Einstein: AI in the Salesforce Platform | Salesforce

Tell me a secret.

How do you do that thing? You do magic? Gotta do magic when you.

Sam.

Hello. Welcome everyone. Thank you for coming out here. Last session of the day and I love coming to New York. One of the things I love about this event is we do it twice per year. Was anyone here at our last world tour in December? Anybody? Yes, over here we have some. So I was here, I was back over there for the Einstein session. But what I'm excited about is so much has happened with Salesforce Einstein over the past six months. We've released so much new products that we have a completely new session for you today. So for my friends over here that did see our last session, don't worry, this is entirely new content. We're going to talk about Einstein and artificial intelligence in general. But I want to start by taking a look backwards at technology and we're going to start with this ad here. So we've seen all these trends in technology over the past century and every time marketers like me, they're going to make sure you know about it. So we see right here we have this ad for the electric toaster. And you know what, I'm sold. I'm always going to buy electric toasters from now on. But every time we see this, whether it's with wireless or with color, with the invention of color in computers, we're always talking about that technology. And today we see everyone is talking about AI, whether it's Google with self driving cars or Watson competing on Jeopardy or something all of us have in our pocket with their iPhone, with Siri. Now that's natural language processing, that's artificial intelligence. And really what we see is there's a gold rush in AI. Last year there was $5 billion invested in in startup companies for AI technology. Also predict that by 2020 the AI marketplace will be $47 billion in revenue. So it's huge, this impact that AI has had. But a lot of this technology has been around. So we ask ourselves, why now? Why is this happening right now in particular? There's three reasons for that. And it starts with the data. 44 zettabytes of data by 2020. Does anyone know what a zettabyte is? Without reading my slide, I didn't either. So it's 44 trillion gigabytes of data. It's huge. So the data that we have and then if we think back, we have these algorithms and we think back to my slide. With all that money that's being invested in artificial intelligence and the talent that's working on artificial intelligence is incredible. They've made huge advances in algorithms around things like deep learning, which we'll get into a little bit later with algorithms. And finally the compute power to go through all this data and create these predictions. And the result is the power to discover insights about all your data, to predict outcomes of what's going to happen, and then to automate those processes so you can deliver this at scale across your organization. And what we'll see is going back to my example here, AI really is the electricity of our era. And if we think about electricity and wireless and color, we don't hear those buzzwords essentially anymore because it's just expected. We're going to get to a point where every application will involve AI. And if it doesn't, it won't be a smart application. It will essentially be dumb. We see this with an example of a company like Uber has the power to disrupt this industry, like the taxi industry. So in this example, Uber isn't doing anything different with the cars or the drivers. It's really that customer experience that you get when you request a car on your phone. And what's powering that is the data science and the artificial intelligence that makes sure you have an Uber outside your house within a certain amount of time and giving you that free, frictionless, amazing customer experience. So how do we bring it into our business? I talk to customers about this all the time. And one of the first things we see is we have this issue of perception with AI. So when you think of AI, do any of you think of this guy, the Terminator, or maybe it's Pepper the robot, a more modern example. And if that's your perception of artificial intelligence, you might be a little confused about where we are with Einstein and where we are today with AI. You might think AI is amazingly inconsistent or awkwardly invasive, or perhaps it's almost irritating. And one I get a lot is almost implemented if you're expecting these humanoid robots to make all your sales calls for you. Maybe you're saying AI is not really here today, it's that issue of perception. But going back to our iPhones, for consumers, AI is all over the place. We have it on Amazon when we're shopping with predictive recommendations. Again with Siri and natural language processing. If you've tagged a photo on Facebook and it already knew who was in that picture, that's AI. And then if you've seen a self driving car they're driving around San Francisco sometimes that's AI with deep learning. So it's here today, but it exists on a spectrum. So yes, we have self driving cars and we have Pepper the robot, but we also have things like when you search and it already knows what you're trying to type after you've got just a few characters in there, or voice recognition or facial recognition. And in all these cases, what's special about AI is it's that secret ingredient. It's the magic in the customer experience that makes it frictionless. So when I'm typing, it already knows what I'm searching for or when I'm tagging photos, I don't have to go through and manually tell the system who's in every single picture. And what we found is that AI is this magic ingredient the secret to customer success. So I want you to change your perspective on AI from perhaps almost being implemented. There's something a little different, and that's augmented intelligence. So it's not going to be a Terminator robot that does your job for you, but instead it's these tools that make the customer experience that much more special. And to talk about that in Salesforce Einstein, I have here Katie Hurley from the Einstein team. So, Katie, why don't you take it away?

