[Music] [Elliot Sedegah] Welcome everybody to Accelerating Content Production with Firefly Services and Custom Models. My name is Elliot Sedegah. Thank you, sir.
And you can't do anything without a team here. I'll talk about this team, incredible team that we're talking about here in a moment. But we'll start off with a short video which we will hit here to play. Yeah.
Great. So thanks again to all of you who've been a part of this journey, fantastic journey we've been on in the last year. Just quick introduction. My name is Elliot Sedegah. I lead product marketing for our Creative Cloud for our enterprise customers as well as for Firefly for our business customers as well.
Today, we're going to talk about some incredible things that you and some of our partners are doing, with some of the innovative technology coming from some of our product team, so thank them as well. Good to have Ben Vanderberg here as well, who's technology evangelist extraordinaire. He's going to be showing lots of demos and capabilities that we have to show. So if you're looking for demos and trying to understand the value of what's coming out, Ben's going to showcase that. And we're going to be joined by Matt from The Clorox Company. So we're going to talk a lot about what The Clorox Company is doing. And when Matt talks briefly towards the end, you're going to see some really innovative ways that his team is leaning into GenAI. Lots of lessons I learned that you can take back into your organization. So how you can take this technology and be incredibly productive, personalizing at scale. So with that, since my...
Clicker is not working that well-- We all know that content is a fuel for customer experiences. It's effectively the way that you communicate to your customers, your different segments, understanding how they can connect to your brand, how they can be loyal to your brand, how they can be aware of what's there. Content is the key to that. In many cases, it's also the bottleneck to that process. And here comes Generative AI.
Many of you are forming into this room right now because it's a massive, massive opportunity for us to take advantage of the new technology so we can reach that potential of doing personalization at scale.
If you look at the-- So if you're looking more broadly, we know that mostly everybody probably-- You think about it across the board. We know this is something that's going to help us be productive. We took a survey last year, and we found that only less than a quarter have actually connected Generative AI to their larger business transformation initiatives. So that doesn't mean that you're not doing it. We know it's happening in small pockets within the organization, but how do we unlock that, unlock it more broadly? That's the key. How do we change the business? How do we change the people? How do we scale up? How do we take advantage of the new technology that's coming at a very, very rapid pace? We've seen the productivity gains. In some cases, we've seen a 12x productivity gain in individuals, especially the creative users that are creating content. It's amazing, amazing opportunity for us to push forward and innovate.
So...
It's been an exciting year of innovations. So our clickers here is having some issues. But we've seen a lot happen. When it comes into precision and control...
Putting more tools in the creators, in the marketers' hands to guide these models so we can get exactly the output that you need that's relevant to your brand.
We know many are using this technology as well to generate new ideas and new concepts to help your brand stand out even further. And we're also seeing interest of new modalities, audio, video, 3D, many other form factors coming in so you can express your brand in new ways.
So as we continue throughout this evolution, I'll explain to you a little bit more about how you're going to unlock this throughout our technology stack. So at the bottom here, you can see our Firefly generative models. These are all designed to be commercially safe. So you have the confidence to use those in your core content, whether it's going out to a different channel or during a campaign. You have the confidence to know that you can use that, say, for business. On top of that, we're going to talk about our customization layer. And the idea there is you have your own brand, your own elements that are there. That's your unique special sauce to express your brand. And how you can bring your own content and train on Firefly so you can get that output.
And then on top of that is where our applications come into play. So many cases, we have a lot of content coming out from the platform. You're going to use the tools and technology like Frame.io to review and approve that. And then those get unlocked directly in the workflows that you use, whether you're creative team using Creative Cloud for Enterprise, your operations and marketing operations team using Firefly for services. We're going to talk a lot about that today as well. Or using Express or GenStudio for Performance Marketers. All the different marketing teams out here as well who need to activate things on social channels to advertising platforms as well. We need to make sure that we're all leveraging the same technology cohesively. Again, this is connecting it to that broader digital transformation initiative.
So we'll talk about a few concepts. And then the first concept I want to land here is, you know that creativity and your brand matter more than ever before. Everyone's getting the technology of Generative AI. What's going to distinguish you and set you apart is how well you can use that technology to stand out.
So with that, we're going to think about how we can creatively leverage the power of Generative AI to stand out.
And one of the key technologies that we've introduced for this, and some of you have been along with this family...
Part of this innovation as we unveiled this last year, is Firefly custom models. Now this ability for you to take your own brand IP, your brand assets, content for different campaigns, whether it's for your brand styles or seasons, bring that into the model, train it uniquely in a safe area, and then unlock that for your creative teams or your marketing teams to use and generate everything on style...
Across your different teams.
So with that...
