and where we are all one with the machine and we’re energy beings, right?

Steven (03:37):

No, but maybe

Spencer (03:39):

one step closer every day, baby.

Steven (03:42):

That’s right.

Josh (03:44):

That’s the exponential growth of AI. Maybe not this year, maybe next year, but I’ll take that topic. And what I’ve seen a lot of recently is first I guess struggles to figure out what those AI use cases even are, but then once they figure it out that it could be cost prohibitive, right? It’s very expensive to run a lot of this AI and you have to be really thoughtful about how you build it out and how you scale it up. These businesses that are blowing through their whole budget in the first two months of the year and then don’t know what to do with it. So I think that might not be the most sexy part of it, but that’s foundationally you need to be able to figure that stuff out if you’re going to do that for a business. So I’ve seen actually a couple of different businesses mentioned that recently and you sort of have to think about the consumption upfront to make sure that you are not overpaying.

Steven (04:39):

That’s a very good point. Yeah, there’s definitely, I would say the Cal cost have gone up. If I’m going to throw a stab in the dark there, which I think is an excellent topic or an excellent lead into,

Spencer (04:51):

You want to get into the topic number one, which is 

Steven

I want to get into number one  

Spencer

for our listeners out there. I am going to put us on a timer, we’re going to see how it goes and I’m starting it. Alright.

Steven (05:01):

Wait, are we putting you on a timer or putting me on a timer? We’ll see, I don’t want to share my timer with you. I’ll never get to talk.

Spencer (05:10):

No, I have the timer. It’s a timer for all of us.

Steven (05:14):

Okay. A shared timer.

Spencer (05:16):

Topic number one, what is your take on the two to three year future state for the CDP world?

Josh (05:23):

What I see, I think it’s going to be similar to basically all the different software in general MarTech software. It’s going to be just weaving that AI into every step of the process. So in the case of the CDP, you have sort of data collection and enrichment and profiling and activation, and each of those stages you’ll have the AI maybe scanning for anomalies and errors upfront. You’ll have it suggesting the right segments and attributes in the middle you’ll have it sort of or guiding you where to send it at the end. I mean, whether you embrace AI or not, whether you’re a heavy user of chat GPT or not, it’s all going to come to you regardless, right? We’re all going to be walking around with this thing in our pocket that’s filled with it and you’re going to use it on a daily basis talking about your smartphone there. But yeah, it’s the same thing, right? It’s going to be in your refrigerator, in your microwave oven at some point soon too. So it’s going to be everywhere and the CDPs that can’t figure it out are going to be in trouble.

Spencer (06:26):

I don’t recommend putting any CDPs in your microwave. I assume they’ll have some sort of metal in them and you don’t want to start a fire.

Steven (06:34):

Yeah, actually I’ll piggyback off you there, Josh. I think when you mentioned AI in regards to being able to look for anomaly detection or being able to use as have AI do more of governance, I think that’s a very clear value prop for CDPs in the future because data warehouses will have really any stake in the ground in terms of the quality of this data collection, right? Because they’re just a point of collection. So you think you’re Snowflake and you’re newer more, I would say mature providers in space, Snowflake and Databricks. I can’t imagine that they would really care about the type of data you’re collecting as long as they’re getting it and you’re calculating off of it, there’s no motivation for them to have a quality control on that. Whereas with the CDP, your value statement is in the accurate interconnectivity between different ways that you’re pushing that data out and so you have a bigger stake, well you have a pretty major stake to make sure that the data that is within your platform that somebody is collecting and using to either aggregate or to do some enrichment on is very accurate.

(07:40):

And if you can do some pattern recognition on that from whether it’s say, hey, this is transactional data and this is what is good transactional data and this is what a schema for good transactional data looks like, and this is what you’re sending us and this is where you’re missing something that can just automatically do a quick assessment of that without necessarily needing a human. And then you do that at the scale of as you’re entering new data, there is an AI that sort of evaluates it for you. Now the reason I say that is because even at Ragnarok, when we sit on data councils or we do these sort of governance workshops, there is always in some way, shape or form some engineering team somewhere that is going to break that process. And your only quality control is essentially having humans who either from a permissioning perspective allow or disallow that data to come through and then through that end up slowing down adoption.

