 We're broadcasting. Welcome to the Center for Global Enterprise Global Scholars Expert Connect series, leading the digital supply chain. My name is Ira Sager. I am vice president of global learning initiatives for the Center for Global Enterprise. This is the fourth in our series of six expert connects on the digital supply chain. In today's session, the second on digital supply chain technology enablers, we explore another important technology trend, artificial intelligence, and machine learning. Before we start, a few housekeeping people will be, all right, now I might mute button off. Welcome to the Center for Global Enterprise Global Scholars Expert Connect series, leading the digital supply chain. My name is Ira Sager, and I am vice president of global learning initiatives for the Center for Global Enterprise. This is the fourth in our series of six expert connects on the digital supply chain. Today's session, the second on digital supply chain technology enablers, we're going to explore another important technology trend, artificial intelligence and machine learning. But before we start, of course, there's a few housekeeping notes. Next week, there will be no expert connect because of the US holiday. We will be back on July 10th at the same time for discussion on another critical component of the digital supply chain, and that is people and your talent. But today's session, as with all sessions in this series, will be recorded and posted on the CGE YouTube channel. At the end of the presentation, you'll see a slide with the upcoming schedule for additional expert connects, the two more that we have in this series, as well as a link to the digital supply chain institute website for more information. We encourage you to both come back for the other two webinars and also to go to the DSCI website for additional information. We will leave approximately 15 minutes at the end of the session for audience questions. You will also be able to give us questions throughout the presentation. You will see on your screen a little button, a feature of Q&A that says Q&A. Please use that feature to submit your questions. We'll try to get to all the questions time permitting. As I mentioned, today's expert connect is about the effective use of artificial intelligence and machine learning. Our presenters will discuss how to use AI and ML to anticipate future demands and to predict the likelihood of future risks and why all that depends on those algorithms and software but on breaking down organizational silos and barriers. Today we're fortunate to have two experts with us that have experience with this important emerging technology. Our presenter is Shawn Numan, technology research leader for CGE's Digital Supply Chain Institute, where he has been leading our work in this space. Joining Shawn as co-presenter is Pablo Zegers, founder of Aparatix, which is based in Santiago, Chile. Pablo is leading a team in charge of building advanced commercial artificial intelligence algorithms for all types of applications but he's been involved with AI projects for the agri-tech sector and for contract life cycle management. With that, I'll turn it over to Shawn and Pablo. Thank you. Thank you, Ira. Good morning, good afternoon, good evening, everyone. Thank you for joining us on the second technology session. Next slide, please. What we're going to talk about today is the impact of artificial intelligence, machine learning and why you need to consider it and how it may fit into your business model. We're going to talk a little bit about what's the difference between artificial intelligence, machine learning, deep learning. Those are all words that they get thrown around in the press interchangeably. They really are very different topics, so Pablo is going to take us through and talk about the differences and talk about breakthroughs and artificial intelligence in the machine learning arena. And then we're going to talk about the impact that the technologies will have on the digital supply chain. Should be a good session. I look forward to active participation and hopefully we'll have a lot of questions as we did during the blockchain session last week. Next slide, please. Going back to our data model that I think has been a common theme through all four sessions to date. There are many new sources of data as a result of the advances that technology has made, the availability of sensors in the Internet of Things, social media, text data, images, videos, audio that produces both structured and unstructured data than that needs to be organized and cleansed and maintained. And you need to look for patterns across that data that affects your business, could impact your business model. So going forward, data courses becomes extremely important, the reliability and trust of that data, knowing what you have and how to make sense of it is important. So what we're going to do today is talk a little bit about how you make sense, how you organize that data, how you make sense. The yellow ring really in the middle of the slide that talks about artificial intelligence and machine learning. And of course, what pulls it all together and what differentiates you as a business from your competitor's business is how you employ the data that is available to you, make sense of it. And that really is by creating the algorithms that allow you to manage your business and differentiate yourself. So in the center here, we've got the algorithm console there in blue on this slide. That is something that is very important as you move forward with analyzing the data, assessing the impacts on your business, changing your, adapting your business models to the information that's available, changing your workflows. You really need someone to govern and guide that information flow and the changes that are created as a result of the information. We think the algorithm console is a name that we've come up with, but it's something that you will find, I think, going forward