Salesforce Einstein Keynote: AI for Everyone | Salesforce

Ladies and gentlemen, please welcome Senior Vice President and General manager Salesforce Einstein, John Ball.

Good afternoon, Dreamforce, and welcome to the Einstein keynote. Is everyone fired up to hear all about Einstein? Come on. All right. Well, we certainly are, and that's because we're actually at the beginning of a revolution in AI. And this revolution in AI is already transforming the consumer world. Sometimes in simple ways like product recommendations, sometimes in very profound life changing ways like artificial hearing for the deaf and hard of hearing. And what's amazing is what powers that are these little devices called cochlear implants. They're about the size of a Bluetooth headset. And with that, someone like my daughter, who was born completely deaf, can now not only hear but enjoy all the music that she loves from bands like Coldplay and of course, U2. And so this AI revolution, it's not only going to transform the consumer world, it's going to transform the work world. It's going to transform how all of us work. And that transformation is going to be for the better. So we couldn't be happier, prouder, more excited to be here, to go deeper than ever on Einstein and share our vision for AI here at Salesforce. And there's really nothing that sort of shouts out excitement like our safe harbor statement. So with that, this AI revolution, it's leading to a smarter world. And this smarter world has been built on sort of the mega trends that we've all witnessed over the last 20 years, starting with the web and of course the cloud and social and mobile and IoT. And now everything and everyone is connected. And why that's important for AI is that all of these connections are generating orders of magnitude more data than ever before. And thanks to the cloud, we have virtually unlimited compute and storage capacity. And so it's really that combination of massive data and massive compute that's leading to this revolution. And you experience this revolution every day as a consumer. When you go to Amazon and you see a product recommendation, when you go to Netflix and you see a movie recommendation, when you see that photo that's automatically identified and tagged in your Facebook feed, you're experiencing the power of AI. Now, wouldn't it be great if our business apps had those same AI smarter powered experiences? Wouldn't it be great if our business, our sales app, could tell us which leads are likely to convert, or our service app could tell us which cases are likely to be escalated? Well, with the power of AI, we can do that. Unfortunately, for the vast, vast majority of companies, it's just too complex. It's out of reach. They can't apply AI to their business process. And that's because it's hard with AI. First it starts with data science. And to do data science you have to have data, so you have to collect and integrate all this data. And then for AI you have to do this crazy stuff called data wrangling, which is basically transforming the data so that you can use it for machine learning. And then you have to hire these scarce data scientists and then they build these predictive models and then they have to maintain them and refresh these models and they have to have the infrastructure that is trusted and secure and scalable. And then after all of that, after all that work, you then have to take these predictions and these recommendations and put them into the context of the business user. And so for all of these reasons, the vast, vast majority of companies are just not applying AI to their business. And so we asked ourselves a couple years ago, how can we change all of that? How can we simplify all this complexity? How can we democratize AI just like we democratize CRM and social and mobile in the enterprise? And so we embarked on a journey and we basically said, let's take our CRM experts, our UX experts, our world class data scientists and let's simplify this AI for our customers. And so along this journey we also then identified some amazing companies with amazing talent. Companies like MetaMind, Breaking New Ground in the area of deep learning. Companies like Implicit applying AI specifically for the sales process. Companies like Prediction IO leading the way with open source machine learning platforms. And we all came together and we built Einstein. We all came together, we built Einstein. And so we couldn't be prouder to introduce Salesforce Einstein. Einstein is AI for Salesforce and it's built right into the platform. It takes the world's number one CRM and makes it the world's smartest CRM. It empowers sales, service, marketing and IT professionals to to be their best by making every customer interaction faster, smarter and more predictive. And since it's built right into the platform now everyone can build AI powered apps faster and easier than ever before. Now let me take a step back and explain how all this magic works. So earlier I said to do data science, you need a lot of data. Well, one of the fantastic things about doing AI here at Salesforce is, is we have tons and tons of data. We have structured CRM data like accounts and leads and cases and opportunities that you all know and love. But we also have communication style data, things like emails and calendar events. And then we have what I call interaction data, things like web clicks on e commerce sites or marketing campaign responses. And of course, we have social data coming from community cloud and chatter and marketing cloud, social studio. And what Einstein does is it takes all of this data, all these rich signals coming from the millions of users and the billions of transactions that we process every single day. And we apply deep learning and machine learning and natural language processing to really do four things to discover insights, to predict outcomes, to recommend the next action to take, then tying all that together to automate a process. And since it's built right into the platform, we can then surface these insights and these predictions and these recommendations in the context of the business user so that the business user can be their best and connect with their customers in a whole new way. Now, I said that Einstein is AI for Salesforce, it's AI for CRM, and it makes Salesforce the world's smartest CRM. Einstein is all about transforming the way that you work to focus you on what matters and give you time back. So, for example, for the sales professional, we help them prospect more efficiently, sell more efficiently, forecast more accurately, and we give them time back by automatically logging all of this activity during the day. Now, I don't know about you, but I've never met a sales rep in my life who doesn't want to have more leads. Everyone wants more leads so they can close more deals, but they also want to have better leads. They want to have the good leads. And Einstein delivers that with predictive lead scoring. We can predict which lead is likely to convert into an opportunity. Then once that lead becomes an opportunity, we can then predict through opportunity insights, which opportunities are going to close and become customers. But not only that, Einstein is advising and helping that sales rep throughout the process. So we can, for example, detect that a competitor was mentioned on a deal or that a deal is trending up or trending down, or that they're not engaged at the right level in the account. Now, for marketers, marketers want to engage with their customers. So Einstein can predict who's going to open, who's going to click, who's going to buy, who's going to unsubscribe from a campaign. That allows marketers with predictive journeys to basically send personalized one to one campaigns to the right person at the right time, increasing their engagement and increasing their roi. But of course, it's not just for sales and marketing. It's for service and for communities and for commerce. And in the winter release, coming around just around the Corner Here in October, we're shipping 17 Einstein powered capabilities across the clouds. And just a note on scale, we're already now producing hundreds of millions of predictions and recommendations every single day. Now, I just showed you Einstein in the context of CRM, but here at Salesforce, we look at everything from a platform perspective and Einstein is no different. So with Einstein, our vision is that every user, every admin developers of all skill levels will be able to infuse AI into their apps. They'll be able to embed AI into their apps. And so since Einstein has built so deeply into the platform, admins can take these Einstein powered fields or these AI powered fields and they can use all the goodness of the platform that they already know and love. Page layouts and assignment rules and workflow rules built with Lightning Process Builder, and they can infuse that AI right into the app and customize it to their heart's content. Now, for the more technical folks, we're also introducing new APIs and services like the Predictive Vision service and the Predictive Sentiment service. What this means is that the millions of Salesforce developers can use the same environment they already know and love and they can now access state of the art deep learning to do image recognition, image classification, sentiment classification, and they can do that without becoming deep learning experts. That's how you democratize AI. Now lastly, for the super technical hardcore data scientists, we married the best of Heroku and Prediction IO so that you can build these custom predictive models and put them into production without worrying about scale and reliability and DevOps. Now, I've said several times, Einstein is built right into the Salesforce platform. What do I mean by that? So we start with a multi tenant trusted infrastructure. Everything we do here at Salesforce starts with that, so you don't have to worry about reliability, scalability and security. And then we add in that amazing asset of data coming from the millions of users and the billions of transactions, CRM data, IoT data, social data, email, calendar. And what we've done is we've layered Einstein right into the platform. Right there you've got your deep learning, your machine learning, your natural language processing, and we use that to then build these intelligent apps in sales and service and marketing and communities with our lightning development platform. Now, I think now it's time to see Einstein in action, to bring it to life. And to do that, we're going to walk you through a couple of customer stories and a couple of awesome demos. And to walk us through that, I'd like to invite Jim Sinai up On stage, our VP of marketing for Einstein. Please give Jim a warm Dreamforce welcome.

