 All right, I can't do it. There we go. Now everybody can hear me. Excellent. So thank you for the introduction. That is exactly what we are going to do. We are going to talk about real use cases for artificial intelligence. And the reason is that there is so much hype around artificial intelligence and the things it can do and machine learning and impact a society and things like that. And all of that is great. However, a lot of customers have the question, what can I do right now? What can I do today? There's a little bit more challenge with AI because the technology is complex. It's not well understood. You have data, you have algorithms, and you have to deploy those algorithms. It requires not just one person, it requires not one type of skill, but many different types of skills coming together. So it's complex. So the best way that I have found when customers come and say, hey, what can this technology do for me? I feel like I should be doing something with it. I just want to get started. It's really good to start talking about, well, here's a customer in manufacturing. Here's a customer who's in government. This is what they are doing. And once you go over those use cases, I feel it is easier for customers to now start to understand what their possibilities are. So that is what we will do today. We'll talk about five customer stories that are using artificial intelligence. I will name all of the customers, and you can go to our site and learn a lot more about them. Just a quick question. How many of you know about SAS Institute? Just a quick show of hands. Okay, there's some. That's fantastic. For the rest of you, just a quick introduction to who SAS is. So we are not very well known as evidenced in this room, but we're big. We are number one in advanced analytics in the world according to IDC, which is an analyst firm. We are number two in the world in revenue in 2018 for artificial intelligence software platform. When it comes to Gartner and Forrester, they'll tell you we are a leader in AI and machine learning. And finally, in March of last year, we announced that we are investing a billion dollars, US dollars, sorry, into artificial intelligence research over the next three years. However, that's a quick introduction to who we are. More importantly, let's talk about our customers. So this is just a quick slide that shows you a subset of some of our customers who are doing really, really cool things today on artificial intelligence. You see Honda's up there. You see Lockheed Martin's up there as well. Unfortunately, we will not have time to talk through all of their stories. I would like to. You can go to sas.com slash customers and see all of them and read a lot of detail about what they're doing. Today, I think we will have time to cover five. So we will talk about Epipoli. This is a customer who is doing micro-targeting in Italy. They're located in Italy, but they are all across Europe. We will talk about a company called Volvo. I think everyone has heard of Volvo, a really, really great brand, and how they are using IoT with machine learning and what kind of benefits they are getting. We will talk about Swisscom, a large telecom company in Switzerland, and they are using natural language processing and text analytics for better customer experience. We will talk about Amsterdam University Medical Center. To me, this is exciting because this is the future of where artificial intelligence technology is going. I will tell you a little bit about that. Finally, my favorite is Wild Track. Just curious, how many of you have heard about Wild Track? I do not see any hands go up. It is a fantastic story. Customers are located or operate in Namibia and Africa, but now they have spread to different parts of the world. I will show you how they are using computer vision and how they solved some of the problems they were trying to solve. The way we will tell these stories is we will talk about who the customer is, we will talk about what problem they were trying to solve, we will talk about how they solved that problem, what type of data, what type of algorithms, what type of results. And because every good story has a good lesson that you can take away, we will talk about some lessons that I learned. You may learn a different lesson, but I think that is okay. We can apply these lessons to other AI projects. So if everyone is okay with that agenda, then let's go ahead and begin. And we will start talking about Epipoli again. They are a customer who focus on advanced analytics and data. They believe that that can provide them a lasting competitive advantage when it comes to micro-targeting. The problem again, they were trying to solve, and it's just really interesting, how many of you recognize the picture that's in the background of this slide? Okay, here we go, a couple of hands go up. For everyone else, this is the island of Capri in Italy. And when you sit down with Epipoli and they tell you about their business model, they say, we want to be just like the lemon cello makers of the island of Capri. They say, okay, what does that mean? Well, they say that if you think about this drink that started in a small island, in a small country, but it got so popular because those lemon cello makers, they knew their customers taste so well. They knew their customers' preference so well that the drink got very popular. It started being exported to the rest of the country and then now to the rest of the world. We want to be like that. We want to know our customers just as well as they did. Very interesting