What's better than a demo? So we're going to introduce Ben Vanderberg here. He's actually going to walk us through some of the innovations here with Firefly custom models, and we can see how that can be unlocked to create personalized content at scale. Ben. [Ben Vanderberg] All right. Thank you very much. So how many people in the room here have used Adobe Firefly before? All right. So my hope is, for everyone who did not raise your hand, that by the end of this, you're going to firefly.adobe.com and checking it out yourself. But what we're going to do here is we're going to see what it looks like to create a custom model in Adobe Firefly. So I've gone to firefly.adobe.com, and I can go in here to custom models. And this is where you can see-- Oh, thank you very much for-- We're going to switch to number four. There we go. So here we can see, I have a variety of different models that I've created. Firefly custom models is great for things like your brand styles, like we can see for CitiLights here. It's great for things like characters, if you're using that for, say, like your mascots and those types of generations. We have those, like the kids club and the alien here as well as for some of the other types of visual content that we have here. Now I have a visual that we have with the CitiLights here. Let's go ahead and what it looks like to train this. So if I go in here, I'm going to call this CitiLights Demo. We can give it a description. And we have a couple different training modes. So we can train on a style. And if you imagine, if you have a number of different images that are existing of your style, you can upload those into there, and it will train a model based on that. The other option is what we call a Subject. So if you have a very specific subject, like a character or a certain item that you want to be able to train that on and reference, I specifically want this generated in this image, you can train on a subject. So I'm going to choose, in this case, a Style, and we're going to save this to a project that we have called CitiSignal, and we're going to click on Continue here. And at this point in time-- At this point in time, hopefully we don't have too much of a base here. At this point in time, we can then take some of our different images and we can drop them right into here. So you'll notice that I have a number of different images that are uploading into here. I'm just going to have a look here. So just imagine this as our brand style. We have the general color of the different lights in here. And we want to create new content based on this style. Now what happens is, you upload 15 to 30 images up into here, and it's learning from that. Then what it does is it uses AI to try to create different captions to describe, to tell the AI what's actually in this image. So that it knows that and you use your own terms to help it understand how to create that content. So for example here, it did a pretty good job on the neon lights description here, but I'm going to say, "Triangle streams of light...
With a reflection." You don't have to be extremely specific in that actual caption. You can keep it relatively simple in that caption. And you would go through and you would review every single one of those different captions just to make sure that that description is right in there. And then what you can also do here is we can go in and give it a sample prompt. So we're going to say...
"Streams of light...
With city skyline in the background." And that will be used just as a reference for us to be able to generate. So when we click on Train, that's going to take all those different images. You do have to make sure that you are properly licensed for every single one of those images that you're uploading onto there. And that will take just a little bit of time for it to actually process that. Now we don't have all that time while we're here, so I'm going to jump ahead and show you an example model that I have already trained within here. So if we go in here, I'm going to have a look at one of the models we created before here. And before it gets published out to anybody, we can preview and test that to see what that is actually going to look like. So in this example...
I did neon streaks of light with Seattle skyline in the background. And here we can see it took the elements of that style, took my prompt that I had of Seattle, and it generated these different variations in here. So I can then go in and say instead we're going to go with Los Angeles and click on Generate. And it'll generate variations. Now as we're generating this...
One of the biggest complications that people have when generating content is, how am I going to write this really, really, really long prompt? Well, I don't have to think about it in the concept of a really, really long prompt because actually directly within here, we make it easy by giving you references. So you can upload a reference image if you're trying to say, "I want it in the shape of this." Or, "I want it in a certain style." All of these are things that can help you inform your actual stuff that you generate. Even things like camera angle, whether it's shot from above, wide angle, the color, the tone. These are all things that you can then use to reference within that. Now another way that you can also approach this is also within tools like Adobe Express. How many people use Adobe Express? See, once again, everybody needs to go to express.adobe.com and check that out after this. So within here, what we can do is we can actually use this in a practical sense in the context of something that I might be working on and designing here. So I can click on Generate Image. And this allows me to be able to then adjust my prompt in here. So once again, I'm going to use Seattle as my reference in here.
Click on Generate. And that will allow me directly in the context of here generate my actual image. So here we can see just from before here, this was what we generated. It still keeps the vibe and the style of my actual light streams in there. And we'll see that sure enough, it brings some of that style into here as well. So making it really easy for me to create a bunch of different variations of that. Another example is, we actually have an example with food here. So maybe I'm at a grocery store that needs to quickly show some of the different things on the menu here. Instead of it being something that I would have to go and figure out how to capture that, I can go in here and change this to, let's say, steak with-- We're going to go with something really, really random just because you know I won't have this as an actual image in the model. So with blueberries and-- Give me something else. Something else. [Man] Purple cauliflower. Purple cauliflower.
Is that right? Did I-- All right.
