AI Can Help Me Personalize That?!

Media Thumbnail
  • 0.5
  • 1
  • 1.25
  • 1.5
  • 1.75
  • 2
This is a podcast episode titled, AI Can Help Me Personalize That?!. The summary for this episode is: <p>If you’ve been online at any time this past year, you’ve most likely heard of how revolutionary many find generative artificial intelligence (AI)&nbsp;to be. Being able to generate written works and images in distinct styles, it’s easy to see how it can dramatically improve marketers’ productivity.&nbsp;But is AI something you can truly “set and forget?”&nbsp;</p><p>In this Level Up episode, our hosts, Joe Kaltenthaler (Sr. Manager, Data/Identity Practice) and Laura Madden (Sr. Manager, Marketing Strategy &amp; Services) discuss the current benefits and limitations of generative AI and how marketers can become better AI prompters to help automate the brainstorming process, strengthen the relevancy of their messaging, and personalize the customer experience. &nbsp;</p>
🤔 What is Artificial Intelligence?
00:49 MIN
📈 AI Can Help Your Marketing Mature Quicker
01:06 MIN
👩🏽‍🏫 AI Won't Do Everything For You; It Needs a Teacher
01:12 MIN
🤖 An Introduction to ChatGPT
00:22 MIN
💻 DALL-E: AI-Powered Image Generation Tool
00:46 MIN
🧠 Generative AI: While Great for Brainstorming, Don't Trust Everything It Outputs
02:18 MIN
✉️ ChatGPT Example: Brainstorming Email Subjects
03:18 MIN
💪 Generative AI Won't Automate the Prompter
00:19 MIN
🤔 DALL-E: How it Benefits the Image Brainstorming Process
01:24 MIN
😲 Generative AI Pro-Tip: Ask the AI Why It Provides the Answers It Does
01:49 MIN
⏰ Other Power AI Tools: Email Send Time Optimization
02:11 MIN
🕸️ AI Can Even Help You Personalize Web Content
02:10 MIN
👾 The Bottom Line of AI: Don't Set It and Forget It
00:54 MIN
📚 Embrace Learning How to Prompt AI
00:26 MIN
💡 Curiosity Is Your Best Friend When Working With AI
01:06 MIN

Joe Kaltenthaler: Thank you for joining us for this episode of Level Up the podcast, for marketers by marketers, created by Lev, that distills best practices and strategies focused on helping marketers increase their EXP, one up their strategy and grow personally and professionally. We're your hosts, I'm Joe Kaltenthaler, Senior Manager of our Data and Identity Practice here at Lev, and I am joined by...

Laura Madden: It's me, Laura Madden. I am the Senior Manager of Marketing Strategy and Services. So Joe and I will be your strategic and technical arm as we navigate these discussions today. So Joe, what are we here to talk about today?

Joe Kaltenthaler: Yes, so today we're diving into the captivating world of artificial intelligence and its impact on the marketing landscape. But before we delve into the deep end of AI, let's break the ice with a thought- provoking question. And that question is, if you could have an AI powered marketing assistant with one extraordinary ability, what would it be and why?

Laura Madden: I mean, other than the genie answer of, can I have more wishes? Can I have you make more AI marketer babies to do these things for me? I think mine, if it could magically tell me why people aren't converting, I feel like that's the age- old question of we're driving traffic, we're getting qualified traffic, we're getting people, and they're just not converting, and we make our best guesses of if it's e- comm, is the checkout process not clear enough? Or is the offer not good enough? Is the price not right? What that is? If somehow my little AI buddy could pinpoint at the user level, Joe, why aren't you buying my stuff? I'd probably be a millionaire.

Joe Kaltenthaler: Yeah. Or a bajillionaire maybe even. Yeah, I really like that. I think my answer gets pretty meta because I want my AI to be so smart that if I ask it this question, AI, what ability should you have? It would be able to tell me what it should have and then it would have that. And then you see where I'm going.

Laura Madden: Stop it.

Joe Kaltenthaler: It's Skynet, it's Terminator. It's the end of the world. So that's what I-

Laura Madden: Joe took it too far. You took it too far already. Well, Joe has given us a glimpse of what some people think or expect AI to be, and basically being monsters who take over the world, robots, whatever that might be. But for the purpose of this discussion, let's define what AI is and what we mean when we say artificial intelligence specifically as it relates to marketing technology.

