[MUSIC] [Ash King] Good morning, everyone. How are you doing this morning? Did you go see DJ DIESEL last night? Show your hands. Alright, so I'm going to try to keep you guys from falling asleep, and we'll talk about something interesting today. Content strategy and architecture: Design to drive velocity with GenAI. Now, if you've seen this session in past Summits, I'm keeping a lot of the slides from last year. And there's a reason for that. I want to show how we build upon content strategy to get to these personalized experiences, to get to realizing all of the capabilities of GenAI. Content strategy is at the core, just like a programing language is at the core of building an application.

So, today we're going to talk about the evolution of content strategy and architecture. We're going to talk about content strategy and its impact on velocity. And we're going to talk about assets in the GenAI content supply chain. And what does all that mean? But first, let's start with a definition. Content strategy. Getting the right content, to the right user, at the right time. Sounds a little daunting because there's so many components to that. What is the right content? Who are the right users? And what is the right time? And how do I model that in a structure to be flexible. But it continues. For the right reasons. What's the intent of this content and how does it fit into the right call-to-action, within the right budget? That's a key factor there. I see some of you wincing. I feel that.

Through the right device and medium. So being able to understand the channels and activation and how do you do that to attract and retain a clearly defined audience. So now we're talking about all of the context around which our content is fueling and experience with personalized messaging and with the resources allocated. So not just the budget, but what do you actually have now that can drive this experience in this strategy. And the systems currently available. As much as I would love you to have everything and all the latest from Adobe, I know that everybody's in various stages of implementing on their DAM. And of course, to drive profitable customer interactions. And how do you measure that? Using the right data and capturing analytics, in real-time. I think this captures pretty much everything about content strategy, but it also shows the importance of this in the overall experience.

So let's dive in.

The evolution of content strategy.

There are core practices that we like to think of in content strategy. Of course, with user experience. Right now, we might think of that as a separate discipline, but it's intimately a part of content architecture. And of course, technical content architecture. How are you interacting with your systems, with your content, with your content supply chain, and building a structure that can be shared and provide and interoperability in the platform. And of course, content marketing, and editorial and brand strategy. User experience broke off when on its own, went to college, graduated, went to an internship. Now, is an executive within the world. And of course, content marketing, and editorial and brand strategy. There's another set of this that has been very prominent within roles in the enterprise, but now it's becoming even more of an offshoot with generative AI. But when we're talking about the primary role of technical content strategy, it's still many-faceted.

There's lots of Venn overlap in the four areas of this diagram that we like to think of as four ways to approach the center of technical content strategy. Of course, there's the user experience, the presentation itself, the content architecture, the structure, and how you're putting that together, your metadata, content marketing. So what are you doing with this content and how is it influencing and working and building valuable experiences? And of course, the conversation within which all of this occurs. The narrative. And it sounds like a way to really, from the center, as a content strategist, think about looking from the outside. And again, there's more neurons. And the more roles that you're interacting with. So of course, you've got UX/UI specialist, content designers, architect interaction developers, everything on user experience side. Of course, in the editorial strategy front, we have our copywriters, our branding experts, content leads, graphic designers, video producers, SEO, content marketing, multi-channel marketers, campaign managers, social media managers, SEM specialists, industry marketing, content strategist. There's so many things that are interacting with our content strategy.

But within content architecture that we like to talk about here, where we're thinking about content architects, library scientists, taxonomists, metadata specialists, information architects, marketing technology specialists, and of course, content strategists.

So we're going to take this access here and look at how we pull together these elements and think from a standpoint of what are we trying to do overall within our enterprise neuro strategy? What are the technology underpinnings that we're working with? What structure can we provide within that, and how can we share it across the content supply chain and be able to think about hydration, about use and being able to optimize and iterate, improve, learn, and move forward as the strategy evolves over time.

