And actually, it's the same principles that we apply AI to the power of four that we learned about, also to infuse AI in the support process. And it's actually the same team which is working on SAP Cloud ALM and on AI in support. We would like to provide you with a balanced view by this presentation. The presentation is structured like follows: a quick intro, what does it mean, customer support powered by AI. Then we would like to take a deeper dive on some of the selected scenarios or strategic themes that we are covering with AI in customer support.
Then, of course, we would like to illustrate how it looks like to give you some, let's say, visual impression. You know the support process, how it used to be, how is it developing now. So seeing the latest developments out of the support kitchen, and we will have some time for questions and answers.
Customer support powered by AI. We get this question very frequently when people read, let's say, articles in media, and you will hear me talking about a concept which is called invisible support. Yes, do you do this, like, do you infuse AI in the customer support process to reduce your support organization? And we know, like, we also saw it in the industry, the effort, like, this typically backfires, and that's definitely not our intention. We would like to assure you there is a very, very skilled and well-oiled and very present support team actually around the globe. So with the eight support centers in more than 30 different countries, we are literally in your corner wherever you need us. So this is there, and this is going to stay like this. We want to be present there where you are as our customers.
So why do we do then, why do we do then AI? Well, the answer is very simple. This very stable and very well-oiled and trained support team is coping an increasing number of perceived customer issues and tickets. SAP is on a growth trajectory. We have more and more product solutions in our portfolio. Sometimes also the complexity is increasing. So how to make sure that this great support team that we have at our disposal can cope with the load and also make sure that you have your questions answered and your problems solved whenever you need this? When I would describe it, it's rather these are superheroes, and we have very, very many of them. The AI is actually the power-up that is giving them the superpower to be there where you need them.
There is nothing that can replace a skilled support engineer. When it is about a very complex problem, when it is about assessing a situation, when you are, as a customer, approaching a specific business event, specific situation like Black Friday or some go-lives, there is no agent in the world that can cover that in that complexity. But these agents and AI functionality can give the support engineers the support power needed.
This view is important because it kind of sets the stage. On the left-hand side is actually the strategic themes which we are working on, and on the right-hand side is how we would like to be visible and appeal to you. So what are the strategic themes fueled by AI? The first topic is the invisible support that I already mentioned. This is a response to the well-known challenge, 'Why SAP did not react?' You know me, I'm your customer. You know my support history. You know that I'm approaching a business event, for example. So this is all about kind of being preventively there where you need us, figuring out about risks that you are facing internally, escalating when we need it, and solving the problem maybe before it occurs to you. A very, very important topic, and the whole kind of AI strategy is shifting to this gravity point.
If something slips through the crack, of course, you are experiencing a problem. Then, well, you want your questions answered and answered fast. You want to be able to find solutions fast, not being connected maybe to the human operator. You want the solution. So that's why the next theme, when there's already an issue or when you are having a question, that we empower you to find the answers by yourself and also to get the solution instantly at your fingertips without having to create a ticket. This is more convenient for you and less complex for us. The empowerment, the self-service theme, is actually a very, very omnipresent and very, very important accent in whatever we do.
The last thing is the assisted support. We had already great questions before we started. AI agents, that's the new theme, and believe us, we are experimenting heavily. There are actually already agents in place that you can see. I will talk about it a little more. What we saw in our lab is that one AI agent already can do great things. But if you have more of them, each specialized into doing something special, and then orchestrating them, they will be reasoning, they will be working together. An orchestration of a system of agents is definitely something that you will see is a very, very major improvement in the support interaction. So these are the three, let's say, strategic themes that we are working on: the invisible support, the preventive right, the self-service empowerment giving the power to you to find your answers, to find your solutions, and if something needs to be done in the background, we will be relying on the agentic superpower as an orchestration.
How this should appear to you then, I mentioned that support must appear to you as fast. So all these things, solving a problem before it shows to you, the effortless support. We would like to get rid of this, let's say, old offline model, submitting a ticket, having to wait. Fast problem solving. You living in a dynamic world, we know it. So we want to be fast by addressing these three things. We want to be rich. We have all the information you need in terms of possible solutions, knowledge base, and things. So you must be exposed to all the channels that you need to have rich knowledge empowerment, seamless experience. We know we have multiple support access points. You can access us via SAP for Me, right, the Get Support area, which is your one-stop shop. But you also experience built-in support, for example, in your LOB solution being already there. So support is there in the cloud tenants. You can contact us, you can look for information. We also have a mobile app for Get Support. So giving you a seamless experience, that this is kind of the face of support, like an intelligent front door. This is also something that will become more and more visible to you as we progress working on the three themes.