And once again, you can use that as the reference for you to be able to then generate here. But this is the same concept that you can then apply that directly in your actual designs to be able to place that into there. So give it a quick second, and let's see, sure enough, steak with blueberries and purple cauliflower. You know what? I think we could use a little bit of a drum roll. Can we get a little bit of a drum roll? This better be a really good looking dinner.
First time you ever use the model for the day, it does take a little bit of a second for it to load. It's not every time you generate this image. It's the first time. It's cold opening for today. All right. So wait. Wait. Wait for it. Wait for it. All right. One more drum roll.
Little louder. Little louder. Little louder. There we go. And sure enough, we have our steak with our purple cauliflower and blueberries in there. But still, it keeps that style of the slate in the background, the plate looking from above. It's keeping that to give me that consistency. I'm not sure if purple cauliflower is the most photogenic food in the world but we did get it. Sure enough there. And so with that, that is Firefly custom models making it easy for me to take my existing images and then be able to train custom models to be able to create new fresh content based on that. There you go, Elliot.
Let's give Ben a round of applause on that one. Purple cauliflower. Thanks for that one. So if you think about the power there, you effectively took one of your most creative people and then take the different styles that you have. Many of you probably have tens of thousands of assets, maybe in your DAMs, for example, that you have for over the different campaigns, different seat styles, different types of brands that you have across different regions. You can train each and every one of those elements and then actually make that available to your creative team, to your marketing team. And everybody's singing off the same sheet of paper when you're talking about the spring campaign, for example, or Christmas campaign, or maybe you're doing something for the Super Bowl. Oh, we're just announcing today as well, we're making it available in GenStudio for Performance Marketers as well. So you'll start to see custom models appear across the Adobe ecosystem. So you can train them centrally and then use them across the different platforms.
Okay. Of course, it's not working, but we'll-- Oh, it's working now. Wow. And this is part of a spectrum of customization controls that you have, that you get down to your creative user from-- Many of you probably use Generative Fill and Generative Expand, compositing different objects, Style Kits, for example, sharing different styles across your organization, and Custom Models is the peak of that as well. So you're meant to use these in combination together, not in isolation. Generative AI, there is no magic button to get your unique style out there. We're giving you as many tools and controls that you need to express your unique brand as possible, so you can be creative and have content that really impactful.
Okay. Well, still, this thing has a mind of its own. Sometimes it wants to work, sometimes it doesn't. So I'll just hang out over here.
So happy to announce we've actually crossed 20 billion assets generated from Firefly.
It's like an amazing time to pass over this where not too long ago, some of you were just experimenting with this. And the beauty of this is I'm highly confident a lot of this content is actually being used in production in different campaigns because we are taking a lot of effort to make sure we have commercially safe models across the board. So we didn't stop there on imagery. The next step was to move much towards broader multimodal experiences. So imagery, vectors, design, video, 3D, audio, all of those we're putting into our Firefly application. And what sticks out to me most is video. And the ability to take an image and turn that and animate that into a new creation is almost like a superpower if you will. This is something that was reserved to only maybe a smaller group of creative teams. Now this power is actually extended out to anyone who wants to be creative. We're actually giving you those tools to use right away. So this is an example of what that image to video technology looks like. Again, if you haven't tried this, firefly.adobe.com, you can go there and experiment with this as well. This iguana that you see here is actually on our web page. We use this on our own marketing content. We're on the marketing team. We're often begging the creative teams on the on the websites, like, "Hey, we need some more video. We need some more animations." And they look back at it and be like, "Where's the budget? Where's the time? We don't have the ability to do that." Now we're starting to see much more animations in here. Even this presentation, you're going to see a lot more videos and animations because we're empowered to do that. We're moving into a new generation of this. So to actually see this a little bit closer, we're going to have Ben showcase what you can do with this technology. So, Ben, over to you. All right. So to start off here, I am going to be working on an ad that I want to create for, say TikTok or Facebook and so forth. And if I have a look at this, we have two different visuals I created. I'm in the Adobe Express right now. And something simple here, we have our gradient in the background, we have, say, some headphones, and we have a logo here. But what I really wanted to do is I want to add a little bit of zazz, little bit of animation to make this a little bit more attractive as we're going through this. So what I can do here is I went here and I exported these out as the individual images that we have here from Adobe Express. And I can go now into Firefly, scroll down here into text to video, and within here, what I can then do is I can then take my actual images-- Sorry. Give me one second here.
We can actually take our images, and I uploaded them into here. So if we have a look at what actually got generated from this, we can see that the video here, this is headphones have sound waves coming out of them and then spins and morphs into a green icon. So let's have a quick look.
