How to Prepare Your Data Foundation for Artificial Intelligence (AI)

Media Thumbnail
  • 0.5
  • 1
  • 1.25
  • 1.5
  • 1.75
  • 2
This is a podcast episode titled, How to Prepare Your Data Foundation for Artificial Intelligence (AI). The summary for this episode is: <p>In this Level Up episode, Jordan Kraus and special guests Brian Esham, Director of Technical Services, and Christy Glesing, Sr. Marketing Consultant, explore how AI has evolved throughout 2023 and its future for business applications. Together they discuss topics like building consent and trust into AI models, the steps to prepare your data foundation for successful AI integration, and how you can ensure its marketing use cases align with your company's goals.</p><p><br></p><p><strong>Listen in for:</strong></p><ul><li>00:45&nbsp;-&nbsp;03:55 📈 The evolution of AI in 2023 and continuing to 2024</li><li>03:51&nbsp;-&nbsp;06:07 🤝 Data and building AI models, ensuring consent and privacy in the trust layer, policies</li><li>06:09&nbsp;-&nbsp;08:26 🔧 Preparing your data foundation</li><li>08:30&nbsp;-&nbsp;10:52 🔄 Understanding the customer lifecycle and the data gaps, narrowing focus</li><li>10:57&nbsp;-&nbsp;14:02 🤖 Traditional AI, Machine Learning, and Generative AI: is your data set up for these technologies?</li><li>14:10&nbsp;-&nbsp;17:05 📀 Data and breaking down silos that ultimately benefit customer experiences</li><li>17:06&nbsp;-&nbsp;18:26 🤦🏽‍♂️ "Do I really need to go through all this work for AI?"</li><li>18:22&nbsp;-&nbsp;20:59 ⏰ Do you need to have a majority of a data foundation before pursuing AI?</li><li>20:59&nbsp;-&nbsp;23:34 🙋 If you don't know where to start, consider bringing in a partner to consider and build an AI integration strategy</li><li>23:40&nbsp;-&nbsp;25:50 🏆 Small wins that lead to big wins with data cleanup and adopting AI</li><li>25:50&nbsp;-&nbsp;26:39 ✅ Use cases that align with company goals</li><li>26:39&nbsp;-&nbsp;29:36 What are the possibilities with AI once a data foundation has been created?</li><li>29:37&nbsp;-&nbsp;33:11 Salesforce announces free Data Cloud and Tableau trials. Considering other Salesforce tools and Einstein</li></ul>
📈 The evolution of AI in 2023 and continuing to 2024: navigating customization and data creepiness
03:09 MIN
🤝 Data and building AI models, ensuring consent and privacy in the trust layer, policies
02:12 MIN
🔧 Preparing your data foundation
02:18 MIN
🔄 Understanding the customer lifecycle and the data gaps, narrowing focus
02:21 MIN
🤖 Traditional AI, Machine Learning, and Generative AI: is your data set up for these technologies?
03:06 MIN
📀 Data and breaking down silos that ultimately benefit customer experiences
02:55 MIN
🤦🏽‍♂️ "Do I really need to go through all this work for AI?"
01:19 MIN
⏰ Do you need to have a majority of a data foundation before pursuing AI?
02:33 MIN
🙋 If you don't know where to start, consider bringing in a partner to consider and build an AI integration strategy
02:34 MIN
🏆 Small wins that lead to big wins with data cleanup and adopting AI
02:08 MIN
✅ Use cases that align with company goals
00:48 MIN
☀️ What are the possibilities with AI once a data foundation has been created?
02:57 MIN
💰 Salesforce announces free Data Cloud and Tableau trials. Considering other Salesforce tools and Einstein
03:34 MIN

Jordan Krause: Thank you for joining us for this episode of Level Up, the podcast for marketers by marketers created by love that distills best practices and strategies focused on helping marketers increase their experience, one of their strategy, and grow personally and professionally. I'm your host, Jordan Kraus, client success provider, and I am partnered here with a couple of special guests from our Lev consultancy. Joining us is Christy Glesing, our senior marketing consultant, and Brian Esham, our director of technical services.

Christy Glesing: Hello.

Jordan Krause: Hello, hello.

Brian Eshram: Thanks for having us.