story. Okay, this is how they did it. So I've sort of defined who the customer is, the problem they were trying to solve. Let's get into the detail. So with all artificial intelligence, machine learning, analytics, projects, it starts with data. And they're in retail, they're collecting data on sales and products and performance and promotions and things like that. Everyone is doing this. This is not something new. Any retailer who is operating in this century has been collecting this type of information. Where Epipoli did something more interesting is they're collecting what I'll call more strategic data. So they're finding out where customers' preferences. Where does a customer make a buying decision? Is it in their bedroom? Is it while they're waiting for the bus? At what price point is the trigger? Do they make it over the phone? Do they do it face-to-face? Do they do it over email? How are they making these decisions? So these preferences, they're collecting this information. And they're in Europe, in Italy, so they're complying with all the GDPR requirements. So there's some opt-in type of elements that have to happen here. The next step, and I'll be honest with you, many of our customers, they get excited about artificial intelligence. They want to talk about computer vision and natural language. And what I give a lot of credit to Epipoli for is spending so much time in investing in the data part of the entire project. They decided to establish a data warehouse, a brand new data warehouse. Now this is not required. If you want to do it, I think that's great. That's what Epipoli did. But as long as my advice will be as long as your data sources are covered under a single data policy, they can be in AWS, they can be on the cloud, they can be on premise, but as long as you're covering under the same policy of governance, of security, of GDPR, then I think your data practice is good. Epipoli decided to create a brand new data warehouse and they partnered. So with about 250 partners across Europe, they now have accessed about 7 million customer profiles. And they keep these very updated in terms of changes, in terms of new purchases, et cetera. So after they've established really, really solid data practices, they are now using machine learning to figure out the patterns. And they're also using some really fundamental concepts such as just behavioral modeling, segmentation, economic impact analysis. And the combination of all of that is they send the promotions and the offers directly to the customers. They say that they're operating in stealth mode because they never publicize in big letters their promotion, so their competition never sees it, but they're able to go out to the customers and give them the right product at the right price, at the right location in a way that the customer feels comfortable making that purchase. So hopefully this starts to help you understand how they did it. The results, 29% increase in average value of transaction. That is huge, because if a customer is giving you one or 10 euros last week, they're now giving you 13 euros. That's a big increase. And they overall increased the base customers by 11%. So really, really good result for Epipoli. I think the lesson for David Turin at least is if you want to be successful with artificial intelligence, you have to look beyond the hype. You have to look beyond AI. And you establish a really strong and a really well-governed data practice. And I think that was the key for success for Epipoli. And for a lot of customers, I think that is going to be important as well. Okay. So now I think you understand how we're going to tell these stories. So I'll start going a little bit faster. Next up is Volvo. Volvo is a great brand. It does not need any introduction. But specifically in this use case, they are manufacturing trucks. And these are vehicles that go long distances and they haul lots of freight. And when you think about Volvo's business, they are just manufacturing these trucks. They are not operating these trucks. They have big logistical companies that operate these vehicles for them or for their customers. And if you're one of their Volvo's customers, you want to make sure that those trucks and those vehicles are on the road, hauling your freight, and they are spending less time in the repair shop. So the problem that Volvo wanted to solve was increase uptime. Have the trucks spend less time in the repair shops. So they created a new service offering and they sold these service offerings to some of their customers who were interested in buying them. So let's see exactly what they did. Starting in 2017, Volvo started adding a lot of sensors to their truck, to their vehicles. And let's just take a look at what these are. So they are collecting information on the vehicle such as what is the engine RPM? What is the gear position? What is the load that the vehicle is pulling? So in addition to this, they are starting to collect data on the environment around the vehicle. So what is the temperature? What is the humidity? Is it an incline? Is it a decline? And collecting all this data, they establish a pattern of what does normal look like for a certain type of vehicle, for a certain type of load, even for a certain type of... for a certain path that the vehicle is taking. So once they establish what normal looks like, now every time there's an anomaly, they can use machine learning to understand what is going on. But before we get to the algorithms, let's just pause and think about the data, the amount of data here. So they