 I'm going to begin, you know, as I said, it's been a 30-year experience roughly for me. I began my AI journey as an undergrad where we built a Bayesian expert system for medical diagnosis in 1990 and I got me hooked then. So I'm going to talk about use cases of custom experience. We've heard a lot about them already today. So I'm going to speak about it from the perspective of a software company delivering product to execute on those use cases. So first of all, has anybody here heard of Genesis? Hands up if you've heard of Genesis, okay. Good, a few. Okay, so I'll introduce you to Genesis. So we make contact center software infrastructure. So we're a world leader, a recognized world leader in customer experience software across markets all around the world, high-end customers, mid-market customers, on cloud, on premise. So we're a pretty big brand in this area and we process, roughly speaking, I think last year we did about 25 billion customer experiences through our platform. So we handle a lot of volume and we're mediating that critical relationship between the brand and the customer and we're in the middle of it all. So that's a very interesting place to be right now. I'm going to take a bit of a digression into a little bit of a story around AI in the course of my career, the key milestones that have really struck out stood out to me, rather, and why Genesis is really, really investing heavily in AI now because of the level of confidence we feel in the technology. And I'll go back to 1996, when it was a big landmark when Deep Blue beat Kasparov in the chess game. And that was an example of old-school AI with explicit knowledge, logic, search, heuristic search, and lots of computation, brute force search against a person. Fast forward to 2002 and that was really a time when Google's search engine had really become dominant. And at the time I was working in NASA, and we were just across the, we were sharing a campus with Google in Mountain View, and one of their leaders, a guy by the name of Peter Norvig, who used to work at NASA and wrote the book on AI literally, was giving a talk to all these researchers who spent their careers in symbolic AI and old-school AI. And he basically said that Google is figuring out all this stuff through a bunch of statistics. And, you know, just number crunching and data make all the difference. So that was quite a telling comment to me. A couple of years later, still at NASA 2004, has anybody heard of the DARPA Grand Challenge? No? Okay. So it's a thing funded by the military in the U.S. The challenge was to have a robot car traverse through 150-mile terrain in the Mojave Desert completely autonomously. And it was a race. So it was very exciting. And I remember the first year when this happened in 2004, it was a complete wipeout. I think the car that got the furthest was seven miles in. It was a total disaster. And it showed really the limits of AI to be able to autonomously control the system. It was also the time where the MapReduce algorithm emerged. So that was the beginning of a lot of interesting things in data. One year later, that epic failure of the DARPA Grand Challenge was completely modified, where several teams completed the challenge in grand time with their Sebastian Throoms team from Stanford winning. And that was basically good sensor data, so better sensors using statistical and probabilistic techniques to deal with the uncertainty in the world. And that was proved. And some good, solid engineering. And that was a massive improvement. Also, we found Hadoop emerged for processing data at scale. Still, there was some old school kind of AI techniques. That picture there, if I can show you this here. That's a satellite that was designed by a genetic algorithm. So, and it's a very, very performant little satellite that no engineer would probably ever have come up with those kind of antennae. Moving along further, we know Jeopardy won by Watson. We've had new devices like Siri, so virtual assistants emerging on the scene. Another key thing that emerged around 2012. I was not really focused in this area, but I learned about the incredible results coming out of image classification. And in particular, one video that many of you may have seen around demonstrating deep learning, it basically took all this image data and using unsupervised techniques and deep learning, it had developed all these features and it basically learned to figure out like features of faces like eyebrows and noses and things like that. And it basically developed automatically this representation that became the basis for subsequent algorithms to be very highly performant. So that was where sort of the performance results for these benchmarks really started to transform with deep learning techniques. And then, you know, we have more and more platforms coming out, more and more source of data. And we're here now in 2016, just a couple of years ago, AlphaGo won using deep reinforcement learning in the Go game and beat the best in the world, I think. And, you know, we're seeing robotics happening with Google Translate made a quantum leap in its performance. So this is all sort of the march of AI. So what we see when I think about AI, I think about the old school world, which is not to be discounted, which is, you know, handcrafted knowledge, logic-based heuristic search. It works well on narrow domains. It's very poor handling uncertainty. There's knowledge engineering bottlenecks. This is sort of useful, but it has its limitations. We also see robotics, it typically has operated in very uncertain domains. So it has driven the technology to create these intelligent agents working in these complicated dynamic uncertain domains. So it really developed a lot of probabilistic techniques for decision-making. There's a bit of, it drove explosions or benefited from explosions in computational power at low prices to drive sensors. And this is really, this explosion has really led to this proliferation of machines that we're seeing now. So lots and lots of platforms out there as well. And many of them are new ways for customers to interact with companies. And of course, big data, statistical learning at scale, the tooling around this stuff is amazing. There's tons of open source available. So