Salesforce Einstein: Blaze New Trails with AI | Salesforce

Ladies and gentlemen, please welcome senior vice.

President of product and general manager for Einstein at Salesforce, John Ball.

All right. It's hard to believe that just one year ago we were up here launching Einstein. And wow, what a difference a year makes. We shipped 20 AI powered features across the clouds, driving success in companies of all shapes and sizes, from Fortune 500 financial institutions like U.S. bank, to small high tech startups like Astronomer, to retailers like Adidas and Black diamond, even to small nonprofits like College Ford. And we're seeing tremendous business results. We're seeing improvements of 300% in the lead conversion rate in sales cloud Einstein. We're seeing improvements of in revenue increase in revenue of 7 to 15% in Commerce Cloud Einstein. And now after just one year, we're cranking out close to half a billion predictions every single day, which is amazing. Just one year into this journey. Now, we're not on this AI journey alone. And we couldn't have done this alone. We couldn't have done it without. You are trailblazing early adopters. We've had this fantastic year thanks to this great partnership that we've created with you. And so, on behalf of all of Team Einstein, quite a few of them right here. Big, huge shout out. Big, huge thank you to all of you Einstein trailblazers. It's really been great. Yeah, you can give yourself a round of applause. Absolutely, Absolutely. So I said that we're on this AI journey and we are on a journey. And we all heard about this yesterday in Mark's keynote about the fourth Industrial Revolution. Now, the fourth Industrial revolution is transforming economies, entire industries, even our daily lives. And at the heart of it, at the heart of this fourth industrial revolution, is AI. And really, there's no industry that is being transformed, being sort of disrupted more than software and technology. And in just the last year alone, we've seen breakthroughs, transformations that would have been considered to be science fiction even a decade ago. Now we've got these little earbuds that can translate 40 languages in real time. We have machines that can mimic specific individual voices in healthcare. AI and computer vision is helping us to detect skin cancer and tumors and eye diseases like macular degeneration. And of course, we see more and more AI in our apps. Google tells me when to leave to make it to a meeting on time. Waze gives me the best route to my destination. Although I've got to say, it got a little confused this morning on the way to Moscone. Sort of red everywhere, just one big red glob. And so these are just a couple of examples of how AI is transforming and impacting our personal lives. So what about business? How is AI impacting business? Well, first and foremost, it's all about helping us do our jobs better. AI can really help us do our jobs better by focusing on the customer. And if we take a step back, if we think about why we're all here at Dreamforce, it's because we want to think about, we want to learn about, we want to be inspired to connect to our customers in a whole new way. And at the end of the day, it's about doing these core things better than ever before. Sales and service and marketing and commerce and application development. And that's where AI can really help. AI can help us to focus on what matters and that's the customer. If you're in sales, you want to focus on those leads and opportunities that are going to convert. If you're in service, you want to focus on those cases that are going to escalate or those customers that are going to attrition. In marketing, you want to focus on those the audience or the folks that are most likely to engage with you. This is the opportunity for all you trailblazers in the room and all the trailblazers online is to embrace AI, to use AI to transform your companies to focus on the customer. But we all know that building AI from scratch is really hard. And that's why we built Salesforce Einstein. Einstein is AI for CRM and and it's built right into the platform, takes all that great customer data, applies state of the art AI algorithms to surface predictions and recommendations and insights right there in the platform, helping you to be your best in making Salesforce the world's smartest CRM. Now, the great thing about Einstein is Salesforce just keeps getting smarter and smarter. We just released another 16 new and pilot features in the winter release. We went ga with marketing cloud with Einstein engagement scoring helping marketers focus on the folks who are most likely to open, click and buy. We launched super exciting pilots and service cloud with Einstein agents driving agent productivity with predictive triage. And there's more coming in just a minute. We of course continue to invest in sales cloud. Excuse me, with Einstein opportunity scoring and predictive forecasting. These are really cool apps and you'll see them in the sales cloud keynote. And of course we're going to continue to invest and add Einstein to all the clouds. But you guys told us something. You said there's something missing from this picture. You said, what about all my custom apps? What about all Those customizations I've made to sales cloud and service cloud. Well, it turns out you guys love to customize Salesforce. It's super important to you and we've made it really easy to do with our metadata driven platform. You've gone to town, you've built millions of custom fields, millions of custom objects. In fact, 80% of all records in Salesforce are on custom objects. It's kind of mind boggling. 