There we go. We could see the animation there, brings that all together, and then morphs into my actual logo there. So all I did with this is I uploaded two images and put in a prompt. I put in that image that I mentioned before, which was the image of the headphones. I uploaded the end frame that I want for this video, which is with this logo, brought that together inside of here. It will generate from this first frame to the last frame. Use that prompt to generate in between, and that's what it generated from that. So then when I download this, the great thing about this, makes it really easy for me to then go into something like Adobe Express to add, say, my titles that I want on the top and the bottom here. I can then take that. I can generate that animation. And then directly from even within here, I can download that and I have my video that now I can then take that, use that with TikTok, with Facebook, and so forth. And it's quick and easy animation on that. But let me show you one other quick other example. So this video right here is a video that came from Adobe Stock. So this is one that-- It looks like we got a little bit of a beach ball. Hold on a moment. It is an Adobe Stock video that is essentially these guys in the kitchen cooking. But what I really want to do is I want to augment this a little bit more. I want to add a little bit more of something else to it. So to help me with this...
What I can do is if I want to expand this frame, I don't have the opportunity to go back because this is a Stock image. I don't have an opportunity to go and say, "Go reshoot this," because I don't have the original. But inside of Photoshop, I can actually take this image and I can actually go in and I can use what's called Generative Expand to then, if I want to add a little bit more space to this...
I can actually go in here...
And I can say Generative Expand.
And this is going to expand that video based on that still frame. So if you imagine if it is a camera that is not moving at all, we can actually expand that out and it'll fill in the area around that actual image there. So if we have then a look, we will see-- I'll jump to this part. We'll see that we had the part that it generated of the actual image here. Then what I can do is maybe I wanted to also add another element which is, say, like this, this carrot.
Yes. Because maybe this carrot is my mascot. And so I want to take this, I want to drop this carrot into here, and not only want it to drop into here, I want it to dance.
So the way that I can do that-- Sorry. Give me one second here. Let's try this again. So the way that I can do that-- All right. Little bit of a Photoshop challenge, so let's try this one more time.
So what we could do here is then, inside of Firefly video, we, once again, can take our reference image of our carrot. I put it in the background of that actual video. I then told it a carrot dancing on a counter.
Here we go. Sure enough, we have a carrot dancing on a counter. And then I actually blended the things together into a final video, which is now a dancing carrot along with our original video, which was actually a Stock image. So I can actually take my existing content. I can expand that out using things like Generative Extend inside of Photoshop and use things like Firefly video to create content. Now I am not a video effects guy. But the fact that this helps me be able to quickly visualize and create some of that content is a huge time saver because it empowers me to create stuff that I never would have been able to create before. And so that's really Firefly video, making it easy for me to take my content, my assets, be able to generate new video based on that, based on text prompt, based on my images, and it's super powerful. Great. So thanks, Ben, for that demo as well.
So we're going to move on and think about all that amazing technology and what has that given you. And the key thing that we're thinking about here is speed. Speed to content creation, speed to market. If we're reducing time spending on some of those mundane tasks, that frees up time for us to go after new opportunities. And many of us have seen something like this before where we're stuck with this doom loop, if you will, of all this content that you need to create for all the different variations and all the different channels, many different languages, many different regions, all the different variants that are there. This multiplies very, very quickly. And so in many cases, you probably just have one core asset that you're going to use. And the idea of even asking that creative team to create all these variations is something that you don't even bring up because you know your budget doesn't support it and the time doesn't support it as well.
This is one of the reasons why we invested very heavily to taking the power of Generative AI and our creative automation technology and making those available as APIs. So one of the key things we wanted to do is we know that a lot of our creative teams are spending a lot of the time doing a lot of repetitive production work. So think about some of your most valuable creative teams working with you, from your internal agencies or your broader agencies as well, and they're spending time doing cutting, cropping out assets, for example, or generating something new for a new campaign. Or they won't even have the ability to change, create something that you're going to A/B test because you don't know what content is going to work, for example. So the idea is taking back power of that, building those core workflows so that you can automate and scale production across the team. So what are we seeing since we released this? What are our most popular use cases? So this sort of a little bit of a framework that you can use to bring back into your organization of areas that you can use to automate.
By far, refreshing campaigns very quickly. You saw exactly how we're using them. For example, we do annual conferences. So our next one is coming up with MAX London. We know we're going to do this every year. It's going to have a new style. Thousands and thousands of assets that need to be created. New styles and new content that we need to expand out to new regions because we're growing.
We're tapping into that to refresh campaigns quickly. We're seeing this happen at a seasonal rate. Some people are doing things weekly. Some are running different social campaigns every other day as well. So you can think about different ways you can automate and engage.