(08:34):

And so if there is a way to speed up adoption, but have an AI do the intercept, clean it for you and then push it forward, you really kind of make the CDP easier for really most data teams out there to adopt that need to think like, oh, lemme go through data council force. I don’t think it eliminates the planning process, but at least sets a standard of what’s acceptable and what’s not. And that can be things even like is this Camel case or snake case? What’s our standard around casing? I think on the other side of that, which is more on the activation side, and this is where I think CDPs, I’m excited to see them play a bigger role here is to being the central point of the data and having also the, we’ll call it the telemetry data and the exhaust that comes out of your Braze or your Iterable or your AppsFlyer, your Facebook, anything where you’re getting some of that data coming back out of those tools and is landing within the platform that the CDP would be able to sort of standardize or look at that and say, here’s the types of use cases you should be activating, right?

(09:34):

Like, oh, I see you have engagement data from Braze coming in here and then I also see you’re getting Facebook lead ads coming through the CDP, right? Two great examples, right of data. And then you could say something like, did you know that people who use Braze to create a custom onboarding experience for Facebook lead ads experience a XXX percent return, higher return in their ROAS compared to people who don’t? And here is a set of data that you can use when those people come in that you can then activate in Braze. So I think it’s almost like you become a bit of a hub for the marketer to understand where their opportunities lie. And it might be, even if you miss half the time to be at 50% of good recommendations, I think is pretty powerful. So maybe that’s at the third year end of that three-year run, but I can see a world where you’re using more of the brain power of the AI as opposed to just trying to do, which I think governance and stuff like that is almost like your first, your next year, year and a half around

Josh (10:42):

That. I saw an experiment that we did with the help of a tech partner where they built basically three different bots, three different versions of, I forget which large language model they were using. They pointed it at our documentation and then they pointed it at a synthetic dataset of customer data and one was sort of like you were saying before, suggesting new configurations and things within the UI. One was a data analyst that was looking at the data and basically speaking to us about the data. And then the third was, I guess they were a combination of the two, but it was really interesting. And at this point you have to have these separate instances so they each can sort of get their own expertise, but soon enough it’ll get plugged right back into the UI and it’ll actually do those configurations for you once you sort of say, go ahead and do it. But yeah, it’s amazing to see the progress and that you just sort of mess around with these bots for a couple of days and ask ’em questions and they really learn. They get smarter.

Spencer (11:44):

It actually rolls nicely into the second topic, which is AI’s impact, but it relates to what you said before, Steven, about governance. Do you think that because you won’t necessarily always have to have a human checking every single, is it snake case or not? And you can have the CDPs AI component do a lot of that for you, do you think that means marketers could actually get more useful data into the platform because they’re spending less time on governance basically? Can they actually activate more stuff than they would’ve been normally able to?

Josh (12:18):

Just back to what you said, I think the question was really about time savings and efficiency. And I think AI is absolutely going to explode when it comes to those sort of things when it comes to boring, repetitive tasks. Would you rather I’ve done it before, sit there writing SQL queries or staring at spreadsheets and analytics reports, or would you rather let the AI do that for you? And then you get to do the more strategic things. So when it comes to something like governance, that’s one of those topics, it’s sort of difficult and boring and repetitive and you’ll have the AI to sit there and scan and to look and to check to make sure everything is going the right way. And then you go onto more exciting things like creating campaigns and managing across channels and personalization at every customer touchpoint. I think the AI is going to help there the exact same way by automating the boring steps that people’s brains aren’t really built to do. The answer is yes. I dunno. Steven, if you had anything to add to that?