So let's take this in a view of the content supply chain. And normally, this would be a little bit of a build-out, but I figured I'd leave it up here. So that we can talk through all of the elements. In the content supply chain, typically, we are talking about starting over on the left hand side with a tool like Workfront, where we have a brief and associated with that brief, we have some idea of the intent, might have some visuals and examples. We have some metadata, and we start with that brief to move into the creation process. Creation process. We'll be working with our creative teams, and we'll be iterating to get to the right asset that we need. So there might be a photoshoot. There might be some selects from that. And then we pull those selects back in and we think, well, maybe these aren't the right ones. Maybe we iterate back. And that's an expensive process at that point. Think of the dollars adding up as you go from left to right. And then once we have a finished asset that we've all approved, we're going to move into the collaboration and then evaluation stage into enrichment. And we're going to take the metadata from that finished asset and include several elements of it within our structure. At this point, we should be pulling metadata from the brief. We should be thinking about where all of this started. But we also should be thinking about what's the end in mind as we move through enrichment and ensuring that this content can not only be tied together within an experience, but that in the future, a gift to yourself in the future and everyone else, you'll be able to also find and repurpose this asset. So how can it be found outside of this experience? And of course, through there, and delivery publication, activation. We are going into the end users. So this might be multiple channels. We might be using AEM sites. We might be using Content Hub to deliver assets to end users. We might be using HAO, we might be fueling an app through APIs. There's multiple ways that we can think about this, even going out to paid media, email, etc. And then bringing back all the information from that experience and being able to do something with it. We have to think about that within our structure. How do we update and understand the performance of the asset, not just from a data standpoint, but from the structure standpoint? How do you say this performed well, we should do it again? Maybe we should invest more in content like this. And if you look all the way across from left to right, typically we've been thinking about $10,000 to create a piece of content. And if you lose that content, if you can't find that content, if it's sitting on someone's thumb drive somewhere, that's $10,000.

A whole team of experts has spent all of this time to create. So, content architecture. And the role of the content strategist is valuable in everything that you do. You can count in that way.

But what about end users in the world of GenAI? What does that mean for us? When you can take an asset and instead of sending it to a creative to change it into a banner, you can apply an API so that it fits the channel. Or you can change backgrounds, you can do lots more. You can assemble experiences.

That's what I want to talk about more today.

So I've just put up, just a really barebones, high-level way of thinking about how content strategy relates to generative AI. So with content strategy, we have our assets, we have our content architecture. And that really rolls up into and supports our brand strategy and our workflow in order to accomplish that. And if we're thinking about these in concrete terms, you probably have a finite set of assets that you're working with, in a DAM, that you're organizing in a certain way with your folder structures, your metadata, etc. And then you have guardrails as to how that structure and those assets can be applied as part of the strategy. And then you're organizing the teams to accomplish a goal.

The GenAI side. There are some analogous ways to think about how you apply and utilize this technology. First, you have your models. So when you're thinking about models, we might think, an LLM, ChatGPT, Firefly. We might think of other common ways that we interact with generative AI models. But there are multiple vectors or axes that you're thinking about with generative models. In some cases, a model might be multiple models. They are actually working together to accomplish a result.

And once you have these models, they're going to evolve either through real-time feedback and interaction with users, with the results. You've all seen the thumbs-up, thumbs-down buttons. Those help improve models. Or through some other type of feedback in those models, or even because maybe the data sets that were used to train those models change over time because of licensing. So these models are not a static concept. They're not a monolithic concept. There's something that you, as a content strategist, need to have an understanding of, in terms of which ones are being applied and how. I'll give you a couple of examples.

Have you heard about ChatGPT refusing to do things? The lazy problem. All of a sudden you're using ChatGPT and saying, hey, could you write this out and give me a deep explanation of this subject, and it says, well, here's how you can do it yourself.

