Hi, everyone. This is Ram Sivarajan. I'm part of a group called CXAI, or Customer Experience Artificial Intelligence. I'm a product manager in that group.
And in today's topic, we're going to talk about how we approach AI problems in our customer experience portfolio of products. Broadly speaking, we're going to cover three main topics here. Basically how do we operate as a central team, and solve our customers' problems in our customer experience portfolio. And then we'll talk about how our approach different, and how do we operate with our LOBs, and how we are tackling the problems of our customers in this group. And then finally, we'll also talk about how our customers benefit from our approach in solving these AI problems.
All right. Let's dive in and talk about, we'll start off with our approach to AI and in our customer experience portfolio products. All right. Just to set the scene, this is a picture of how our product is situated in this portfolio of solutions. So right in the middle, you see CXAI, our customer experience with AI machine learning, which basically connects our different LOBs that you can see here like CPQ, Sales Cloud, et cetera. And the way we operate this basically we are a central team who solve AI problems for our customers.
And primarily, we have a AI machine learning platform, which is a centrally located platform, which is utilized for solving our AI problem for our customers. Additionally, what we do is we work with different LOBs and solve the customer's problems end-to-end. So we work with-- the LOB understand customers' problem, pain points, use cases, et cetera. And then finally, we integrate our solution with the LOB. For example, if we see here, we have CPQ, we have certain product recommendations, which are available with our CPQ portfolio LOB. Sales Cloud, or C4C, we have a product recommender cross-sell coming up beta customers later on this year.
We have Commissions customers who are making use of our plan and team optimizer AI solutions. We have Territory and Quota colleagues who are using our territory and account alignment optimization solutions. We are working closely with commerce colleagues.
And we'll have certain hyper personalization use cases satisfied sometime next year. And then also finally, we are also working to get our AI offering available in the Marketplace. In the SAP's Marketplace, which means third parties and partners can make use of our AI offering.
And in terms of the offering itself, on the bottom part of the screen that you see, we have recommendations, which are primarily product recommendations or cross-sell, upsell, pricing, et cetera. We have a little administration portfolio that we offer here as a headless service and also as a UI, which allows our customers to monitor and schedule their training, et cetera. And they can also look at the quality of model and the data as well as part of this solution.
And we talked about optimization use cases, which are also certain things we offer. And then we talked about hyperpersonization. So these are some of the portfolio of offerings that we have. And this is how we work very closely with each of the LOBs.
All right. So before we go further and very high level bird's eye view of the different AI use cases that are out there in customer experience, as you can see, there's quite a lot of use cases out there. I won't go into each of these. But just to give you a sense of where we stand, on the left hand side you see some product recommender use cases used by our sales pillar. You see some optimization use cases used by our sales performance management commission strategy and quota, and so on. We have certain use cases for service, commerce, and marketing as well.
So this kind of a very high level a bird's eye view of all the use cases that are out there, just to give you an idea of how many cases we are talking here. All right. So before we move further, a little bit of background about the different approaches to AI.
So on the left hand side, you see AI platform, which basically is nothing but a set of platform solutions which are out there, which allows customers to bring in their data scientists and develop AI solutions for their problems. That's one approach. And on the right hand side, you see AI solutions, or basically which are completely end-to-end developed solution, which the customers can directly make use of.
And we fit on the right hand side. So we don't offer any platforms. We are actually a AI solutions organization. So we work closely with our customers. We closely work with our LOBs, and make sure their AI problems are solved end-to-end, rather than bringing in a data scientist and trying to solve the problem from scratch. And in terms of how do we approach this, if you break it down into four parts, on the left hand side, we basically understand the customer problem, work with the LOB, try and make sure AI is the right fit for this solution that we're talking about here.
Once we identified this is a real AI use case, we'll work with them. We connect our CXAI application with the corresponding LOB, typically via API to API integration. We make sure the training models are appropriately selected and they are automated end-to-end.
The training and scoring are available for our customers to monitor. The scoring is basically also taken, scoring or inference, or product recommenders optimization use cases that we talked about are easily consumed within the end user application. So the customer need not have to worry about going into different places to figure out and solve, look at the AI problems. And then finally, we also make sure our models and the whole infrastructure is monitored, and made sure it's solving our customer's problem, and we continuously improve it. So that's the approach we take in terms of the end-to-end integration that we talked about. Now we'll focus a little bit more on the product pricing recommendations.
But before we go into that, a little bit of a quote from our CEO on the AI-powered CPQ that we offer cross-sell and pricing recommendations to our CPQ customers. So a great quote from our CEO on that. I'll just quickly read that, and then we'll move forward. "SAP's AI-powered CPQ can support your sales teams to stand out and optimize being relevant. And that is a good place to be when customers need fast, accurate quotes."