80%. And so what you've done is you've taken Salesforce and you've made it your Salesforce and you've asked us to do the exact same thing with Einstein. And so we are super excited to announce My Einstein. With my Einstein, every trailblazer can build their own AI powered apps with clicks and without being a PhD in data science. Now let's pause there for a moment. We've all been talking about the fourth industrial revolution and how it's transforming our lives. And at the heart of it all is AI. But there are simply not enough data scientists in the world to build all the AI that we need. We need a different approach, we need tools, we need my Einstein. And so with my Einstein, the very first thing is super exciting and it's brand new. It's called Einstein Prediction Builder. Now Einstein Prediction Builder gives the power to any admin to predict any field in Salesforce on any object right there from setup. So for example, maybe you're in financial services like our customer AXA and you've got millions and millions of accounts and you track those on the person account object. So you want to predict churn and why do you want to predict churn? You want to predict churn so that you can track and find those folks who are most likely to leave your service. And why do you want to do that? You want to focus in on them, to maybe send them a specific marketing offer, a special offer to win them back. And now with Einstein Prediction Builder, it's as easy as point, click and predict. Now next up on My Einstein is Einstein Bots. And this is all about driving service efficiency through case deflection and better customer experiences. And what we're doing here is we're taking all this great natural language, processing all the crum data in Salesforce and combining with the Salesforce workflows. And what we're doing there is really building an AI powered self service experience. And we're super excited about this. We saw this yesterday in the Adidas demo and there are a couple things that are really cool that I get excited about. First, if you Think about Adidas. They run all their commerce and their service on Salesforce. So we have customer context. This is super important. We don't have to sit there and prompt. The bot doesn't have to prompt the end user for all this information that the system should already have. This is one of the most annoying things in bots today. You keep getting prompted for information that you're sitting there saying, the system should already have this. It knows me well with bots, Einstein bots. We're taking care of that. The second thing that's really cool is that the bot is in Salesforce. And since it's in Salesforce, anything that Salesforce can do, the bot can do. You can look up order status, account history, trigger a workflow. It's all right there. The last thing that's really cool is that we have all this historical data in Salesforce. We have historical chat logs, which cases got resolved with which knowledge articles. And that allows us to really turbocharge the bot building and training so that we can deliver these faster than ever before. Now, next up, we've got Einstein Vision and Language. And this is all about bringing the power of deep learning to our trailblazing developers with Einstein Language. Deep learning is powering the natural language processing that you'd find on all the unstructured data in Salesforce. Things like the notes fields that sales reps and service reps enter, all that unstructured data. So now you can predict sentiment, you can classify intent on all this free text that's in Salesforce. And with Einstein Vision, we can classify images, but we can also detect specific objects within those images. And that's what our customer Crossmark is doing to optimize shelf space in retailers. So they have tens of thousands of field personnel out there setting up displays, ensuring that they're compliant, optimizing them. And now they can simply take a photo and we can detect if there's a problem. And if there's a problem, we can then create a case and take care of it. And this is saving them thousands and thousands of hours because they don't have to manually sift through all those images. And finally, we've got Einstein Discovery. In Einstein Discovery, it's all about insights. It's about finding those previously hidden insights, the things that you didn't know about your data. Like our customer here at KeyBank with opportunity data and finding sales insights, what we're doing is we're looking at the millions of potential outcomes and surfacing the insights that really matter. And we're going to hear A lot more about this tomorrow in the analytics keynote, so I don't want to steal their thunder. You should come and check it out. Now, what is the point of all this? Well, my Einstein, what we're doing is we're opening up the platform, we're opening up Einstein so that all of you trailblazers, regardless of your industry, regardless of your use case, you can build the AI powered apps that you need that are important to you. That's what this is all about. Now I think it's time for me to stop talking and for us to start showing some demos. And we're gonna do that through two customer stories, US bank and College Ford. And I love this part of the presentation because you get to see how Einstein is being leveraged, how it is helping a really big company like US bank as well as a small scrappy nonprofit like College Ford. Now we're going to start with US Bank. We all know US bank, fifth largest bank in the United States. Multiple divisions, thousands and thousands of branches, tons of customer data. And we've put together a short film to show you how they're leveraging Einstein to present one US bank for their customers. Let's go ahead and roll the film.