And it hadn't been doing that before, but because of interactions and with human feedback and learning and evolving, it got to a point where it wasn't as performant and the results weren't aligned. And of course, that had to be corrected. So the model changed again. And all of the results over time are different because of these types of interactions. We call that Model Drift. So it's really important to understand if you're generating content or modifying content with models, which models were you using and when did you do it? So that you can have some idea as to whether or not those results are replicable. And sometimes, even if they're not replicable, that could mean that that content is more valuable because you can't recreate it in the same way.

Thinking of model customization. Now you have the ability to take your data, to take additional information and be able to train the model or fine-tune it for a purpose. We saw that with Firefly custom models, where you were able to take 10, 15, 30 images, be able to curate that set, provide some actual context with it, in terms of your captions, and be able to create a language around this for the model to understand how all of the elements, in that set, are meaningful in terms of getting an output and a result that model customization could take the form of maybe a style, a photographic style or a concept of an object like a backpack or maybe an illustrative style. And if you are creating multiple custom models for multiple purposes and using them to modify or create content, you're going to want to account for that in your structure so that you understand which models were being used. It's just the same concept of, if you wanted to take that really cool image that performed well from the Back to School campaign, last August, and you want to use it again this year, what if it was generated by a custom model and that custom model was built last year, and who has access to it? Where is it in the ecosystem? How can you get that model back? These are things that need to be accounted for in your content architecture. And of course, there's context. So when I say context, I'm thinking of what campaign is this being used in, and what brand is this campaign a part of? And within that, what products are a part of this campaign? What personas and segments, and profile attributes are being used to assemble this experience or to target this experience? How did it perform? There's so many different things that you want to think of in terms of context. And really, if you're thinking about context, this is probably one of the most important areas of GenAI because so much of what we do has been moving over towards the delivery side or the marketers or actually assembling in real-time for a particular user. And that context is the language. It's the underlying way to pull together elements within that experience.

And of course, there's automation. We've had automation in the past. We've had ways to do like smart cropping and other types of things in AEM, but with automation, with Firefly services, you now have a broader array of tools to be able to work with and change imagery. So instead of thinking about in the upstream, in the actual brief, when you're working with creatives and you get an image with a red background and then you say, that's not what we need, I need to send it back to the creatives and change that to a blue background. Now, with automation, it's pretty easy to just run an API pipeline, get the mask, do a cut-out, change the background or just use Gen Fill. There's so many different ways to apply automation to modify content, but that then introduces a multiplicity of images that may have the same structure, the same subjects, one that would have nuanced changes within the image. You need to account for that within your content structure.

So let's look at a new drawing of how this works within the asset lifecycle with GenAI. So now going from the left, we still have the brief.

It is ultimately generative AI, doesn't change that. Humans are in the middle. Humans are driving the process. But this brief, instead of having a finished asset, might create a template, and a set of images and some text copy and some brand guidelines that are now the core elements of the experience. The core drivers of that experience. And then once you have your context applied to that through your metadata and you understand which personas you might be interacting with or targeting and which channels you're going to be delivering that, you can now take that first category, that first column of images, text copy, brand guidelines and templates and start to generate variations that are going to be specific to these categories, these personas and channels. So for example, you might have an image with a blue background which targets a certain segment. But because of the channels, you also use Gen Expand to change it to a banner, a square, and other aspect ratios so that you can deliver to the channel. And of course these variations then need to be assembled and they are assembled based on the template, based on the personas, using the metadata for the channels that they're going to be delivered to. And then how do we measure and what do we do with those insights is all part of this process. If we're looking at the experience and we know from the metadata that a particular image is just a banner for the decoration of the lower portion of an email that people probably will never scroll to. But as part of the design, why would you measure that and weight it as much as a hero image in that email? It very much feeds into how you measure. So thinking about that as a vector, metadata powering and informing how you measure certain experiences.

So all of this should be a flywheel. And again, the experience assembly is built upon the language of the content architecture and the measurement. And the insights are informed by the content architecture. And if we're doing this right, then every time we go through this loop, every time we do something, we learn and we apply. And it's easy to make fine-grained changes to your approach and the experience. That's what we're looking for in this evolution of content architecture with generative AI.