Joe Kaltenthaler: Yeah. I think it's a really good place to start. And so yeah, let's just define that right now. So AI or artificial intelligence, basically think of it as a way to make machines or computers think and learn similar to how humans do. So just like you learn new things by practicing and getting better over time, AI is all about teaching computers to do the exact same thing. Simply put, think of it as having a really smart friend who can solve problems, play games, or even help you with different tasks if you ask it. And honestly, you've already probably come into contact with it through things like a voice assistant, video games, or even chatbots. So it's really all about making machines or computers really smart and helpful as the primary purpose. So nothing scary at all about that.

Laura Madden: Right. Right, right, right. No, I love your point about that you may have already encountered that. I think from a marketing perspective and a consulting perspective, when we think about how we're trying to move our clients forward and have them be more advanced, we operate off of a marketing maturity curve where the end all be all and the tip- top of that has always included an AI element. And up until maybe the last three to six months, yes, that's always been the goal, but it's been out of reach. It's like, well, yeah, of course I would love to be able to have AI power everything that I do, and that's ultimately what I'm trying to get at, but I'll just stay over here in my segmentation lane or make sure that I have some things automated and I've got segmentation running and that's going to be good enough for me because AI is not really achievable. But with even just those items that you mentioned and some of the newer technologies that we're going to talk about today, it's feasible for you to get to that ultimate point of the marketing maturity curve sooner rather than later. And it's not just a pipe dream and you say, " Oh, yeah, sure, I'll go tell my C- suite we're aiming to be AI driven," knowing that we're never going to get there. It's like, oh, you're going to have to put your money where your mouth is, and you might actually get there, but that's... it's a little bit scary, but it's more exciting because it could be a thing and it's going to free you up to do other things once you train the machines to do more of the mundane tasks. So with that in mind,

Joe Kaltenthaler: Yeah. Real quick, I think something that dovetails off of that is I think a lot of people hear AI and immediately think, "Oh, the model's just going to do everything for me, or I just buy the thing and then look at our revenue grow 10x" or something to that effect. When really, like I'd mentioned earlier, the model is only as good as the data that you give it. Essentially there's a need to be involved and to not assume that you buy the thing or you implement the thing, and then it just works. You have to... And we'll talk about this a little bit throughout the episode today... But you have to be willing to be a part of the process and be there to provide inputs or course corrections and ultimately provide that oversight. So I think that's just a thing that might get overlooked or people may not realize that there's the human element still, but I think it's just something good to just bring up.

Laura Madden: Yeah, AI has to learn, but that means it needs a teacher. And spoiler, that's probably going to be you.

Joe Kaltenthaler: Yes.

Laura Madden: So be careful what you sign up for.

Joe Kaltenthaler: That's right.

Laura Madden: All right, so as we talk about some of these things that we're going to need to teach to get smarter from an AI perspective, what are some of the programs that you're seeing out there right now, Joe, that you've been playing around with and you think that our customers could play around with, from an AI perspective?

Joe Kaltenthaler: Honestly, if you've been on LinkedIn or read a news article in the last six months, you've probably heard of either one of these programs. And so I think it's probably good to focus on two main ones. And those are ChatGPT. And I'm probably going to get some eye rolls because it's talked about so much. But ChatGPT is one. And then DALL- E or similar program to DALL- E is called Stable Diffusion, but we'll just be talking about ChatGPT and DALL-E. And so first off, I'll just start off with a quick little intro on ChatGPT. So ChatGPT is essentially a computer program that allows you to have conversations with this computer similar to how you would have a conversation with a human. It uses a ton of information that it has learned from reading and learning all about the world to understand and respond to what people say. And it's been trained by analyzing just loads and loads and loads and loads and loads and loads and loads of data. So that's ChatGPT. So think of it as text input, text output. So ask it a question, get an answer, that's ChatGPT. And then the second item is DALL- E. So similar to how ChatGPT operates with a text input, DALL- E is the same, but instead of that it provides a generated image output. So pretty awesome. So think of it as you provided a prompt like, "DALL- E, I want a image of a cat wearing a hat," and it will produce an image. A very realistic image based on that text input, that's a cat and a hat. The more specific you get, and this will be a theme that we talk about today... But the better you get at prompting these two engines, these two programs, the better your output and the more specific your output will be.