Localizing videos and personalizing assets and videos and content as well. My favorite one is actually on down to the bottom as well. Gatorade. Look at those Gatorade bottles that you see there. They're making unique bottles that are tapping into the user designated and saying, "Hey, what kind of bottle do I want to create?" A real physical bottle, and it's generated using Generative AI. So you can start to think about these personalized experiences of new opportunities to grow, and this is where the real opportunity is with Firefly Services. So with that, we have to show a demo. So this will be our third segment of a demo. Ben's going to showcase how you can tap into Firefly Services. And, Ben, over to you. All right. So to start off here, I'm going to go over here. And a lot of times, there's a variety of different type of content that you need to produce different variations for, whether it is web-based banners that you need to produce or even print-based content. And that localization can be based on the language, could be based on the specific region, it could be based on the specific products. And every one of those becomes a multiplication factor of how many different variations. Now as a creative, going through and manually having to key that stuff in is very, very, very tedious. And that's where Firefly Services can help me quite a bit. So let's walk through two examples that I have here. So the first one is, let's say we have these different posters that we want to use for, say like a retail store, and we want to change out things like the discounts, we want to change out the language, and even the products that are promoted based on data. So what we can do is inside of tools like Adobe InDesign, we can go ahead here and we can actually set up tags for what we call mail merge. So we can say, this is where we want a specific bit of text to go. This is where specific images are and where we want to take them. And we can create a template in the tools that my creative teams are using every single day. But then, what we could do is we can use Firefly Services to help streamline that process. So I can take all of my different templates, and this is an app that I have here. This is internal app, and this is using our Firefly Services APIs to run this automation. So what this is doing, it's taking the templates that I designed inside of InDesign. It's going to take some of my different linked images and assets. And then if you see inside of here, we have different data for different disclaimers, different locations, different languages. So we have German here. We have Spanish. We have different discounts. All of these are different variations that I otherwise would have had to rekey in manually. But when I run this, this is now taking all the information from my data. Now merging that into those templates, getting the right images, getting the right visuals, bringing those altogether one by one, creating all those variations, we could create thousands and thousands and thousands. And what we get then as results is two things. We could get PDF files, that could be print-ready PDF files I could send directly to the printer. Fantastic! But I also have InDesign files. And why is this important? Because your creatives might go, "Hey, you know I love this automation stuff, but sometimes there's last mile edits that we still want to make." And so we still can have InDesign files that can give you that full control. But then as we look at the actual outputs, sure enough, here we go. It gives us all those different variations.
Looks like I have a slight font issue on that one, but we have all the different variations with different products, and I was able to create all those variations at scale. Now this isn't just with Photoshop either, because you can also do this with things like-- This isn't with the InDesign Sorry. You can do this with Photoshop as well. So you can take your Photoshop files...
And you can create templates in Photoshop to dynamically change out different images, different text, turn on and turn off different layers, use Firefly to use generative capabilities inside of there, run actions that you have as part of those. So then, as you create all those different variations that have, again, that's masking, generating, replacing elements, replacing text, creating different variations and sizes, it can then, once again, create all those different variations and generate even dynamic backgrounds based on that with different text and different variations from that. So this is something that you can take something that would have been manual and be able to automate many of those different actions.
That's really the power that you can do with the Photoshop and the InDesign APIs within that. You guys want to see one more thing on this? Yes? All right. Cool. So one other thing that we've also done is brought things like our video APIs into this as well. So if I go in here and have a look at this video that we have.
Let's have a quick listen. [Jess] Hi, there. I'm Jess, a Lead Designer here at Star Designs. We are so excited to announce we've got our three latest clutch bags in pink, green and mustard yellow for every occasion coming to you next week on our website. All right. So if we take that now if we wanted to use that, that would be perfectly fine. But we might want to expand our audience into different various languages, which again, might mean that I have to reshoot that with different people as voiceovers. Or you might have subtitles or something that is like a really bad overdubbed movie, with some lack of lip syncing there. But what we can then do is with translate and lip sync, which is available both through Firefly as well as through our APIs, we can actually translate this. How many Spanish speakers? All right. Let's have a listen. [SPEAKING IN SPANISH] Sound pretty good? Yeah. Yeah. All right. But notice that it also matched the actual lip sync of that as well. So then we can also do that in another language. We'll do-- Let's do Korean.
[SPEAKING IN KOREAN] Or let's do-- We'll do German.
[SPEAKING IN GERMAN] So helping scale that ability and be able to create those variations gives a whole lot larger audience to some of your video-based content. And as you saw from the Keynote, also doing things like reframing and stuff now available to really help you scale content variations across your different sorts of mediums and channels. That's really what Firefly Services is all about, being able to help take those different variations and be able to scale those. Now with that...
Just to give you an idea of some of the different things that are within this. We've taken many of our capabilities from Firefly, from Photoshop, Lightroom, Substance 3D as well as InDesign, and some of our other AI capabilities, and brought them directly into those as APIs so that you can take those, bring those together to create your automated workflows for your content creation. So your creatives can create the greatest templates and focus on the design and creating those variations and that you would have never been able to do before. You can now do that at scale and in the formats and tools that people use to create content with Firefly Services. So with that, I'm going to hand it back to Elliot. Wow. That was amazing.