Jordan Krause: Thank you for coming. So today our topic is something I feel like we have been inundated with for the last year, which is how to prepare your data foundation for AI. And just to start the conversation, one of the things that Christy, Brian and I were talking about was to talk about how specifically this year in 2023, how much AI has continued to evolve throughout the past year. I think we've heard a lot of buzzwords, but now even, it's hard to say this, but it's almost Q4 and looking back over the last nine months of this year, what are your observations of the way that AI has continued to evolve and change?

Brian Eshram: It's a great question. Yeah, I think so much has changed, especially if we're thinking about in the world of generative AI. I mean, back in what, January, February, I think most of us had never even heard that term, didn't even know what it was. And now we're at the point where it's in a lot of our daily language of, "Oh, yeah, let me just put that into chat GPT, and we'll see what comes out. So I don't know, Christy, what you think, but I think it's like a world of difference in the last six to nine months."

Christy Glesing: Yeah, I would agree. I mean, it seems like it just came out in the conversation and everyone's trying to figure it out and organizations are launching their own GPTs and people are getting really good at prompting questions, but it seems like a brand new playground to be on for sure.

Jordan Krause: I feel like in 2022, there was so much conversation that we were having, especially as a consultancy about personalization, creating an improved experience, not a creepy experience. Even preparing for this podcast, thinking about new ways that I see AI being used. I mentioned that I saw a British Airways advertisement on a billboard, and it was an electric billboard next to an airport, I believe next to Heathrow, where whatever airplane that had taken off and flown over the billboard was casted on the billboard and creative that said, "BA flight one, two, three, four departing to Singapore now." And at first I thought like, "Wow, that's really fascinating that they were able to put all that tracking data together to be timely to show the time of that airplane taking off." But then I had a second thought, "If I'm on that airplane, would I want them to know where I was headed? I feel like our conversation about is it creepier, is it cool, has evolved from that to is my privacy protected? Am I secure?" And so in some ways we're still having that conversation, but now I feel like in the topic of AI, privacy is so much more critical.

Brian Eshram: Yeah, I think anything related to personalization, AI, I think consent, creepiness, the creepy factor that always plays into, and I think there's always a fine line there. And I think that maybe we get into this a little bit more as we talk about data, but having a lot of data is helpful to build out these AI models. They're based on data. We need to make sure that it's done in a consent and privacy- friendly way. I think that that's important going forward. I think the industry has seen that too.

Christy Glesing: Yeah, and I think in this case, and in many cases with technology, the laws have to catch up with the technology, but I got to go to connections earlier this year and then with Dreamforce here recently with Salesforce, and they couldn't talk too much about their trust layer and how they are really taking that into account. So they're putting the protections in place on your data, but that doesn't mean that you don't have to. It's to protect your own data and keep that in mind of how you're going to gather and share and use it to make sure that you do honor people's privacy.

Brian Eshram: Yeah. That also goes back to just the original question, Jordan, you asked about how have things changed in the past year. Again, we didn't really have a lot of policies. Most companies didn't have a policy on what data can you use for AI or what can you use, especially when it comes to generative AI, how does that fit in? So I think, again, we've seen that as an evolution. I think we're also seeing, again, from a tooling perspective, so go back to March, April when a lot of these tools started becoming, I guess, center stage, it was really just the raw technology. Take a prompt, generate some content, whether that's an image or text or whatever. But now we're seeing tools from Salesforce and others where they are putting those wrappers around it, making it a lot more friendly and easy to use in the context of a particular business as opposed to just a general purpose technology.

Jordan Krause: I think that is a really nice segway into the very first topic that we're talking about today with preparing your data foundation, which is to look at your data first. You must have quality data to train those AI models on, and that's also where you start to talk about how are we protecting our data? But the thing that we hear all the time is marketers are struggling to connect disparate data sources. The concept of a CDP has been a very hot topic, but it certainly is not anything new as a concept. It's something that marketers have always been trying to do, whether it's from an analytics perspective to a personalization perspective and targeting perspective. Data is the one thing that is going to power all of those recommendations and your AI, so it's certainly the place where you need to start. And that also is really intimidating as a marketer as far as where do I start if I have all of our disparate data sources? Where would a marketer even begin as far as looking at their data foundation so that they can start to use AI?