in all companies in the not too distant future. Next slide, please. In fact, we just published the results of our annual survey. I believe they were published either yesterday or late last week. We surveyed 112 supply chain and C-suite executives that responded to the survey. And a couple of the questions we asked is, do you feel that you need to get better at collecting and analyzing your data? 90% of respondents said absolutely, either agreed or strongly agreed that that was something they needed to do, needed to look at what was out there and categorize it. Next slide, please. However, 88%, you know, almost the same number that says we need to be better says we lack capability today and we need to dramatically improve our ability to analyze the information that's out there and to deploy it in our business model and enable us to manage our business better and to shift our focus to the right places. What we call the front side flip, looking at your customers, their needs and wants and ensuring that the supply chain is properly positioned to respond to their demands and needs. So there's a big gap. People recognize the information's out there. They're not quite sure how to use it, how to organize it, and they certainly realize that they're not making the use out of it that they should be and need the algorithms probably need to change their skills a little bit, change the organizational structure in order to remain competitive in the future. Next slide, please. So let's take a look at artificial intelligence and machine learning in a perspective here. This perspective came from Irving Waldeski-Berger. He happens to be a CGE fellow. This is from an article that he published in the Wall Street Journal last week. His view is that that artificial intelligence really is a general purpose technology. It is as meaningful, as important, and as generalized as as the internal combustion engine, electricity, computers, and over time we will we will see the the shift in importance of artificial intelligence. What AI really does, and if you look at if you look at computers, you know, I'm packing up a step here. Computers enabled digital operations by enabling arithmetic calculations in a in a quick and predictable manner. It reduced the cost of performing those calculations. It enabled us to do things faster, enabled us to automate those calculations. If you look at the internet, the internet really was a reduction in communication costs and in access to information. No longer had to pick up the phone and call people. You could you could shoot quick emails. You could you could research things quickly. You could communicate far more effectively than than you could before. Another game changing technology. AI is is is no less of a game changer. What it does when you think about it is it enables better forecasting and predictions, which enables more rapid innovation. Today, there's a proponent is a big data it's ubiquitous. Collecting it's a challenge. But things are instrumented out there in every business process you have is connected to big data in some way shape or form. Machine learning has has really made taken AI out of the out of the research room and and enabled it to become a mainstream technology. Because it's with machine learning, the your the artificial intelligence component becomes self learning. We're going to talk more about that. You need to you need to influence the machine learning algorithms, you need to direct them, you need to make sense sometimes out of out of what the information is telling you. But once you establish a set of rules, then then the machine learning process has become iterative and improve over time. AI will likely be one of the most significant general purpose technologies of the century, although we're very early into the century. But you know, much like blockchain, I would say blockchain still in the research phase at this point. AI is more advanced, is more general, the use cases are more broad. But AI is also in the early stages of deployment. Most of what's out there and Pablo will talk about this as we move forward in his section is is what I would call narrow AI use cases, very specific tasks API interfaces, you you feed it information and and it adapts and and learns on a very specific and focused area. So we're in the early adoption phase there. Next slide, please. Go go ahead. Okay, thank you. So if you look at if you look at the technology and the technology that's become available over over the course of mostly the past decade. And I think Hans Thauber from SAP says it well that to win in the digital economy, we have to reimagine how we do everything, you know, whether it's design, deliver, operate our products, what the assets are, creating digital assets. And one of the things that has enabled us to do that is cloud and mobility. So if you think about the the three Vs that are continuously talked about in big data, you've got you got volume, which is the quantity of data and where you store it. You've got the velocity of data, which is how fast it it comes in. Is it static? Is it? Is there a time delay? Is it real time? Is it streaming data? So you got to deal with the velocity? You got to deal with the variety of data? Is it? Is it structured? Is it unstructured? Is it? Is it a spreadsheet? Is it a database? Is it? Is it email? Is it social media? Some sort? Is it a video? So you got you got a wide variety of data that you now have to deal with in the future. But really, when you look at cloud and mobility, what that has enabled is the volume of data and in the velocity of data. It allows you to access data very quickly, categorize it, move it around, analyze it, and then storage is ubiquitous. You store it in the cloud, you got apps that are born in the cloud. Pablo will talk a little bit about this in a few minutes. So you've got a lot of AI interfaces, you got a lot of machine learning capability. Those things are all being born in the cloud, where the data is is