The future of commerce (Gen AI and 3D)

Transcript

All right, we are back and ready for Eroka O’Connor talking about the future of commerce. Take us away. Hi, everybody. Thanks for joining us today. I’ve been asked to talk about the use of artificial intelligence in Adobe Commerce and what we’re doing today and where we’re going with it in the future. So to start off with this presentation, what I did was I actually went and popped into Google Bud. Can you give me an agenda of using artificial intelligence in commerce? It gave me a bit of an agenda. I put that together. It gave me an outline for the presentation. I then took that outline, threw it into ChatGPT and asked it for the VBA script for the designs for my slides, which is what we’re going to be using today as we go through. We’ve got a long history of artificial intelligence within Adobe and we are one of the market leaders in using artificial intelligence. And that’s coming down into commerce as well. The three main areas and the benefits of using AI in commerce are around improving that decision making, getting recommendations from analytics packages on what merchandisers should be focusing on, enhancing that customer experience by being able to deliver personalization at scale without having to build a rule for every single customer. You know, there’s three point five billion customers out there. You don’t want to have to create rules for every one of them. And the other benefit of using AI in commerce is around increasing efficiency. Marketers and merchandisers can move to focusing on campaigns and features that differentiate their business as opposed to that day to day management of the categories and the content and the product and such like. The Adobe Sensei is Adobe’s AI tool. It’s our core technology that’s used in every product across the creative, the document and the experience clouds. So before we begin and jump into commerce specifically, I wanted to define the different AI capabilities that Adobe is bringing to the experience cloud applications. Firstly, we firstly by natively integrating AI features that work specifically with experience cloud application. These are features like anomaly detection in Adobe Analytics or personalized recommendations in commerce. These machine learning AI tools, we’ve had them within Adobe Experience Cloud for several years now. The second area is Adobe Sensei’s generative AI. It’s a co-pilot for customer experience teams to use across the experience cloud applications from a variety of use cases like asset creations or personalization across the customer journey. Customers will be able to move seamlessly between Sensei Gen AI services and existing features right within the workflows they use every day. And then lastly is Adobe Firefly, Adobe’s new family of creative generative AI models that generate visual content. We feel that these are the things that really set Firefly apart from other image generation services. It’s designed to be commercially safe. It’s not trained by scraping the internet for data. It’s trained on our contractually agreed Adobe stock assets, openly licensed content and public domain content. And we’re actually compensating our content contributors as well. Firefly is ethically minded and it’s designed to be creator and artist friendly. For example, you can’t request an image in another artist’s style as other image generating platforms will allow. And from an ethical perspective, anything we’re doing is using our content authenticity initiatives provenance technology to bring more transparency to digital content through content credentials, which allows creators to attach important information to a piece of content like their name, the date they created it, what tools they use to create it or if it was generatively created. And that information travels with the content wherever it goes. So to get more specific, I’m going to dive into commerce now. So what are the actual tools in commerce and what are they designed to do? The tools in commerce are designed to identify the right customers to target based upon their needs and their intent. For example, we use AI to understand customer behavior and improve segmentation and targeting for better personalization. This way you can see the likely you can use the likelihood of someone’s someone’s likelihood of purchasing a product or their likelihood to cancel subscription. It’s used to engage the customers on the right channel at the right time. And we’re using it to deliver the right experience for each customer, whether it’s the right offer or the perfectly composed email or the subject line with the perfect persuasion strategy. And then finally, it’s about insights and measurement, measuring how much they experience is driving the right business outcomes. Today I’m going to focus on to start with, I’m going to focus on the three products that we have product recommendations we’ve had in the tool for a while. You may be familiar with it. I’m going to give you a bit of a demo today in case you haven’t used it yet. Search AI driven search we call live search. I’ll give you a bit of a demo of that and show you how it works and how your merchants can benefit from that as well. And merchandising, our new merchandising services for building out category pages. So first off, let’s jump into product recommendations. I have a couple of different merchants set up for us to look at today. I’ve got a clothing brand called Vina and I’ve got an auto parts brand. I’m going to jump straight into our admin screens and I’m pretty sure everyone here is familiar with Adobe commerce admin. When enabled product recommendations will appear under the marketing banner. And this is where your merchants can start to build out any type of recommendation types that they want to show to their customers. So let’s have a look around the actual recommendation engine. First off, there’s some very good reporting on what recommendations are actively working on the site. Because as we know, good retailers and good merchants will have several different types of recommendations all working at the same time in different parts of their site. And they want to know at a snapshot which ones are working really well and which ones aren’t