Brian Eshram: Yeah, good question. I'm going to answer I guess as the technology person, and yeah, Christy, I'd love to know what you think because you're the real marketer here. So yeah, from a data foundation perspective, right, agree with everything you said, you got to have clean data. It needs to be harmonized in a way that's meaningful. Ideally, you scrub that data to remove anything that might have an implicit or undesired bias, especially when you get to things like equality and gender and all of that. I think there's a lot of natural just bias and data. So getting that data to a clean spot is really, really important. And again, that's where things like, sorry, not to go back to the tools I suppose, but Salesforce data cloud or any other type of customer data platform where you can start to pull that data together and have that picture of who a person is. Then you can really start to build, I think, more accurate models and then apply those models to those people. Christy, what do you think from a marketer perspective?

Christy Glesing: Yeah, I mean, I think in order to know what data you need, you really need to lay out your whole customer life cycle. What are you doing throughout that life cycle? How are you communicating? What is each customer's experience? And then you can begin to dissect like, okay, I can personalize this, or I can segment this group and put this messaging in front of them and move them through the cycle naturally and really at their own pace. So I would look at the customer life cycle and see what data you need and where are the data gaps, and then how am I going to fill them? How am I going to get that data? And then with a tool like a CDP, you can bring it all together and see the data for yourself and know, okay, now I can see the gaps and start to go back and fill them. So it's iterative, but I would always start with the customer experience, I guess, the whole lifecycle.

Brian Eshram: Yep. Yeah, I completely agree with all of that. So I think, yeah, that's where having that marketer or business viewpoint on it is just really, really helpful. Again, I think that the natural tendency with data and AI is to just say, "Hey, we've got a data lake out there. It's got a ton of data in it. Let's just dump that all into some AI model, and then we spit it out, let's see what the results are, and you're going to get results." The question is, "yeah, are they accurate? Are they good? Are they really predictive?" So starting with something cleaner and more put together like a CDP rather than just a raw data lake, I think is going to give you a lot better results.

Christy Glesing: And I think going through the practice of looking through all of your data helps you narrow your focus because we see people trying to boil the ocean all of the time. We've got data all over the place, there's so much data, but let's just focus on the pieces that we need specifically. Let's get started and then we can grow into it later, but I think that's another gotcha that we see happen a lot, and it's no wonder because there's just so much data out there available.

Brian Eshram: Yep, 100% agreed. I also wonder too, In terms of historically when we're talking about AI, we were talking more about, I'm going to use the word traditional, I guess traditional machine learning. So things like creating a propensity score or clusters of individuals, I feel like the generative AI has really taken center stage and shadowed poor little traditional AI. But I think what generative AI has brought to us too is it's not just about raw data points. Now we have to start thinking about, well, how do we organize all of our content? So do we have that foundation in place? So maybe we've got an asset management system out there with all of our images, but are they appropriately tagged? Do we have anything to even load into those AI models to understand marketing images and content, and then be able to turn that into meaningful results? So I think there's potentially a lot of work to be done from a data foundation perspective on the non- text content.

Jordan Krause: Yeah, it's not sounding as fun as it was in the beginning of the podcast. Now that we-

Christy Glesing: There's a lot of homework to do. Yeah.

Brian Eshram: Turns out it's not magic. Yeah.

Christy Glesing: But truly, because the AI technology is so new and companies are still figuring it out, now is the time to focus on your data. Get that in order so that when you're ready to use that technology, the data's going to be there for you. And these are simple things like do you have an enterprise data dictionary? Do you have naming conventions? Are you using them? Do you enforce pick lists? Or do you not allow null values when you're entering data? Do you have an owner for each data source who is ultimately responsible for its quality and maintenance? I mean, these are a little checklist of things that you can be doing now to prepare, to make sure that when you do lean into that data, it's going to be there and it's going to be usable.

Brian Eshram: I think we've probably all done the fun traditional corporate training where you talk about data privacy and security and data stewardship, and I've been seeing some of those in the past where it's like a knight or some other cartoonish character where they try to make it fun, but that's exactly what we need here. This is where you really need data stewards that own that data. So just to your point, Christy, enforcing that it's clean, that everything's properly tagged, an owner. So I was laughing about those trainings, but it turns out actually they're probably beneficial right now. We need that knight.