Every time I see those demos coming from Ben, you see the possibilities there of personalizing at scale. At first, you're thinking, "Oh, yeah, I don't think I can do that. What about the videos? What about these different languages, different regions, different devices?" It is now possible. The tools are there. This is your list of ingredients. The team have been making different recipes, if you will, that you can take and leverage within your organization. And when you think about it, there's two things that-- When we're thinking about from Firefly Services that we announced. So these new audio and video APIs we'll be able to do at scale. So previously, it was just imagery, and now we're moving to these different formats and creative production. We're creating new interfaces on top of that. So if you don't have the API resources or development resources, you have those creative production teams be able to stitch together those workflows for you. So that's something that we're actively working on as well.
So with that, I'm going to continue this conversation, but we're going to get very interesting now.
So I'm going to introduce Matt Harker from The Clorox Company. So Matt is leading customer experience orchestration for the broader teams, working across the board with many, many different powerful brands that maybe you use probably every single day. So, Matt, over to you. [Matt Harker] Thanks.
Thanks, everybody. Hi. Yes. I'm Matt Harker, and thanks to the Adobe team for letting me join you at Summit. I lead an area at Clorox called Consumer Experience Transformation, and we're basically a center team in service of all of our different views and brands, focusing on things like media, commerce, digital shelf, technology that powers all of that. So it's a really great and fun intersection of where all those come together. And as you can imagine, there's a constant demand for assets and content and variation in media on digital shelf, and especially to meet our ambition of personalization and as close to one to one as you can get. So I sit with that challenge a lot. And so we're exploring a number of use cases that can be powered by GenAI with partners like Adobe. And this is just some of the content that was in market this year that was created harnessing the power of GenAI. But today, what I wanted to do was share five thoughts, five perspectives about GenAI that we're working with, wrestling with, fumbling with at The Clorox Company. I know you all are in different industries in a different stages of the journey, so some may be obvious. Some may trigger a new thought. Wasn't quite sure what the makeup would be, but we did decide to do it in a fun way to show off the power of Firefly. So we're going to go-- I'm going to give five thoughts, but we're going to follow a repeating pattern here. And that pattern is I'm going to show 15-second ad, no GenAI, for one of our brands in the Clorox family. I then made that ad available to the Firefly team who chose a few still images from that 15-second ad and made a whimsical Firefly transition for me to then make my point that I can share with you. So through this, I hope you'll meet five beloved brands in the Clorox family, and you'll get to see the power of taking a still or two from that, adding a simple guiding prompt, what you can do, because it's pretty incredible. So with that, we'll get started. You want to kick it off? We'll get kicked off with Fresh Step.
[Woman] Every cat acts like their...don't stink. And now with our strongest litter ever, it actually doesn't.
Get 30-day odor control. Step it up to new Fresh Step Heavy Duty. Okay. So that's a 15-second ad that could find itself into many places, CTV, online, etcetera. Pitch that over to the Firefly team, and what did they do with text to video? So I show you this because it's fun. These get increasingly funny. I especially love the last one. But, again, you're like, "Where do I start?" And you may already have the assets, either in video still or otherwise form to just jump into this space. Right? So the first message is jump in. If you're not trying GenAI, get in the space.
You don't have to overthink it. You just have to get started. And I would say now this room is probably largely bought in. So again, I'm saying this more so you can take this message back to your bigger companies and enterprises, but it takes time to just soak in the space and get a sense of how it works. What are the terms? What's the language? How do I think about creating a good prompt? And GenAI is going to start to show up everywhere in company processes. So at Clorox, one thing we've done is we've created teams' channels where certain user groups are given access to different tool sets, and they can just go in and play. They can just share. Sometimes people are like, "Oh, there's a new holiday coming up. Look at this new ad I created for this." Or, "Oh, my son had a soccer game this weekend. I made him this thing." And it's just a free fun sharing thing internally, but it gets everybody excited about the power of possibility. And more importantly, it socializes GenAI in the company so people aren't afraid of it. They understand it's something that empowers people to do things quicker or faster, they couldn't have done otherwise. So definitely jump in if you're not in it. Don't wait. Don't wait for it to get mature. Don't wait for it to get the best because it is rapidly moving, and it will overtake you even if you start now. So I'd say jump in would be one of the first lessons. All right. Next brand, 15-second ad.
[Woman] Burt's Bees moisturizes with a tingle of peppermint oil that feels like if an Arctic mountain whispered peppermint secrets to your lips. [Woman] ASMR-style. [Woman] Yeah. Feels kind of like that. Burt's Bees, a Touch of Wonder. All right. And then over to the Firefly team.