Let's get to some concrete, if not light experiences, that we can ground ourselves on. So last year we had the 'Where's Valdo' exercise to talk about the stages of content architecture. Today we're going to expand it. So it's easy to find Valdo when we have a single image.

When we're working with large enterprises, we start to think of where people keep images. Most marketers or people that have a job to do, we all have our set of files or images, the things that we need to do. But even on our own hard drives, in our email, how many times, how much part of your day do you spend searching for something that you created? Just a couple of weeks ago, maybe yesterday, and you can't find it and you're thinking, was that in Slack? Was that an email? Was it a PowerPoint? Who sent that to me? And if we start to think about enterprises, the default state is that, well, we don't even know if Valdo exists.

It could be on somebody's thumb drive in another department. It could be on an attachment to an email that you sent, but then because of data retention policies, you don't have a copy of it and you can't retrieve it. So it's often easier to just recreate assets. And again, as you're thinking about these really high-quality assets that drive experiences in that $10,000 metric, how much money is sitting in this stack here? And can you even find Valdo? Anybody see him? Okay, we got one.

He's right there in the little corner. So, phase one of content architecture, we start to adopt a DAM.

We start to flow content into this DAM. It's a mess, but at least we know it's one place. It's a famous story of my old desks before. In the startup days of web, I always had a stack of papers on my desk, and if anybody messed with that, the whole business would fall, because I knew where everything was in that stack. That's about the stage. It is for phase one.

And there's Valdo.

Phase two. We can start to develop a taxonomy. Well, we have a folder structure. We have some metadata so that we can start to organize this a little bit. And if you're thinking about, again, just your hard drive, your laptop, I'm sure you've gone through multiple iterations of trying to create some kind of structure for all the PowerPoints you have to have or client documents, and then you start another structure, and then another structure. And your backups have five different structures of storage that worked in the day but don't work today. So at least you have a source of truth. In some organization, but it's still hard to find Valdo all the time.

But, we move on to phase three. And of course, we have a well-oiled machine now, the center of content. One DAM to rule them all.

And of course, when there's no taxonomy sourced or provided by your users, you get it from that business process creating the content. You're in a good state where Valdo starts to be a parent. It takes less time to find Valdo. And this is where I think most of the customers I talk to are in a lot of ways. And we'll talk about why. Yes, you can find Valdo most of the time, but I would challenge you to go into your enterprises and talk to people who are interacting with their DAM and ask them how much time they spend searching. There was a customer who was really, really deep workflows that cut across multiple disciplines to be able to achieve outcomes, and when we interviewed the teams, we made a cut video of people saying, what do you spend most of your day doing? "Searching. 50% searching. 60% searching. 40% searching. I'm searching. That's most of the time in my day." That's about the state of a lot of enterprises.

But phase four, we know more about Valdo enough so that we understand how important Valdo might be within the structure. So it makes it easier to say, I can locate all of these experiences that Valdo was in. I can see how Valdo is helping me drive conversions and valuable experiences. And because of that, I also know how to surface and when to surface Valdo. So that instead of me looking for Valdo, Valdo is saying, hey, where's Ash? Ash has a job to do. These types of content architecture can be fueled with assets, with insights, and even throughout the content supply chain, being able to pull those assets back in to modify the structure.

So again, let's go back to this. A system without content architecture. A lack of metadata provides issues of findability and discoverability and ability to target the personalized content to known audiences. Inability to track the efficacy of messaging. Sorry, it's too early in the morning. But again, these are things that can be fixed if we're thinking from the outset about why we're doing this. Meaningful data to contribute to audience profiles, role identification, personality traits, communication preferences, all of these things are not present if we do not have a proper content architecture. And then there's that reuse. Not being able to request reproduction of existing buttons from last year. Apologies that I didn't change that. But there's so many different things that, it happen where you don't have an effective content architecture and this costs money, this costs time, this costs opportunities. And it also makes it harder to adopt new generative AI capabilities to drive exponential experiences.