You talk to people and you ask them what's most important to them. They'll tell you, of course, their families and their friends, but the next things they'll say are my health and my finances. When you think about what banking can do for people's families, for their children, for their futures, for their retirement, all the most important things in their lives are touched by what banking can support them with. What does being a trailblazer mean to US Bank? We have a responsibility to think five years for our customers and think, what is the world going to look like? What's the technology in making sure that we're going to be able to support our customers to get to that place with us.

When you think about the amount of data created in the last five years, it's equivalent to the amount of data created up to that point in human history. So the complexity of managing all that data has gone up tremendously. We need machine learning and AI to help.

AI is a way to more rapidly bring to market insights into the data that we already have that we don't have. Know that much about. Einstein can tell you more about the relationships between data.

One of the big benefits to Einstein is the data is already there. It's actually already in Salesforce, so we already have a unified platform.

Our approach to implementing Einstein has been first to take A look at leads management. What leads are we putting in front of our employees and what data is wrapped around those?

Machine learning and artificial intelligence is about enabling our frontline employee to. To do what they do best and do it better.

When our employees have a good experience, then our customers have a good experience.

We have 2000 plus users that are going to move on to Salesforce.

Trailhead has given us a tool to be able to communicate and train in a fun environment. It adds a little competition, a little gamification to the training.

Somebody says, oh, I have this many badges.

I have this many badges.

I have 17 right now.

With Salesforce's help, we are able to really be intelligent, proactive, and not just understand information about our customers, but understand the insights.

If we're able to leverage better insight from Einstein in a way that's going to help a customer make a better decision about should I start that small business or what do I do to make sure that my kids are going to be okay in college? If we're able to leverage this capability in service of helping our customers, that is exciting.

All right. What a great film. Now to see this technology in action, I'd like to welcome up Allie Witherspoon, director of product marketing for Einstein.

Who's ready for Alicia Keys tonight?

Right.

That song just puts you in a good mood. We're super excited, and I promise you just have to get through through a few more sessions before the party starts, including this one. So, as you heard John mention, U.S. bank came to us with a vision. And their vision was to create one US bank, which was all about giving every single one of their employees, not just their executives, democratic access to insights about the company and about their customers. And we knew instantly, as the Einstein team, that this was the perfect task for us to help them solve. And so in this demo, we're going to walk through two different use cases. We're going to talk about how we're revolutionizing the way that U.S. bank's wealth team sells as well as their commercial banking team. But before we get into all of that fun and excitement, I want to give a big, hearty Dreamforce round of applause to our live demo team in the back of the room.

Oh.