And now the messages. So inside, outside. So I personally divide the world into what's happening in your company that a consumer or customer will never see and what is going to go out in the world that your consumers and customers are going to see. So first, you should think about risk tolerance for your company and establish some guardrails. When I say jump in, I don't mean jump in recklessly and get everybody using every third-party AI tool and create a disaster for your brand. So you need to have a little bit of discipline. Understand your risk tolerance. It may be the same for your company, or it may be different brand by brand. Then inside, don't worry about having it all perfect or approved by your brand suitability, legal process, etcetera. Just go. You're practicing speed internally. As an example, what if you have a marketing leader, an innovation leader, a product leader who has an idea and trying to socialize it or galvanize people behind them? Use GenAI to bring that picture of life and get everybody on the same page. It doesn't matter if it's perfect. You can run it 1,000 times. You can use it later as the beginnings of the storyboard. You can brief your agency with it. You don't have to be so precious with it. So don't be so careful. Kind of run, break things. See how fast you can go. See what used to take 4 weeks and get it down to 10 minutes. So that's the inside view. Now outside is very different. You're thinking about brand outcomes that you want out there and protecting your brand and everything you've built behind it. So there, you do need to be more careful. Use commercially safe tools and IP like Firefly where you can feel confident in what you're putting out in the world. Make sure you're using your internal gate processes. We are not necessarily doing one person looking at every asset going forward because it becomes overwhelming, but you're grading those assets based on what they are, what's risky, where is it going to show up, how many people are going to see it, and then you can take a risk-adjusted view. So definitely inside, you act one way with a little bit of reckless abandon, I would even say is like poke yourself to do that, and then outside, treat it very differently. So just another perspective. All right. Third brand.
[Man] All right. What are the rules of the tailgate? - Burgers. - [All] Flip once. - Ribs. - [All] Low and low. And? [All] The center is always the hardest. You guys. [Man] Kingsford, King of the Tailgate. And then over to the Firefly team.
All right. Third message, rapid mining. So I think what's really interesting is there is data everywhere, and of course, it's proliferating. Some of it you may have access to. Some of it may be trapped behind walled gardens. But to make sense of any data previously, you had to get it all in the same location. You had to normalize it. You had to get it in some structure, and then you had to spend time analyzing it. Now a lot of these tools can work across data where it sits. So it could be data on your own estate. It could be data with a partner. It may be in different formats. These tools are intelligent enough to make sense of things, even if you don't spend years bringing all that together and polishing it all.
The next is that you can think about things-- Something we do at Clorox is we scrape customer reviews that are left on all kinds of sites or unstructured data left on social sites, and we bring it together to quickly learn what are people talking about, what are their pain points with our products, what do they love about our products. It goes right into our innovation road map. What should we focus on? What should be a line extension? What should be something new? It comes direct from customer data. Normally, that would have been difficult to access all of it. Then you would have had to have teams spend time combing through it, professionals that know how to boil up insights and then share them across. Now the tool that we have in place is a self-service tool that sits across all that data that by no matter what facet you want to adjust for, brand, customer unmet need, whatever, it instantly surfaces and generates those insights for you. Now again, it's not meant to replace insights teams. They still have decades of practice, and they know how to help people understand how to apply the insights. But the drudgery of bringing it together and combing through everything and trying to simplify that down from somebody who sits in a different department, it's a lot of time and effort and a lot of back and forth. So some of these Gen AI tools now immediately generate the "so what" when they sit on top of data. So it isn't so much about creating copy or images that consumers will see, but generating actionable intelligence and insights from large datasets. All right. Fourth brand.
All right. And Firefly team.
All right, fourth message. Agents, which we've heard about a lot, going to fundamentally reshape experience, whether it's an agent working on its own in a certain space or a bunch of agents working in concert with each other. So when agents are working, and I'm saying this a little bit more from-- Some of our leading retail partners are creating shopping agents. And when those shopping agents act on a consumer's behalf, as a brand, we have to understand how those agents are going to work. What are they looking for? When you ask a shopping agent, put something in my cart to clean up afterwards, how does that agent know what to go look for? I mean, we're essentially then marketing to bots. We may need to create totally different content or 100 times the content in order to be ranked in relevance. Or the words we need-- It's almost like an SEO game, but to an agent and a shopping agent. So agents are here not only in workflows and for speed and efficiency, but also in how consumers are going to live their daily life. And as a consumer brand or product group, we need to understand how to get the best out of those so we can help consumers find the brands they love. The other thing with agents is rethink end-to-end. I think one of the dangers in AI right now that we found is that when you start about, you say, "How do we do things today? And who are the people involved? And what are all the processes?" And then you can break that down and, like, "There's this step, then there's this step, then there's this step." And then within any step, you're like, "What tools do I use to do it?" And if you ask yourself, "How do I use GenAI to make that step faster?" you're missing it. You have to start complete end-to-end. What's the starting point? What's the end goal? Because these things are so disruptive, those eight steps may completely go away. You may have two totally different steps and have to retrain your teams.