So I stole this from last year. It's super scientific. It's just really anecdotal evidence, but I'm sure we can all relate to this. So, the usefulness of content and the time of metadata application.

A lot of times we have a metadata, content architecture built out and we think, oh, great. Yeah well, we're not going to apply the metadata right now. We'll get to it. We have things to do. We're not going to enforce this. Oh, wow. We could pull this from the brief. Well, we'll implement that in the future. But really what you're doing is, again, burying Valdo back into the papers deep within the structure. The time of creation, or even before creation is the best time to apply that metadata. Because again, you understand the intent from the start, you can always enhance and hydrate that metadata. But it's harder as content goes through the lifecycle to remember why it was created. What was the original intent? How many times was it used? What persona did it do well with? And as you get down through that lifecycle, things become obsolete that maybe shouldn't.

So I'll talk about some things with generative AI and add in considerations for content strategy. Where Valdo can be. So imagine that a new creative director or a new marketing director come in, there's a whole new set of opportunities that we want to take advantage of. But we found that overall, we made a big mistake with Valdo. Valdo should not be an Easter egg. Valdo should be a gnome.

And we want to target this gnome Valdo to all kinds of new segments that we've been able to pull together from new data sources. We got an agency working with us that has provided us with our own customized model that allows us to flesh out new segments and target Valdo because they understand our market. So we're really going all in on Valdo.

Well, how do we manage that within our content structure? What are the considerations around us? Well, of course, there's simple variations. Let's just say that we're talking about occasions for Valdo. We want to also target segments with Valdo. So we created some images. And again all of this is structure match with Firefly. So I took the Valdo image, used it as a structure, and then created prompts to be able to change and make those variations. So very straightforward to do. Psychedelic holiday, winter ski, colorful spring, traditional holiday, other things, of course, Spring Caturday, you have to have Spring Caturday. And we want to target these to segments. We have to think about how we took that original asset, and then how we have taken the variations and understand the meaning, the intent of each of these variations and how they were created. So it's not just about the fact that we created them. They were created with Firefly Model 2, with structure match, and wow, we have some prompts, and some keywords that we can also include, so that we understand how they were created.

Another way of thinking about this is instead of taking an original photo and just changing different elements of it, what if we got so sophisticated that we just knew what content needed to be, from a structure standpoint, and we wanted to decorate it with GenAI. Again, structure match, but instead, we start with a 3D object and render that to provide the structure. And then we can provide different scenes. So now you're referencing not only the image that was used to create the variations, but also the 3D object that was used to render it.

Also the prompts and the intent, of course, the context around the different variations.

More complex example. Let's say that you are targeting multiple regions with a food campaign, and you went through this whole process of doing the Superbowl shoot, and you had this really well-performing asset that you wanted to use.

How do you transform that? Again, here's structure match coming in again. We've changed all the foods over different holidays and cultures to be able to take that well-performing image and use it again. We want to say, all the metadata that's been utilized for this first image and the event that was used, and then where it's being used and how those foods have changed. You can look and see that you've changed different foods within there. You've changed even some of the utensils, the bowls, the crockpot, those are all elements that are relevant in terms of your content architecture.

Even the structure itself is relevant. How many dishes in this image can I change? What's the proximity of one of the dishes or the plates to the crockpot, for example? That's meaningful neuroscience in a lot of marketing campaigns. So taking advantage of that in your metadata.

So here's one way to think about it. Let's say that we have different personas that we're targeting and we have different occasions. And we just matrix it and say, wow, if I'm looking for a spring break image of Valdo that targets gardeners, maybe I could find that pretty quickly from a variation that performed well in the past. Or winter holiday, I can do a gardener's version because they're looking at those sticks in front of Valdo and they're thinking, you know what? They should have planted better in the fall. Or there's creatives that want to have something more colorful. Outdoor enthusiasts. Maybe you're selling really nice plaid outfits. There's lots of ways to think about how to take these variations and matrix them.