So hello, everyone. Good afternoon. James mentioned this is the last session of the day, so thank you so much for coming out. We're really excited to talk to you about Einstein. And without further ado, I'm going to switch over to our safe harbor statement. So James did a great job about talking about how we think about AI here at Salesforce. And now I want to switch into product to tell you what we've been focusing on with Salesforce Einstein. And so I like to start with this slide because it gives us a view of what's been happening in the last 10 to 15 years when we think about the technology transformations we've been experiencing. So we have an increase in applications, which is changing the way that we communicate. You can think About Facebook and LinkedIn, this whole new world of social. And almost every one of us carries a mobile device wherever we go. So we have a phone on us at all times. And most recently, we've seen an advent of connected devices. But what's important about this is the amount of data that these new transformations are creating. So much data, as James mentioned, 44 gigabytes, if I got that right. And the reality is that we as sales individuals and people in marketing and service, there's no way that we can analyze this data at scale. And so we need something like artificial intelligence to help us create these next level customer experiences. So that's why we believe that we're entering the age of the customer. And so when you marry artificial intelligence with enterprise applications, you get something wonderful. You get more productive employees, and as a result, you get enhanced customer experiences. And in 2018, so just one year from now, Gartner predicts that over 500 million business users are going to see an increase of 2 hours of saved time through AI powered tools. So we won't spend our days doing things like entering data. And what that's going to allow us to do is have more tailored relationships with our customers and spend time on tasks that improve those types of experiences. So we spent the last three years here working on Salesforce Einstein and Salesforce Einstein. It's not a new interface, it's not a new cloud. What we're doing is we're bringing artificial intelligence into our platform. And we're able to do this because we have tons of customer data. We have an incredible research team led by Richard Socher, he was actually recently featured in Forbes, and they're working on cutting edge AI technologies that we're bringing directly to you into the Salesforce platform. So the best part is understanding how it all works. And James mentioned how everyone is focusing on AI right now. So what makes it different for Salesforce? And the best way is to think about what does it take to implement AI into your business. A lot of steps. First of all, you need a team of data scientists, you need experts, and then you need to decide what models are best for your business use cases. And you need to do something called feature engineering, which we won't get too nerdy because that's not the point here. But it's time consuming and it takes a lot of resources. And most importantly, you have to bring these amazing insights into the context of your business. And so with Salesforce Einstein, we're bringing you a data scientist, both literally and figuratively. And we are bringing you automated machine learning. So all of your data is prepped for you. And we're delivering insights through these AI techniques that I had mentioned, so deep learning and natural language processing. And we're bringing it to the context of your apps. So it's basically click and go. There's no heavy implementation required. And it's going to bring you parity with what you experience in your consumer applications like Siri and when you get that custom music playlist given to you in an app called Spotify, which I was just listening to when I was on my way over here, as I had mentioned earlier. So it's not a New cloud. So we're infusing AI into our platform and what this allows us to do is to bring intelligence to all of our applications and it gets really exciting. We should probably. Wasn't that a fun sound? I always forget about that part. Anyway, so what's going to get exciting is when we can customize these experiences and so that you can bring machine learning and different AI techniques into custom objects. And as James mentioned, we've already seen incredible success. So at Salesforce we have three releases per year. And this is a list of all of the fun Einstein features in our February release. There's a ton here, so I want to focus on a few that I find impactful. So sales cloud, we are bringing predictive lead scoring. So what this means is now we're going to add machine learning to your historical lead data and let you know the likelihood of a lead to convert Einstein Data Discovery. So I actually I was in sales strategy before moving to product marketing and I would get asked to pull a SQL report of a big data file and then analyze it to understand why revenue may be increasing or decreasing and what was contributing to these different factors. With Einstein Data Discovery, it does that analysis for you. What's really cool is it provides narrative stories so it will actually give you natural language processing output of all of these results in your data and you can instantly put it into a PowerPoint. So if you think about a quarterly business review, that's pretty amazing. And it also goes back to what we said when you're saving two hours per day of your time. And also, why don't we talk about AppCloud Einstein? So this is a really exciting feature for our developers and our admins. It's called Einstein Vision. And when you think about Tesla, they're able to bring us self driving cars with a capability called image recognition. So Einstein Vision is doing something similar. It's abstracting the complexity from building these models into our workflows. And it uses an API call. So when we think about what this means, because that sounded like a lot of technical jargon, you're able to do things like object detection to see where your brand is showing up across marketing channels. Another use case may be visual search, so you can enhance enhance your e commerce applications and provide a new way of shopping for your customers. And there's no better way to kind of show this success than to see it in action. So now I want to bring up James again and he's going to give us a demo of both Sales Cloud Einstein and then also Einstein Vision.