have about 175,000 trucks that are equipped with these sensors. That means they're collecting about 20 million measurements per day. Now they're not interested in all of these measurements. They can't really manage 20 million alerts every day. But what they wanted to do is, again, find the pattern and understand where the anomalies are. When is something operating out of spec... out of spec... specifications? And because all AI projects also have to combine some kind of process change, they established a new group called the Volvo Group Telematics. And these are the folks who hold all the customer contracts. These are the folks who coordinate with different repair shops. So that when a truck rolls into getting repaired, the people who are going to do the repair, they already understand what are some of the problems with it. If there's a part that needs to be ordered, that is already ready to go. So this is how Volvo did it using IoT and machine learning. And they're getting, again, great results. 25% reduction in downtime of a vehicle. So that's big. If my truck was in the shop for four hours, now it's only going to be in for three. It's big-saving for their customers. And as a logistics company, that's big news. The lesson again for me is that if you're thinking of AI projects or IoT projects, think of combining them together. Think AIOT. Especially for the manufacturing customers. We have a lot of folks who are doing manufacturing and they say, I want to use IoT on this factory floor because it's not performing as well as an identical factory floor that I have in another location. And I think IoT can help me. Yes, it can, absolutely. All that data, data is expensive. You have to put it in the cloud. You have to pay for it. You want to be able to very quickly run machine learning algorithms, find patterns, and get rid of the rest because it's going to cost you money. So again, if you're planning IoT projects, think about using machine learning on those projects very, very quickly as well. Okay, let's talk about natural language processing. Really, really cool topic. Our third story is from Switzerland. We're a telecom company in the country and they get a lot of phone calls. They get a lot of service calls from their customers, about 10,000 calls every day into their call center and it's across four different languages. What they wanted to do is instead of listening to auditing just 3% or 4%, which is what the capacity that they had, they wanted to listen to all of them 100%. And understand what the customers are saying. And actually, the previous speaker talked a little bit about that. This is sort of taking that to a data source that is not public, but that is something that you have proprietary to yourself. And by the way, oftentimes, the data sources that you have to yourself, if you run analytics on those, you will get much better insight, much more meaningful data, and that competitive advantage will come again from data sources that are yours. You can absolutely combine them with publicly available data, and that competitive advantage only on publicly available data is a little bit more difficult in my experience. So this is how Swisscom is doing it. Again, their data sources is call center transcript. Someone calls in, they record the call, they transcript it, and now that becomes part of their text analytics. They're starting to do a little bit more on social media as well and connecting new data sources. Then they do text analytics. They're doing categorization. So for example, if a customer calls and say, hey, I was watching football, I was in Starbucks in this neighborhood of Madrid and I had some quality issues. So the machine learning tool will automatically understand Starbucks as the location, it'll understand football as sort of the event, and it'll categorize all these things automatically. The next thing that it does is that based on all this data, it assigns a severity to the problem. So no longer is the customer service representative have to say, well, I think this is a big problem, I'm going to give it severity one. Or I think it's the same problem, I think I'm going to give it severity two. You take that away and you create an experience that is uniform across all of Swisscom's locations because they have the algorithm that is making that decision and that decision is aligned to how Swisscom wants to treat their customers. One of the important things for Swisscom is to surface all of this information on where these problems are happening to the executive team as well. So we created a visual interface for them and they were able to again understand much better what the customers are saying. I don't have a specific result here, so I'll just read off a quote from our customer Albert who is with Swisscom and he talks about the fact that when you have a fully automated view and a daily view, a report that gives you an understanding of where these problems are, just creating that visibility gives our teams a lot of data to be able to now solve those problems and get to a much better customer experience. The lesson quite frankly for me is natural language processing is underused. If you have been thinking about text analytics or NLP, you can hear everything that your customers are saying and a lot of things that not yet your customers are saying as well. You can use natural language processing and get a lot of really, really great results. If you're operating a call center in multiple languages as well, absolutely think about using NLP. I think it has a lot of capabilities