more frequently residing and enables you to analyze things really, really outside your your infrastructure boundaries of your corporation. Next, please. With with with the ubiquitous data, we've got sensors out there. We've got automation capabilities. So you take all those data sources that we've talked about. And it's really this is where the variety of data comes in, right? You can think of cloud and mobility as volume and velocity, you can you can think about the sensors, the information gathering out there, the ability to to mine things that that are posted, or or across the internet, that really is the variety of data that is being the collection of data, which can then be automated and transmitted back to back to the cloud. Next, please. And then you've got you've got a collaboration. We talked a little bit about that last week with with blockchain. And, and, and there's really a fourth element to data. People talk about the three V's just really a fourth one, which is velocity, which is how how reliable is the data, how clean is it. And moving moving forward, that becomes increasingly important. Blockchain enables you to share data across enterprise boundaries, as we talked about last week, it enables you to collaborate. But as important as the veracity that you've got some faith in the data, because of the consensus driven algorithms that that enable you to rely on that data is as being being relatively clean or high velocity. And then with AI and ML, I think if you hit her again, there's a another header that shows up. You're really talking about augmentation. So it's augmenting the the the information that's out there. It's enabling you to optimize it. It's enabling you to predict what's going to happen with a with a high high degree of likelihood. Next slide please. The point between automation and augmentation is is important. And artificial intelligence and machine learning are much like we talked about with blockchain last week, you got to you got to make sure that you you have a business pain point, you got to make sure that you've identified the problem that you want to solve. And you have to make sure that that you know the outcome you want to achieve when you when you solve that. So are you really improving human productivity? Or are you looking to minimize the involvement of of humans in the process? And that's really the difference between automation and augmentation. So automation, you got things such as chat, chatbots for operational procurement. You've got defined well defined business processes such as those for compliance or auditing. You've got autonomous vehicles that move around the warehouse, these sort of things that you're looking, you're looking to minimize or automate the process. The case of augmentation, you're looking to you're looking to make the the humans in the process smarter, give them more information and allow them to make decisions more rapidly. Allow them to to guide the the the information and and make decisions from it. And of course, over time with machine learning, more and more of those those decisions get get moved into a a automated learning mechanism of some sort. So demand and inventory forecasting. And this is why the the algorithm console is important. How do you want to prioritize things? What's important to your business? What are you correlating? How do you how do you feed the the algorithms? And how do you direct the the machine learning and artificial intelligence capabilities that you have at hand? So again, complex, but really think in terms of, are you automating or you augmenting? And in how you're going to approach it? So as as we move move to Pablo's section, he'll talk about supervised unsupervised reinforced learning and how they're applicable to to managing your business. Next slide, please. And it's over to you, Pablo. Thank you, Sean. Hi, everybody. First, it's important to define what's artificial intelligence. The problem with the definition is that almost nobody knows what intelligence is. So that's why I rather use the term invented by Alan Turing, a universal machine, a machine device, hopefully to solve all possible problems. That's much clearer than artificial intelligence. And just to clarify, a robot is not the same as an AI, a robot can be a very clever mechanism that does a very complex, complex task. But it doesn't necessarily have an AI. So it's important not to confuse. Next slide, please. Now, the thing is why it's important. It's important because it's the first science that studies all sciences. When you when you work in artificial intelligence, you are somehow trying to make a machine that does philosophy, but also behaves like the philosopher or works with law, but also behaves like a lawyer. So you're trying to mimic to work and mimic everything related to the human knowledge and humans themselves. So it's what's called a meta science. And of course, with AI, we can produce, we can make machines that are able to perform very complex tasks. Everything that can be measured, everything that can be reduced to symbols is subject to the study of artificial intelligence. So that's basically almost everything that's in science. Next slide, please. Just to convey an idea of the complexity of the field, this is much more than mathematics, much more than psychology, sociology, economy. We are trying to model human beings. So if you start from the left on this diagram, you see that we are trying to model perception, which allows us to work with present, with the present, almost everything that's happening now in machine learning and artificial intelligence relates to this. Then with memory, we can we are able to work with the past. Research now is focusing on memory, how to use the past and mix it up with the present. Also introspection or imagination, because there it's where we plan and we produce alternative futures. So past, present and future, that's the main focus of machine learning, artificial intelligence right now. But