Great quote from our CEO. So let's move forward. All right. So in terms of the product pricing recommendations that we talked about, broadly speaking, we have these four types of use cases that we satisfied.
Cross-sell is basically when you're selling a product A and you also want to bundle in product B. And that's the type of recommendation we provide, and that is right now available with our CPQ solution already. And also, we are also working closely with our C4C colleagues, and this should be available for beta customers by end of this year, 2020. And we should have a generally available solution by early 2021. So we are expecting to have some GA customers by early 20221.
Pricing is another use case where we provide pricing discounts. So we figured out based on historical data and past data points, we figured out the best appropriate pricing discounts that a salesperson should offer to the customer. And again, this is available right now with our CPQ solution. And also, we are looking to expose this further in our marketplace portfolio around the middle or end of next year. Upsell and Config are also other use cases that are already available and ready to go.
In terms of Config, if I talk about that, that is primarily used by CPQ, a type of solutions, and it's one of those solutions which will be available, hopefully by middle of next year with our CPQ colleagues. We are working closely with them. In terms of upsell, that's again, available right now. And we are looking to incorporate that with our CPQ solutions sometime next year.
All right. So we had a quick look at the various data points. Now let's jump in to a quick demo and see how our product recommender is available, and built-in, and available within the CPQ solution, and how it helps our sales person to close a quote quickly. So I'm here in the login screen of a CPQ application.
I'm going to go ahead and log in. Let me go ahead and select a particular product. I'm going to select this, my mountain bike.
And I'm going to go ahead and add it to my cart. All right. So the product is added into my cart.
You can see the product in the bottom half of the screen. And then on the right hand side, you see a small light bulb. When I click on that, it basically shows the CXAI recommendation. So there are two types of recommendations that you see here. One at the top is basically the cross-sell recommendation. So with this bike, we are also recommending a helmet for the quantity of 1, and a discount of 0%.
Now a salesperson has an option to add this product or decline this recommendation. And additionally, we also show a little bit of description of why we are recommending this product. Similarly, we are also recommending another item here. This is basically the same primary product, mountain bike, with a discount of 10%.
Now so this is, again, the salesperson has a choice whether they want to change the discount, decline the discount, or accept it. So these are the two types of recommendations that we're showing. And we call this what is called a compound recommendation, because what we're doing here is we are showing multiple recommendations in one screen, which basically allows our salesperson to quickly decide and add these cross-sell products, and close the quote quickly with appropriate discount in price recommendation. All right.
So that was a quick demo on how the hosting fits together. So broadly speaking, we saw on the recommender popping up, which had that compound set of recommendations. So in terms of the types of recommendations that we are offering here, it's obviously the pop-up screen which shows the recommenders on the right hand side. And then broadly, when we classify that, we show the cross-sell recommendation to start with. Along-- we also show along with that the quantity and the percentage of discount that we should be offering. So that's one set of recommendation.
And then we also show on the bottom side if you see, we show what we call an augmented intelligence. Basically, we are showing a appropriate discount, which the quote should include, along with the other cross-sell product that we're talking about here. And then, plus, we also provide a set of explanations as to why we are recommending this.
So that the salesperson is well informed as to why these recommendations are popping up. And based on that, the salesperson has a choice whether to take these recommendations or reject them. So either way, that's OK. So this gives them a choice and then they are able to quickly close the quote with these recommendations. So we saw how this works in the front end in the demo, and we saw the screenshot as well.
Now how does this whole thing fit together when you look at it from a back end angle? From a back end angle, the way it works is we have a CXAI, which you see at the bottom part of the screen. And we basically have what we call a mediated integration, which allows us to pull data from our CPQ application, and also from a CRM application. The CRM we support are C4C Salesforce, or any other cloud based CRM.
So we are able to pull data from any of these applications, and basically transform, clean it up, and then push it further into the model. The automated models run the ingested data run and then get trained, and then finally spit out the inferences. The inferences are available within the application, within the CPQ application in this case as a recommendation, a cross-recommendation and pricing is what we saw. So this is the back end of how this whole thing fits together, just to give you an idea of all these back end complexities are hidden away from the customer.
And the customer ultimately reaps the benefit of the inbuilt AI product recommender within the application. All right. So before we move further, a quick screenshot of our C4C cross-sell recommender that is coming up for beta customers later on this year. And then GA early next year. In this case, what we are doing is basically we are pretty much utilizing the same cross-sell recommender that we exposed in our CPQ application.
We are making use of the same recommender in C4C, which is a good status to be in. Basically, we developed it once and reusing it in across multiple applications, which is a great thing. All right. So in terms of the business benefits of cross-sell recommendations that we just saw.