Laura Madden: So the better teacher you are, the better your student will be.

Joe Kaltenthaler: Ooh, yes. Put that on a T-shirt.

Laura Madden: Ooh, buddy. Coaster, all the swag.

Joe Kaltenthaler: Yes.

Laura Madden: Well, I mean, Joe, I know we had time to prep for this, but those explanations were spot on from my POV and were super helpful even though I've been in the space and understanding it. So I just want to say kudos to you for putting those definitions together, and I hope that they were helpful for our audience.

Joe Kaltenthaler: Well, you are so very welcome and you should actually be thinking ChatGPT, because I knew what we were going to talk about today and honestly, our little intro icebreaker question was generated by ChatGPT, the artificial intelligence description was generated by ChatGPT, I asked it to describe it to a third- grader. And then the definition for ChatGPT and DALL-E were all... All of it's ChatGPT. So my job today was to ask it questions and then read off the answers. So you're welcome.

Laura Madden: I feel bamboozled, I feel hoodwinked, I feel like I don't even know you anymore. Are you even Joe? I'm unclear. I think you are ChatGPT.

Joe Kaltenthaler: Beep boop. Yes, Laura.

Laura Madden: You are the Terminator. Well, I think whether I liked it or not, Joe just gave us some really great examples of how you could leverage some of these AI functionalities. But I think what we've been seeing is they might not be ready for prime time in terms of like you were saying earlier, Joe, just input some stuff and it's like, "Cool. I don't even need to look at it, shoot out this email to my couple million subscribers and I'm just going to trust what ChatGPT and DALL- E say." We're probably not ready for that and we'll talk a little bit more about why in a bit. But what it is ready for and can be really useful for is just brainstorming. So especially in... sounds so cliche, like the fast- paced world, but a lot of you might just be a marketing team of one, and it's on you to generate all of these subject lines and generate copy. In my early marketing days, I remember I wrote SEO copy for a promotional products company and they had about 30 different types of promotional pens. And I was like, I'm running out of ways to cleverly say, you should put your company name on this pen and sell it and then it was just a hot mess. So I'm thinking ChatGPT could have been very helpful for me as I'm getting burnt out writing about promotional pens. Or since we're all remote, you're sitting in your home office trying to figure out what the heck am I going to do for this presentation for all of these things? ChatGPT can be your brainstorming buddy. It can help with that. It might not necessarily give you a really great ROI to start with. It's not necessarily going to save you a ton of time and money maybe at the very outset. So it's another expectation you should have because again, you have to teach it. You have to refine your prompts. But it's a great starting point instead of just our normal, "Well, let me Google this." It's like, "Oh well, let me ask my ChatGPT buddy". So I think from a brainstorming perspective, that could be really helpful. And Joe, I know you probably played around with this even more because you just went nuts with the ChatGPT ahead of this, so you probably have even more examples for us.

Joe Kaltenthaler: Yeah, no, I totally did. So I used Northern Trail Outfitters as an example. So Salesforce's outdoor gear brand that they often use for a sample e- commerce retail example. So I literally wanted to put my money where my mouth was. And since this is an episode about AI and we wanted to provide a couple examples of some ideas or how to use ChatGPT as a brainstorming tool, I went through and exercised myself. The prompt that I gave ChatGPT probably went in way too much detail at the start, but here is the full paragraph and here's an example just right off the bat of how you can provide a prompt and then get an output. So here's the prompt that I provided: "Provide five email subject line options for a retail company providing outdoor gear to people with differing levels of expertise around the outdoors, ranging from hobbyists to professional outdoor athletes. The user persona I would like to target is a hobbyist outdoor camper between the ages of 20 and 32 years old, female, owns a dog and goes camping one to two times per month. The email content is intended to provide personalized gear recommendations based on recent purchases this type of customer has made before and should be no more than seven words."

Laura Madden: Well hot dang.

Joe Kaltenthaler: Yeah,

Laura Madden: That's pretty specific.