Steven (13:16):

No, I think you nailed it. I think it also, when Salesforce had their whole Dreamforce event or whatever they call it these days and they talked about the future of the platform being agent driven, you’re not as in the UI as much. You’re kind of at the building block stage when you’re talking about governance or agent doing that kind of bland work, but you’re still activating in the platform I think is what would be amazing to see is that you actually don’t have to interact with the UI that much. You essentially just have the agent where you’re giving ’em all of the prompts in there and the agent is sort of building out the audiences or segmentation for you to your point around refine this even further. So I have a different version for each person where you’re getting more of the personalization or you’re able to apply more granularity without necessarily having to go in and not just repetitive on the analysis side of what data’s in there, but also repetitive in terms of creating hundreds or thousands of audiences within the CDP to then activate downstream.

Josh (14:15):

I think we’re all going to feel like we have these new assistants, these new work friends, and you’re going to be slacking with them, you’re going to be sending them emails, you’re going to be talking to them on the microphone and it’s going to be a bunch of software. Maybe it’ll make the day feel a little more friendly. I don’t know. But it’s definitely going to be a change to the way we’re all working for sure.

Steven (14:36):

I did see somebody post they were looking for, they were hiring for a manager role and one of the qualifications was that you had experience managing AI staff. They were basically saying AI agents as your staff besides human and non-human staff. And I kept thinking to myself, the Jetsons, the 

Spencer

Rosey 

Steven

Maid Rosey. Yeah, Rosey the maid. It was like one of your like, oh yeah, I managed Rosey for a couple of years. I’m good with non-human assets.

Josh (15:07):

Well, that’s what you hear, the CEO from Nvidia talking about that, and every employee in that company is going to have many, many agents working for them to distribute that work. I mean, they’re ahead of everyone else, but we’ll all be doing the same thing soon enough

Spencer (15:24):

And bring it to orchestration and channel optimization side. If you’re working with an agent, how does it know to optimize for the right channel? Isn’t there some element of human input there, or do you think there’s a way to actually train it to like, oh, this should be SMS versus a Facebook post or something like that? 

Josh (15:46):

I think you have a goal you optimize for, right? Is it profit? Is it you want to move that last product that’s on the shelf, the particular product that you’re not going to have any more of next month? Is it spend on your ads? Whatever your target is, the AI can optimize for one or more targets.

Steven (16:06):

I think the interesting thing is where does the CDP play in that, right? Because the CDP is probably not necessarily selecting down to the, am I going to send a text message or an email, right? That’s probably always going to be the role of your market automation system. But what it will say is something like your point, Josh, should I be sending this to Facebook as a Facebook audience? I might increase my conversion rate by having another channel that I’m paying for, but is it going to move the conversion rate enough that the incremental cost on that is more efficient than just sending it to my retention platform to do the actual marketing? And then think about that at, no, not just Facebook, but Pinterest or TikTok or anywhere else where I am trying to reach that person, even if I already own them as a customer.

Josh (16:52):

Yeah, absolutely. I mean, if the marketing automation tool is going to have sort of control over its domain and its channels and it’s going to be able to make those decisions, I would think at the very least we should let the CDP know what signals will make it perform best and make sure we’re tracking those signals and passing it on. And then other channels, the call center, other channels in digital advertising, other channels in whatever the next is, they’ll all have some of the same signals that they’ll want. They’ll have different signals that they’ll want. You just got to make sure that you’re capturing all of them and sending out exactly what’s needed to each one.

Spencer (17:28):

And in terms of the human element, how do you see it? There’s still a lot of work to be done right now for a human in the CDP, UI. How do you see that affecting practitioners in the next few years? Will people be out of a job or will it change their job? At Tealium, what’s on the roadmap? And I think that’s something that a lot of people have fear over is what will this mean for my job in two years when this new feature comes out or it gets adopted?