Great. Thank you, Katie. So we're about 15 minutes in and we've got to our product, which I'm really excited about. Want to spend a lot of time in here talking through sales cloud Einstein. And to do this, I'm going to take on the role of a sales rep. I'm going to be a sales rep at a company I created called Solar sun. And Solar sun is a solar company and we have tons of leads coming in. So let's say there's all these government incentives, people are motivated to get solar. And then also I have this marketing manager who's created some awesome viral campaigns on Facebook and we have tons of leads pouring in. So when I get into Salesforce, my lead list and what I need to do today looks something like this. Now, not all leads are created equal, so I don't know where I'm supposed to spend my time, how I'm going to prioritize my day. But with Einstein, what we're doing is we're using machine learning on all the leads, all those fields on the lead object, whether it's your standard or your custom fields, and we're coming up with a lead score based on all your past conversions from the past six months. So what I can do is right in this lead list, I can prioritize based on the Einstein lead score and we can work through that throughout my day. Now, because this is in the Salesforce platform, as we go to our home screen, you'll see I'll actually have a dashboard in here so we can go through this data on wave. So if I have a hunch that maybe those leads coming in from social media aren't quite so good as the ones that are coming in from our conferences or from other sources, we can take a look at that right here and look at that lead data. But Einstein doesn't stop there. So we also have insights on our accounts and our opportunities, which we can leverage as well. So let's take a look at this one here. We have an opportunity that's boosting. So we'll drill into this and we'll see. Why does it think this opportunity is boosting? We have an increase in email volume, which is suggesting, based on the past patterns in my org, that the opportunity is more likely to convert. But I'm a sales rep and I hate logging email in there. It's actually true. I never log emails in Salesforce. So how does Einstein know this? And does that mean it's gonna work differently for different reps? Some people get insights if they're on good behavior and others just don't. What powers this feature in both of our insights, Opportunity and accounts is Einstein Activity Capture. So I set that up and I just authenticate my email, whether it's Gmail or Office365, and that's all I have to do as a rep. Now all my emails that are relevant to Opportunities and accounts get logged automatically into Salesforce on the activity history and I still have control over my sharing settings in there. So that's great. Now we have really good data, all my contacts, and then Einstein can use that natural language processing, go through these emails and serve as insights just like this one. What's also great about that is this is positive. So awesome. I'm doing a great job. But what about the flip side? What I might be more concerned about is, is there a negative insight? Is my pipeline looking trending down? Is there an opportunity where I need to take action right away so Einstein surfaces that too, and then I can take the recommended action right from here to resolve that. Everything in here actually looks good. So why don't we focus on drumming up some new business? So we have this new lead in here, Mr. Julian Sauvage, and he's a lead score of 70, which is great, or, sorry, 92, which is great. But as a rep, I need to build trust. And in this lead, the way we do that with Einstein, and you'll see this in multiple implementations of AI from Salesforce, is we expose to the rep what those predictive factors are, what's in that black box for artificial intelligence. So if we look in here, we see these indicators and they have a strength associated with them. So we see here the roof type is pitched, which for solar is really important. So if you have a pitched roof, you're more likely to buy solar panels. If you have a flat roof, those solar panels don't work so well, so you're less likely to convert as a lead. The point I want to make with that is this is specific to your org. Every org has their own model, every customer has their own model, and it's going to go through and do all that machine learning and find the best model for you. What's also great about that is if something changes in your industry, maybe there's an invention and now we can do solar panels on flat roofs. Your model is always retesting itself, so it's going to give you the best model and it's going to make sure it's updated. So if something changes, it will rescore all your leads with the New model, and you'll always have the most current version. Why don't we drill into this lead and on here we actually have a little example of how you can use Einstein Vision. So if we look at this here, how do you think we determined that this house had a pitched roof? Well, our solutions engineer over here, Mark, who's having my demo. Hi, Mark. He actually created this lightning component using the Einstein Vision API. And what it does is it's going to pull the address right here from the house, and it looks like he's flipping around there, but it's going to pull the address and it's going to actually go through the Google API, pull out the Street View image, and I think Mark's just refreshing it here so we can see if it does it again. It's going to Google pulling in that image. And then we've trained this model to recognize from a Google Street View image, does the house have a pitched roof or does it have a flat roof? Once it recognizes that, it's going to go ahead and populate the lead, and then we get the lead score. So there, you see right there? That's exactly what it did. What's powerful about this, and what I want you to take away from Einstein Vision is not this particular component, but that you can train it for any type of model. You can build your own models with your own image libraries. So, Mark, why don't we show them how you built this component? Does that sound good? So in here, Mark has a few different models he's trained. Looks like some types of cars, scenes of beaches and mountains. These are all examples we've used throughout the day, showing customers how this works. But I'm interested in showing them how we actually built that. So to train this model, all you need is an image library for the different classifiers that you have in there. So while we're showing this, think about how this could work for your customers. This could be embedded into any app. This is a developer service from AppCloud, Einstein. So Mark's uploading all these images of pitched roofs, which is great. And he's also uploading into our other category, these images of flat roofs. So what we're doing is now we're creating that Einstein visions, going through all those images that we've told it. And if you think about it, that's unstructured data, right? So that data, we haven't given it any metadata or any information that's text. It's actually finding patterns throughout all those images, essentially the pixels, and it's using Deep learning to make connections and learn using a neural network how to recognize a flat house versus a pitched house. So let's have some fun, Mark. Let's find a good example. We'll just go on Google and find something to test this model. It looks like Mark's doing the Salesforce tower, which is our new building in San Francisco. So why don't we see if this is a good fit for our solar customers. So all we do, we save that image and we're just going to test it. So drag it over there and it looks that's pretty good for that model. So 96% sure that it's flat. So really happy that our model works. Got a flat top on that building. Unfortunately, I don't think we'll be cold calling Marc Benioff anytime soon and selling him solar panels to put on top of his tower. But again, this could be in any one of your apps with Einstein Vision. Really powerful technology. And with that, Katie, I'm going to pass it back to you.