and potential. Okay, my last two stories. This is the second last one. Amsterdam University Medical Center. They're located in Amsterdam. They're a small clinic and they focus on using computer vision to understand disease progression. And this is an important differentiator here. Again, there's a lot of hyper on artificial intelligence that says that you can cure cancer. You can detect cancer faster. I want to be very careful with this topic because cancer is something that it's a frightening disease. It's touched all of us in terms of either we've experienced it ourselves or we know someone who's experienced cancer and it's a destructive disease. Where we have seen hospitals use machine learning for cancer is not in detection but to understand how this disease is progressing. And I'll show you exactly how Cancer Center Amsterdam or AUMC did that. So the data here is different now. The data is image data and there's three types. The first is x-rays. The second is MRIs and the third is CAT scans. These are the data that they use. They bring all of this in and they use an object detection method. It's pretty simple, convolutional neural network to understand how the tumor is. And once they've established that baseline now when that patient comes in in one week, in two weeks, in three months or whatever time that is, the algorithm is able to detect three changes in the tumor. What is the shape change? What is the size change? This is really a measure of the difference in color of the tumor. So now instead of having someone take a look and measure this tumor manually which may or may not have some errors the machine learning algorithm does a really good job of measuring against these three specific variables and put that report out to the physician. But this is why the next step is why I think this is so exciting to me. Rather than just stop on computer vision let's say we're incorporating AI in here. AUMC uses natural language processing as well. They pull in data such as clinical data. What is the patient's medication history? What is their diet history? Any family history as well? So they talk about, yeah, computer vision gives this great insight on how the tumor is progressing but we want to look at the whole patient. And by using computer vision and natural language processing they can get to a much deeper insight to the physicians to make the final recommendation to the patient on what treatment options they should pursue. And again the reason this is exciting to me is because business problems, real world problems will not be solved by one piece of technology. That is a fact. Again a lot of times our customers get excited about computer vision or natural language and they want to see what potential this has but I'm here to tell you that it's going to take multiple AI capabilities to really be able to solve a problem. This is an image in the SAS Viya which shows three medical images and these are CAT scans that have been stitched together in 3D. And using data and analytics the tool recognizes the organ which is the liver in this case of the patient. And in orange it is able to identify some of the areas that have tumor. Now it's able to identify it, it sets the baseline next time this patient is in here it's able to say what's the change in shape size and density of the tumor. So a physician can understand what is going on with this patient. Again I'll read off a quote from Dr. Garit who is part of AUMC and he talks about the fact that yeah a scan shows you a lot of data but once you start combining that with other types of information like natural language processing and clinical data you get to a really really good result and you start to understand what's happening with that patient. And again like I said real world problems are messy, not one technology is going to solve it you should look for solutions in artificial intelligence that not only have computer vision but have optimization and forecasting and natural language and machine learning and deep learning and create that whole tool box because it is going to take all of it to solve some of these problems. The last story which quite frankly is my favorite the companies called or the customers called WildTrack they operate out of Namibia and Africa in addition to other parts of the world and the problem they were trying to solve is they wanted to come up with a non-invasive way to monitor wildlife. What does that mean? So the way you track an animal today is that you can bring the animal down by either introducing some type of sedative in their food or you use a dart gun and you bring the animal down. Well that's disruptive to the animal anyways but there's been some studies that show that if you do that to a female rhinoceros then it becomes more difficult for that animal to have little baby animals. So the species that you're trying to save you end up hurting a little bit using the current method. So WildTrack had a different idea. They said can we use computer vision to look at the tracks of an animal and see if we can identify what it is and what they're doing. Here's how they did it. The data they're collecting is collected very differently. This is collected via crowdsourcing and I'll show you at the end of this presentation sort of how exactly they're doing it. But every time there's a footprint they invite people on their website and I'll show you how that looks like to collect that data in a certain way and send it to WildTrack that becomes part of their overall data source. They then use an image classification algorithm to