Joe Kaltenthaler: I'm going to take a deep breath because I'm a little lightheaded after reading the prompt, but that's an example of this is how specific you can get and if you see the output at the end, which I'll read a couple of these off... If you see the output and you're not happy with it, feel free to go in and tweak it. And that's what we'll talk about that a little bit more. So here are a couple examples: "Discover Perfect Gear for Your Outdoor Camping Adventures!" "Tailored Gear Suggestions for Your Outdoor Escapades!" "Get the Best Camping Gear, Just for You!" So a couple of those examples I think ChatGPT went very heavy on the exclamation points. Which again, nothing wrong with being excited about camping gear and getting outside, but I think it's a great example of within that you've got a couple keywords that you might really like. So having the word tailored versus just for you or camping adventures versus outdoor escapades. I think this is a great example of just immediately coming up with some topics or ideas or bullet points that you can just look at, and then that just starts to create a lot of brainstorming opportunities from that output right away.

Laura Madden: Yeah, I mean, I feel like this serves the purpose of a thesaurus. Just what you were saying of, "Oh, I got stuck, and I'm always saying adventures. Maybe escapades isn't exactly the word you're wanting to use, but it gets you thinking in that other way." I don't know. Escapades has me all in a different head space. That one's a little weird to me, but it can act as that thesaurus. It can even just things like tailored versus personalized or just for you, if this really helps you get stuck in a rut. And while your prompt was very specific, if you think about it when you know your brand, that's what's going through your head all the time anyway. You know what your brand is about so the idea of we market from hobbyists to professionals and we want to give professional recommendations, you probably even loosely have your personas together of saying most of the people that shop my company have a dog, or this is how often they do. So that's not something that you have to create out of thin air. You probably already have all of that information. It might just be in your brain or it might be written down in a presentation that you have on personas or whatever. But this also just helps it be a little bit more scalable. So we say if you're a marketing team of one and you win the lottery, and we're like, "Well, there goes all of our knowledge about how we write this content because we didn't have any of it written down." If you have all of these prompts together, that could be helpful. So it might be a little bit more work upfront, but feels like it could be really beneficial in the long run in your marketing escapades.

Joe Kaltenthaler: Well, I think even to your point, even going through the mental exercise of creating the prompt and then the mental sweat that it takes to review it, iterate and adjust, I think helps you as a marketer further define and hone in on exactly what fits, what doesn't, through this almost conversational process, you're starting to get a better idea of who... even in these prompts, there's nothing in here about dogs. So maybe I go back to ChatGPT and say, " Okay, well maybe I really want this to be a dog focused subject line." And now you have to get into the mindset of what's the right level of creepy? If we have the data, again, we've talked about this a lot on earlier episodes. If we have the data, should we use all of the data that we have? What's the right level of personalization where if we know that this person or this audience has made purchases of dog related items before, is it okay for us as the marketing team to tell the person that we know that fact? Is that the expectation? But yeah, I mean, I'll just run through a couple of thoughts that I have as I was going through this prompt format, which so as I said, we got a lot of exclamation marks, we didn't get any dog content. Maybe we want an AB campaign or an AB test of a dog related subject line or not. Maybe we want subject lines that instead of exclamation marks, maybe we want something that is a question. These are quick, easy tweaks that you can ask the AI model to adjust the output and give you even more options. You can change the max number of words, you can adjust the themes or the keywords that are in there as well. So these are some ways to just think about, even if that first round wasn't great or was overly excited, you can just use that as a jumping off point to iterate and adjust based on what you're looking for.

Laura Madden: I mean, I'm going to argue you can never be too excited about personalized outdoor gear. So first of all, how dare you? And I think exclamation point is always necessary. The last thing I'll say on this is you saying it's conversational. I can just imagine, maybe this is just me as a crazy marketer of you put these prompts in ChatGPT spits them out, and you're like, ChatGPT, you're crazy. This has nothing to do with it. And then you're like, oh. You have to hold yourself accountable and be like, well, maybe I wasn't specific enough or whatever. You just think inherently, well, yeah, of course you would do this. Of course, I wouldn't put an exclamation point after everything, or of course I would say X, Y, Z... But when you have to be a little bit more literal and very definitive in what you're doing, it forces you, like you said, to go through that exercise. So I could see some people getting in arguments with AI, but I would hope that we would win, but could be a fun exercise there for sure.

Joe Kaltenthaler: Yeah, and it doesn't always listen, because there were a couple times where I used the 7 word limit and I would see 8, 9, 10, 11, 12, like ChatGPT was just forgettingthat part of the prompt, which at the end of the day, that's why we have the prompter.

Laura Madden: Right.

Joe Kaltenthaler: Right.