Josh (17:59):

Yeah, that’s an excellent question. And going back to when I was a teenager and my very first real job, I was in an office scanning papers to put stuff that was previously in books into a digital repository. I’m really happy I don’t have that job anymore. I’m really happy that most people don’t have that job anymore. So I think it’s going to get rid of either very repetitive and boring jobs or it’s going to get rid of very repetitive and boring tasks that you have in your job. I think that the people who will be most affected are the ones that don’t embrace the AI in the first place. So get out there and log into chat GPT or Gemini or whatever, which one you prefer, and try and figure it out and try and learn it. I use it all the time. It makes things really so much easier. So I would just encourage everyone to get comfortable with it. I know my kids are, your kids are, they’re going to be okay, but maybe an older guy like me, you got to stay on top of it or else you might get replaced. But I don’t think that that’s really, that’s not the goal. The goal is to make everyone’s life easier.

Steven (19:06):

Oh yeah. And Josh, we refer to us older guys as seasoned or experienced. Yeah, I have to agree with you on that, Josh. Obviously getting in there and seeing it is, if you’re afraid of it, how much of that fear is just fear of the unknown versus, because if you used it enough, you’ll realize, yes, it can do a lot and there’s a lot it can’t do yet. And so I think understanding where its limits are is just as helpful as where it can support you. I think on the other side of things, it’s so much of it will be domain-specific. If you think about Iterable over the past, I think a couple of months ago, they put in a general release their AI journey builder, and it is phenomenal, so much better than I would’ve ever expected it to be, right? I was expecting something kind of janky and maybe wouldn’t get it right.

(19:59):

But with a couple of, you’re writing a couple of paragraphs, you don’t just put a couple bullet points, you put a couple bullet points, you get a couple bullet points of material back, but you’re writing a brief and think of it instead of like, oh, now I have to go and do this. I can actually have Iterable build out a majority of the journey for me, and that just saved me an hour or so in connecting little module blocks and putting things together. So that was incredible. It doesn’t replace me. Somebody had to write that prompt. And then somebody also has to go through and validate, does this actually work the way that I intended to? But how much more valuable am I or a marketer if the AI is taking away this redundant piece of my job and I’m like, oh, I’m actually thinking here more and more and more.

(20:47):

And then that thinking that more time that you’re spending thinking it will eventually spark really good ideas. But if you’re stuck in mundane activities, you don’t get the opportunity to think. And so I almost think of it, if you look at an eight-hour day and four hours of your day is spent in, we’ll call it repetitive or redundant or low-level tasks, if you can scratch two hours or three hours and call that back, even if that’s, you’re not necessarily doing 30% more activity, even if you’re doing just 10% more activity, you have 20% more room to think, right? I think that ability to think is what leads to more impact. If I’m a marketer, I’m more afraid it’s more like, we’ll try it out, because again, you’ll find where its limits are, but then you’ll also be able to see like, wow, I actually have more time in my job to do the things that are more interesting, and I’m not putting things and putting them in places like following this McDonald’s burger flipping type activity.

Spencer (21:49):

I mean, my take on it is it’s probably not going to stop being buzzword AI anytime soon.

Steven (21:57):

True.

Spencer (21:58):

And investors will keep throwing money, ridiculous amounts of money at it. And so whether or not it’s a good thing for the world doesn’t matter because it will continue on because people will keep putting a ridiculous amount of money into it. So that’s pretty much the reality I have here real time. What does it take to make it work? I don’t know if I have context into what that full sentence is supposed to mean or I don’t remember.

Steven (22:24):

I think this is a great prompter, Josh. This is the Tealium value statement almost, but

Spencer (22:30):

Alright, topic number three, real time. What does it take to make it work?

Josh (22:36):

Awesome. This is something Steven and I have discussed before, and he’s right where Tealium’s real-time data platform is one of our unique or relatively unique aspects compared to a lot of the other ones that are out there. I think there’s a couple other things we do really well in addition, but being able to collect the customer data in real time, turn it into a profile, send it back out to wherever it has to go, like a braise, like an Iterable, like a Snowflake, and then maybe learn something from that as well. And then get that feedback loop back. In the case of Tealium, that’s about 200 milliseconds. So you tap on that app right now, you call that call center right now that can influence a personalized experience on the website that can influence what the message is that gets triggered to be sent to them milliseconds later.