So how does that help our salesperson? So basically, the whole point of this is to eliminate guesswork. So salesperson need not have to worry about whether any money was left on the table, whether they could have sold any additional products. So all those guesswork is completely eliminated. And the recommendation pops up and the salesperson is ready to go, and close the deal quickly. So that's the basically gist of this type of recommender. So in terms of the pricing recommendation, again, the same concept eliminates the guesswork, it provides ideal discount.
So the salesperson need not have to worry about whether they're giving too much discount or too little discount. The discount it's pretty much uniform across the portfolio. So all the customers benefit from uniform type of discount. You don't have to worry about the salesperson.
You don't have to worry about any doing any manual calculations, et cetera to figure out the discount. All these are eliminated. But again, there are able to very quickly close a deal with the appropriate accommodation that are popping up. All right. So we have-- we saw some use cases, we saw the demo.
Now let's take a look at our approach and how do we approach this solution. And how is our approach different from others? OK? So broadly speaking, we have three categories where we are unique and differentiate from others. To start with, customer first approach. That's our motto. We work closely with the LOB.
We work closely with the customer. Make sure the solution that we're talking about here is a completely, a solution which is AI element. If it is not something that can be solved by AI or it can be solved by a rule-based engine, we would rather recommend that rather than going with an AI solution. So that kind of our first step. So make sure the problem that we are trying to solve is a AI fit. That's kind of the cost of everything.
And second thing that we do is we make sure the AI is once built is automatically available. The R&D that goes in is democratized. The customer need not have to worry about all the ins and outs. Bringing in data scientists, engineers, et cetera, everything is hidden away from the customer.
It's right out of the box available. And in terms of the technology, we have multiple highly scalable data centers, and technology we are talking about here. Certain technologies that we utilize are patented as well. In terms of the scalability, we use Kubernetes clusters, which we can bring up and down, depending upon the demand. Low cost is one of our mottos. We make sure the cost associated with infrastructure is pretty low.
And we keep a tab on that. Cloud Native is another approach of ours. We are cloud agnostic, which means we can host our solution in any of the cloud vendors.
And then finally, we also support GPU for a certain deep learning use cases. And one of our standards is we make sure that all of these solutions that we offer are also available as a headless API service. Which means it's pretty easy for us to integrate with different LOBs that we talk about here, and also in future when we expose it to marketplace, it's pretty much easy for our third parties, and partners to make use of these headless services.
Now when we talk about model side of things, our approach is to run multiple models and evaluate the best. That's a norm with all of our AI solutions. And in terms of the types of models, if you talk about CPQ use case, or C4C use case, we have a history based models, cart based model, popularity based model, different types of models, and you can begin choose the right one depending upon the use case that we are talking about. And then, in terms of the recommendations, we saw an example of a compound recommendation. So basically we make sure to jump back the customers, the customer gets all the benefits of those multiple recommendations in one go.
So s that they can make a consolidated decision on whether they want to go ahead with that recommendation or not. And then finally, in terms of the models itself, we have an option to spin multiple models, which can be kicked off pretty quickly and easily with just switching on a particular parameter. So that's another use case we support. An example would be when you want to run a model for North America, and a model for Europe, you can spin it, spin both pretty quickly by just passing in a segment parameter models. All right.
So, so far, we've talked about the use cases, and we talked about our approach, and how are we different, and unique, et cetera. Now let's talk about the infrastructure, and the infrastructure side of things, and how we are investing in infrastructure, and how our infrastructure is used by our AI solutions. So this is our vision for 2020 and 2021. As we speak, we already have our solution running on kubernetes clusters. It's low cost, highly scalable, all pretty much already there. We operate on consumption based pricing, which means our price is baked into the LOBs and we try to keep it as low as possible, and make sure we charge the LOBs based on consumption.
So the customer need not have to worry about any AI cost as such, because it's baked into the core application like C4C, or CPQ for example. In terms of the cloud-nativeness, we currently support GCP. We have them hosted on our Anthos clusters. Apart from that, the Azure. Is something that is coming up later on this year.
We are pretty much there I would say with that. That should be available around November, December time frame. And that would be a fast cloud native solution in Azure. Apart from that, as I said, our solution is cloud agnostic, which means in future, if you want to move to Alibaba, or AWS, us you can pretty much easily move further on those lines. And then finally, in terms of the expandability of our solution, as we talked about, we are a central AI/ML solution, solving the use cases of our set of portfolio of products in CX. And basically, the whole point of this is we can pretty quickly and easily expand additional use cases as needed, utilizing the central platform.
And then as I said, we will be supporting local interfaces in future as well. And again, comes under expandability side of things. So we'll be supporting that sometime next year for our marketplace use cases. All right.