Laura Madden: Rude. Maybe ChatGPT is really just a bunch of toddlers who also don't listen. Maybe I'm being triggered about my four- year- old. Anyway.

Joe Kaltenthaler: More exclamation points.

Laura Madden: More exclamation points. Okay. So that was our fun with ChatGPT. I would say that DALL- E, from the imagery standpoint, is really pretty similar. And I'm sure my creative team would come for me on some of these things. However, I know that they would agree that a big part of the design process is just getting over that brainstorming hump and figuring out what that starting point is. I think there's the blank page, insert cliche quote here, but just figuring out where to start and what to do with that, that DALL- E... and similar programs can serve a very similar function from a brainstorming standpoint that ChatGPT can. We've talked about some of the limitations with it right now, and it needs to be better defined. I know we did some tests internally at Lev that we found that maybe having it produce stock imagery of people not so great, because Joe, didn't we see some people had extra fingers or your nose might be where your ear is or something like that? But if you're just looking for stock imagery or a background image or something like that, it could be really helpful and could take some of that manual work off too. So lots of great possibilities for brainstorming and the like with imagery as well.

Joe Kaltenthaler: Yeah, unless you're looking for some Salvador Dali abstract impressionist. Yeah, like you said, Picasso ear where your nose is kind of deal. I would definitely have someone double check before you copy paste that into an email to your subscribers inaudible

Laura Madden: Yeah, it can be a mess. I know, Joe, you were trying to get super meta with this too. Did you end up having ChatGPT create a DALL- E prompt? I don't know. My brain started spinning like crazy when you were talking about AI on top of AI and how that can work.

Joe Kaltenthaler: Yeah. Yeah.

Laura Madden: inaudible

Joe Kaltenthaler: And I'll give a little teaser. That's something that we're going to share an output of this whole process end to end where you can use ChatGPT, you come up with this... In this example that we were talking about for email subject lines... So I kept adding onto this. It's just not a great format when I can't visually show you. I had this whole thing where it outputted this five or six column table where I asked ChatGPT to basically tell me why it came up with the subject line that it did, given the prompt and the user demographic that I provided. And I also had it give me a three sentence section of body copy based on the subject line that it provided as well. So it all dovetails. And then the last part, the pro-level move was asking ChatGPT, again, for the same subject line that it provided, the reasoning why it provided the subject, the body copy, and the reasoning why for all of it. I also had it generate a prompt for DALL-E to come up with relevant imagery that was related to all of that as well. So now you're having AI talk to AI like you were saying, and you're layering it. Which again, I think you need someone in the driver's seat to broker that and just double check that everything's kosher.

Laura Madden: Yeah. Teacher might become a PhD by the end of all of this.

Joe Kaltenthaler: That's right.

Laura Madden: Hopefully not. I don't think so. I think it's fine. I love how you made it show its receipts of like, no, you tell me why you said this ChatGPT.

Joe Kaltenthaler: Exactly. Which again is-

Laura Madden: You don't get to spit this out with no explanation.

Joe Kaltenthaler: ...Well, it's interesting to think about there's a reason it provided this response and it's... yeah, it's asking show me your homework, show me your work. Which again, is an additional layer of insight in the brainstorming process where now you have even more things for you, the prompter to look at to be, "Oh, I didn't think of that component." There was one of the subject lines that said something to the effect of, " Oh, we're using the fear of missing out to make someone really desire clicking into this email." Which that's one last thing that I'll kind of use to wrap up this AI related section for this email example, is if you have data on your top performing click- through rate, email subject lines, that's something that you can provide as context to an AI model like ChatGPT. So you can say, " Hey, I want an email subject line for this audience that the body copy is about this, and here are my top three or top five performing click- through rate subject lines," to essentially start to train the model specific to your business and what people are clicking on for you, and then see how ChatGPT responds and how that adjusts. So you can use... Again, data is the currency, training the model is a really important component if it hasn't been clear already. That's been the main thing we keep talking about today. But I thought that was just interesting, is testing that out.

Laura Madden: That just makes me think, I bet there's going to be a job in the future that is prompt trainer.

Joe Kaltenthaler: Oh, yeah.

Laura Madden: Do you think that could be someone's job?

