[Music] [Pierre Charchaflian] Okay. Hello, everyone. I realize this is the session right after lunch. We're going to try to keep it high energy. Right? Not sure we'll keep it interactive, but if you have a question, please go ahead and ask a question. My name is Pierre Charchaflian, and you probably heard all conference long, the words thrown out, purpose, experiences, purpose built. You've heard gen AI. You've heard orchestrating experiences. We're going to combine all these things together to really talk about what's next in the evolution of customer experiences. So, we'll talk about where experiences came from, where they're heading and what you can do to really adapt and adapt to this new mandate. I am lucky to have an honor to have with me today, and they'll be coming on stage to talk about their perspective experiences in delivering purpose-driven and future-driven experiences, Wealthy Desai, and Wealthy is Head of eCommerce, B2B and B2C, Go-To-Market for AT&T. He is all about experiences. He's been living it, breathing it every single day. So, he'll give us the B2C perspective mostly in his examples. We also have Sam Yeung from IBM marketing. If not the largest, one of the largest B2B marketers in the world.
Sam is responsible for global, web performance and personalization. She'll give us the B2B perspective. Also joined with, Thomas Schäck. And Thomas is a distinguished engineer in our watsonx organization, Watson, the OG of AI. Right? He's got tons of years of experience in building AI models and generative AI models. He'll share with us what our generative AI companies, providers of this technology, are doing to increase trust. So I was told the best way to start a presentation is telling a story. So do you guys know where this is? Costa Rica. Right, 2022, coming out of COVID, we're all been cooped up, not going anywhere. We decided to take a family trip, and I love to travel, and Costa Rica is not a typical Caribbean destination where you go plop yourself on the beach, drink, have dinner and go back to the room, do that like seven days in a row. It's a very active diverse geography, Atlantic Coast, Pacific Coast rain forest. And I went to my typical hotel destinations, my airline destinations, online, recently. Everybody had good content but I just didn't know how to curate a trip for myself. Then a friend of mine says, "Hey, have you heard of Costa Rican vacations?" I'm like, oh, that sounds relevant. So I went to their website. I talked to a person for 30 seconds. They said, "Yeah, fill out a questionnaire." Time that I wanted to go, who's going, the ages of who's going, what they like to do, what we want to do, how we want to spend our time. And literally spent about a couple of minutes filling the survey out, and I got three different itineraries of activities, of different options for accommodations, hotel versus a home, airlines and so on and so forth. And after a brief conversation, we ended up with a great itinerary that not only delivered on our experience but delivered on our purpose for the trip. We all had fun. We all did something everybody enjoyed. And I started thinking why can't every brand do that. Now I know it's an agent example. We're all used to that. Why can't every brand deliver not on the product but on the purpose for why we engage with the product? And I think this is a very relevant topic because two important points. One, the technology is finally here. The possibility for us to curate purpose-driven experiences on a one-to-one level, not content on a one-to-one level. Curated experience. Right? Anil yesterday talked about orchestrated experiences. And I don't think that's enough. I think we need to curate the experience and go outside of Adobe, connect Adobe to our supply chain, connect Adobe to our different systems to really curate that experience. The reason why this is very important is if we don't do it today, somebody else is going to do it. And I think there's a whole new level of disruption of disruptors that will come into the industries and really create new kinds of brands that really can collate together different products and offerings that deliver on your purpose. So, let's talk a little bit about why this is important right now. I don't want to throw stats at you guys, but I think we're all here because customer experience is really important. Right? There's nothing more important. There's nothing that's more competitive and crucial to the survival and to winning in corporate world today for our brands than experiences. And we've had tremendous innovation in our supply chain and the way we deliver experiences to our clients and deliver services. I mean, to pick a retail example, 20 years ago, you couldn't return a product you bought online. Today you can buy a product online, pick it up in the store, buy, have it delivered to you in two hours. You can look at all your purchases and see how much money you save. I mean, we've done incredible, incredible amount of transformation and innovation, and that is allowing us to really deliver complicated and personalized experiences. Two, we have a whole new generation. My kids, some of your kids, that are digitally native. Right? That's all they know. The iPad generation.