So let's take a look at the back end infrastructure a little bit more. So in terms of the back end infrastructure, we currently support Anthos. So we have customers live on that already.
And it's hosted in our legacy Callidus data centers. Apart from that, we also have the Azure first approach that's coming up. As I said, November, December time frame, we should have Azure ready to go. And then, as I said, this is cloud agnostic, which means in future we will be supporting Alibaba and AWS well. Now in terms of the internals of this infrastructure, kubernetes cluster, everything is highly scalable. We can spin more pods as needed.
And then we have a micro services based architecture. All of our services are micro services. We have training and scoring running as micro services.
We have collected services, which basically allows us to pull in data from external sources. via APIs. We have AI services. We have gateway services, which basically allows us to route the requests. We have tenant and master data services to allow automatic provisioning as needed.
So those are some of the kind of back end services that we offer within our platform, which allows us to scale quickly, and be nimble, and adaptable to the changing needs of our customers. And then in terms of the external applications we integrate with, we obviously integrate with CX applications. We also integrate with non CX applications.
So if we take the case of CPQ, which used to be part of CX. Now they are part of S/4 HANA, but we still continue to work closely with them. And then finally, marketplace is another piece that is coming up later on this year, later on next year I should say. And then finally, on the Kernel services side of things, we are integrating with them as well for authentication authorization purposes, security reasons. So again, that is something that is going to be available later on, late this year, I would say. So those are some of the external applications that we integrate basically with them.
Right. So we have seen the back end infrastructure. We have seen the use cases. We saw you know how the whole thing fits together. Now let's talk about a little bit about how our customers are benefiting from these AI use cases. So let's talk in terms of if you talk the case of, if you talk about the case of upsell and cross-selling opportunities that we talked about in the past, they offer our customers basically our salespersons are not worried about missing any opportunities.
They are able to sell the product without worrying about whether they are missing out on any opportunities. So that's one benefit of our offering. But from that, we saw use cases where we eliminate the guesswork. Same applies to plan and team optimizer, which is available with our commissions application.
The sales operations, you don't have to worry about whether they are selecting the appropriate right quota, whether the rules are set up properly or not, whether the salespeople are bringing the appropriate bookings, et cetera, or not. So they don't have to worry about that. Optimal sales quota, in size again, a typical commissions use case we make sure the quota is appropriately set, and the bookings matches those quota setups. In terms of pricing, we saw that before. Again, this eliminates guesswork, and the discount that we offer is automatically available and ready to go.
Territory, assigning territories for peak performance, this is a typical Territory and Quota use case where we allow our customers to optimize territories and account pretty quickly in an automated, very automated fashion. And that eliminates a lot of man hours, manual man hours that they used to spend over a number of quarters. And then finally, in terms of quota optimization, again, this is a typical commissions use case where we make sure the quotas are appropriately set. So these are all the benefits that our customers are reaping based on where we stand as of now. And hopefully, looking forward to satisfying some more such use cases in the future.
All right. So we've come to pretty much to the end of the session. But before we close, a quick closing statement. As you see, AI is evolving. It used to be nice to have now it must have and SAP is investing a lot on the AI side of things. And we, as an organization, CXAI are right in the front, and we are making sure our customers' needs are fulfilled as much as possible.
And wherever AI fits, we try to make sure we come in and chip in and help out our customers. We-- this approach obviously helps to automate the solution end-to-end. A lot of our customers are seeing the benefits of this. They can shift from tactical work that they do on a day-to-day basis. They can move to more strategic focused work, and eliminate a lot of man hours that they spend on manual tasks.
And then finally, we also make sure the AI is embedded within the core application. No data scientists required. It's easy adoption. That's our motto. So that's the kind of summary of what I wanted to close off on. So as nice to have do a must have use case.
All right. So before we close, a little bit of-- there a couple of sessions, which you may be interested in. The CX113 is a session, which is conducted by one of my colleagues who is a data scientist.
We talk a little bit more on the territory alignment use case that we talked about. They go deep dive on that. You may be interested in that session. And then, CX116 is another session conducted by our C4C colleagues where they talk more about the intelligent technologies and the AI use cases. So definitely feel free to check out those two sessions, I would say. They are related to what we talked about here.
And then additionally, I've included certain URL links further down. You may want to check out those to get a little bit more insight on our AI offerings. And then finally, obviously, feel free to make use of our learning hub where we have thousands of titles in multiple languages available.
So you get the opportunity to not only learn what you learned right now, but also look at other titles that are out there and help you to become more familiar with our SAP applications and portfolio of products. So with that, I would like to conclude this session. And I would like to thank you all for attending this session.
Feel free to reach out to me on the below email address. And I'm happy to answer any questions. Thank you all.
2020-12-19