it's a mixed up. Machine learning is just a subfield of artificial intelligence. Artificial intelligence has have existed as a field since 1957. And machine learning, it's a rather new subfield of artificial intelligence. And the two of them are complementing each other. To add to this, you need to add all the what's called the apiary knowledge, for example, two plus two equal four, that doesn't depend on our experience. It's true anyway, anywhere. To that you have testimony, which is communication between everybody, how to grow knowledge. And finally, you need to guide everything with ethics. So that's the most complex thing in artificial intelligence right now, how to mix ethics with these machines, we need machines that are able to behave in an ethical way. You see all the discussions about biasing in companies, gender issues, everything related to those topics is extremely complicated, and somehow needs to be reduced to machines. Next slide please. This is for those that are going to see the slides later. A description of all the main fields in artificial intelligence, specifically machine learning. You see lots of names like K means, K nearest neighbor perceptrons, boosting residual neural networks, then you have a statistical learning of Avni and Chervonenkes, lots of things that are very, very mathematical. This is not like programming where you, for example, you learn Python and you start building up an application. And once you finish, the application will behave as you expected. This is a totally different thing. It's very mathematical, but you need to know about philosophy about psychology and mix everything into something that has to be devised in a process that it's more like alchemy. It's more than art than a system. Next slide please. Just to clarify things, you have artificial intelligence. Within artificial intelligence, you have machine learning. And within machine learning, you have deep learning, which is what's famous now. And deep learning is using a metaphor, like our brain is composed by neurons, biological neurons. You have a sketch of them on the right side of a screen. Deep learning uses artificial neural networks, which are depicted in the lower diagram. And then you mix up millions of them, and you train them. Train them as just to a euphemism, to describe a process where you find the parameters that make them work together. Next slide please. What's important is to understand why this is happening now. Artificial intelligence has existed as a field since 1957. So how come now is making all this buzz? Well, something happened. In 2011, things that were extremely difficult were solved in a matter of months. And three months later, things that were thought to belong to science fiction, science fiction had already been solved. So what happened? This was understood recently in last year in April 29. Two researchers from the Hebrew University of Jerusalem discovered that there was a mathematical theorem that explained how to build complex systems out of simple ones. So this explained how we had been able to build these such wonderful and complex systems with such easiness. And what's happening now is that everybody is adjusting to this discovery. Because this is not, this is very important to understand. This is not like a mode, for example, the color of the clothes we use or the type of car we use. That totally depends on our decisions. But when you're based on a mathematical theorem, this is like we've been operating on two plus two equals five. And it was discovered that two plus two equals four. So everybody's switching to two plus two equals four. Because if you operate under those grounds on top of those grounds, you will be safer. So that that's explaining the breakthrough we're experiencing. Next next slide, please. Just to give you a brief sense of the things that have happened. On the right side of the screen, you have an image that has been labeled by the computer. On the right hand of the image, you see a glass of water with ice and lemon. This was labeled by a computer that had never seen the image before, without human intervention. And it was achieved in 2015, three years ago. Now this can be done with video. And they have totally surpassed the human capacity for doing this kind of things. Next slide, please. Here you see how in 2016, a year and a half ago, December 2016, researchers were able to produce fake imagery. So on the right hand of the screen, you again have, for example, flowers. On top of the flower column, you see a text. This is an artificial intelligence that just received the text and produced the bottom flower. So the capacity for producing totally fake images was reached one and a half years ago. And less than half a year later, the researchers were able to produce totally fake high definition video that was undistinguishable from the real ones. Next slide, please. In terms of manipulation, things are more complicated. Because even though artificial intelligence is ready for handling very complex manipulators, the manipulators as capable as the human hand still doesn't don't exist. So we cannot have the same human capacities. Next slide, please. In terms of language, this is sort of a boring example, but it's very impressive. On the right hand of the screen, you see a large paragraph on top of it. But on the bottom part of that column, you see a summary that was produced automatically by artificial intelligence. Artificial intelligence now can understand text, not only translating word by word, but understanding their semantic meaning so that they can produce useful summaries. This is from May 2017. So a year ago, I think it's an example from Salesforce. Next slide, please. This is a very simple, this is Google Translate. On the right hand, you see a diagram of the neural network, the deep learning architecture that's handling only all the translations. Each of those boxes has thousands of artificial neurons that need to be trained in very complex