Christy Glesing: I think we will knight you, Sir Esham. You can be our data master, I guess. Well, I think the other thing that can come out, The positive thing that could come out of preparing for your data for AI and could maybe make it fun for you, Jordan, is really we can finally maybe break down those silos because data is the great equalizer here. Everybody in the organization needs it. Everybody realizes its value and everyone can contribute to the quality of the data, and you're going to have to work together to get it there. So there is some good opportunity there if nothing else, to work across teams and work together. You may have a data collection team, you may have that person who oversees it, that owner, but then the marketer might be just the end user of it, but you all need to work together to make sure that you've got the data you need and that it's in good quality. So there's hope there, break those silos down eventually.

Jordan Krause: I think, Christy, you always do a really nice job of bringing it back to goals, bringing it back to the customer experience. And a lot of time too, and we are guilty of this as well, being in a consultancy, is we can explore technology for technology's sake. We just want to see what it can do and see what's out there. So I think Christy's advice is so sound in saying that data is the great equalizer. It's the thing that everybody in your company needs to be successful. So if you can paint the story of how it will improve your customer's experience, then you can create some buy- in to make those updates to the data. And so maybe it's not as boring as like, Hey, we have to go do all of this really important data practice work and eat our Wheaties to prepare for the use cases and the customer story. You can paint those stories in advance of the work and create your collection data strategy based on the customer experience that you want to create.

Christy Glesing: And it can become a challenge of change management too. Because you're going to be working across teams, it's going to be important to show the value. Why do we need to do this? And it really does take a champion to keep teams informed and always be showing them the value and the why and moving the projects along, because these types of projects are the first ones to stagnate. They just fall away. People forget about them, unfortunately. But if you can have a good change management plan to support all these new data practices that you're putting in place, then you're going to be able to get it, get the data and use it for these newer and better technologies.

Brian Eshram: I, again, completely agree. I think one question I could see people asking as well here is, "Do I really need to go through all this work just for AI, or am I going to do all this work and get my data clean and then great, I can run AI on that? What are the other benefits?" And I think, again, this isn't really just for AI. I mean, cleaning up your data, getting things into that proper foundation, building that foundation is really the goal for any, I guess, future, I'd say initiatives, for lack of a better word. So whether it's from a personalization perspective, a segmentation perspective, analytics, I mean, it's not just to power AI. So just to bring it back to a larger picture, having data clean is beneficial in a lot of different areas.

Christy Glesing: I mean, it's really one of the company's best assets, and you think about your staff and your people, but it's also your data as well. So you would want to take care of it and invest in it. And Brian, for all those reasons you just mentioned, we're talking about AI, but it powers personalization, segmentation and analytics, all the things in between too. So yeah.

Jordan Krause: Do you think that marketers need to have the majority of their data foundation in place before pursuing AI?

Brian Eshram: That's a good question, depends on what you mean by the majority. I mean, more is always better, I suppose, but no. Do you have to have every single piece of data? No. I think if you start with the core of it, so focus on who are our customers first? What data do we have on them? So what is the root of it, some of that base level data? And then start adding on from there. So do we need to bring in our e- commerce data and know what order information and things like that? Maybe app engagement, web engagement, all that, layering those on one at a time, I think that's a very valid strategy. So I think, Christy, you used the point earlier of you don't have to boil the ocean, and so I would say no, you don't need to bring that all in at once.

Christy Glesing: And that's something that we help our teams do is prioritize. When we're talking about strategy, we lay out what all of the use cases are, but then we narrow that down, we prioritize them, and we try to get a mix of quick wins so you can see the results, but also the things that are going to drive the biggest results, give you impact in the longterm. So we help our clients prioritize the data. Anyway, going back to there is so much of it and throw in all of the different technologies. I mean, we're talking a little bit about CDP and AI, but all of the marketing automation tools that organizations use. I don't know, what is the average now? We're up to 30 different marketing platforms for any organization. So really focusing in on the right ones to get you up and running, and then before long you know what to do and you can build out your data practice.

Brian Eshram: Yeah. Yeah, that priority is definitely important. Otherwise, yeah, if you try to bring everything in and clean everything at once, I mean, it'll be three years until you're ready for that. So yeah, I think the priorities and quick wins, those are a great way to focus.