Thanks, John. So Salesforce Einstein is all about helping you be your best to do things like discover insights, predict outcomes, recommend next actions, and even, in some cases, automate tasks and business processes, all so you can be closer to your customer. Now, when I think of Einstein, I like to think of Einstein like my own personal data scientist. I don't have to understand how to wrangle the data. I don't have to understand even how to use it or why it's working. It just works to help me be my best. Now, we're going to look at a story today about a company that we all know and love. And this company is called Square. And just like we're in the business of democratizing AI for all of you, Square is in the business of democratizing financial services for thousands and thousands of small businesses with devices like this, their new contactless card reader, and numerous other products. But they have a real hard problem. Why? They have thousands of businesses that their reps have to interact with, and that's a huge prioritization problem for both sales and service. So we're going to take a look at how Square is using Einstein to help their sales reps be their best. Now, if you're anything like me, you start your day by grabbing your phone call and checking her email. Often, sometimes before the feet even hit the ground. And here's an email from Salesforce Einstein with an alert. And it's telling me that I'm gonna miss my quota. Now. I don't need coffee. This gets me up. I am awake. You have my attention, Salesforce. And I slip right into the Salesforce Wave app. The Sales Wave app here. And I can take a look at what's going on. Look, I'm a good rep. I make club. I don't need to miss quota. But I'm trying to understand why it's forecasting that I'm going to miss and so I can go in and I can start exploring. Well, if I'm targeting all industries, I'm going to miss by 25,000. But what if I explore and see if I just focus on a certain industry and Einstein is going back and looking at all my historical data and putting together this analysis right on my phone again. Remember, I haven't gotten out of bed yet, and here it's told me 5,000 is the best I can get it if I focus on restaurant industries. So it's with that frame of mind, I'm speeding into the office Well, I take Muni, so speeding is. Use your own interpretation. And I get into the office and I pull up a list and this is great. I got thousands of leads. Look at all these leads.

These are great.