And prescriptive guidance, the power of SAP Cloud, and of course, that we heard in the keynotes, that you are seeing also with your installations of SAP Cloud ALM, there is knowledge there, there is good data, there are events, there are signals. So connecting them as the execution platform for implementations, for operations, for services to the whole support conglomerate is definitely a great value that you will be getting out of it.
What are the key AI scenarios? Those of you that attended my presentation last time in Mannheim maybe recognize this picture. One big difference is there that the, let's say, agentic and generative AI, that's the violet boxes, that they are now omnipresent and they are kind of spreading through the process. The process is very simple, and you are very familiar with that. We would like to prevent issues. That's the preventative part. Solve it before it happens. When something happens and you are recognizing there is an issue, I might have a problem, might have a question, we would like to give you the answer straight, possibly giving you the solution. So the AI-assisted knowledge search, you might say, 'Oh, knowledge search, yeah, just search.' No, our measurements show that when we train it using AI, it will be able to understand your question much better and it will be able to find the most relevant solution. So there's even AI and tons of AI in our knowledge search.
Incident solution matching, that's another very popular topic, a highly, let's say, performing service. This is also what you can connect into your ITSM, right? That's the same kind of solutions proposals, and we are making it available through the different channels. You will see me then... okay, if you haven't found your question or haven't found your solution in the self-service, in empowerment, case prevention space, you will proceed, of course, with case creation. Here also AI can help to classify the ticket much better so that there is no ping-pong happening, giving also the support engineer the superpower to solve it fast because the operational efficiency is what is visible to you as fast ticket resolution, as ticket resolution up to your satisfaction.
Where we have many, many scenarios like case summarization, that's something that we use behind the scenes. There are many handovers of tickets, for example, between processors, and then the next one doesn't have to read the whole long text. We have AI that gives him or her a nice summary. You might think, 'Well, what a great idea,' but if you count all those minutes and hours spent with this simple task, you will get a very significant amount of time that is freed up actually for the engineer to really look into the problem and solve it. Of course, the feedback loop to the product organization is important. If you figured out there was a system with a product, it must be extremely fast handed over to the product organization so that they make their product better. Here even we are getting into the phase where AI will be able to code a solution actually, or let's say isolate the technical objects that are affected by the problem so that the programmer can orientate very fast and implement a correction. Many, many exciting things, and kind of showing them on the logical flow of the support process makes pretty much sense.
You see also the preventative customer mitigation on the bottom. I will talk about it in a second. So that's kind of the world map. We have much more, but this is kind of what tells the story. Now let's, let's say, deep dive into the three main themes. Those were, those who were on the left-hand side: invisible support, empowered support, and accelerated support. So invisible support.
A logical thing, yeah, you would say, 'Why haven't you figured it out earlier, dear SAP?' Yes, we are aware of it, working on it, and there will be more and more coming. What's the issue? We are connected, we know what your cloud tenants are. We are going to have an agentic framework, I'm talking about the AI agents again, and this is the awareness of your system healthiness, of your business events, and of your support history. And how does it work? It's very simple. Many, many situations that maybe you experience, we experience them as well. A customer is approaching us and say, 'Didn't you know about our go-live? Didn't you know about our period-end closing? Didn't you know that we're going to have a Black Friday?' Some, like, significant business events that are impacting the load on our systems which goes beyond the, let's say, usual traffic. Then we say, 'No, no, but I told you.' Yeah, how? 'Yeah, put it in the response to my ticket. I wrote there, there is a go-live happening.' 'Oh, I'm so sorry, my colleague might have overlooked that.' But then, why, like, you knew there was a go-live happening, you saw the support history, you saw the tickets happening around that time, and you didn't escalate and you didn't solve it for us. So you see, let's say, what a human might overlook, these careful and diligent agents which see more than a human eye could see when they are connected to the right sources of information, they will be able not to solve the situation, that's not what we want to, they would rather highlight risks being connected to the system health, to the business events that we can actually extract using AI from your tickets. There's a go-live on October 30th, then let's, okay, let's extract it and let's put it somewhere, and any tickets happening around that time frame will be escalated because there is a problem happening, you will be unhappy. So that's kind of how we would like to do it. The final discretion is, of course, with the human operator, with a mission control employee for escalation, support engineers, they will review these risks and take the respective action. So we are not prescribing them, but we are giving them the superpower to be connected to critical risks happening with you as our customers. Preventive, solving things before they happen.
If something happens, as I said, if you have a question or if you're experiencing, seeing an issue, we would like to be there in the support access point, SAP for Me, built-in support. As I said, you will be searching for something, there is AI assistance. You approach built-in support, there will be different AI scenarios already happening in your LOB tenant. And this gives you the ability to get your answers fast and get your solutions very, very promptly.