computers that many times take months of training. Next slide, please. And finally, to summarize the advances, here you see the Alpha Zero paper, where they show, this is from last year, they show a machine that was able to learn chess and top the best humans in less than four hours, which is absolutely impressive. In just four hours, being able to dominate the field, discipline and game that had been developed by humans in thousands of years. The machine automatically discovered most of the main combinations you can learn. It could take, for example, to me, years to learn that kind of things. It only took four hours to the machine. Next slide, please. So now what you see on the commercial side, you see all type of AI engines. Everybody's trying to grab a piece of the market. You see TensorFlow from Google, Azure and the Lewis from Microsoft, recognition from Amazon, Watson from IBM, Core ML, Apple, Einstein, Salesforce, the EPU guards of NVIDIA. Many of them are springing up, appearing everywhere, focusing on hardware, different problem, et cetera, et cetera. Next slide, please. So in terms of business applications, you see how they're pervading the market. But first, one of the most important ones are personal assistants. And here you see two major pleasures, the Echo and a plus Alexa of Amazon and Google Home from Google. Next slide, please. In terms of the companies, you see two trends increasing revenue. And here you see the ideas developed developed at the DSCI, where you have to model the customers. And based on that, you measure demand, you forecast demand, you stimulate the demand, and then you control it, you manage it. Next slide, please. And we're using cost. And this is taking mainly the following approach. They're using a system called RPAs, robotic process automations, where you can mix them up with artificial intelligence, and then you're able to process to, for example, device machines, those processes that are able to handle documents like people in companies. There are three major players, automation anywhere, blue prism, UI path. Next slide, please. The problem is the high level of the expectations. We've seen so many movies that we expect everything from artificial intelligence, but now we only have narrow AIs, AIs that are able to solve very specific problems. So before you introduce them into a company, you need to have people determine where in the company, then you need people to adapt the artificial intelligence, and then you need people to operate the artificial intelligence. So we are really far, if will be ever possible, to replace all the people. We need hybrid teams. Next slide, please. This is just to exemplify the hybrid teams, and an experiment done in 2013 by the Carnegie Mellon University at the Robotics Institute, where they were able to reduce the time needed to weld the chases of a handi from 89 hours to less than a third of a time. And in the 89 hours, they used three people. And on the other side, just a computer and artificial intelligence, a robot and one person. So that's an example. Next slide, please. That's your time, Sean. Yes, why don't we stop here and see if we have any questions? Or do we have any? No, we don't have any questions yet. But let me remind people that the Q&A function is available. If you have questions for Pablo or Sean, just write in that feature, we will come there and pull out your question and ask Sean and Pablo. Okay, we'll give it just a minute here. See if any questions come in. I have a question for Pablo, since we're waiting for others to enter. Pablo, are there any issues with standards in the AI field that we will need to see companies hash out? That's a very good question, because now there is a huge fight between the companies that are building the AI engines, and each of them is pushing for their own systems, trying to make them the standards. It's very complicated from our side companies, because now we're in a scenario where we have to use all of them. There are no standards yet. Very good. All right, we have two questions here. And when I got a lot, Lee asks, how are AI and ML related to the supply chain? And I'll ask a follow up. Dimitri also asked a similar question, do we have some examples of application and supply chain management? Pablo, would you like to take that one and talk about what Anastasia is seeing as you talk to your customers? For example, what we're seeing here in Latin America is that with artificial intelligence, you can measure forecast the behavior of the customers and predict it. So immediately you can start using the results to manage the entire supply chain. It's very, very interesting because normally companies have not thought about this in these ways. But with this, we are able to forecast the demand of people, the customers, and also the demand from the supplier point of view. So definitely, unless you're able to, you need to build these very detailed, very precise, artificial intelligence models of the dynamics of the companies in order to correctly manage the supply chain. Good. All right. Sean, why don't we continue there as the other questions come in? We'll, we'll get to them at the end of the session. Okay, all right, we can do that. Although I see, let's answer Greg's question here. He's, he's asking, are RPA tools using AI which is doable today? Yeah, definitely. Yeah, I'm saying that here in Latin America, I see that people are implementing RPAs and adding artificial intelligence to handle, for example, documents, they receive a document. The RPA extracts the information that goes through artificial intelligence that takes decisions concerning the process. So yeah, definitely. That's a yes. All right, Sean, do you want to take this next question? Let me, before we do that, let me, let me add to Greg's RPA question, in that I recently saw a demo by Budweiser, where they, they use our RPA to, to manage the last minute loading of their trucks when they have an order cancellation and they're