I can prioritize them by title, company, lead source, maybe operating system. This is not helpful to me because even if I'm doing my best, even if I'm closing 20%, I have to talk to 100 leads just to close 20. What I need is some help. I need salesforce Einstein. So let's take a look at what being a predictive seller is all about. Well, there's that forecast. We don't need to talk about that. We know that's right in my face. But then there's also these new insights up on the top, right? And we can see these insights. We can see this new one. Julie Chavez came in and came in just 10 minutes ago and it scored it as a 9 to 2. Einstein is saying, hey, you know, this is a lead you need to focus on. But why? Well, Einstein has looked at all the leads that have ever converted. It's looking at the standard fields, but also my custom fields. And it says, you know what, restaurant and industry and title have a high probability to increase the score. And it also things like custom data like what page on the website they looked at. So this is great. And when I go back to my lead list, I can transform that whole lead list and prioritize it by this score. So now I'm giving my best energy, my very best energy to the very best leads. But you know what, as great of a rep as I am, I can't do it without my trailblazers and sales operations, My trailblazers, all of you who help customize the sales cloud for me and we can go right into process builder and extend this lead score because it's just part of the platform. It's just a standard field in the platform. So I've had my trailblazers build a little simple process here. If it's low, send it to Pardot Nurturer. If it's high, you can send it right to me by way of chatter, or you can create an email or whatever alert you want and we'll send all the rest to the bullpen and let the junior AES take care of it. So we've looked at how to discover insights and predict outcomes. What about next actions? How do we recommend the next action to me? Well, let's step over to the opportunity page and we're here on our Opportunity Kanban board. And what's new are these little color coded arrows, up, down, and the gray ones horizontal. And I've been working this company, this deal, this company, the Sock Depot, they make these really cool socks. If you've seen these Einstein socks, sorry, here for the camera, figure that out. But I have this red insight and I'm like, what's going on here? So we can open up this opportunity, we can take a look, and we see that the prospect has raised an issue. Now, what's going on in this scenario here? Einstein is looking at all my email, it's looking at all my calendar data, and it's applying natural language processing to identify positive and negative sediments. And these words unusually slow and won't work. Sales have been unusually slow and won't work. We need to get Julian up and running on a mobile system, and we need to get his sales up because at Square, we make money when our customers make money. So I'm thinking to myself, how do we help Julian? And, well, in this case, I'm going to send him an email. But that button, of course, is customizable by you. And so we're going to send him an email. I'm thinking, let's get Julian to Dreamforce, because if we can get him to Dreamforce, he can sell a boatload of socks and I can give him this mobile reader. So I send the email and I'm all about automating things. I don't have time to log in Salesforce and Einstein's got my back. Einstein has said, you know, we put that right into Salesforce for you, Jim. And guess what? When he replies, that's gonna be automatically. Oh, and there's the reply right there. That's automatically in. He's on his way. This is great.

Hey.

Hey, Jim. What's up, everybody? This is me. I'm having difficulties with my bike. Hi. Hey, good afternoon, everybody. That's me. I'm selling socks. Einstein socks.

How's that? Like, who wants some socks?

Julian, I got your reader right here. You want some socks Right here. This is phenomenal. Socks. Yeah, you want some socks? Julian's taking his business mobile. Yay, Julian. Julian, I got your mobile reader right here. It also has a new chip card functionality, so you can. You can accept those new chip cards.

That's amazing. Yeah, I love it.

Great. So can we close this deal and move it forward?

I think we can close this deal.

Move forward.

All right.

And thank you so much. And you have some socks for my fans here. You guys want some socks? Who wants some socks?

Let's Go grab some socks.

Thank you, guys. Throwing out some socks here. We're back here on the homepage. I don't have enough you. And. If you want more socks, you can go downstairs to the Einstein Zone. And so look, a day's work with Einstein. Einstein's got my back, and I'm back on track to hit my number. And so with that, we've looked at how to discover insights, predict outcomes, recommend next actions, and automate business processes. And so I want to welcome the head of sales for Square Taylor Casino to Dreamforce. Taylor, come on up and join me on stage.

Welcome. Thank you.

All right, so I talked about the reader, but you guys do a lot more than just the credit card reading. Tell us about what Square's up to now.

That's right. You mentioned we are democratizing access to financial services. It really started with accepting payments. But as our sellers have grown and their complexity and needs have grown, we've had to expand the products we offer. So we offer point of sale now. We offer customer engagement marketing tools. We offer payroll. In some states, we even offer small business loans through Square Capital.

Now, you guys are not just a financial services company. You're also a technology company. And so you're looking at AI holistically and how you're bringing AI into your business. How does Square think about this AI revolution?

We love AI. We use machine learning already. It's actually a big part of what made Square possible in the first place. We accept 95% of people that apply. That was unheard of in the industry. And we accept them in five minutes. It used to take five weeks. And we do that through machine learning. We take a very inclusive approach to accepting people that apply to us. We assume you're a good actor, and then we use all of our machine learning algorithms to make sure that you actually are a good actor. And if you're not, it'll get flagged. More recently, we've actually been using all the data we get. We have millions of sellers on our platform to think about new products we can develop for them. So a great example is Square Capital that I mentioned, where we can extend small business loans through Square Capital to businesses, and we can do the right amount of money to the right seller, and we can have really conf. Sorry, a high degree of confidence in the underwriting of that through our machine learning.