They don't even call it frictionless omnichannel experiences. It is the experience that they want. They're much more willing to share their data. But they expect relevance with that data and they expect to trust with that data. So how do we deliver that to them? And betraying trust to them is, if they give you a piece of data and they don't act on it and you don't act on it. So another really key indicator in how we evolve in the customer experience. Number three, we're always talking about data. The truth is there's going to be more data tomorrow than we have today. And the question is how are we using that data? But most importantly is where does the consumer fit within that data process? If that customer has given us personalized information, how are we protecting it? How are we protecting the privacy of that customer? How are we dealing with that consumer privacy and all the regulation that, trust me, is coming. Okay? You're going to be more limited in what you can do with that data than ever before. So, building that trust and that contract with that customer is going to be really important. And obviously, security, I mean, how many spam calls do we get every day? How many spam emails do we get every day? It's incredible. And lastly, generative AI. There's this technology that understands the context of a customer, that understand the meaning what I'm trying to say, what I'm trying to communicate that can automate content generation, not just for one version of copy but in really creating curated experiences and assembling experiences that mean something to each one of us. So I can tell you right now the true business impact of generative AI is not in creating content, it's in really creating experiences, and I'll give you an example. Let's say you're a $10 billion brand and let's say you spend on marketing $300 million, that's a fairly accurate metric, 30%, 40%. And let's say you spend 10% of that on content generation, 30 million. And let's say you cut 50% out of that through generative AI. That's 15 million in saving. What if I told you I can get you 5% more revenue with better experiences? That's half a billion-dollar incremental revenue. So again, we're heading down that. We don't have the operating model, we don't have the practices to do it, but ultimately, the true business impact is not in content generation in one copy but in creating many versions of experiences. So to put the next stage of customer experience in context, I thought I would share where we're coming from and where we're going. So this is a little bit of a maturity model, and for those of you, and I was there, who were around in the early 2000s, the birth of the initial of the digital experience. The focus was on the what. It's the product, it's the transaction. And the data was about first party transactional behavior data. It's about the click through, what you clicked on the email, what pages you-- And we still measure that today. It was all about, let's get the MarTech stack to talk to the Commerce stack, to talk to the email, so we know we closed the loop, and yes, I sent someone an email and we got a response from that email. And they clicked and they didn't buy and so on and so forth. And everything was batched and push. Then we moved to the next phase where we talked about omnichannel journeys, and omnichannel, now I'm measuring and I'm looking at not only the transaction but how you buy the product, how do you service the product, how do you track all the delivery of that product, look at all the things that I bought. Whether that's financial services, your bank, how you manage your credit card and installing the card and locking the card and asking for credit, everything became digitized. So we got into this how or the process, and nothing has done more to that in creating advancement than COVID because we couldn't see anyone. So everything shifted to digital. We owe that to them. We still focus on first-party data, collecting a lot more behavior. But this is when we started integrating front office with back office. Right? And then it's a lot about journeys and omnichannel and all that, 99% of companies are still in the middle of the stage. There's a lot more room to go in terms of improving that experience. I believe where we're heading, today, tomorrow and the future is really what I'm calling the purpose-driven experience. It's really delivering on the why or the context of why consumers engage with your brands. And that's whether you are a retailer, whether you are Coca Cola, I mean, we engage with Coca Cola as a beverage very differently. Some of us use it as a caffeine substitute. Some of us use it once a day. Some of us use it once a month and some of us use all kinds of different products. I think we all deserve a different way to be promoted to by Coca Cola. That was a great session, by the way, yesterday, wasn't it? So I think it's not about just the name on the bottle. But it's about how do you create the entire curated experience. That's what we're talking about. The context. How do you understand the context and how do you deliver on that context? But what do you need to do to deliver on that context? How do you understand the context of a customer? You can collect some of that data, like Apple, they know you're in your car, so they'll tap into your calendar, they see you have a flight, they'll give you the exit to the airport. That's beautiful, right? But a lot of times, we can't decipher context unless the customer tells us. Zero party data. Right? And I think most of you guys know what zero-party data is. It's data that a customer gives us willingly so we can do something with it, so we can deliver value. So the question here is how many of us are collecting zero-party data? How many of us have a plan to collect zero-party data and how many of us know what to do with that zero-party data? I think that's one of the key pivots that we're talking about. And when you talk about what industries and what networks and technologies I need to do, think of the example of the agency of the Costa Rica example. This little company does not own an airline, it doesn't own a hotel chain. They don't actually own anything, but they got 90% of my share of wallet for that trip. I think us as brands, if we know the context of our customers and we have that relationship, we can go connect industry networks. So a hotel chain can connect with OpenTable, so you can make your hotel or your restaurant reservations right there. Nothing's preventing us from that. So there's a flip in the operating model that we'll talk a little bit about. So this is really the next evolution of what we're talking about in delivering customer experiences. Now, zero-party