(23:31):

And we’re living in the world where all our expectations are at Facebook and Netflix and Amazon, and we want the good products recommended to us and the good shows recommended to us. That’s where we’re at. And if your business can’t do that, people are going to look at you and just move on to the competitor. So if you have thousands of engineers and you want to build it all yourself, that’s absolutely an option. But there’s a lot of other businesses out there that can’t. So that’s what it means. It means being able to reach the customer at the right moment with the right message. Now, what does it take to make it work as a whole other sort of story?

Spencer (24:08):

Josh, Steven, what advice would you have for anybody listening as we think about the future state of CDPs, how the AI’s role in it? What would someone listening to this makeup, all this, what should they go out and do? What are you recommend that they do with this information? So whichever one of you wants to go first,

Josh (24:26):

I think that there are some amazing resources out there free to learn more about ai. Nothing Tealium related at all. I think it’s fascinating and we all have to keep up on top of that. Go to these free sessions, go get a training or a certification. The certification will probably be obsolete in a year, but you’ll learn so much doing it. That would be my first thing. And then the second is in life, try and be a scientist. Get the facts, examine different people’s opinions, make sure that, make sure that you understand the reasons why things will and won’t work. And there’s just a lot of confusion out there right now. No one knows exactly what the answer is going to be 12, 24 months down the road. I think that we’re all going to have to pivot and change and adjust to things. So keep an open mind and keep learning and just use that education to make the best decisions you can.

Steven (25:22):

Yeah, mine are a little bit different here. I’d say first and foremost is as you sort of look at your tools that are within your marketing capability right now, which of those are, I’d say adapting AI or including components of AI as part of their feature set? And do you feel like that what they are adding from AI capability is helpful? So kind of give those a try and see if it’s helpful because it’s always an opportunity to say, am I in a platform that is working its way towards more of a legacy platform as opposed to someone that’s keeping up with the trend? And then the second thing I always encourage, especially if you haven’t used AI before, telling my parents this as well is like, go on the chat GPT and have it teach you something. Ask it. I had it. There’s a fantasy book series I was reading and I had asked the questions, 

Spencer

Nerd, nerd, nerd, nerd, nerd.

Steven (26:10):

I had it explained to me the different historical things that they reference and how it connects to the current character’s experience. And I was pretty surprised at how I was able to, I dunno where it found this information. I would say probably 90% accurate in terms of being able to not only condense it down to me being able to read it in a few paragraphs, but also to give it some analysis as opposed to me trying to do that analysis myself. And I just think that’s really interesting when you’re like, wow, if it can do that for a literal fantasy world, what else could I use that for? And I think just being able to stretch your mind a bit on what can it teach you as opposed to what can you teach it. I think that’s an interesting takeaway for folks as well.

Spencer (26:53):

Wow. It’s not, no, not what can your country do for you, but what can you do for your country, for your AI country. Steven, I’m going to challenge you to come up with a question for Josh and it has to be marketing related. It can be silly or it can be real. Would you rather give Josh a would you rather scenario?

Steven (27:16):

Alright, Josh, would you rather write a brief and have to put a whole Dunning or credit card breakage series together or optimize an abandoned cart campaign, an e.stablished abandoned cart campaign?

Josh (27:30):

Yeah, I like to do new things and I’ve done many abandoned cart campaigns, so I’ll choose the brief. They both sound kind of tedious, but at least I’ve learned something new that time. I’ve done plenty of abandoned cart stuff.

Steven (27:46):

Ooh, like it.

Spencer (27:49):

Alright, so there we’ve got it. I’ll be expecting that brief posthaste. I’ll do a fast follow on that. Thank you Josh. Alright everyone, thank you for listening. Thank you Josh, so much for joining. We’ll catch you next time here on the Ragnarokast

Steven (28:04):

Ragnarok!