so they can move them around and adjust them as they go. And they can see how many impressions they’re getting, whether there’s revenue being generated from that product recommendation and what the viewability is like. To create a product recommendation, your merchants can pop in a name. We’ll do this one as you may also like. Pretty standard. Depending upon where in the customer journey the product recommendation is going to be it defines what product recommendation types would be available. View this, view that would not be available on a homepage because there’s no actual product. And that would be a recommendation type that would be down to products. I’m going to choose products so that we’ve got the most types of recommendations. You can have recommendations that are personalized and this is based upon the customer’s own browsing and purchasing behavior. There are cross-sell and up-sell recommendations. Getting that wisdom of the crowd and popularity. What is most viewed? What is most purchased? What are people most adding to cart? Or there’s your high performing items. What are the items that you always get sales on and do you always want them in your product recommendations? As you can see from the recommendations, there’s an activity status and active units. Activity status on the actual recommendations and this tells you when that recommendation type has the most information to be able to be of most benefit to your merchants. So you know, 0% means that this is a new environment hasn’t really learned a lot yet. So what you do is when you’re starting you probably start off with personalized recommendations and then move to the other types of recommendations as the sensei engine is gaining more and more information about your customer base. There’s a display label for the front end and you can choose how many recommendations will go into the banner. So depending upon your style, you can define how many recommendations you’re going to be showing. You can also, if you’re running multiple recommendations, which is what I was talking about before, define where it will show on the list of recommendations. So if you’ve got three different algorithms running on a product detail page, you can define what the order those recommendations are going to be showing. I’m going to show my new one at the top and then I’m going to be putting a you may also like just underneath it. AI engines are great for building out those recommendations and taking the grunt work away from your merchandises, but you still want to have granular control over what products that go in them. You may want to add certain products from certain ranges because you get a better margin on those or you may want to exclude out of stock products from your recommendations. And you can do that by using the inclusion and exclusion filters where you can include and exclude based upon price or stock availability, whether it’s low stock or out of stock or a specific category could be included because you get a better margin on the products in that category or you could exclude certain categories or products. If you start to think about running a marketplace, you could actually start to monetize the product recommendations by saying to some of your suppliers, hey, I will ensure that you are in every product recommendation that gets rendered onto my front end if you reduce the price of the stock for me. And once that’s saved, you can actually put in a SKU or a name of a product and you will see what the recommendations are going to be for that product on the product page. I’m just going to cancel this one for today. You can also build product recommendations that will be page builder content. So if you want to add those product recommendations to some dynamic content or some widgets anywhere in the customer journey, so you’re wanting to add product recommendations onto the shopping cart page or you’re wanting to add product recommendations on the my account page or in the B2B scenario onto a requisition this page, you can use page builder with product recommendations to do that. That’s a quick look at product recommendations. So we offer 13 unique types. Collaborative. What are other people what have other people viewed or bought based upon what’s currently being looked at personalized product recommendations. So that one to one recommendation content based or item based recommendations. And these are recommendations with similar characteristics. One of the ones I find really exciting, especially if you are a retailer and you’re dealing in fashion or footwear is visually similar. What the recommendation will do is engine will do is it will pick up items that are visually similar to the one in your that you’re wanting that you’re looking at. And then there’s that wisdom of the crowd or popularity based recommendations. What is the most viewed? What is the most added to the cart? Product recommendations are a SaaS service, so they are separate from your Adobe Commerce Core code base. You enable the commerce services connector and then you’ll be able to see and use your product recommendations. There is a web or headless SDK with two services. That’s the events for sending the recommendations to the storefront. You know, someone views a product or add something to the cart and the fetch recommendations SDK, which is fetching the results from the storefront. What has been engaged with and sending these out to the SaaS servers. It’s important to note that product recommendations is also compatible with our B2B shared catalogs. So that means if you have a unique one to one pricing for your customers, the pricing and the products that that customer will see on the front end will be reflective of their shared catalog pricing. The next area of AI tools within commerce I want to talk about today is live search and live search again is it works with B2B and B2C customers. So when a B2B customer searches, they will search and see products and pricing that is relevant for them. It is fast and intuitive. It’s an AI powered feature set. And again, it’s a SaaS based service which is included in your license. So none of these services we’re looking at are any extra licensing cost. And there’s a really flexible framework for developers. So