data, where's our zero-party data today in most companies? It's sitting in binders. You know why? Because the zero-party data that we have today is like the focus group that we did. And we go and do the "Why do you engage with our brand?" We get that data. We segment customers on that. And we look at it and we understand it and then we move on. But this is the phase now where we can actually activate and create profiles of that zero-party data about our customers in the database. And that evolution, guys, happened also in all the other phases. Let me remind you, in early 2000s, retailers didn't know what you bought from them. Some of them had a loyalty program, some of them had a private label credit card program but nobody had that data. But now they have it. Healthcare companies, insurance companies, banks, nobody had a 360-degree view of the customer. Now, we take that for granted. I think the next wave is taking zero-party data and putting that into the database. It's collecting it, it's observing it, and it's using it to drive personalized curated experiences. Does that make sense so far? Okay. So what is a purpose-driven experience? We've got three pillars for that experience. Number one, trust. And trust is not only in safe keeping the data, which is obviously is the number one element in doing that, but it's really delivering value with that data. Very important. You guys understand an element of that trust, and Thomas is going to talk to us about how generative AI is actually delivering on that trust through different elements. Number two. You've got to understand the customer where they are, their context. We've talked about that. Where they are at a macro level, but where they are at a specific given point moment in time. They're in their car right now. They're heading somewhere. So Apple shows you directions to the airport. But it is also the context of the grand of why what kind of customer they are. So segmentation, the personas, all that needs to come together. And finally, we've got to deliver value. It's not only important to be relevant but it's very important to deliver value with that data. So how are we making the experience easier, providing maximum value to that customer? It's the intersection of these three pillars that we define as a purpose-driven organization. Okay.
Let's see. Three pillars delivering purpose-driven organization. What are the things that you need to do in your organization? And I know that Wealthy Thomas will come up on stage here, and Sam, to talk to us about what they're doing relative to all of these three things. Data, we'll talk a little about that. We touch upon it. Your operating model and then finally your gen AI. What kind of gen AI engine were you using with that data? Fast forward three years from now, everybody has all the technology that they need. They have all the gen AI models that they need. What could be the strategic advantage that one company would have over another? It's the relationship that they have with the customer and the trust that they have with them. So zero-party data, in my opinion, will be the competitive advantage and the currency of the future. So for each one of you guys, what I would say is start thinking about what data do I want to collect about my customer and how do I drive value with that data? When do I capture that data? So I'll give you an example. Discount Tire. Who knows Discount Tire? Okay. Really great brand. Right? Great service, great customer experience. They probably have one of the first most successful, scalable, online retail sites where you can go buy a tire and go to the store and install that tire. We've taken them through a shift, where it's not just about, hey, plug it in my model, my car model, the year, the trim, and get the right tire. Everybody can do that. But we're asking the questions online of where do you drive? What kind of driving do you do every single day? Are you driving around town taking kids to school and busing them around or are you driving 50 miles to work every single day? What's most important to you? Is it noise reduction? Is it longevity of the tire? Is it smooth ride? Are you a car aficionado? Or are you really into driving or you just want to get from point A to point B? We're taking all of that input to deliver a recommendation to you for a tire. That's purpose-built, that's purpose-driven. We're adding to that now for people who are willing to change tires, hotel services. Did you guys know that? They actually have hotel services for tires. Well, they take them, change the tire, store them for you. I call it hotel services, they don't call it that. And then put them back on the car, and they do that all for free, by the way, which is incredible. Right? Just an example of the shift that we're seeing in experiences. So zero-party data, collecting that from your customer so you can deliver relevant-- And by the way, it doesn't cost a lot more. You're already collecting data. They just really go into the strategy of operationalizing that. Operating model.
On your left is the current model of a really good data customer-driven organization operating model. We've got our products and services, and then a good experience-driven led organization says, what are my customers' needs relative to my products and services? And then they go and design really good customer experiences. And there's nothing wrong with that, and I think that is a really good model. But I think the model of the future is, we start with the customer needs and services because once you have that relationship, which no one else has, it is your responsibility to maintain that relationship and expand your value proposition as much as you can in order to maintain that relationship, deliver the experience and then the products and services. I mean, who thought Amazon would go beyond books to sell us entertainment, content, cleaning services? You name it. I think because they've got the trust, like they've built it differently, and I think they're working on the context. But that's just an example of once you have the relationship, you can sell them anything. And once you have a friend, you can sell that friend anything. So very high level in here, but we're working with clients on to say, okay, let's put this model around. Let's figure out what are your needs are first, who your customers are, what are their needs, and then deliver on the expanded value proposition. So I do want to invite Thomas on stage, and like I said, Thomas has been working on Watson models for a very long time. And he's going to talk about what's the next level building trust in the gen AI models that we have. So, Thomas. [Thomas Schäck] Right. Thank you. Thank you, Pierre.