Great job, James. Let's give James a hand. Awesome demo. So seeing is believing, like we just showed you with the demo, but I also want to talk about some customer references. So how many people in the room here are in sales? Is there any salespeople? Awesome. So Silverline is a company. We just saw a wonderful demo of sales cloud and Silverline has implemented sales cloud Einstein, and they're using things like predictive lead scoring and opportunity insights, all those interesting features that I covered in the release slide. And they're seeing a 30% increase in their close rate. And anecdotally they're saving one to two hours on data entry because they're using automated activity capture. Okay, who's a marketer in the room? Any marketers I know, James and myself. Awesome. So fanatics is another customer who's using our marketing cloud, Einstein, and they're using predictive email capability and so they're able to provide their customers with tailored emails that match their interest from a sports perspective. And in this segment alone of AI powered emails, they're seeing an increase in about 15% in click through rate, which is pretty powerful. Lastly, how about since Black diamond is a sports company, any sports fans out there or big skiers? So I grew up in Park City, so yeah, awesome. So they are using our commerce cloud and they're using product recommendations both in store and online. And so if I visit their website, I'll be probably shown a pair of slalom skis. That's the brand that I prefer. And so they're targeting their product line to individual customers and they're seeing an increase. And again, our lucky number, 15% in revenue. So pretty exciting. And we also have two customers here from American Music Supply that I'd like to welcome up Doug and Eric. Awesome.

Hello.

How are you guys today?

Good.

So thank you so much for coming. So I want to start by. So I don't know who's familiar with American Musical Supply. So Doug, would you mind telling us a little bit about your company and what you guys are focused on?

Sure.

And thank you very much for having us here. American Musical Supply is one of the top five musical instrument retailers, e commerce retailers in the United States. We're in the Internet 500 and we are celebrating our 31st year in business. So we've actually been preceding the Internet with a catalog that has been mailed since 1986. But the story of American Musical Supply dates back decades prior with a very well respected family music store business in northern New Jersey. So you're really looking at a four generation family business that has gone in to become one of the largest online retailers of musical instruments. Live sound, dj, recording accessories, guitars and basses. As you might expect, everything you need to outfit a band and more. So what we do different in that we're in a very, very competitive vertical. And I'm glad everyone's sitting down here because you really want to be sitting for this. We lend money to musicians. I know that sounds shocking, but we are actually a direct lender. What I mean by that is all of our competitors, they rely on Synchrony bank or another third party retail lender who actually denies 70% of the applicants. And many young musicians may not have the ability to pass muster with a synchrony bank. But we are the direct lender so we can bring people on and we find that they become some of the most loyal customers that we have in our catalog. So we also do something rather different unlike anyone else. And really outside of Amazon, we're the only retailer who has invested in three state of the art, strategically located warehouses around the US so we can deliver 90% of the US is getting their packages from us in two days or less. And we're also highly customer centric. We've just opened a new state of the art call center in Minneapolis, Minnesota. We have a very, very diligent and vigilant group of people who look out for our customers every day. And we're really on top of our game and excited to be able to leverage some of the technology. So it was about two Years ago, the owner of our company charged me and a group of people to go out and look at the very best in email solutions. And of course, as we vetted all of the big names and we looked and we saw features and advantages, we really realized without question that the ecosystem that Salesforce Marketing Cloud represented to us and possible future enhancements that we might need for our call center with the service cloud or other parts of your and social studio, of course, which we licensed right away, was really the way to go for us. So we jumped in feet first. We are just past our full year mark in terms of our cutover to the marketing cloud and very happy to see that. And by the way, if anybody here looking for an instrument, I'll be around afterwards to talk with you. But in any case, we do. We're very happy to be here and that explains who we are and what we do.