looking to, to backfill a shipment. They, what used to be a very much of a manual process, they have now implemented a, an RPA solution that, that enables them with a, with high, with a high degree of probability to, to fill the, the space in the truck with what will likely be ordered while the truck is in transit. So that, I thought that was a pretty good, pretty good example. All right, let's, let's move on here, because I think there'll be some more questions coming up on, on a future slide here. But summarizing what Pablo had to say about, about AI and what it is and how it's being used and what, what, what the state of the art is, really, you need to think about outcomes that you're, that you want to be focused on. Are you talking about automation? Are you talking about augmentation of an existing process? What is the, what is the right solution? What is, what is the right platform for implementing that? We talked about, about the four Vs, the last being a veracity, as you, as you embark down the, the artificial intelligence and augmented machine learning, start with good data, start with data that you've got a high degree of confidence and high veracity seed data, that relatively clean data set, and make sure you've got consistent feeds and in downstream supply visibility, so that as you, as you improve your machine learning algorithms, as you improve your business models, that, that you're going against a relatively clean data set to start with, so that you can, you can start recognizing the, the anomalies in the data going forward. And many of the platforms can assist you with that, but it's important to start with a, a, a data set that you know is, is relatively clean. And then of course we talked about the human interaction, reinforcing learning models, feeding the, the machine learning algorithms, adjusting, adjusting for, for your, your business changes. If you don't know what vector auto regression is today, you probably will sometime in the near future, because it's certainly an important tool for, for correlating data in, in optimizing the machine learning algorithms. Keys to effective implementation, operationalize one process at a time. As Pablo said, most of the use cases that are out there today that are generally available are, are, are specialized, they're narrow AI cases. Look at your, look at your process, your workflow, experiment one use case at a time. As you, as you find ones that are valuable to your business, then I can continue to, to improve those. I continue to take advantage of the data that's out there, leverage the platforms. This is a, this gets into an interesting conversation. And that, you know, with, with AI, and the various platforms that are out there with machine learning and the various platforms that are out there with, with deep learning capabilities. There's a lot of tools that a lot of companies have spent a whole lot of time making investments in. Think about where, as a company, your competitive advantage comes from. Because that's what's important. And that's really in how you use these tools, the algorithms that you create, how you, how you feed them. There are also open algorithm sources out there today. You know, I was talking to one of the members not too long ago in, in their view was that the algorithms are a commodity. I don't necessarily share that view. Their view was what's important is, is the weighting that you give to the, to the data that feeds, feeds the models. But certainly as you move into the AI platforms and the machine learning platforms, those are things that are out there. People are making investments and take advantage of them. That's probably not where you're going to get your competitor advantage from. You, you don't need to be writing the AI platform. And, and you probably don't need an army of data scientists to do that. Pablo is, is a data scientist. He's from Chile. It's a fairly, fairly progressive country. Pablo is, is one of a handful of, of data scientists with the PhD and in Chile. So those, those folks are, are scarce. They're going to be leveraged across multiple companies, such as, such as what Pablo does. And so think about where you're going to drive your competitive advantage from. And, and how you're going to make use of the investments that, that others are making. Expectations are high. We mentioned that it's, it's early in the, in the adoption cycle at this point. But tremendous gains are being made very rapidly as, as Pablo showed during his presentation, as he said, this was three years ago. This was two years ago. This is where we are today. Major, major inroads are being made very quickly. Next slide, please. So we, the, the Institute published, published a paper, I believe, back in March, talking about, talking about AI in a machine, machine learning and how that will apply to, in effect, your supply chain. It's, it's a, it's a well written paper, if I do say so myself. And it's, it's got a lot of good info and valuable information in it. I would, I would direct you to the, the Institute's website. If you haven't downloaded, read the paper, read it, please, I think you'll find it very valuable. A number of things that we came up with words, our top 10 list, you know, what are the top 10 ways that AI and ML is going to transform your supply chain. Now, first of all, is, is data, the 80-20 rule. You know, today, your data scientists are spending things that, spending the majority of their time doing the things that are least satisfying to them, and probably not a productive use of their time, which is a cleansing data. So we got to reverse this. Instead of them spending 80% of their time trying to come up with high-varosity data, need to, need to take advantage of the machine learning tools that are out there and reverse that, so that they're spending 20% of their time teaching the algorithms how to evaluate data and dealing with the anomalies rather than the 80% that they're doing today, and that they're spending, and