what you can do is you can change the look and the feel and the styling or you can use live search in a headless deployment. So let’s go and have a look at live search. Again you’ll find live search in the marketing banner under live search. I’m going to jump onto the front end and give you a bit of an example there. If a customer comes along and they’re looking for, let’s say what live search will do is it has fast searches you type capabilities and it will bring up product recommendations. When your customer, it will bring up recommended product based upon the AI algorithm. When your customer goes through to a search landing page, the navigation is dynamic and the AI engine can help drive what the navigation items are going to be. And we’ll have a look at that in the back end. There’s also infinite scroll available on search result pages. So I’m going to jump into admin and we’ll have a look at the features within live search. You can see how your search is performing. What is the conversion rate from your search? What’s your click through rate? Your average click position. Are people clicking on the first, the second or the third item on average? What are my unique searches that customers are searching for? And what are my zero results? Unique searches and zero results are great because they can use that information to then create your synonyms. You can also see what are the popular results so that you know what people are actually searching for on your site. And all of these, all of this information can be exported out to CSV. There’s the faceting, which I was talking about before. Now you can have pinned facets and that’s what I’ve done on this front end here. I pinned these three facets and said, I always want these three facets to appear. But you also have this concept of dynamic facets. And what a dynamic facet is, is that is a facet that will only appear on the page if there is a product in the search results using that facet. So I like to think of the example of a tool company that will sell drills and have some merchandise clothing. If someone comes along and they search for drills and only returns drills, it will only show the facets that are related to drills. If someone comes along and searches for red and the red pulls up the red drills and it pulls up some red merchandising jackets, then size, because size is a jacket dynamic facet, would appear in the faceted navigation fields. You can set bi-directional or unidirectional synonyms on the actual site. And what’s really exciting, I’m going to jump into settings now, is that we have multiple languages out of the box and available for your Adobe Commerce environment, including Chinese, Japanese, Korean and Hindi. We’re looking at our regional market here in APAC. So if you’ve got a storefront in English and then you’ve got a storefront in India, you can actually have the Hindi search terms working. Like with product recommendations, I can create rules that will boost and bury product based upon my search terms. So a merchandiser may come and go, OK, if someone comes along and they searches for red, I can add conditions. They search for red and they search for pants. And I want them to get, maybe I’m going to go personalised. So I recommend it for your personalised algorithm every single time someone comes along and searches. I want to add some more intelligence through to that. I can do manual ranking of certain product so I can boost or bury or pin other products in the results. So when someone comes along and does do a search for pants. I’ve boosted the Selena pants and I have pinned the Bella Eilat Capri, so they will always appear in that position. And I can use that analytics and information that we looked at in the front end at the beginning of search to actually define what my boosts and buries and pins are going to be. Or I might be running a campaign for a certain period. So my boost and bury rules will be part of a summer campaign. And I can have specific dates that campaign is running from for. So my search rules will work for a specific period. Once I’ve done save and publish and that will reflect onto the front end straight away. So we’ve had a look at search and merchants can use these top five algorithms that we looked at, the recommended for you. So much like product recommendations, the algorithms are collaborative or personalized or using the wisdom of the crowd. We talked about live search and product recommendations being SAS based solutions. That means that we’re running the actual indexing on a separate SAS service on a separate server from your actual storefront and from your back end. And we are syncing that information across for you. But the great thing about that is it’s ultra fast. There’s no infrastructure headaches for you. There’s no regression testing and you can wave goodbye to some of those previous manual work or investments in those third party services that you’ve had to use for search. What we’ve recently used, we’ve recently launched is using the same AI tools to manage the category landing pages. We’re calling it personalized browsing or merchandising services. Your merchandisers and your teams can set up the same algorithms to define how the category landing pages are going to work. So let’s jump in and have a look at how easy it is for them to do that. I’m going to in my search environment, jump into category management. I’m going to jump to my auto parts store. And now you can see under category management, I can pick up a category, any of my categories that have product. I’m going to go to my parts category. And I’m going to go shop parts, add a rule. And again, much like search, I have the algorithms that I can use to define what the sorting is on that page. So you could now have every single category landing page personalized for your customers. If I want to just set the little ones and have it apply to all subcategories under that category, I can by applying intelligent rankings to the subcategories. And I still have the ability to manually boost, bury, pin or hide product. This is really exciting. And any merchants that were already using Live Search and had it enabled in their environment have this turned on automatically because as a SaaS based service, we