Perfect.

We really do pride ourselves in helping our customers get the gear they want when they want it.

That's great. And so I understand that you brought Eric on board to kind of help with that email marketing journey, especially as it relates to bringing AI innovation. So would you mind telling us a little bit about how that's going?

Sure, absolutely. I mean, I came on board, like Doug was saying, a little bit afterwards. We signed on with Salesforce here. So I've been there for about 18 months now. And what we started out with is we had a real challenge in terms of our segmentation strategies, where segmentation typically for us was built around our main taxonomy categories of our products. So if you know, this group over here were guitar buyers and this group here were keyboard buyers, we would go back and we would look at our purchase history from our customers and say, okay, for keyboard people, we're going to put you into the keyboard group. For guitar people will put you into that group and we'll send our messaging out accordingly. Had very good results with that. Everything was wonderful. But there's a better way to do everything. One of the things we decided to do was try and leverage some of the AI and try to leverage some of the real time data that we're getting through the marketing cloud and try to build out those segments to create more overlap and ultimately reach people who are trying to purchase particular products based on what they're looking at right now. So rather than saying, you folks bought guitars a year ago and saying maybe you'll buy another one, it's how many people from this group over here are interested in a guitar? How many people from this group over here who bought keyboards are interested in a drum set or something different. So we leverage that information where we're collecting data in real time from our website, and we're taking that and building those segments on a consistent basis to create new looks at our customers. All the time we've taken that, we're bringing that down even further down the level so that we're taking specific offers and we're looking at exact items and saying, there's a very small segment of people here who are interested in this item, so why don't we send out a special offer to those people? And we're trying to build that out and automate that today so that we are delivering the holy grail of marketing, which is the right offer to the right person at exactly the right time.

Perfect. And then so for the marketers, especially in the room, do you have any, like, best practices that you want to share or any early kind of success metrics since it is early in your.

Absolutely. I mean, in terms of best practices, I mean, this is. It's not, as we might say, rocket science in some respects, but it is also looking at the data as deeply as you can and trying to understand your customer's behavior. It's not always the easiest thing to understand, but when you have the data insights like this, then you can really start to dig pretty deep in terms of success metrics. We have seen engagement rates that are three times to four times better than what we were getting with our standard segments. We're getting much better sales figures as a result of that, and everything is on the up and up as a result.

It's been terrific, amazing, great. And then I want to, before we wrap up, I want to hear from you, Doug, about some of your future plans with AI and what you plan to do next, especially as it relates with Einstein.

Sure. Well, buying a musical instrument is really an emotional process. You don't need another guitar. You want another guitar. And we do find that it's fascinating to observe our customers browsing behavior, collecting that data, and really starting to deliver offers at the moment that a customer is really in the mood and ready to push that button. And we think that the promise of AI and Einstein is that we can really harness that and get to the point where we can identify that moment and actually try it in a bunch of different channels. You know, we have predictive intelligence product recommendations on the site. We have a call center. At some point, we could trigger a phone call out to our prospect, who's about ready to, hey, maybe talk about that guitar amp. Keyboard, drum set or whatnot. So it's our intention to maximize the use of the Salesforce marketing cloud, really embrace AI, embrace Einstein, and use this for all it's worth so that we can, in a sense, help our customers get the most out of their relationship with AMS and continue to grow as a valued partner with Salesforce, and of course, improve our business profitability and ultimately want our customers to get the best product at the right time.

Awesome. Well, thank you so much for joining us today. I love to hear your story and I can feel the passion. So it's really exciting to be able to share some of the success that we're seeing with Einstein, with everyone here today. And so if you want to stay up here while I do my last slide, you can. Or you can.

Thank you.

Thank you very much.

Awesome. So I hope that everyone got a chance to visit our campground so we have more demos that you can take a look at. I know we're about probably closing down at this point, but if you didn't have a chance, I highly recommend visiting trailhead. So we have four amazing trails on Einstein and we'll be adding more sales cloud Einstein, Einstein Data Discovery and then Einstein Vision, which is what we saw in James demo. And thanks again for coming out. James and I, Doug and Eric will be around for any questions that you might have. Thank you.