We're starting a thing right now, content supply chain. Again, a bit of a buzzword, but an important step for us to map all the places we create content today and how, some in-house, some with agencies, some with influencers, etcetera, all the places that our content lands, media, websites, shopping engines, etcetera, and understand those value streams. Some of them take a lot of time. Some of them have a lot of asset variations going through them, etcetera. Once you have that map and you can look end-to-end, you can think about what tool can get you the most expedient way end-to-end, and then help your teams work totally different with those tools and agents doing the drudgery work. We would not have got there had we just started to talk to one department and say, "What do you do today? Can GenAI make your job go faster?" Whatever, you would miss the real value unlock. So our big discovery there is always start end-to-end, understand how disruptive you can make it, and then help teams adjust to that because, hopefully, their job's even more exciting, enriching, whatever, but it's different than potentially what they were doing today. And then the rise of synthetic, again, that's back to the marketing with bots. There's so many things that are just going to be-- You just have to think different. I think we also see it in consumer testing and innovation. Previously, when you need to get together focus groups and in-person things, a lot of this now, you can basically have synthetic consumer groups react and give you reactions very quick to what's working, what's not, does it meet their pain points. So just totally different ways of working.
And so that's the big thing. You got to think really big first to get the value out of it. So okay, our final one here. [Man] Hey, man. Ever since you tried Zesty Hidden Valley Ranch, I feel like you've been a bigger fan of that than the closures. [Man] How could you say that? - Well, it's fun. - [Woman] Hidden Valley Ranch. [All] Only Serious About Flavor! All right. And over to the Firefly team.
All right. Final message. Pretty amazing. Right? I didn't have to do anything other than provide an ad that already exists. And they're like, "Yep. We'll take this image, this image, we'll give it a guiding thing," and then can create this whimsical stuff, which is just awesome.
But so last message, individual made possible. We talk a lot about personalization, individualization, people have been talking about forever, easy to say, super hard to pull off, getting more and more possible. All these barriers are coming down. The data to understand and the triggers to know what consumers want when and be able to serve exactly that to them. There were so many problems before. All of those barriers are falling away. We're finding it most frequently about the variation of assets, this atomized content approach when you work with templates and bits and elements and they're in a DAM. How it can all get assembled much quicker, including video now with some of these products is going to just supercharge it. And we've done tests already that when we add in a personalization layer, either to better target an audience or to better serve that audience content, every time we get a statistical lift. So we're pretty sure about personal relevance matters. It's just a matter of how often can you do it and how much is it going to cost you in time and money. And so these tools are just totally shaking it up. So I would just say, don't abandon that Holy Grail idea because it's becoming more and more possible with some of the new tools. So with that, we'll turn it back over to Elliot and see what the future holds. Great. So Matt, thank you so much for that.
We're having way too much fun with this video model as you can see.
But Matt shared some incredible insights on things that you could do today. You could bring back to your organization. They're doing that now. And how many of you knew The Clorox Company has so many incredible brands that were there? For me, I just pretty much knew the Clorox, the wipes, cleaning, things of that nature, but they're pretty much affecting many different facets of your life in a daily basis. And the way they're taking technology and their organization together, there's a lot you can learn from that there. So, Matt, thank you for that. And also, thank you for letting us play with your brand a bit and have a lot of fun with that. Great.
So let's wrap up here. A couple of things that we can do to think about how we can unlock Generative AI within our organization. So I think the one key message that's coming across today is get started quickly. Start to tailor your brand. Start to look at the core assets that you may already have. How can you leverage that? How can you leverage that to jump-start creative ideation? How can you use that for new campaigns, existing campaigns to bring a refreshed way to leverage your brand? Start to think more broadly about how you can use that to boost creative productivity as well. So if you think about all the different people who are using, your core creative teams to your marketing teams at the end, start to blur the lines a bit about who creates content. So you need to make the technology available and also enable them to use it productively at scale. And then, of course, automating what's in bulk production.
Many cases, you have these recipes that you've been running for years and doing over and over again. And trust me, when you talk to those creative teams who are going through that drudgery of that process of all the resizing and the different campaigns that they need to do, if they want to get towards that creative, new, impactful work, this is the key. Take that, canonize that into core workflows, automate those through APIs, and then truly get it out to scale. So one other thing, where do we go from here is you pull out your phone and you can take a survey to win. So we're going to wrap up here on this one. So thanks everybody for joining us today. This wraps up our session for today.
If you do have the Summit mobile app, please do fill out the survey here. Your feedback is incredibly valuable to us, and it's even more valuable if you give us a five on everything. That's the only rule that I have with that one. And we will be taking questions but we have to take them outside, so for that. So thanks everybody. Yeah. We'll be hanging out outside if you have any questions. - We'll be out there. - Right. So thanks to the presenters today, and I'll see you all later on.
[Music]