And last but not least, and as I said, this is a topic very close to my heart because we're doing a lot now with the agents. Some of you maybe saw that something changed for certain products. We are rolling out already the first agentic scenario called auto-response, and this is a fantastic thing because some of you, if you are on the right component, might have experienced you created actually a ticket and 24/7, within like a split of a second, the ticket got responded.
What is happening behind the scenes? There is an agent which is reasoning with itself, saying, 'Is this something that I feel confident answering to?' And if yes, then it gets the solution, wraps it up, and sends it back to you. This is already happening, and we had many discussions also with, like, legal and, 'Yeah, how will our customers love it?' But the first signals from you is like you like it. If you've opened a ticket on a Sunday afternoon and the agent can solve it, you will get your solution. So we will have a very, very high success rate of these interactions. Legals are very concerned. You need to provide the customers with, like, an emergency exit. 'I want to be connected to an operator.' Yes, it's there, right? If you don't like it, if you don't feel comfortable working with an agent, sending tickets back and forth, you can always switch to the manual mode. But we tested it, and for those thousands of tickets in our pilot, we had one situation like this where a customer reacted, 'Oh, I don't want the agent. I want the support engineer.' What the support engineer did, copied the answer from the agent, sent it back to the customer. That was exactly the solution. It worked, and customer confirmed the ticket. So we have a very, very high rate. I will talk about it. This is something which is mind-blowing. And imagine this is one agent, that's the kind of solution finder and responder. Imagine the power if you have very specialized agents in orchestration. There is a case coming in, then the whole conglomerate, the orchestra, starts playing, and they will start reasoning with themselves. They will classify the case. Then, let's say, they will decide, 'Do we feel confident? Do we auto-respond, yes or no?' This is something that, let's say, a human support engineer needs to solve, but we already recommend the business action. So recommendation, is it an outbreak, whatever, so that this guy is then informed and can solve the case very, very rapidly with all the kind of agentic analysis happening up front. So in the upper situation, you are very happy, you got a very reliable response because the agents knew that they can answer. If they are not, they're unsure, still they prepare, they give the superpower to the support engineer, and the case is then solved hopefully very promptly. This is really working very, very nicely.
So talking about the actual value, and this shouldn't be like a theoretical presentation. We really, let's say, brought some metrics or some statistics that we would like to share with you to show you there is really value in infusing AI in support. That it's not about firing support engineers, but it's rather about tangible value for you because we realize that's kind of the blue box is your mindset. Yeah, you don't want ping-pong. You want fast solutions, you want proactive solutions, preventative solutions, and that's your good right. You're paying support, right? So absolutely no discussion. We know you, we feel you, we know your persona very much, and that's why we infuse the AI little helpers and agents into the support process as we saw it on the process flow. So you have the GenAI search, you have the incident solution matching, you have categorization predictors. I will show it in the demo. The agentic thing working there Sunday afternoon 24/7. We have intelligent swarming. What combination of support engineers worked well in the past for a similar ticket? These guys will be invited to solve your ticket, and they will solve it much better than, let's say, any newbies who pick it up and tries to solve. So knowing support history, working behind the scenes also to orchestrate the human engineers, that's definitely what AI can help with. Outbreak detection. If 20 customers are reporting the same issue with the same root cause, we don't need 20 engineers, we just do the swarming, we pick the two which are the best, and they will solve it for all 20 customers because it's one issue, and they will also be able to help those customers who are affected because they are similar to those who are already reporting tickets. So nice preventative feature which is out there, and this and a lot more is already turning into very tangible benefits.
I spoke about the AI-powered knowledge search. So you know it in Get Support in SAP for Me, you can search. Our statistics show that eight out of 10 such queries deliver already the right answer that the person is not searching anymore or not creating a ticket. Now we kind of would like to measure it even more. Now the PhD experts measuring this are arguing, 'Hmm, if somebody searches before lunch and then comes after lunch and searches, is it still the same search or was the first not successful?' So we are kind of here in the complicated area, and we don't want to do any, like, personal monitoring of our users, that's not what we want to do. But we really thinking heavily about these statistics, does it appear right to you? Incident solution matching, this is really deflection confirmed by you in the surveys, 'Did it help me? I got a solution recommendation when I was describing my issue in SAP for Me or built-in support.' Yes. 10% get really, really deflected. Yeah. So you were close before, kind of pushing the button, 'Submit ticket.' You didn't do it because you got the right solution. With, let's say, many millions of tickets every year, this is a very significant, let's say, effort that we save to you and to ourselves. One-third of tickets get submitted already with a recommended product component. This is also important for us to be able to classify and put to the right team to process.
