Name: Using SAP CX AI Toolkit in SAP Commerce Cloud
Uploaded: 2024-5-17
Duration: 709 s
Description: Learn how to use the CX AI Toolkit for SAP Commerce Cloud use cases. Discover tools for generating product attributes and descriptions, as well as implementing visual search. Improve product discoverability and enhance the customer journey.

Share feedback

<p><u><strong>Planned maintenance of Practice Systems</strong></u></p>

Planned maintenance of Practice Systems

Why create an SAP account?

Enhance your learning experience

SAP Learning

Choose your subscription

Manage your subscription

Live Sessions

Hands-on Practice

Certification

My Learning

Ready to start learning?

<div><div><div>Learn how to use the CX AI Toolkit for SAP Commerce Cloud use cases. Discover tools for generating product attributes and descriptions, as well as implementing visual search. Improve product discoverability and enhance the customer journey.</div></div></div>

Using SAP CX AI Toolkit in SAP Commerce Cloud

Hello everyone. My name is Ken Fajardo and today we'll be doing a quick demo overview of our CX AI toolkit for Commerce Cloud use cases.  So the first thing we want to start with is just an explanation of what's gone into this overall demo landscape.  So first and foremost, we have an implementation of commerce here. You can see we're in the product cockpit.  You can also see our different items and products that are in our catalog.  For the sake of this demo, it's mostly retail fashion items  and we can then jump over right to the CXAI toolkit.  And so taking a look at the toolkit,  obviously we have some of the tools you may have seen in previous videos, things like the ability to search for answers or type into questions and get responses.  We'll skip that today only because we've covered that in a couple of different videos already.  And really what we'll focus on today is some of the key tools that are listed here in the middle of this page as well as some of the additional items over on the left side of the menu.  Now, if we are going to start at the most basic fundamental part of a commerce operation, I want to start then with product attributes.  So we have this predict attributes tool and we can go and search for any product.  For the sake of this demo, we're going to say that we're working with this Nicci dress.  And what it'll do, it'll show me any existing attributes that we already have as well as the ability to go and generate new attributes.  And the way this works is it's going to take a look at different facets of the product, whether that's the image or the description, etc.  And from those things, it will try and generate some attributes that might be relevant.  So in this case, it's taken a look and said, OK,  this has a color and that color is black.  This is for an occasion type.  And in this case, that would be for formal occasions.  This has a fit and that might be slim, et cetera.  So very neat, very straightforward.  And the reason this is powerful is if you think about a Commerce Cloud implementation,  one of the things that any Commerce person is going to be most focused on is basically how many orders they're getting, their conversion rate and things of that nature.  But tied to all that, of course, is you know, how are people finding the products they're looking for?  Now, there's only so much a human can do in terms of determining what these attributes should be, what the values of these attributes should be, and it's a very, you know, time intensive, tedious process to determine all of this.  And this is just for one product. Imagine catalogs that have 100,000 unique items that have to be managed.  So this is really where we're stepping in providing the ability for the system to generate attributes that make your products easier to find and therefore shorten that customer journey.  So ultimately, what we're really doing is increasing the likelihood of a customer finding the right product and checking out.  So that connection between attributes, discoverability of products and conversion rate is really, really straightforward.  So this is one piece of this,  generating new attributes or generating values for existing attributes that may not have enough descriptors there.  So one key piece here.  Now the next piece. Once you have those attributes, you can start generating product descriptions. So let's find the same Nicci dress.  And you know, what you'll see here is this is the default description. There's one that's already generated from last time we played with this.  But I can pick and choose and say I want to mention that this is black. I want to mention that  you know this is suitable for romantic dinners, that it's sleeveless, that it's for the summer.  I can give it extra, you know directions when it's writing this.  You know, I can say something like use a Shakespeare quote just to make it fun.  I'll change the tone of this to be something maybe witty, and I will hit, oh, and I'll make the length of this maybe, let's say medium, and I'll hit generate.  So the system will take these different attributes It will take the existing description. It will take my directions and come back and give me a new description.  So introducing the Nicci, modern sophisticated dress, you know, blah blah blah blah blah, Perfect for romantic dinners.  And then it even has at the very end though she be but little, she is fierce a Shakespeare quote.  