And you have all the analytics around that small business.

That's exactly right.

So as a head of sales, you obviously are dealing with a high volume of customers.

Yes, indeed.

You need your reps to be their best. How is this AI revolution in Salesforce Einstein, really, how are you thinking about using this for your team?

It's going to be huge. One of the biggest challenges we have is lead scoring. We see thousands of new leads every week and they're very, very diverse. We have coffee shops, full service restaurants, doctors, lawyers, hair salons. You name a business that accepts credit cards, they're on square. And so we also see a very diverse range of sizes. We've got small businesses that do hundreds of thousands of dollars a year in revenue, all the way up to hundreds of millions of dollars, even billions. And so no two leads look the same. So we really excited to get smarter about our leads through Einstein, and that's really going to be about lead scoring and helping our AES know what leads to work, when and how. So that's going to be huge. The other thing I'm really excited about is just finding the insights. We have so much data that having Einstein actually show us what are some processes that maybe we aren't thinking about that have been really effective that you can raise up and show us so we can get much better at coaching our AES and just making them better at their craft.

And that's what we're here to do. So this is great. A big round of applause for Taylor.

Thanks, Jim.

So we looked at AI in the context of CRM, but I'm sure we're all asking ourselves about the platform. And here to talk about AI and the platform is a woman that after yesterday needs absolutely zero introduction, our own head of data science, Shubha Nabar. Shubha, welcome to Dreamforce.

Thank you, Jim. So you're all excited to find out more about Einstein? Yes. Okay, well, as you can see from that demo, the entire customer success platform is getting smarter. And as my boss, who's sitting over there likes to say, intelligence is the electricity of our era. And I thought about this when he first said this. I thought about it and he's right. I mean, just not too long ago, people used to talk about their electric toasters and their electric ovens, but now people just talk about their toasters in their ovens. And the very same way nowadays we talk about our smartphones and our smart watches, and Mark likes to talk about a smart toothbrush. But very soon we'll just have phones and watches and toothbrushes and everything else will just be stupid. We're going through an AI revolution. And our vision at Einstein is to bring AI to everyone, enabling all businesses to be smarter, more effective, and connect with their customers in a whole new way. But first, let me tell you a little bit about AI. So AI provides us with statistical learning algorithms that can be applied to massive amounts of data to learn patterns and then generalize from these patterns to make predictions about the future. So for instance, here we have these images of Parker Harris, our co founder, dressed as lightning man and as Emmett Brown. And we can apply learning algorithms to these images to learn the pattern of Parker's face so that the next time we see him, we can still recognize him even if he's rassed as Einstein. And we can apply these same learning algorithms to all your data, to your business data, to your CRM data, your email, calendar, social data, to learn patterns and make predictions about customer behavior in the future. So, for instance, how likely is your customer to churn? Or, you know, why is he likely to churn? How should you reach out to him and even automatically generate what the text of your communication should be? So let's take a look now at how Einstein touches each of those three things, the data, the models, and the business context, making for a completely unique and transformative AI platform. So first, Einstein is data ready. All, there's hundreds of thousands of businesses and millions of people who use Salesforce every day to manage all kinds of customer touchpoints. All that data, the CRM data, the email activity calendar data, social data, even IoT signals, all this data already resides in Salesforce and you no longer need a team of data engineers to wrangle it for you. It's just there. But solving the data aspect wasn't enough for us, because the biggest challenge for us is that we want to bring AI to everyone. And every business has its own unique business processes, its own data, and its own unique use cases. The Churn prediction model for fanatics is likely going to be very different from the Churn prediction model for square. And we simply do not have enough data scientists in the world to hand tune the models for every single customer. So we had to build something that was game changing in a very principled way. We had to build AI to create AI. So feature engineering, feature selection, model selection, hyper parameter tuning, sanity checking, sampling, recalibration, all the things that a data scientist would ordinarily have to hand tune for every single model she builds. We had to automate. And you know, after yesterday, a lot of people were asking us, you know, do people, Does Einstein use SVMs or logistic regression or random forest? And the answer is yes, it uses all of them and it automatically finds the best model personalized for your business, your data and your use case. And if you didn't understand too much of that, that's fine. The point is that Einstein is modeling ready and you don't need a PhD to use it. But more importantly, Einstein is CRM ready. And that's because it's integrated with the CRM platform. And so it's really easy to take the results from your predictive modeling and and integrate it back into the business workflows where your business users can make use of them. And so if you think about it, Einstein touches each of these things, the data, the models, and the business context. And it enables predictive experiences that are deeply integrated with the entire CRM platform. But most importantly, because Einstein is integrated with the platform, it's production ready. Which means that you don't have to worry about a whole host of production headaches such as scaling, scheduling, security. Einstein is data science as a service. It just simply works. And with that, I think it's time for another customer story and I'd like to call Ali.