that these, they flip to spend the majority of their time on assessing the, the data, and, and figuring out how it impacts your, your business model. You got, you got the ability now as a result of, of the data and the information feeds to do real-time demand simulation targeted advertising. You can do dynamic pricing much as the airlines do today. You can change, you can change the price of your goods and services based on the demand that you're seeing out there. As you analyze your data, you're going to say, you're going to find new product opportunities. You're going to find some surprises and, and deal with those surprises. Generally, it's, it's good information. You may want to adjust the way that you market. You may want to adjust your supply chain. You may want to target a new product. But that's the capability that you have in the, in the digital world as you, as you transform your business. The data visibility will reduce your risk. Algorithms are, are important. You know, no bones about it. That's, that's where the battle is going to be in the future. You're going to need an algorithm console or some sort of governance structure to, to deal cross business units, cross, cross business processes to ensure that, that you're taking a holistic view of your business and adapting your models as you, as, as you digitize your, your business models and discover changes that, that can be made. You have to prioritize that. You have to adapt your business models. You have to ensure that you're, you're, you're prioritizing the information in the, in the correct way. And that'll become increasingly important. And then we haven't talked about it yet. I think we will in the next session. But all this will drive changes in the, in the skill sets that you need. Those people that you have on board today will have to change and adapt their, their talents. And you're going to need need some new skills. You're, you're going to need a Pablo or two to you're going to need a data scientist. You're going to need, need somebody who is, is skilled in managing the, the day-to-day information flows. Next slide, please. So really, you're going to see an intersection between your supply chain specialists, your data scientists. This is, this is what next week sessions all about. You're going to have to look at organizational impacts. Do you put your data scientists, your data information specialist, you put them in the business units? Do you, do you organize them centrally? Do you create a, if they're in the business units, how do you create a community, cross business units of the data scientists that allow you to, to leverage best practices and exchange ideas amongst them? How do you integrate the new skills with the old skills? And, and, and find that intersection that enables you to, to optimize your, your HR performance. So challenges, challenges to think about next week's session is, is all about this and it will go into, into a lot of detail. Before we move on to the, to the closing comments, let's, let's take a look and see if we got more questions here that, that we might want to talk about. All right, let's see, Raj asks, what is the spectrum range of applications starting from business intelligence to cognitive computing in the context of supply chain? Okay, that sounds like a, that sounds like a master's class. That's a broad question. Yes, very broad question. Pablo, you want to take a crack? For is, for example, what I'm seeing in the market are recommendation engines from the business intelligence side. For example, demand forecast as well. And on the other side, cognitive computing, you see everything that's related to personal assistance and handling human, human language. I don't know. I guess we can talk about this for hours. This is a cool question. Yes, yes, we could. Let me, let me ask a slightly different question, Pablo. Pablo works for a AI firm called Anastasia. And he has a lot of, a lot of commercial interaction with with clients around, around Latin America, Chile, Argentina, Mexico. When you go out and talk to your clients, Pablo, what are the things that they are most interested in hearing about in, in learning how to apply to their business? Well, that's interesting. The first thing they're secretly hoping is to have a sort of a useful terminator that solves everything. They've seen too many movies, like all of us. That's the first thing. The second thing is that they, the language is universal. They want to increase revenue and they want to decrease costs. So if you somehow translate the technology, artificial intelligence into that, then we, all of us are talking the same language. So now, for example, in concerning a recommendation engine, we are running for an important company in Latin America. We managed to get a 20% increase in the revenue in the channels that where the artificial intelligence is operating. That's the kind of things they want to, to listen to. But this is important because we have to work hard in order to produce numbers. Because if not, it turns into a political thing and ends up in, I believe you or I don't believe you. So they want to have numbers, objective numbers that have been measured, so they can prove that whatever technology, in this case, technology, artificial intelligence is useful for them. So that's the kind of things they want to hear. It really is proof of value once again. Folks, folks aren't interested in proof of concepts. They want to know what the value is that can be delivered from a, from a project or a use case. Yep. Pablo and Sean, I have a follow up question that I want to ask. And as you look at artificial intelligence and machine learning and you apply it to the supply chain. What do you see specific jobs, functions, tasks being automated? And the flip side of that, do you see new roles being created? Yeah. On my side, I see, well, firstly, the simple ones, not simple, sort of obvious is that you need people that know more about artificial