Agentic statistics. I'm really very proud of these. This is what I told you, this auto-response scenario. You're sending a ticket, agent responding. You happy? This is not science fiction. You are happy because 40% of these get really closed after maximum two SAP replies, and we are counting, like, 'Thank you' as a reply, right? So think of this, 30%. And with more replies, like, we solve pretty much everything because you are giving us a good note. The customer app survey is actually what makes a difference for us, and also my bonuses and everybody's bonus is tied to this. So we are kind of watching it very, very carefully for the agentic interactions. If you as our customers are asked, 'Give us the rating on the scale from one to five stars,' the agentic interactions, 90% of it get rating four or rating five, four stars, five stars. So this is not any illusion, any satisfaction. The agentic in the first responder situation are responding very nicely, and now we are kind of building it up as you saw on the previous slide. This is a nice anecdote.
Sometimes as a customer you complain the language used by the engineer was technical, it was blunt. He or she sent me some bullet points. It was not empathetic anymore. Well, we use now AI to give it a human touch. So the engineer can use the convenient feature of kind of polishing up the answer. The technical information remains the same. There's no change to the actual reply, but giving it a nice wrapper, putting it in a nice language, figuring out any grammatical errors, and making sure that it appears very empathetic. In many cases, a machine does that, and with surprisingly good impact. So this is what we do, this is what we measure. We're now starting also a research project with Technical University Munich where we would like to go to some of you as our customers and really measure whether you can confirm these statistics. And last but not least, I brought the demo with me. The screen resolution is not optimal, but this is how the interaction which is AI-powered looks like in SAP for Me. You know it, and there are some new things already shown. So that's why things still might change. This is why you start describing your issue. It's very critical, the description part here, the text that you put, the better you formulate your issue, the better the AI can kick in and give you a good reply. Also, you select your system, this is giving AI the context, and we know, okay, what installation is related to other, let's say, problems with similar installations as well. This is why also this is very important. So you make sure that you give us a good text and you isolate the system where things are happening, and you see already AI is kicking in from the side because when you scroll down, you will see there is already incident solution matching giving you a solution. So you haven't created a ticket, you're just describing an issue, AI kicking in, and AI, this we are very excited, there is GenAI now working. So this is not any GenAI model hallucinating, we really kind of ground them to the max, and we control, let's say, the creative power of the model that we get from the supplier. The knowledge which is coming is coming from the SAP knowledge base, but it's nicely put as a guidance, as a series of steps that you can follow up on. What a difference, just to mention, in comparison, just 'Hey, KB, look it up.' This is already giving you instructions. Recommended solutions are shown below, but this is not the first sentence of the KB article that can help you solve the issue. GenAI is already giving you a very nice summary in two lines. Major, major improvement, because maybe the first sentence doesn't describe what is in the article. Now it does, and you can see and you can select and you can do things. Imagine this didn't work. So there is, let's say, you would like to proceed to the next step, and this is where you can hit continue.
Already, this is what I mentioned about product and product function. This is very critical to put the ticket on the right product component. So AI can predict this for you. If you are happy, then good, you confirm or leave it as it is. If not, you can still change, but this is already where AI is, let's say, entering the workflow for the second or third time. There is really AI running behind the scenes. Then you would like to proceed, and AI again is already proposing the course of action knowing the nature of your ticket. It knows what worked in the past and can say, 'Okay, in this situation, the best solution is to create a ticket.' That's highlighted here as the recommended channel, but you can also use alternative channels below. So you decide to create a ticket. Here, for some of the tickets, you can enter, this is not AI, this is you, enter, like, a processor which is there 24/7 if SAP needs you. But then you are able to proceed to the next step where again AI is kicking in. It put the ticket on the right component, making sure that it lands with the right support team instantly. It also suggested priority where you can look at it, 'Oh, it's medium, I agree, I disagree.' So there's always a way to change, but already AI prefilled it so that it is processed with the right attention. So you confirm it's medium, and then you get the final, let's say, button here where you can create a ticket, and you can follow up on your mobile app in SAP for Me or whatever support access point you have selected. Simple workflow. You saw there eight to 10 AI interactions, and now the ticket is with us, and there will be many more AI interactions happening to solve it with the right speed and the right precision. This is what AI and support is about, and I'm very glad that you've been here with us and looking forward to keeping you updated through the next sessions, maybe next time what we developed else to help you. Thank you.
Are there any questions or comments before we proceed to the break? I'm happy, of course, to answer. Thank you so much.