Thank you so much, Subha. Thank you. It is my privilege to be here today and to introduce you all to the Fanatics to. So for those of you who don't know, Fanatics is the global leader in licensed sports merchandise. So chances are, if you've ever purchased a jersey or a baseball hat from your favorite team, it's probably been from a site that's powered by fanatics. And I'm particularly excited to deliver this demo today for a couple of reasons. One, I'm a marketer and you're about to see this is a really transformational marketing story. And two, I'm a big sports fan and I've been a Fanatics customer since I was an undergrad at Stanford, and Fanatics supplied me with all of my Giants gear once I moved to the city, which I got to bust out last night. How about that game, right? Yeah, but whether it's college sports, the NFL, the NBA, fanatics knows that the relationship that a sports fan has with his or her favorite team is a really, really personal one. And so that's why the marketers at Fanatics are hyper focused on getting the right content to the right customer on the right channel at that exact right time. And they're using Salesforce and Marketing cloud Einstein to do just this. Last year alone, Marketing Cloud powered over 4 billion emails and 27,000 campaigns for fanatics, all with only 5 full time CRM employees. Maybe 6 if you count Einstein. But let's take a look at how they're Doing it and pull up the demo. So our story starts with the fanatics marketer, Katie. And like any great marketer, Katie is extremely data driven. But unlike most marketers who are looking in their rearview mirror at analytics in the past to determine what to do next, Katie is actually looking into the future with predictive analytics powered by Einstein. And what this allows Katie to do is to meet her customers exactly where they are on their journey with fanatics. Better yet, Katie can be a step ahead of her customers. And Katie starts every campaign on her favorite dashboard. It's her predictive scores dashboard. And this gives Katie every metric that is important to an email marketer. She can see her opens her unsubscribes her clicks, and probably most importantly, her conversions. And right away, Katie's eye is going to that red right there. She's down since last week. She's only getting a fair rating. And Katie wants to be her best. So she's going to click into View Details and see what's going on with her conversion. From the second dashboard, Katie can see all of the attributes that are leading a customer to either convert or to not convert. Things like their frequency of opens, their email domains, whether or not they're bouncing. All of these are potential correlations just waiting to be uncovered. But the amazing thing is that as marketers, we no longer have to uncover these correlations ourselves. We're going to let Einstein do it for us. And the first correlation that Einstein is discovering is in the form of a predictive audience. And what this means is that Einstein is simply segmenting our audience for us. And I think those marketers in the audience can really appreciate this. Segmentation is one of the most time consuming and manual things that we have to do as a marketer. And now Einstein's going to do it for us. And right now, Einstein is grouping our audience into four buckets. And the window shoppers are particularly interesting to Katie because they have a high open, so they're engaging with fanatics content, but they have low clicks so they're just not converting and making a purchase when they get to the fanatics site. And Katie knows that this is the perfect group to nurture in a one to one journey. And she's going to do that with Journey Builder, the number one platform for customer journeys. And with one click, Katie can send that audience to Journey Builder. And this is Journey Builder. This is our journey canvas where every marketer can bring together that audience, join channel, timing, content, all in one place. Katie's gonna start with her audience. That group of window shoppers. Now it's all about targeting each of these window shoppers on their favorite channel. And using Einstein's smart split technology, we're able to look through that group of window shoppers and to determine which channel each of these customers wants to be targeted on. Right now, Einstein's even recommending the distribution, email, mobile and Facebook. This is incredible for marketers. So now we can drag this onto our journey canvas in one click. So we've optimized for our audience, we've optimized for our channels. Now it's all about timing. Katie wants to make sure that every window shopper is targeted by fanatics at that exact special moment when they're most likely to make a purchase. So with send time optimization powered by Einstein, we can do just this. We can figure out when that magic moment is for each customer and target them and engage them at that time. So we'll drag our timing onto our journey canvas. And now we're ready to go. We have our journey optimized for our audience, our channels and our timing. So let's see what this journey looks like when it comes together for one of our window shoppers. What do you know? It happens to be me. As I mentioned, I'm a big sports fan and I've been a Fanatics customer for a while now. And this is the email I received when I was getting ready for Dreamforce this morning. It's recommending ladies Steph Curry jerseys for me because I think Einstein is smart enough to know that I love my dubs and I know without a doubt that Tim, our amazing demo driver in the back, is also a fanatics customer and an avid sports fan. So Tim, what does your email look like from fanatics?

Hey, Ali, I don't know how you.

Could tell, but I feel a bit underdressed.

I actually didn't get an email, but I was on my Facebook feed and I could see this ad from fanatics and it's for Pacers apparel. Clearly the ad worked well.