All right. So let's talk about building trust with generative AI. This is a multi-layer problem. The first layer that we typically have to look at is trust in data and making sure that we have correct data, we have relevant data, we have data that is complete and current, and very importantly, that has proper terms of use around it so that it's known what the data can be used for. That's especially important in Europe but also in other places to basically adhere to what customers expect and not do something with the data that customers would actually maybe not like. The next layer up is trust in the use of the data, how the data is processed. In the context of gen AI, this means choosing the right Large Language Models, applying appropriate guardrails so that around what the Large Language Model does, there is measurements of HEP and other factors. Also, things like adhering to certain expected outcomes. And that is something that can be monitored in real time and where it's possible to ensure that the model operates within expected parameters and does not go off the rails. The other part is monitoring performance consistently to ensure that the model performs well and does give timely responses that are high quality. And, all of that, again, at that level also honoring the terms of data use. So these two levels are, on one side, technical and on the other side, governance-oriented. The next level up is trust through value. And this is probably the most important level. This is all about collecting user feedback, which may be implicit or explicit. Sometimes, it's just observing what the user does. Sometimes, it's asking the user for feedback and evaluating that. And in doing that, observing interactions, outcomes and making very deliberate improvements based on the feedback and based on the analytics of the feedback.
Now, one example that we implemented this is a Q&A solutions pattern and solutions component that we provided. This is something that now many of our watsonx customers can use. And this is one example. There's a big software company that also is working to use this. We are currently deploying this for them. And it's basically going to work in a way where the whole documentation for all products is ingested. This is something that is millions and millions and millions of documents. And on top of that data, we basically integrated it with Q&A with RAG pipeline that uses a Large Language Model to automatically generate the answers to user questions. And here's an example of a question that I could ask and then the answer that is given. And then, the next picture shows the feedback mechanism. So in this case, we did a simple feedback mechanism. A user can indicate whether the answer was perfect, good, okay, bad or very bad. And if not perfect, can give a comment why it was not perfect. And this is something that we record as part of the solution. We record all the questions, all the answers, all the content chunks based on which the answers were generated. And most importantly, the end user feedback is also recorded along with that. Now this, over time, gives a body of knowledge that can be analyzed. And in the next graph here, the bar diagram, you see the analytics that we do over top of that data. So it basically starts with automatically detecting the topics that users are asking questions about. Then we report by topic what is the user feedback? Is it mostly positive, mostly negative? How much red is there? This is what can inform for which topics enhancements of the knowledge content should be driven. And hence, over time, the more users use this, the result will get better and better. And then, when the same user comes back or when other users come, they will already have a better experience, which is enabled by the feedback that other users have given. So this is one example where trust can be increased and increased and increased over time by having this virtuous loop of users using it. And the more users use it and also give feedback, the better and better it becomes.
Another example that I want to show is going in a somewhat different direction. This is a POC that we built to build a gen AI pipeline that uses multiple gen AI models to do something pretty interesting. This is a picture that is a real picture. It's actually a photograph. And it has a real car in it. It has a real brand. It's a Lamborghini. And then, it also might have people in it. It might have things in it that are not appropriate to, for example, use for publishing. And what we did here in this gen AI pipeline that we built, you can actually also see this on the show floor. We basically, at the first step in the pipeline, we use a Large Language Model to do a check whether the image has any harmful or offensive content. And in this case, it checks okay. And then the next step is that we use a different Large Language Model to generate a description of the image. So this description that you see here on the left-hand side, this is what was automatically generated by that second Large Language Model based on the picture. And this description now it's something that we could not immediately give, for example, to Firefly for image generation because it would not necessarily yield a good result. So what we did here is we used a third Large Language Model and combined with it or provide to that the Adobe Firefly prompt guide. And that together is able to generate a new prompt that is optimized for Firefly generation and corresponds to the description that was initially created. And this is the end result that we get. So this is now a picture that was completely generated by Firefly. It has no real people in it, no brands and is something that could be published as part of a campaign. One thing that we also actually did more recently, it's also something that we show on the show floor is we even introduced a further step to refine this based on brand guidelines. For example, we tried it out with IBM brand guidelines. But it could be other customers' brand guidelines just as well. And then it's possible to also automatically further refine this picture one more step and then make sure that it's also on brand.
That's what I had prepared for today. Thank you very much, Thomas. - Thank you. - Yeah. Thank you, Thomas.