And then there's your custom models, which include another vector. So here we're looking at the assets that are being used that have been curated to train a custom model. So we've got a set of images. We need to have diversity in those images. So multiple angles, lighting conditions, backgrounds, so that the image can be understood well by the model and various states of this. So we've got straps, we've got the straps behind it, we've got it in different positions. And all of these are required for you to have a really good curated set of images to train a custom model. So how did you get those curated images, first of all? Hopefully, your content architecture has helped you to be able to pull out that curated set. And once you have that curated set, you will want to be able to say, this was that curated set. So being able to account for the images that you use to train a specific model, and then what model, what is it that you trained, how do you reference that? And of course, these curated assets and the captions that were being used to train that. And of course, your captions can be informed by your metadata, can now be used to create more images that if you have upstream thought about the curation set, you've thought about the captioning, you've thought about capturing the model itself and the intent of that model, the context as to why, what personas was that model intended to serve? In terms of the imagery that it's generating. Now, you can, easily start to inform the metadata of the images that you generate from that model. So again, you have a custom model that is allowing you to generate backpacks in various settings or different prompts in different settings. Even you can use this to generate an image that you then do structure match on or Gen Fill or Gen Expand, or you do background removal, you do object compositing. There's so many tools that build upon this. But if you're at this stage and you don't have a strong content architecture that takes all of these tools into account, then you're probably not going to be able to take advantage of the experience and the tools. So here's an example of an assembled experience using Firefly services. Very straightforward. Let's say you already have a rendered digital twin. You've got some images from a photo shoot. You've got a background that was created by Firefly. You can utilize APIs to pull these together into a single assembled experience and harmonize that so that you can then utilize that in real-time. Because these APIs, the Photoshop APIs, can actually provide a nice experience for you when you're, for example, creating paid media that you need to post or an email that needs to go out. But there's lots of ways to think about this. There's lots of ways to flip through the elements of this, to be able to create variations in a very targeted way.

So let's talk through some use cases here and see how Adobe products infuse GenAI capabilities into the content supply chain.

One of my favorite drawings here. This is something that we did for a customer called Henkel. And this was presented in a session on Tuesday. So if you go back to the recordings afterwards, if you weren't in the session, this is the core of what we did. But it also is a broader view of the content supply chain. So again, going from brief, creating a project, bringing in creatives, bringing in copywriters, approving the experience, getting your assets into the DAM and using profile attributes, Journey Optimizer, being able to deliver that to a channel, pulling insights from Customer Journey Analytics.

But in this case, we wanted to do something a little different. So on the right-hand side, these emails were targeted at people who had gone into the store and actually had a real-time measurement being done on them, so that that measurement, those sets of criteria were now hydrated back into their profile, and they wanted to be able to create a one-on-one experience that acknowledges the effort that they went through. So that included a newsletter. It included emails that they wanted to keep fresh. It included product recommendations and included an actual postcard that was printed out and placed into the box when they ordered something. And of course, they wanted to get insights back from this. So the approach we took was we started with the creatives, we started with the photos, but then we enhanced those photos so that we could target specific profile attributes and be able to make sure that we have a set of content that could be utilized to assemble and experience, and of course, with marketing copy, there's variations for that as well. And then making sure we have the right content architecture so we can apply the right metadata so that Journey Optimizer knows how to pull together the right experience. And then we can measure that on the other side.