Tim, I guess we won't hold it against you that you're not from here, but this just goes to show the power of Einstein. Einstein knew that I was a Warriors fan. They knew that I wanted to receive my ads via email at 7am this morning when I check my email every day. And they knew Tim was a Pacers fan and wanted his content so served up to him on Facebook during his lunchtime where I know he's always killing time on Facebook. Tim, Tim, Tim. Anyways, I don't know about you, Tim, but this email was just enough to push me over the edge and to get me to purchase this jersey. So with just a few clicks, I can tap through from my email, I can go to the fanatics site, order my Steph Curry jersey and be ready for the home opener against the spurs in a couple weeks. So fast forward to that day that I've been waiting for. The day that my jersey arrives in the mail. I also get a push notification from fanatics and it's telling me that if I post a selfie in my new jersey to the community, I can enter for a chance to win season tickets. And those of us from here know that the warriors have been on quite a winning streak over the past couple years. Tickets are pricier than ever. So I am without a doubt entering this contest. With one click, I I can tap through to my fanatics community and right away I'm in this immersive, branded, personalized community for sports fans. I can see relevant information and there's some selfies from other fans. There's Tim, I think he beat me to it. And oh, there's Mr. Benioff himself. I think I actually saw him at a game last year. His seats were a little bit better than mine. We can also see trending articles and products, everything that's personalized and relevant to me. But I want to post my selfie. Like I said, I want to enter this contest. I want these season tickets. And with just a few clicks, I can select that perfect image, may have applied some filters to it, select that perfect image, post to the community, scroll down, hope to see my selfie at the top of the feed, and cross my fingers that I win these tickets. Now let's take a second to talk about what's going on under the hood. Now what you don't see is what Einstein is doing with each of these selfies. It's not just a fun social promotion. Einstein is actually classifying each of these selfies as they come into the community and using Einstein's Predictive Vision service, powered by deep learning to recognize physical attributes of the customers in these selfies and with amazing accuracy. It is realizing that Tim is a short blonde and it looks like we're getting a new selfie. Let's wait. Oh, it's me. Let's see what Einstein figures out. So Einstein knows with 98% accuracy that I'm a long haired brunette. This is incredible, incredible power and technology for marketers to be able to turn images and just noise on the social web into marketing value. And what Einstein's actually doing is serving up the next best product to recommend to me because based on some of these physical classifications and physical classifications of customers that are similar to me and it's recommending this White warriors headband. And the next time I log into my community, what do you know, I see my feed, but the product recommendations have been updated and I see that White warriors headband right there, powered by Einstein's recommendations. Now this is the power of AI. When you put AI in the hands of your marketers, you're allowing them to create predictive one to one journeys that will turn all of your customers into fans for life. So with that, I want to invite up Chris Orton, the chief revenue and Marketing officer from Fanatics.

Thank you.

Thank you for being here, Chris. Thank you for letting us share your story.

Absolutely.

So we had a chance to take a peek at the Fanatics demo. I would love for you to tell the audience a little bit more about your role as cmo, maybe some of the challenges that you were facing and how you thought AI could help with these challenges.

Sure. So, I mean, first of all, getting to work at Fanatics is an amazing experience because we basically are totally dependent on what happens on the field.

Right.

So at the end of the day, what's happening on the field dictates who's going to sell tomorrow. Being able to predict what those sales are going to look like is a huge, huge part of our business. And without the ability to predict it, we wouldn't know what to buy, we wouldn't know what's going to show up the next day. So it's important probably for everybody here to understand the complexity too of our business. So in addition to Fanatics.com, we run NFL Shop, MLB Shop, NBA Store, NHL Shop, the 49ers Shop, 300 different professional team and colleges. It'd almost be impossible for us to figure out a way to scale our insights across all those without some back end engine, some sort of math engine behind it powering that.

Great. And I guess that's Einstein, right?

That's Einstein, exactly.

So tell us a little bit about how you're partnering with Salesforce and Einstein specifically and maybe if you could share some roi, that would be totally terrific.

Sure. So I think one of the challenges for us again is this notion of hot markets. So until about 9pm yesterday, we didn't know if it was going to be the Giants or the Mets who were going to win the game.

We know now and it was the.

Giants, so that's good. But so we didn't know. And because we didn't know, we didn't know actually what marketing needed to kick in right after that. And so we tried to figure out how do we actually set up the campaign in advance to be able to target folks who are going to be interested if the Giants do win. Of course, we had an if lose scenario. I'm a die hard Giants fan, but we were prepared in the event that the Mets had won. But the basic idea is we wanted to be sure that we could actually target folks who've told us either explicitly, which makes it really easy, or implicitly, which is really hard. Have people told us that they might be interested in Giant's gear? And that's really where I think a lot of the pieces of the puzzle come together for us. Because as somebody traverses our site or as somebody engages with an email they send out, we're collecting all that data in the cloud and we're storing all that data so we can go back and use it the next time we need it. And we can actually send an email out 9:01 last night as soon as the Giants win to all these folks that we know are going to be interested in what we have to sell.

I think I probably got that email.

I'm sure you did.

So tell us a little bit about what's next for your journey with Salesforce and with Einstein. What does the future look like with AI?

Sure.