Joe Kaltenthaler: Absolutely. Prompt engineer, prompt architect, throw the prompt rockstar, throw whatever moniker you want on it. But yeah, I definitely think there's going to be-

Laura Madden: Might be a thing. If nothing else, you should add it to your own LinkedIn profile. I think everybody could be a prompt engineer. Yeah, it's pretty snappy. Totally.

Joe Kaltenthaler: And don't forget, don't forget to ask ChatGPT to help you craft what you should put on your LinkedIn response as well.

Laura Madden: So meta.

Joe Kaltenthaler: Yeah.

Laura Madden: So meta. Oh, man. Well, okay. Spent a lot of time on ChatGPT and DALL-E as kind of the new cool things, but we don't want to forget about out of the box baseline AI functionality that you might already have, especially if you are in the Salesforce, specifically Salesforce Marketing Cloud ecosystem that we want to make sure that everybody is taking advantage of. Right? These are some of those things that you might not necessarily think of them as AI because they're not as super cool and sexy as some of this other stuff, but it's still that core definition of data in, output out and it's being trained. So The one that is nearest and dearest to me is Einstein within Salesforce Marketing Cloud. Specifically send time optimization or STO for short. So this I think is so powerful because it really answers the age- old question that so many email marketers have been faced with of when is the best time to send my email? I think that probably gives every marketer a little bit of a twitch because you've gone through all the tests of like, "Oh, Monday is the best day to send, unless it's the second Monday, and unless it's a deal, a Black Friday deal, or unless it's this, unless you're in this persona, then it's on Wednesdays. But it's always in the morning, except when it's in the afternoon." It's just, it's too much. And all of the testing that you have to go through. STO really takes all of that onus off of you, and it learns... Some of it's behind a black box, but it's really looking at what people are doing when they're opening, when they're engaging. And it will assign a personalized send time to each subscriber, and even if they don't have a personalized send time yet. So maybe I'm lapsed, I'm not doing a whole heck of a lot in email, so it doesn't have enough information to create that for me. It will use that wisdom of the crowds as the default, even. So you don't even as the marketer have to pick the default time. Einstein will pick that for you too, based on how everybody is performing. So to me, that's a really great baseline AI tool for you to use if you're using Salesforce Marketing Cloud, whether it's in your journeys and using it for wait steps and saying, "Well, they just need to receive the second email sometime within the next seven days. Einstein, you tell me when." Or if it's in your promotional emails and you say, "I want to make sure they receive it in the morning so that they have time to shop, but let me put a four hour STO window on this and just let it run." So I think that those are some great options that are, again, pretty out of the box that has a lot of AI power behind it from a marketing cloud perspective. And you guys know that I'm the email person, so of course I'm going to talk about email and Joe I think is going to walk us through some web things.

Joe Kaltenthaler: Yeah, I think there's a lot of products out there today already on the market for web personalization and RTIM, realtime interaction management. But yeah, I'll quickly make a plug for Marketing Cloud Personalization, formerly Interaction Studio. So similar to how send time optimization looks at a huge bucket, like captures a ton of data, looks at all the data, and then makes a decision as a result of what the data is telling the model to do. Interaction Studio or Marketing Cloud Personalization operates the same way, where we set up a data model specific to the client's business and are capturing these behavioral triggers for the products that that client has on their website. So for example, viewing, or adding to cart, purchasing, favoriting, things like that, we can capture all of those behaviors and then start to build relevant product recommendations as a result of different algorithms that we can tell the model to output. So for this email, maybe I want recommendations that are a result of things that they have purchased in the last 30 days that are not currently in their cart. That is a rule that can be set up and told to the model and have relevant recommendations output. Maybe we want a web scenario where on a product page we have items in the same category, but not the current item that's being displayed. That's another example of how we build the model, how we train the model, can all come together to then provide our desired output for what is the right set of recommendations in this particular context for this user at this current moment in time. And again, that is not science fiction, that's science fact. This can be done today, which is pretty crazy.