Again, closing the loop with trust, you saw the feedback of the customer feedback, how we improve the focus on the model. And then with this picture in here, the next time the customer comes back, after the first, they're going to see a more relevant picture. So it's really all about incorporating the feedback into the process. So lots going on here, and you'll continue to see tremendous improvement in fine tuning these models and the movement from Large Language Models to Small Language Models. Why? Increase the relevance of that performance. Next, I'd like to invite Wealthy to really talk to us about his journey in delivering purpose-driven experiences with AT&T. [Wealthy Desai] Thank you, Pierre.
Thanks. I appreciate you taking the hour from your day and spend that time with us today. I'm really privileged to share the stage with some fine minds from IBM and showcase some amazing work my colleagues and I are doing at AT&T to improve our customer experience. I'm Wealthy Desai, by the way. So my name, if you are curious about that, I keep coming to Vegas trying to generate wealth, it's not working so-- But we have better bets and better odds on generative AI. So let's talk about data, generative AI and trust.
Yesterday, James Quincey talked about Coca Cola's purpose. He talked about the fact that to refresh the world and make a difference. For AT&T, our purpose is to connect people with greater possibilities. And we are really focused on customer experience to drive that goal. Our strategy is conversions. How do we get customers connected, whether it is fiber network or whether it is wireless network? And over the last four years, we have really spent our energy in simplifying our business. We've become really focused on that. And you can probably see that in our stock price. I'm really proud of that. So if you think about the investments we are making right now, our core four strategies are laying out the nation's largest fiber network that can work as a platform for us to run our wireless business, to invest in customer experiences and products that they love and care and they want from us. Third is around reducing our cost structure and re-engineering that. And then fourth, investing and elevating our brand with AT&T guarantee.
How are we building trust? What I want to share with you is how we are using data, how we are leveraging gen AI and improving customer experiences and ensuring how it's built on trust. And so, I'll show you a simple framework we are using at AT&T and three examples that I'll walk you through.
Here's a simple framework. It's no big deal. Experiment, personalize and scale. So no secret. Right? "Crawl, walk, run" approach, if you will.
So one area that we are really fortunate as a company, we have millions of touch points as AT&T with our customers. We get millions of customers coming into our store, calling into our call center, and coming to att.com. And so we want to leverage that opportunity, ensure that we build the trust every customer touch point. So here is an example of a customer coming to att.com and doing a search, "How do I add an international plan?" right in the search window at the very top. And here's the generic answer we'll provide. And in the past, two years ago, we would have shown a result with a link. Customer would click on that link and go to an FAQ page and get the question answered. Then we graduated to providing an answer card, similar to a Google search. The customer don't have to click to get the question answered. Now what we are doing is we are leveraging gen AI. And especially if you authenticate, you trust us with your data. You log in. In that case, we are going to leverage who you are, the device you have, the plan you have, and make that answer very specific to you. So this is very contextual. Again, so going back to the point that Pierre was making earlier, if you give us the data, we are going to value that, trust that, and then we'll provide an answer which is very relevant to your use case.
Here is another example of customers coming online and looking for an answer to their bill. "My bill went up this month. What happened?" Here, we are using a Large Language Model to interpret customers' billing data and providing an answer which is very, very relevant to that particular phone line or an account and indicating that there are two credits which are pending. It's going to get applied to the next bill. So again, when we do these small experiences and improvements online, the result is reduced calls in the call center.
And that's really working well for us.
In this case, the third example, we partnered with IBM and Adobe to build this experience. So think about a customer who is looking for a phone for their kid. Two years ago, a customer who came online and asked this question, here is what we would do. Show them some result, provide a link, go to that page, figure it all out. Right? But think about a customer who is going into a store and engaging with an agent. I'm looking for a phone for my 12-year-old. Now, the customer is willing to share the data. "I have a 12-year-old son. I'm looking for a phone. Help me." In that case, the agent is going to curate that experience. And that's the same vision we were trying to realize with this experiment we are doing in partnership with IBM and Adobe. So what we did here is we created an experience where a customer has some prompts. They can engage. We are asking the customer to provide some information. In this case, if the customer says, "What's the best smartphone for my kid under 12?" In that case, we would say, hey, if you are buying a phone for your kid, what is probably important is durability, waterproof phone, right, security, the types of apps you want to purchase along with the phone. So we are curating this entire experience and ensuring that the customer is willing to share their data and we are valuing that and creating a response which is relevant to that. And then we're making it very easy. Customer never leaves this experience. They never go to a product list page, product detail page and the configuration card checkout and they immediately can get out. Along with this, one of the things that we built is if a customer wants to engage with a human, at that point, they may have interacted enough with this virtual agent. But at that point, they decide, "Hey, I do want to have a dialogue with somebody." In that case, we'll create a very qualified lead. Now they have shared a whole bunch of things with us. So we're going to get a qualified lead and have a sales rep follow up.