I think for us, if we could, like I said, the on field drives so much of our business, which by the way, makes this super exciting and really, really hard as a career. I think if we could figure out a way to predict the outcome and who's going to actually win the game, we can all go to Vegas and I can just stop selling T shirts, which would be awesome. But short of that, I think what we've got to do is we've got to kind of figure out a way, can we figure out how to kind of take these connections to the next level and actually go drive them deeper into the organization? So we've been on a journey in the marketing side to build predictive models, to build these insights into the way that we operate every aspect of our business. But we've been really focused on the marketing side. Can we push that deeper into our buyers? So imagine the buying side of our business where they have to actually, six months ago, figure out how much Giants gear should I buy, how much should I buy for the Mets? If we could actually push the predictive elements into this that says, listen, if the Giants do Happen to win a wildcard, this is how much product you need. I think that'd be transformational for us. So really it's taking the kernels that we planted on the marketing side and pushing them through the rest of the org.

Well, we couldn't be more excited to be partnering with you on this journey. Thank you so much for being here, Chris, and for letting us share your story.

Great.

Thank you so much. Let's give Chris a big Dreamforce round of applause. So now I am honored to introduce our chief scientist here at Salesforce, Dr. Richard Socer, who is one of the brains behind Einstein and I think he kind of looks like him today, right? Got the same outfit, the hair.

Thanks.

I did not think that this jacket would get an applause. I'm glad I saved it for so many years in my drawers. All right, so you already know that Salesforce has phenomenal engineering and product teams that have been bringing you this great CRM. I'm really excited to announce to you today that we're also starting Salesforce Research. We're starting to refocus on artificial intelligence. And AI is still in a field or covers a broad range of things that go from we already know how to solve this, we know how to do lead scoring, but we could do a lot better with the latest and greatest techniques all the way to things that we can't yet do perfectly. We can't yet ask questions, complex questions that require us to logically reason over a large knowledge base and unstructured text. We can't yet do perfect machine translation to translate from any language to another. We can't yet have perfect dialogue systems about a broad range of topics. Those are still interesting research problems that we work on. Now with Salesforce Research, you can know for sure that the data scientists that you get with Einstein is always learning, is always improving, knows the latest state of the art techniques, but also actually pushes them forward with research. We're publishing. We're excited to engage with the open community in AI, publish peer reviewed papers and then take those insights and actually improve the products that we have. So let me talk to you about one first two applications of computer vision research. So we already saw this great demo for the marketing platform. Now the second one is a little harder to demo because it is in the medical domain. Here we're working together with a company called vrad. They're the largest teleradiology provider in the world. So what does tayloradiology do? Basically, if you come to a hospital late at night and the radiology staff is at Home asleep, the hospitals and VRED is installed in over half of all the US hospitals will send those scans to to vrad. And each scan usually comes with a certain amount of urgency. So if you just have a mild headache, your urgency might not be high and so VRAD will get back to you in about 20 minutes or so, depending on the urgency. However, maybe you had a stroke and some strokes might just result in a mild headache. You do get a head CT scan, that head CT scan gets sent to vrad. Now with the power of deep learning, we can actually use an algorithm and classify whether that had CT scan shows so called intracranial hemorrhage or brain bleeds. And if you have a brain bleed, you actually might die in five to 10 minutes if you don't have a hole drilled into your skull and put a catheter in and let fluid out and so on. That's why we're not demoing it here. And so what we have basically allowed them to do is reduce the patient diagnosis time by over 50%. The algorithm only takes a few seconds. It sees something is really urgent and a real radiologist should look at it very quickly. We bump it up in priority. People, you know, real radiologists see it press a red button and immediately have that be taken care of. So to me, this is a really great example of how we're using AI to make you better. It's not that we're replacing the radiologists, but we're letting them actually, actually take a look at the right patients in the right order so they can really save lives in their jobs. And this is running in their system. It's saving over 20 lives a week. Since it's been running, it's improved over 20,000 patient outcomes. And it shows also the broad range of applications that we can do, all the way from fanatics, shirts and jerseys and haircuts, all the way to head CT scans for medical applications. And that that's also as a technologist, what's so exciting about Salesforce, we built this meta platform and we don't just solve a single problem for a single customer, we solve it in a very generalized way. And so to show you a couple of ways that we enable you to do this, I'll jump into two demos. The first one is for our predictive vision service, which you can actually sign up for now. You can go to metamind IO and sign up and train your own image classifier. Here are just a couple of examples. Now you might have thought, well, they can do the SETCD classification. This Interesting marketing cloud, but I couldn't. Well, you can. It's actually as easy as dragging and dropping a couple images into your browser. So let me give you an example of a use case here. So let's say, for instance, I want to sell solar panels and I want to identify houses that have a pitched roof or flat roof, because pitched roofs actually are much better to put solar panels on. You're much more likely to actually sell to a house with a pitched roof. All you'd have to do is you have a couple of training examples. And this is actually a generally good thing to know and learn about AI. You want to have your data in some format and understand. Basically teach the algorithm the kinds of things you want it to understand. So in this case, this algorithm will only ever look at houses and it will just determine whether it's a pitched or a flat roof. You drag and drop these images into the browser. You click train. Now we deal with all the complexity. Do you have enough training data? We'll warn you if your classes maybe don't have the right distribution. We'll send this to GPU infrastructure graphics cards, really powerful graphics cards that need a lot more cooling and electricity than some data centers have. We deal with all of that complexity and you just can basically solve your problem. Now, depending on how fast our wi fi is, we'll send those pictures off and we can just jump here to what it looks like. After you train it, you can now take any image that wasn't in your training data set. It's a very important distinction. You don't want to test on your training data set. Squeeze in a little data science mini lecture here. And now for this new picture you see, the algorithm sees. All right, this is a pitched roof. Now you can be much more certain that this person is more likely to want to buy a solar panel. And you can incorporate this now very easily in your API. There's a Force.com integration. You can integrate it, for instance, with Google Street View and now just automatically go through the entire zip code and don't have to drive around manually to identify the best targets. So this is how broad this AI technology is. And anybody can really solve their own problems. You can imagine this has so many use cases for service. For instance, you have a service rep. You can just take a picture of that old appliance that your company sold 10 years before he actually joined, and they can still just take a picture and see which documentation is resulting or should be looked at to solve a specific problem that you might have with that Appliance. Now that is the power of computer vision and the underlying algorithms there are based on deep learning. But, but we also use deep learning for natural language understanding. So let me jump here into the second demo. This is a really interesting new model. We call it the dynamic memory network. It's a new type of architecture that we've invented in our research. We have a peer reviewed paper at the International Conference of Machine Learning and this is actually compared, we compared this model on a whole host of different public benchmark data sets against other other kinds of models. Here's an example of what this can do. Basically any NLP problem, natural language processing problem that you can cast as a question answering problem over an arbitrary input sequence you can solve with dynamic memory networks. So here's a question that is, what is the sentiment? Usually sentiment has its own sub community and it's a very different sub community to people want to extract structured information or deal with different languages or do machine translation. So on this model can do all of these different things and it can actually do them very accurately. So this sentence here is the best way to hope for any chance of enjoying this film is by lowering your expectations. And most normal algorithms which actually say, well that seems quite positive because they only look at words in isolation, they see best and enjoying and they think, well that sounds overall mostly positive. There's no really negative word in this, so they would say positive, but this actually understands the sequence, understands how words are put together in their context and correctly identifies the sentence as negative. Now I mentioned that the algorithm is just an input sequence. So here you can actually see that even for Chinese, it's a very accurate algorithm. I'm not going to make a fool myself and try to read that one. But basically the algorithm is also very accurate for arbitrary languages. So we know we have a lot of different customers all over the world and this technology, when we roll it out, will really benefit all our different installations. Now a different kind of problem usually has a different sub community inside the natural language processing research world is named entity recognition. Imagine you go through your emails and you want to identify automatically which company is associated or often mentioned in the context of other companies or who's the CEO of that company and what are personal networks like. This kind of algorithm enables you to do that and you can basically just ask what are named entities? And the model will identify each of the different words and has an answer for each separate word. Now the last example is a great one for showing how general purpose this is. I mentioned that it can be A sequence of words. Well, it can also be a sequence of pixels over which you ask a question. So here the question is, what are the men wearing on the head? Little ungrammatical. And it correctly says helmets. And once it really learns about a domain, it can even go further and you can ask, why are the men wearing helmets? And it will know safety. So these are just a couple of examples that show you the. The power of deep learning when you solve it once in a principled way. And I'm super excited to see not just us pushing this technology forward, but really incorporating it into our various products. And with that, I'll give it back to John.