So you know, kind of a silly example, but again, this is something you can automate.  Now one additional piece here is the ability to not just generate product descriptions, but to do them in bulk. So I can select multiple products at a time.  So let's just show you how some of these buttons come up online.  I can hit Generate and I can hit publish.  So I can do that in a basic exercise with if we take a step back and say, OK, what if I've added a bunch of attributes to a bunch of products and I want to enhance on a kind of bulk basis the descriptions of all those attributes.  That's exactly what this empowers you to do.  So again, yes, saving time and tediousness etc, but also improving the quality of your data, making it more likely customers will find and buy the right products.  So with those two items done, there is third item here and this is what we call our visual search. Now the visual search basically goes and indexes all the different product images that you have in your catalog and then opens this for a front end user to be able to upload an image and find a product.  So what we're seeing here is kind of the administrator view of this for an e-commerce manager to kind of play with it and see what comes up.  But let's go ahead and play and see what this would look like. So let's say I'm going to upload an image,  I have this jacket image here. I'm going to say that I took a picture of a jacket. I'm curious if the store has one that looks like this and I can hit search  so the system will go and look and try to find jackets that most closely match this and so I think it did a pretty good job here. So the first one is a dark puffy type of jacket which actually matches fairly well.  The second one is a zip up black jacket with a hood, which again is fairly close and then the further we go the less close we are But again, in terms of what's available in the catalog, it did a great job.  We can also use a more straightforward picture like this one down here, this lady with a black dress. So if I hit just search,  you'll see a couple of different things.  Now what's interesting here is it's giving me lots of different results because I have lots of different things in this picture.  So we've enhanced this to be a little bit more specific. So if I enable object detection.  Now it will identify different parts or pieces that are being shown to help the user be more specific with what they're looking for.  So in this case, if I only wanted to search for that dress,  I highlight that, then I hit search.  Now I'm getting black dresses and again, the first one is closest. It's actually the Nicci dress we were looking at before.  Shorts, sleeveless, etc. And then as we move down, less and less relevant potentially.  We'll even do one more example over here.  Same thing with object detection, it's already turned on. You can see different parts of the image being highlighted.  So we'll say, hey, show me bags similar to this one in search  and boom, it starts giving me things that are as close as it can find within our catalog, right?  Same thing for footwear.  So search for the heels that she's wearing.  You know, kind of a lighter colored heel.  Yep, that seems pretty close.  And we can even do the dress and see how close we get there, right? So again, sleeveless short.  The second one is more blue, but not the same shape and then, you know, keeps going down from there.  So with those tools, pretty straightforward, but again bringing a lot of value, helping customers quickly take a picture and search, find the right product, check out, Get that order in as quickly as we can.  Now one last thing that we'd like to maybe talk about here is this tool at the very top on the homepage, which is the ability to ask about a product.  And again, the idea would be that you would place this on the front end on your e-commerce site experience on a product detail page and let a user ask a question about the product and see if AI can answer it.  The idea here is just to give an e-commerce manager a way to play with this and see what the results are. So let's do that.  So first we've got to pick a product.  So let's again look for that Nicci dress.  So we'll pick that one. So here we see the dress and I can ask something like, is this good for formal occasions?  And so the AI will break that question apart and try to find the answer based on the data it has. So it gives you a description, it tells you what it's suitable for, and then it starts actually explaining. So this suggests the Nicci dress designed to be worn for special occasions that require a more formal dress code.  Additionally it's categorized as a dress which supports the suitability, so it does a pretty good job of looking up all that information.  It also then even shows you the sources, so hey, it's telling you this came from product information, so the description and features of that product.  Now over time, as we enhance the tool, if there are additional attachments or documents that help describe certain products or for B2B settings, if there are care instructions or operating manuals or things like that, or operating manuals or things like that, all of those could act as sources for this kind of question and answer feature as well.  So with that, we'll stop here.  Please let me know if there's any other questions or if you'd like to demo a different part of the tool.  Happy to provide a deeper demo at a later time.  Thank you.