So again, going back to that framework, we are experimenting with this. We'll personalize it. And then eventually, we'll scale it. One of the important points here, you see the thumbs up and thumbs down, is when we have these interactions, we also ask the customer, "Was this relevant? Was this response right?" And then we collect the data, feed it back to our language models and then continuously improve. So hopefully, this gives you an idea of the types of experiences we are building. And I give you three examples. One was in the care space. One was in the billing space. One was in the sales space. And then I talked about the framework. So connecting the dots back, we're building the fiber network. We are guaranteeing it with AT&T guarantee. We are confident from a macro level point of view. And then at a micro level, we are taking care of all of these small experiences. So hopefully, that gives you a full spectrum of how we are building trust for a period of time.
Well, I think you have one more slide in here. That's really a good slide to talk to if you wanted to share that. Sure. Because I think that really brings it into the visual term with the applications and all of that. Yes. I mentioned about this experience of how a customer would interact with the virtual agent and eventually will end up with a cart. So one of the things that a human would do like, hey, if you're buying a phone for your son, you may want to download specific list of apps. And so those app recommendations is something that we built as part of this experience. And when we saw and we implemented this, we saw 4% improvement in our add to cart experience. So this is a good example of customer willing to give us their data, we valuing it, generating response and then we can see a better outcome from a shareholder point of view. So our belief is that with this, we build trust, loyalty, and in the long run, we'll generate ROI. Thank you, Wealthy. Thank you very much.
And I'm sorry I did a double take with you, because I love this example, and it's such a great example of a micro moment of somebody's trying to buy a phone for their kid and it's not about the phone, it's about the phone and everything around the phone. It's about the security. It's about building that trust. That's a very different interaction versus, "Hey, what's an appropriate phone for my kid?" and just giving them the appropriate model. I love this. Again, small example, but that's where we're heading, and that's what a purpose-driven experience is. Okay. Great B2C perspective. Really nice. It's going to get a little more hairy with the B2B perspective because you know the B2B industry is not where the B2C. So we have a great example here with B2B marketing, and please, right here on the stage with me is Sam.
[Sam Yeung] Hello, everyone. I am so happy to be here talking about a topic that I'm so passionate about. I pretty much eat, breathe, sleep personalization, and I do think that the heart of personalization is the audience themselves. And when you have a cohesive audience strategy, that's where you're able to finally build out a customer journey that fits the audience that you're trying to target. And I think with having a CDP that has the right audiences and the right data sources, that's where you can actually start doing some really cool targeting that can push them down the funnel. So one of the things that I really wanted to ask this group is quick poll, I wanted to ask, what percentage of your website visitors are anonymous versus authenticated? And when I say authenticated, meaning they actually submitted their email address or their first name, last name, whatever it is where you can actually identify them. So by a show of hands, I would love to see around the room if you're in group A, where the bulk of your visitors are anonymous in 90th percentile.
Got some. Some. Okay. We got-- All right. How about 75% anonymous? Okay. Mixed bag there. And then about 50% anonymous, half and half? Okay. And then what about 25% anonymous? Okay. So it's a bit of a mixed bag that I see here. Okay. Interesting. So I'll tell you about the challenges that I see at ibm.com. And unfortunately, the bulk of our visitors to .com are anonymous. 95% of our visitors are anonymous and it's hard to do personalization when you don't really have much signals from your visitors that are coming into your site. So the four main challenges that I see with anonymous visitors is that they're coming into our site through mobile web on their mobile devices or through a desktop. Sometimes they come in through a tablet. So there's a lot of different devices that one single individual can come into our site, so it's hard to stitch them together. The second main issue is just limited signals and queues that we're able to get from anonymous visitors coming into .com. And then the third one is, delays in identity recognition. So it's just taking the anonymous profile and how do we add in more color and depth to understand who they are. So there is delays in that. And then lastly, because we are a B2B business, we rely heavily on account-based marketing platforms, ABM platforms. So this is where we're able to add in a little bit dimension as to who the person is, what job roles, title, company, how big the company is, etcetera. So we can further hydrate the customer profile ID. So there are limited limitations there. So I guess the next question is how are we solving for this challenge? So I think there's five key pillars that we're ultimately looking at to solve for it. So the first one is layering in identity resolution platforms. So at IBM, we are privileged to have two ID platforms that we're able to leverage. So the first one is LiveRamp and the second one is ID5. So the importance in having these ID resolution platforms is that it increases the match rates of your anonymous visitors to pseudonymize. So there is a clear distinction between anonymous to pseudonymize and then to authenticated. So it's just more so a gradient that we want to push in that direction. So I'll go into details in a later slide. The second strategy here is the data layering strategy. So how do we collectively bring in all the disparate different data sources and stitch it together? So having zero-party data, first-party data, second-party, third-party data, these are all crucial to hydrating the customer profile ID. And then the third one is, how do we take AI and drive personalization with it? So I think the beauty about Target and AEP is that there are generative AI capabilities there and we have to leverage that to really enhance the personalization efforts that