Got one. Thank you. Awesome job. Awesome job, Richard. Amazing. And it's amazing to see the research coming out of your group and how we're going to be using it throughout the Salesforce platform. It's just. It's amazing. Now, I wanted to. It's not going to show up on here. I did want everyone to just sort of take a look at Richard and our Einstein character and just notice the similarities with the hair and the jacket. We have our own Einstein hair and it's pretty amazing. And we're super proud to have him here as part of the team, so. And great job, team Jim, Shubha, Ali, Julian, awesome job showing all the power of Einstein. Really amazing stuff. And Chris and Taylor, thank you very much for letting us share your stories. It's really always best to tell the story through our customers of what we're doing on the product side. And I'd like to. I see quite a few folks from the Einstein team. I'd like to very briefly shout out to the team that's been building Einstein, bringing all this, working tirelessly, tirelessly to bring you all of this innovation. Now, now, of course, we couldn't, you know, we can't show you everything. We only have an hour here in this keynote. And I'd encourage you to come and visit us at the Einstein Discovery Center. It's just one flight down on the escalators here in Moscone West. See some awesome demos and sales and service and marketing and lots of different domains. Commerce as well. Come visit us at the campground, see more demos and one last big call to action. Become an Einstein trailblazer. Go and take our trails@trailhead.salesforce.com and with that, I'd like to thank everyone for coming and have a great day.