were able to push that out to them. I was talking to a merchant the other day who went in, saw that, turned it on and started using it in their production environment within 48 hours. So where are we? So far we have Live Search, we have product recommendations and we have browse all running in Adobe Commerce for you. But what comes next? We are just about to launch generative AI into commerce. And I’ve got a quick video here from our product team on how that’s actually going to work. Merchants can come and they can choose a product and they can use the filters. Say I’m running a campaign for a winter campaign. I can choose the product that’s going to be in that campaign and I can come along to actions and I can create a generative AI background on that existing asset. So previously your creative teams would have had to go away, take the assets, apply backgrounds to all of them, give them back to your merchandisers who put them back into the commerce environment. But now what can happen is your commerce merchandisers will have an approval workflow attached to the ability for them to use Firefly, a generative AI tool, to be able to apply the backgrounds automatically to those images. So just think you’re running a New Year’s Eve campaign or you’re running a Christmas campaign and you want to put a Christmas theme on the back of all of your assets in that campaign. You can use this to do that. When the merchandiser assigns the background to them, they go into a pending status and there is an approvals workflow within the commerce engine for them to have a look at it, choose which ones they like. Yes, I like those ones. I’ve accepted that one. This one, yeah, I can accept it. Or I can actually go and choose, you know what, I want you to refresh the image on the back of these ones because I’m not too happy with them. When refreshing the images in the background of the assets, merchants will be able to then go, you know what, I want to define that down a little more and generate out more renditions of that. Yep, there’s more. There’s the one I actually like. And there’s always a content summary and information about the asset and the image and then merchants can accept that. Just going to jump ahead a little bit. One of the things they can also do, excuse me, one of the things they can also do is they can also come along and they can change the prompt on the image. So, you know what, I didn’t like Winter Wonderland on that one. I want it to be winter on the beach. They can change the prompt, decide they don’t like that one, jump back into the approvals workflow, and then once they’ve approved all of those, publish those images up onto the front end. Once the images have been published onto the front end, what they can do is they can jump into the product category and they’ll be able to see that all of those images now have the generative AI background on them. So we’re currently in beta with a few customers on this and expect to see this in general availability very soon. The next thing is I want to talk to you about was sort of 3D commerce and using Adobe Substance to do 3D commerce, both image renditions of 3D, 3D image renditions and augmented reality for your assets, for in your commerce environment. Now, 3D commerce with Adobe Substance is coming. It is general availability early next year. It’s best for customers who already have been using 3D images and then it will become more regular for all of the rest of the customers. So a quick exercise. Can you guess if this one’s real or it’s rendered? This one has been rendered with Adobe Substance. What about this image? Do you think it’s real or rendered? This one’s real. And this image here, real or rendered? This one, yeah, definitely rendered. Real or rendered? Rendered. The ability to use 3D assets for creating your content within commerce takes away a lot of the cost of that content supply chain. Adobe Substance and Project Sunrise, as we’re calling it, enables your marketing team to, it’s a layer that works with your dam or your commerce environment to publish those 3D and AR assets. It enables your marketing teams to improve their e-commerce by using 3D to deliver better visuals and product experiences in web and AR. I love the fact that we’re introducing hotspots on those 3D assets so we can start to think about more technical sales, using schematics or engine drawings with hotspots on them. And the hotspots linked to the products that are sold there. We wanted to put this into a manufacturing or a B2B type environment. And all of this is going to help you increase your sales, increase conversions and reduce your photography and management costs by being able to create this content at scale. Project Sunrise. Ingest and import your 3D tools. Create your 3D images and your 3D images and your AR experiences and publish them seamlessly through to Adobe Commerce. So have you thought about how you can use AI tools within your business? I’m going to stop sharing my screen now and see if there’s any questions that have come through. I do see one question here. And it looks like… I was just trying to remove that line for you. It looks like Scott answered it live but someone was questioning if live search will be available in Thai. Oh, is it available in Thai? Great question. Was it in the list? Let me just very quickly check for you. I haven’t memorized all of them. It is currently available in Thai. So if you install live search and you set it up today, it is available in Thai. Great. What I’d love to see is because there are GraphQL queries for all of these, because there are open APIs and SDKs for commerce and for Firefly and Creative Express, I would have to check that out. Listen to what Mairaz said about App Builder, join Jason’s session today and start thinking about some of the things you can do with AI. Amnesty have just introduced a module which is allowing merchandisers to use open AI’s larger language model to create all of the metadata that they need. Something like that. Using App Builder, not that hard to do. I’d love to see you guys creating something, popping it on the marketplace and monetizing it. All right. Thank you so much. Appreciate you taking the time. It was a great session. Thanks, guys. Bye.