it's really taken off. And then you put all that together with deep learning and you have a massively powerful technology that you can bring to bear to create whole new classes of products. So coming back to the customer experience side of it, we've actually heard a lot about it today. You know, I heard the speaker this morning on customer journey analytics. I heard about omni channel dark data and the problems of unstructured data. Well, and dealing with all these channels, Genesis basically solves the problem of customer experience for a brand. And we're dealing with all these diverse, often unstructured channels, many of them loaded with really, really insightful information. And the challenge basically is to figure out how to get the right experience to the right person at the right time. And that right experience is related to giving them the right content, having them routed to the right person, et cetera, picking which channel. So this is a massively complicated problem. And it's hard to do in a unified way across all these different channels. So I'm going to talk a little bit about some use cases that my group, the AI group, works on. And the first broad category is self service. This is a huge area. You know, it goes from, you know, FAQs on your website to now we're looking at cognitive IVR, right? So this is bots connected to your IVR system to make it much more intelligent. Chat bots, conversational, frequently asked questions. These are all areas that we are actively developing in and have products in market. And agent assistance, you know, we heard this morning about augmented reality or augmented computing or augmented human AI. You know, the idea of all these agents and contact centers using AI to help them be more capable in their jobs, give them more skills, but also to have them teach the AI. Having an agent who's very skilled in how to deal with a customer can give us a moderated incremental approach for deploying AI. And we can change the tasks they do to make them much more about moderating and supervising an AI-based system. The task, you know, with agent assistance, you just imagine you get information in coming from your self-service bot. It's now rich and contextualized. We can route it more effectively because we know more about it. And when it comes to that agent, we've got a better agent for the job. And we can pop up more context about what they should do. We can monitor the conversation across voice and digital channels in real time. And we can surface up actions and recommendations to the agent on the desktop. So routing is the bread and butter of Genesis. We have contact centers that we serve that might have 10,000 agents. And we get an interaction comes in finding the right agent. So I'll talk a bit more about predictive routing a second. But that's like a fundamental problem. It was previously done with very, very laborious, brittle, rules-based systems. And AI is a transformational technology to solve that problem. Finally, on the predictive engagement side, which that didn't come out too well, this, of course, is a huge area where every single touch point, every single event that comes between the customer and the brand is an opportunity to take action. It's an opportunity to figure out what's going on, estimating the behavior of the customer, estimating the behavior of the agent, and figuring out the next best action to take. And this is a critical thing across all your channels. So predictive routing. This is a product that we developed about, we began developing about two years ago. It's now been in market. And it's about a year and a half. The basic idea of this product, interaction comes in. We've trained a model to basically predict the outcome we want to optimize. So let's say it's first contact resolution. We have historical data on which agents had which interactions for which customers and what the resolution was. And we train models to predict for any given agent and any given interaction and any given customer what's the probability of an interaction. And we use that as a ranking function. So we rank the agents on every single interaction. And we can build a matching algorithm that optimally ranks and matches people. We make the routing, our routing engine executes according to this ranking function. We have the call, we get the feedback, and we determine how well it worked. We've been able to generate just on with one telco, we were able to save on average 70 seconds of handle time on every call with no negative consequences just using this algorithm. So it's very, very powerful. And we're applying it for all sorts of measurable target metrics. So, for example, it's sort of a principle of how we operate to do A-B testing so that you can prove the value and ROI to the customer. Another example, this is touching on the agent assist and self-service use cases. We announced a partnership with Google and we demoed in July with Google Next with a main stage demo where we're demonstrating a cognitive IVR and agent assist use case using our infrastructure and Google's contact center API. And we have an open AI or a bring your own AI policy. So we work with our own homegrown solutions. And we also work with leading partners. Another one I just bring up, shout out to Majek. He's probably, where are you, Majek? There you are. So we bought a company in Galway called UltraCloud who is a journey optimization company. They were doing predictive engagement and still are for behavior prediction and next best action. And so we bought that company in, I think we closed at the end of February. And we've been growing it aggressively ever since. And that's becoming our hub for R and D in Genesis in AI. So I spend a lot of time in Galway. So in closing, that's just a flavor of the application of AI in several use cases, some of many of which we sort of alluded to in the morning. Customer experience is a critical issue. We're sitting on tons of great data. We're sitting on interactions at great scale that have a lot of value if you can improve the way they work. And it's a fantastic opportunity to spend your career working on cool problems. So I can recommend Genesis is a place to work. And we are hiring across all engineering functions in Galway. So thank you very much.