Jordan Krause: I think Christy has a really good point too. If you don't know where to start, maybe bring in a partner. Maybe there is somebody else who can help you prioritize that list; at the very least, can help bring the teams together within each data source owner to have a conversation about what is our customer- centric strategy? How do we all contribute to this plan? I think that's really what's challenging in marketing is that everyone is incented so differently in their day- to- day roles that it's hard to see anybody in the center owning an initiative like data foundation or customer experience. So in those situations where you're feeling like maybe you need help with change management, maybe you need help with looking at your data, prioritizing, maybe you just need another voice to start those conversations, to be on track so that you can have an enterprise data dictionary or some sort of data collection strategy that you own longterm, look to a partner. Look to another partner who can be an influential third party in your business.

Brian Eshram: Yep, very fair point. I mean, these are not simple things to solve in terms of data quality, getting into the details of the data and trying to figure out how do we structure that from a foundational perspective. So yeah, bringing in an outside perspective and just that experience, I think that goes a long way. Christy, what do you think? Yeah.

Christy Glesing: Well, I was going to say, I have learned in being on projects with you, Brian and Jordan, the importance of laying out those initial MVP use cases. What do we need to do? But also setting up whether it's... whatever the platform is, let's just say it's a CDP, building it with a framework in mind, building it for the future. We're going to get these first use cases built and done, and we're going to prioritize them, but we want the structure, the data architecture and the data model, the structure of it all outlined and ready to go so that you're ready to move to the next one once you've gotten those first ones launched. So it's a balance. You've got to think for the now, but also build for the future, and having that strategic plan in place helps you do that.

Brian Eshram: Yeah, there's definitely a push and pull there for sure, and a balance. I absolutely agree with that, trying to balance those. What can we do in the short term that's not going to require boiling the ocean, but make sure that we're not essentially creating technical debt? So how do we make sure we've got the whole plan? We know what the long- term strategy is going to look like. We know what the model's going to be, but then which pieces of it can we tackle? So again, back to those use cases you have to prioritize? I think that's awesome.

Jordan Krause: I have to say too psychologically, we just need those small wins throughout a project to keep our motivation going. I think it's really funny because in a silo, these problems seem so simple, and then when you get out into the real world and you have to deal with people and processes, then it's really interesting how those problems evolve. So the other part of me is like you need a strategy as well, just so that everybody has something that they're pointed towards, everybody's rowing in the same direction. There are small wins along the way leading up to those really big wins, because the idea of having to create a full data foundation for every single use case is daunting, and it doesn't have to be. I really like what you're both saying about starting somewhere, making the plan and being able to make progress while the plan evolves.

Brian Eshram: Yeah, for sure. I think as a consultant, again, I've had numerous conversations before about data cleanup, and usually whenever you get to that conversation of cleaning up the data in the source or where it lives today, it's usually, oh, that's going to be way too much work. There's no way, all the stakeholders that are involved and so forth. But yeah, I think that's where if you bring in a partner to help you with that and prioritize, and if we're talking about putting that into somewhere new like a CDP, then I think that's a little bit more appealing. So you can try to take that on. You're not talking about, okay, we got to re- engineer everything. Let's just clean it up as we're putting it into something more modern.

Christy Glesing: Well, and I think another thing that can help motivate the teams that are going to be responsible for this data is making sure that your use cases align to your company goals, whether that's the overall corporate or just your marketing goals, because that way when you go to make the ask like, oh my gosh, you need to clean up this data, you can say, because it's going to help us do this, reach more prospects or convert more prospects. And so I think tying everything to the goals, even to data collection to your goal will help keep the team going and aligned to where you're headed.

Jordan Krause: Well, since you have laid out all of the fun data work that gets to happen, let me just ask for those of us who are looking for something a little more sexy, what is the light at the end of the tunnel? What are the AI possibilities? Once I do all of the work of creating a data foundation, what is the goal? What's the finish line look like for what I could use AI for in the world of marketing? What kind of possibilities does generative AI open for marketers?

Brian Eshram: Christine, I'll let you go first actually as the marketer, because again, I've got some cool things, but yeah, you got the marketing lens here.