We started out by rethinking how we drove the content supply chain and the content lifecycle. The brand managers started out by taking the brand guidelines and that formed the creative space, I like to say. Basically, the guardrails within which all of the experiences could be created. What's the look and feel of the imagery and within the imagery, the lighting, the composition, clothing people were wearing? What are the types of voice that we could use, the typography that we're using to actually write that voice? Logos, components, proportions. And then creatives are creating the design system templates that are driving this whole experience. And some of the assets that are used as either reference or templates or baselines for this experience. And then marketers are now, because this creative space has been defined in the templates and design systems that have been defined, now, marketers can use Firefly services and automation to then drive the content that is actually going to be used in the experience so that customers get that one-to-one personalized experience.

So if you're thinking about it from that standpoint, and maybe you have 20 product images, you have 20 images of people, you have 20 location images, maybe 15 text variations, ten product recommendations, you start to see that if you start to multiply these n factorials, you have a large set of experiences that can be targeted to specific profile attributes or segments. So instead of going through the contents supply chain and the content lifecycle linearly, starting out with workflow and a brief and a project, and going to a creative and the review process, putting it into the DAM, adding it into the experience, now you're saying, wow, I want to create a multiplicity of experiences. And once I define that creative space, I'm free to play within this and see what we need to create in order to drive that level of experience, which then ends up with lots of different experiences for lots of different people. Not just segments. Being able to talk to people. You came into the store, invested your time to work with us, and we're going to reflect that back to you and build a relationship over time with our customers. This is the power of content architecture.

So again, just looking out at the whole set of solutions for Adobe, I just like to map it to some of the roles that might interact with these as well. Too often we might just show things as products and draw lines between them. But there's roles, there's people that are interacting with each one of these solutions in different capacities. And these teams, these roles, they're all depending on the central pillar of the content supply chain, which has assets and assets needs content architecture. But when you're developing your content architecture, you need to also know who's who in the zoo, and what they're doing and how they're interacting with their assets and the content architecture that you've created and continuously evolve this. This is all about learning. I mean, Firefly is a year old, but in GenAI, I think that's seven years. Or is that dog years? Everything is accelerating, and there's so many things that we need to be prepared to iterate and learn on, and improve over time so that we don't go six months or a year and do a review and get shocked by how things have changed. We have a plan for change. We are the change, and we're driving the change. That's how we should be thinking about it from content architecture, because without us, we won't be able to get into that future world of really contextualized experiences.

So takeaways. Content strategy is living. It's breathing. It's evolving. It's changing. It's the center of everything. If you listen to GenAI podcasts whatever you do for entertainment in your life, just think about it. If you look at Netflix and you're trying to figure out what to watch, all of those thumbnails are assembled experiences contextualized for you. You're not going to see the same thumbnails on somebody else's Netflix because Netflix knows who you are. They're going to put a female lead in one Netflix thumbnail versus a male lead. They might show a different aspect of it. Coloring. All of this is content architecture. I hope you are looking at the world as architecture.

Optimizing a content supply chain is not just a 'you' problem, it's an 'us' problem. So you should be an influencer. You should be informing and helping with the change management across the enterprise to be able to think about content strategy holistically. And you're the person in the driver's seat. So take your content strategy even further and really think about what could be possible in everything that you do, that what could be applied into your content strategy. It doesn't matter how simple you perceive your content supply chain. There are complexities that you probably haven't addressed or that you thought of in the shower one morning, but hasn't then applied to your content architecture.

Take the time to find the process that will take you forward into the future.

And of course, if you want to see more about any of the things that I did with the images or how we've been working with other customers in terms of applying metadata for generative AI experiences, of course, then go down to the booth. The one thing that I think is valuable at the booth that I didn't do here is you can actually walk through the whole process and start to calculate savings. I keep hammering that $10,000 number. But when I talk to customers and they say, you know what, I think Generative Expand is the killer app for Firefly services, because every time I apply it, I save 300 bucks, 600 bucks, 900 bucks. I save a week. Really diving into what the value is and understanding the value that you're driving can help you get more budget. It get an enhancement, an upgrade. Things going into your system, importance in meetings, So that you can drive the change that is needed. So, thank you for being here today.