intelligence, not the details, but how to use them to understand basically the limitations and to understand how to use them as a legal break, legal break, to assemble more complex things in the company. Then another field would be people that map typical processes into these new ones where you get these hybrid schemes with artificial intelligence and people working together. So we'll need that type of people. And companies are not used to thinking those ways. I agree with the idea of the algorithm council, because now companies are used to handling groups of people and technology that's fairly simple compared to this one. Because this one is autonomous. And it's handling very complex things. So you need different type of management. Also, you need people that sees the companies in horizontal way, because these machines interact with the customer, immediately create a marketing campaign directed toward the customer, then go to the logistics section, operational section of the company and check for the stock, for example. So they cut through all the silos, the internal silos. So it will be there is a need for people that also looks at companies and horizontal ways. Yeah, I think that would be Okay, and Pablo, we've got another question here that that I think you've partially addressed in your discussion here. But Greg Millen is asking, what's, what's our view of of how AI evolves over time or replace humans as as Jack Ma spoke about? Or will it augment humans as as Junior Medi at IBM talks about? What's your view? I've been reading a lot about that in the last years. Nobody's certain about what's going to happen. But I think we humans are much more complicated in some senses, and that's what we we know. And in other senses are where we're much simpler. So some important aspects of the society are going to be replaced by machines. And I don't think others are going to be replaced by machines. Creativity could be. Yeah, true. A machine cannot behave like a Picasso. But definitely much better than myself. I am not an artist. So normally, we tend to compare machines with the best humans in a certain field. But the problem is that all the rest are not the best humans in that field. So we could be replaced by a machine. So that that brings interesting problems, like the question is posing. It's still an open problem, I guess. But we're headed for a big change. That's my bet. Yes, it really gets back to the, to the discussion on automation versus augmentation and in in how that changes over time. You know, certainly the augmentation automation bucket is going to going to grow dramatically. The need for augmentation is going to become higher level, I think, and more specialized over time. Why don't we move on to the closing slide here? So some closing comments here and Pablo, let's do this interactively if we could. You know, I think that AI and hopefully what we've shown today is that AI and machine learning capabilities are really essential to your digital transformation. It's the only way you're going to leverage the data and the sources that that are out there today make sense out of it. And humans can't do things nearly as as quickly or find that it's find the same patterns that that properly focused algorithms can. Consider the consider your business models, you're going to find some surprises. You're going to find probably many surprises and you need to, you may find new markets, you may find new products that that fill gaps. But you know, one thing I can guarantee is that as you as you organize and make sense of the available data that's out there using artificial intelligence tools, you will you will find some surprises and you'll need to and you'll need to adapt to it before your competition does. Yeah, I agree. You're and I'm you know, I'm a big believer of strategies. You know, you got to have you got to have a strategy, you got to have a roadmap, you've got to have your preferred tools and and you got to consider how you're going to organize initially and then how that organization evolves, evolves over time as your skills, as your skills change. Obviously, when I'm sorry, I would like to compliment compliment with something is that concerning strategy. For example, if you see the strategies of Google, Facebook, all the big ones, Amazon, yeah, they have they have sorted out the very interesting strategies, but they still don't get it. It's even difficult for them. So if you don't have a strategy, you will be totally lost. Yeah, thank you. And then algorithms are going to become a source of differentiation. It's not the AI and the ML tools are going to differentiate you. It's going to be how you use them, how you apply the algorithms, how you how you correlate your your information to your business model and assess the priorities and the impacts, which means that algorithm consoles or some form of governance is going to be necessary. We call it an algorithm console. I think it's a good name, but it's it's it's going to become standard practice. Pablo closing comments. Yeah, I like the the idea of the algorithm council as well. It's a new way of integrating all of this. And I think there won't be other way. And we we are seeing this type of changes starting to happen here in Latin America as well. So that validates it. They have given them different names, but they're heading towards the same structures. I agree. Yeah, Pablo, I'd like to thank you for participating with me today. It's been a been a pleasure. Thanks. It's been a pleasure as well. Look forward to continue to work together. Back to you, Ira. Thank you. And I'll add my thanks. Pablo, we really appreciate your time and your participation in this program. And Sean, again, thank you. A wonderful program. And I want to thank our audience too. And as I said at the beginning on July 10th at this day in time, again, we will have our fifth expert connect series and we will be looking at talent strategy. Again, thank you very much. Bye, everyone. Thanks.