Laura Madden: That is pretty crazy. And I think the possibilities are endless when you get into MCP, Marketing Cloud Personalization. It's scary but exciting all at the same time. So hopefully everybody's taking advantage of that. Oh boy. Deep breath, because we covered a lot today when it comes to AI, artificial intelligence. So we hope you've learned a lot, even if it's just that Joe hoodwinked me into making me think that he did all of that himself. Just kidding, Joe. But The high level takeaways that we have today is first off, don't just let AI run amok. Don't just... This is definitely not a set it and forget it situation. You don't just turn it on and let it run. You definitely want to make sure you have that human check, the human QA. Be your own prompt engineer, add it to your LinkedIn profile and be prepared to maybe lose some of that efficiency at the beginning and that's okay. Just level set your expectations for what you're going to get out of leveraging this AI, especially if it's in a brainstorm scenario like a ChatGPT or DALL- E. It might actually take you a little bit longer at the start once you're refining all those prompts and things like that. So just be ready for it and try to balance playing around without getting stuck at rabbit holes. Joe, I don't know if you're a cautionary tale for that or if you felt like you did a good job. Probably a little bit of both, but-

Joe Kaltenthaler: Yeah, definitely both but-

Laura Madden: can be-

Joe Kaltenthaler: ...I would say, yeah, give yourself enough time to enjoy learning. This is a skill. Prompting... It sounds crazy, but prompting the platform is a skill. It's something you learn to get better at. I am not at all espousing to be a PhD prompt engineer at this point, but just from using it more, seeing outputs, seeing that looks a little weird. The more you do it, the better you get.

Laura Madden: Yeah, for sure. And with that too, I think if you're listening to this and you're like, "Ooh, I am not ready to play around with ChatGPT or DALL- E. I don't have the bandwidth, we're just not ready for it." Whatever. That's okay. It's okay to start small with some out- of- the- box solutions, whatever provider that you're using in email or web, there's probably a little bit of something out there that's already built in, like the STO or product recommendations that we talked about in MCP. So that's AI too. Right? So that's okay. You have to start somewhere on that marketing maturity curve. And with that, just stay curious and keep learning and growing, because I'm sure if we were to do this podcast again and probably a month, we'd be talking about something different with AI, so you're not going to get left behind because AI is leaving itself behind, probably. It's outpacing itself. So that's okay. Just play around with it. Any parting words on that, Joe?

Joe Kaltenthaler: I mean, I think curiosity is probably your best friend in this scenario where, again, as I'm going through, I'm thinking... Through the exercise, I'm thinking, "Okay, how would I use this?" Right? Because even the format that it spits out initially is bullet points. And then the more I think about it, I'm like, "Okay, well actually I want to keep adding on to this. I wonder if..." Right? And that's kind of how I kept going through the process personally through this was, I wonder if I ask you to do this, what'll what'll happen? So keeping that, that kind of childlike curiosity, and I mean, I don't know... I'm just a tinkerer. I love to try and break things. So the fact that it kept spitting out 8 or 10 word responses, even though I kept saying 7 very specifically was a little bit maddening, but that's okay. So yeah, I think approaching all of this with a healthy dose of skepticism, being prepared to say, "Okay, well that was a little interesting, but I don't know if I can use any of this." I think that's okay too. Just have metered expectations as you're going into this type of thing where though it is getting really, really popular and is in the forefront of society at the moment, it is still... I mean, the more it learns, the better it gets, and the more prompts you put in, the more the greater model is built and continues to grow. But it's still... I mean, we're only a few months into this being open source and out there and usable at the scale that it's at. So again, it's exciting. Yes. But healthy expectations are always good.

Laura Madden: Well, thank you guys so much for joining us for this episode of Level Up, looking to continue to level up your knowledge on the latest news technology and marketing trends affecting marketers day to day? Stay tuned for future episodes of Level Up with new episodes coming up every other Thursday on Spotify and Apple Podcast. Until next time, thank you for leveling up your marketing knowledge with us.


If you’ve been online at any time this past year, you’ve most likely heard of how revolutionary many find generative artificial intelligence (AI) to be. Being able to generate written works and images in distinct styles, it’s easy to see how it can dramatically improve marketers’ productivity. But is AI something you can truly “set and forget?” 

In this Level Up episode, our hosts, Joe Kaltenthaler (Sr. Manager, Data/Identity Practice) and Laura Madden (Sr. Manager, Marketing Strategy & Services) discuss the current benefits and limitations of generative AI and how marketers can become better AI prompters to help automate the brainstorming process, strengthen the relevancy of their messaging, and personalize the customer experience.  

Today's Guests

Guest Thumbnail

Laura Madden

|Sr. Manager, Marketing Strategy & Services
Guest Thumbnail

Joe Kaltenthaler

|Senior Manager of Lev's Data/Identity Practice