Christy Glesing: Yeah, I mean, right off the bat, there's content creation, so copywriting, image selection. I think efficiency is a big one because our marketing teams are so tapped and we just don't have enough time in the day. And so anything that we can do to make these processes easier, smoother, faster is going to hopefully free up time to be more strategic, to focus on the goals and the outcomes and not necessarily how the sausage is made and making the sausage. So those are two of the things that I'm most excited about, but also really finally tapping into all that data. You've spent all this time collecting it, structuring it, moving it around, but with AI, now you've got the engine that can interpret it and give you some good and new ideas and approaches. So that to me is very exciting.

Brian Eshram: Yeah, I love the idea of the cost savings, the efficiency angle. I, a lot of times, will tell people, at least in current state right now, we're talking about generative AI, it's not a better, faster, cheaper thing. It's more of an equivalent, faster, cheaper, so it is a cost savings and efficiency play. I knew for me, there's a couple of things that I've talked to clients around that I am excited about when we get to that point, but thinking through product placement. So rather than spending a huge amount of the budget, taking your product out, putting it in different placements and all those photo shoots and all that, if you can have just that stock image ready to go of that product, and then you just generate the background around it essentially. And whether it's a pair of shoes on a hiking trail or on somebody actually running, is the technology there today? Not quite, but it's getting closer. So that's one use case I could see being very popular from a retail perspective.

Christy Glesing: So I would just add, if there was anything else that was really important to say about getting your data ready for AI, Salesforce just announced at Dreamforce that you can get free Data Cloud and Tableau trials, and how cool is that? You can get in there and play around with it, but I would just say just have that strategic mind in place before you just jump right in. What are you trying to accomplish? What are your goals? What are you going to do with these new fancy tools so that you can plan and you're not going to just be spinning your wheels? Or after the trial, you're ready to go. You have a good idea of what you need to do from there. But there were lots of things, lots of exciting things coming from Salesforce with regard to Data Cloud and Einstein, all the AI tools.

Brian Eshram: Yeah, yeah, for sure. Yeah, the free data cloud for sure is great. I think that gets you in there to instead start to maybe play around with it and then you can start to see, this is how I could build my data foundation. And so if you get a sense for that and start to put together your strategy there, then that's going to allow you to use some of the other new tools that are coming to Salesforce demo there. So one of them we saw was Prompt Studio. So I know Prompt Engineering has come up also on the list of things that are new this year, prompt engineer as an actual role. So knowing how do we use these AI tools? How do we write these prompts correctly? Salesforce is really trying to help there with giving you a playground and a foundation for that to say, let's come in here into Prompt Studio. We'll build out our prompts. We'll see what those models are going to spit out as a result, and then be able to really refine those. So I think, again, it's taking that raw technology and then putting it into the hands of the marketer or the business user to be able to rely on that based on that strong data foundation.

Christy Glesing: One of the AI prompts they were talking about was segment creation, where you can just even verbally describe what the segment you want to build and then it'll crank it out for you, so that's a time saver. Plus it might feed you some information that you hadn't even thought of yet. Einstein lookalikes sounds pretty cool where it can just broaden your highest performing segments. So you start with your high performers, and then you can go a little bit further to catch a few more people. So lots of fun stuff coming. Tableau Pulse was another one where you can measure your segment's growth over time or their performance over time. So we're eager to see that one in action, lots of things coming.

Brian Eshram: Yeah, there's is a customizable or configurable propensity score as well. So again, it's outside of the generative AI, but more of the traditional AI. I mean, I think that's going to be great too. Again, just even doing basic propensity is really, really helpful, but it's hard if you're not a data scientist. So putting that in the hands of marketers I think is going to be really cool. So yeah, a lot of cool AI tools coming based on this Data Cloud Foundation.

Christy Glesing: Thank you 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? New episodes of Level Up come out every other week on Spotify and Apple Podcasts. Until next time, thank you for leveling up your marketing knowledge with us.


In this Level Up episode, Jordan Kraus and special guests Brian Esham, Director of Technical Services, and Christy Glesing, Sr. Marketing Consultant, explore how AI has evolved throughout 2023 and its future for business applications. Together they discuss topics like building consent and trust into AI models, the steps to prepare your data foundation for successful AI integration, and how you can ensure its marketing use cases align with your company's goals.