identify four things. The first is what is the species? Is this a hyena? Is it a cheetah? What is it? The second is what is the gender? And these are probabilities. So in all honesty just like we humans make a determination I think this is this type of cat or that type of cat the algorithm makes a prediction that with 90 percent certainty this is a male or with 95 percent certainty this is a female. The third is what is the age group? It doesn't tell you the exact age but it tells you the age group. And the fourth is if there is specific marking on that animal scar tissue or something else it can even show you down to an individual animal. So these are the four predictions that the algorithm makes. They connect this to some overall map and surveys that we have so they can start tracking and creating a pattern of where this animal is going. And they are able to connect any government programs such as anti-poaching efforts such as wildlife conservation efforts into their effort so that the local entities that are now solving wildlife issues have all of this information. And their portfolio has grown and started off in Africa but as you can see on the slide is on many different parts of the world as well. The result that they are seeing here is over 90 percent accuracy in identifying all four of those predictions. Again to me that is really fantastic because if you think about in the wild you have mud, you have clay, you have snow. You are taking pictures at different times of the day when the sun is making different types of shadows on the image as well. So being able to identify 90 percent plus accuracy is really really good result for our customer wild track. And just a quick shout out to them. This is if you go on their website if you happen to be out somewhere and you see a track you can take a picture of it you can use a couple of things to make sure that you are collecting the data the right way and then upload it to wild track. I think it is a fantastic cause and it just helps from a conservation effort as well. The lesson and the last lesson here for me is there is absolutely hype in artificial intelligence and I agree with you. That does not mean that this technology is limited. There are so many use cases for artificial intelligence but if you are wanting to make changes on a big scale like conservation think about crowd sourcing because as you start to collect that data you can get data very very quickly good quality data as well that I think is again so very crucial to some of these artificial intelligence problems. Hopefully that was helpful to you. We talked about five different use cases and I hope to be telling your story someday as well. If you go to sas.com slash ai you can use all of our tools without cost for a certain period of time and again try some of these capabilities that we talked about today and that is all I wanted to show you. Thank you very much again for your time and we have about three and a half minutes for questions so actually what I'll ask is some of my colleagues from sas to come up here as well that way you can ask questions in either English or if you're more comfortable you can ask questions in Spanish as well so I'll introduce Inés and Alberto thank you very much. See if I can help. So if you have a question put your hand up and I'll see if we can see you in English on Espanol shout as well okay we're not seeing any questions am I allowed to ask a question? Absolutely So I'm curious about different use cases and you know are these are you finding smaller companies I have a smaller company so how can I get involved in this because it sounds like it's going to be expensive Yeah yeah good question So do you only work with huge clients yeah yeah good question so that is where I think AI has sort of expanded the market for analytics I'll be honest with you you may have a different opinion but I feel that analytics and AI is very similar it's about finding patterns in data whether the data is image data or whatever it is or speech data you can find patterns and then make a prediction large companies have been doing this for a long long time they've been doing analytics at scale for a long time but AI and just around it has created new customers smaller companies that they want to do this as well and I encourage small companies I think the investment is that you need to make in these type of technologies is not something limited to large companies also and as the technology gets simpler then smaller companies can get some of these same benefits as their larger competitors Sure and do you find there's a lot of companies that maybe have a business problem but don't realize there is an AI solution or assume that it's too expensive or outside their means so there's also maybe some education required about how AI can solve certain types of business problems That's absolutely great and this is how this market is a little different I think talking about some of these customer use cases and what others are doing helps customers understand their possibilities as well the best approach is I have this problem and I want this solution with artificial intelligence what we see we don't know what the right questions are yet to ask so the approach of leading with some examples hopefully helps customers understand what they can do Thank you David Last chance for questions Last chance I think you guys won't have the opportunity We might have one up there Did you see one? No? I did not We'll be around later on so please come by and talk to us if you would like Thank you very much