we're trying to achieve here. So without that, it's going to make our jobs a little bit difficult. And with the gen AI, we are leveraging that as far as lookalike modeling. So I'll go into details a little bit more in later slides about that. And the fourth one is, again, the account-based marketing platform. ABMs are super crucial to the B2B dynamics and audience segmentation, so we definitely do need that. And then lastly is progressive profiling. So again, it's taking the anonymous visitors, hydrating it with as much data sources as possible so that we can get pseudonymized visitors, and then to finally push them down into the authenticated bucket. So this is pretty much a full snapshot as to how we are triaging our anonymous visitors and how we're able to personalize the experience to them. And then once they become a client, then that's where we start to do a lot of nurture and customer management. So I want to say that from the awareness stage, this is where we're dealing with a lot of anonymous visitors coming into .com, so they could be first time visitors that don't exactly know what they're looking for but perhaps they're just getting familiar with the IBM brand or they're just doing some basic research. But essentially, the second time they come back into our site, we have a record on them as to what they looked at in their first visit. And then when they come back in their second visit, we should further anticipate exactly what products that they're looking for. Maybe we can personalize the home page and give them relevant content that they might be interested in. This is where we move from the anonymous stage to the pseudonymized stage through a bunch of different first-party, second-party, third-party stitching. So if we can get them from the awareness to the consideration stage, where they can come back to our site multiple times and then finally arrive at the decision stage, then we can really do some serious personalization. Because once they're arrived at the decision stage, they've already submitted a form or they've engaged with their chat box experience. They've engaged with some sort where we can say, "Hey. We know who you are. We know you're interested in watsonx Orchestrate or whatever products that you're looking into. And then really help the sales org in closing the deal because it's very much helping the sales org be empowered to know the customer and help them close the deal but also how do we supplement that sales experience with the digital experience? So I think there's a good combination there. But I will also say that the bulk of the heavy lifting is demystifying the anonymous visitors and that segment of users is the most poorly misunderstood. So that's the hard part about marketers in general, and we need to do a better job just understanding what are all the different data sources and the tech stack that's necessary to enable all of that. So as we look into how we're stitching all the different data sources together, this is where the magic happens inside AEP. We have a long list of different data sources that we can start adjusting in there and the end goal is to be able to stitch those different data sources to hydrate the customer profile ID. So I put down a quick example. So we have Jane Richards, who works at a bank and we have her email address. And we understand exactly how much revenue this client is pulling in and the size of the company, the position, preferred marketing channel contact, etcetera. So understanding who Jane is, we're then able to further build out similar audience segments that we can start personalizing the content and experience on .com as well as all the ancillary marketing channels because the different marketing channels that someone is exposed to, we want to be able to have an end-to-end journey and be able to measure that on .com because all these disparate marketing channels, their purpose is to drive traffic into .com and we only have a few shots to getting this right and to push them down into the funnel. So we build out these audience segments and then it becomes a bit of a feedback loop with using these points of activation. So we have paid media, social media, email, SMS push, whatever the case is, and to really continuously engage with them where they are in their customer journey. So once they are engaged with our products, we know that they submitted a form. Let's not continue to send them, like submit a form request, because they already did it. But let's continue to nurture that experience with relevant content. So this is where measurement comes in and obviously measurement is key in this whole thing because it provides a feedback loop as to what are we doing correct, what are we doing wrong and is there any room for improvement as to how we should be doing things. So this is where we are able to use gen AI and AEP. I think this is a capability that I've seen and heard a lot of people throughout summit this week not tap into. So within AEP, one of the challenges at IBM, again, is 5% of our visitors are authenticated. So to further demystify the misunderstood anonymous bucket of people, we then take a lot of the attributes and the traits from that 5% that are authenticated and then project it onto the 95% and really understand as to what they're looking for, what they want to read, how do we really push them down the funnel with content that makes sense. So one of the examples that I want to talk about is when you come into ibm.com, there's different job titles or roles, so let's say we have a CFO that comes in and then we also have a CTO that comes in. The content that we serve to a CTO should really be focused on the technical capabilities, where you have the bells and whistles, like what it can do, just the value that it can drive for the business, the adoption, this and that, right? Because that's what the CTO really cares about. How can this platform further enhance my business and drive value? But then if we have the CFO group coming in, they really care about the price. So they're more interested in the pricing model. Is there a way to do some bundling pricing for existing customers that are already using IBM products. So pricing is what they care about. Just the terms of the pricing, is it 12 months, is it 3 years, is it 3 months? Can I bail out of this? Like just having upfront pricing for them is going to be important. So I think when we talk about just the different process and buyers that are involved, it becomes a lot more complex. So we really need to be specific and targeted in how we're delivering the experience to these folks that are coming into .com.
So one of the things that I do hear from folks around Summit is, how do we really pitch this upwards to our executives? And these are the three key things that I tell people, especially if you're in a B2B dynamic, which is, one, you want to increase conversion rates. So top-line key metrics, these are the ones that executives are held accountable to and you need to be able to deliver personalization to help achieve their goals. The second one is closing the deal. So the average time that it takes to close a deal is anywhere from 12 months to about 18 months, and we want to be able to shorten it up. And by shortening it up, we can then layer in the digital experience with the sales organization. And then the third one is increasing operational efficiency, just making sure that we have the right workflow process where we are able to reduce redundancies and overhead and that sort of thing. So this is where gen AI comes in, and it's supposed to help be an accelerator in our jobs and not necessarily replace anybody. So one of the things that I want to close this on is a question that I would like for all you to consider is, how are you unlocking the full potential of personalization for your anonymous visitors? Thank you. Thank you very much, Sam. Really appreciate it. As you can see, folks, how we apply gen AI in B2B versus B2C, the sophistication, the evolution, these two industries are very different phases in their maturity. I'd like to ask you a question, Wealthy, with a couple of minutes left here.
We're seeing evidence that companies are moving beyond POCs. They're starting to talk about rollout and how do you drive adoption of gen AI in the organization. You shared a pretty interesting model for how you innovate and you try. What are some of the challenges that you're seeing in scaling gen AI in your organization? I think, with AT&T, we have had customer data for years and years, right? You could call 911, drop the call, and we know exactly where you are. We have had this trust built over a period of time. So we value that greatly. So it's something that also comes with these challenges. When we were building the solution, there's a lot of data governance. What we are going to use? How much are we going to expose to a language model? How much will it be exposed outside of AT&T? So we value that greatly. And along with that value comes the challenges. So we had to be very careful in how we are leveraging the data. And then we are able to show to the customers that, hey, we are able to service and provide whatever they're looking for. Okay. Thanks. Sam, I have a similar question for you. I know IBM marketing is using gen AI to generate content. You've talked about how you're using gen AI to solve an identity issue. What do you think is next for IBM in driving personalization with gen AI? So audience segmentation is going to be the most important. I think the next step is leveraging gen AI in Target to start automating and serving these contents and these experiences in real time because I think the way that we're able to build out these segments, it's going to be in the masses, as it should, right? Because they're very hyper-personalized and hyper-targeted groups that we're trying to go after. And we have to be able to send these contents and these experiences in real time. So that's where we plan on leveraging target for. Very good. And, Thomas, last question for you.
The evolution and the speed by which models are evolving is accelerating. Can you talk about what is the next in model building to increase the trust and the value and the precision that we deliver generative AI in. Yeah. I think it's a fast evolution that is happening.
People initially looked at models and data somewhat separately. And then data was used piped through the models and that was the result. But I think incrementally, it becomes more and more important to look at both as a system and, basically, gen AI plus data where then the data needs to evolve to better serve gen AI. But also the data being used through gen AI by users with a feedback cycle can make the data better and better over time. So it's a system, not just two separate things. Very good. Well, I want to thank you, all, for being here today. I think you guys really lent the practical perspective to what purpose-driven experience is. I want to thank you again for being here today. Hopefully, everyone here got something a little different about where we want to head with our customer experience. So thank you all for coming in here and sharing this with us. Low plug, if you have any other questions, please, we have our booth on the floor. We just created an IBM Business Value report on content supply chain and generative AI. Please take a picture of the QR code, download the report. It's an incredible report about how do you draft personalization with content. And then lastly, we've got a great session later today about how do you operationalize gen AI in 2025. I encourage all of you guys to attend there at 03:45 at the Community Pavilion. Thank you all for coming, and I appreciate your time. Have a great day.
[Music]