Well, there's a new member crashing the party there, and looks like space is a little tight. So, Einstein, I'm going to suggest that you find a seat, enjoy the keynote, and just relax. So let's start with the Einstein readiness assessor. This is where every company can get their Start with Einstein. It's a simple Org check that tells you which features you're a fit for. And this is exactly how we started with US Bank. So it's as easy as logging into Salesforce and clicking a button. And we will give you a PDF report that tells you which features you're good to go with, and we tell you why so you can understand, is it your data, is it your business processes, what's holding you back from using these features, or what's allowing you to use them. And so this is the PDF report that we got for US bank. And we can see here they were a fit for all five sales cloud Einstein features, which was fantastic news. So as soon as we knew this, we could start enabling these features at U.S. bank. And we started with Einstein Lead Scoring, which is personally one of my favorite features. And if you thought the assessor was easy, enabling this feature is even easier. You can see here it's as simple as clicking a button and you can turn on lead scoring for your company. And what's better is with the winter release, we've allowed you to customize lead scoring even further. So at U.S. bank, they convert their leads to opportunities, like many of you do. But some customers came to us and they said, well, we convert our leads to accounts or contacts. And so now we've built an extra functionality for you to be able to customize lead scoring for your business. So we enabled this at U.S. bank. And what you don't see here is what's going on under the hood. And that's kind of the auto magic of Einstein. He is scoring all of U.S. bank's 4.5 million leads by looking through their data, understanding their business processes, and building the best model for them to score their leads. And that was all done for US bank in less than two hours. 4.5 million leads scored in less than two hours. You didn't see a data scientist. You never saw the modeling. We do it all for you under the hood. So now that this feature is enabled at U.S. bank, we can take a look at what their lead list looks like. And this is a lead list that's powered by data science. You can see here, every one of their 4.5 million leads now has a numerical score associated with it. And that score is a propensity to convert. A propensity to convert to an opportunity. And what's even better about lead scoring, if that weren't good enough, is that we tell you why. So when we ship a lead score, we also explain the predictive factors that correlate positively or negatively with conversion so for US bank, it was things like, which branch does this lead come into? Do they have a mortgage with US Bank? Do they meet a certain minimum level of assets under management? These are all things that are positively correlated with conversion for them. And so for the financial advisors at US bank, it really gave them that extra level of insight to be able to trust the AI, because we hear that a lot from you. Well, how do I trust these scores and these predictions? Well, it's because we're shipping the why so you can start to believe in the AI and it's not just a black box anymore. So with lead scoring, we also give every customer an out of the box dashboard. And we gave this to US bank before we even rolled out the feature. So what this allows us to do is to show them that this feature is going to move the needle for them before they roll it out to every single one of their advisors. So as you can see here on this dashboard, 4.5 million leads scored, as I mentioned, with a conversion rate of about 5%, which is pretty good, but I like the conversion rate to the right a little bit more. 15.2 is a really solid lead conversion. And what that number represents is the leads in the top bucket. So US Bank's hottest leads of about 81 to 100. And so what the difference between those two numbers means is that we can promise US banks a 3x projected lift when we turn on Einstein lead scoring. And so this was all the proof that they needed to roll this feature out to their entire team of advisors. But you heard in the video they didn't really mention sales once. And we've been talking a lot about sales. But US bank is really in the business of service. They want to create amazing customer relationships and deepen those relationships with every interaction. And so sometimes it takes data and it takes finding insights in your data to deepen those relationships. And so that's where Einstein analytics comes into play. It is our data discovery tool that allows everyone to find those hidden kernel of insights inside of their data. And so here we see in this dashboard that U.S. bank found a pretty interesting buried insight. And that was that their customers who were 35 to 44 years old in age and had a mortgage with US bank were more likely to increase their assets under management if they also opened a credit card with U.S. bank. So this is just kind of a random correlation that was hidden in mounds of data that would have taken probably an army of data scientists and a model to figure out. But the analyst at US bank was able to Find this with just a few clicks. And because our analytics platform is built on top of Salesforce, we can take those insights and embed them right back into Salesforce. So now the analyst at US bank can give that suggestion to every advisor on how to improve those phone calls with the customers, how to deepen those relationships. Well, maybe you should start adding a US bank credit card to those conversations. So this is the power of Einstein lead scoring at U.S. bank. How it's completely changing the way that their financial advisors sell. But now we're going to pivot to their commercial banking team and take a look at what Einstein has in store for them. But first, let's check back in with the man of the hour, Einstein. How are we doing back there? He looks like he's enjoying the show.

Yeah.

Oh, we get a little mustache twitch.

I like that. He looks happy.

Okay, so now onto commercial banking. So US bank came to us with a vision, as I mentioned, of creating one US bank, of giving this deep level of insight to every single one of their employees. And Einstein forecasting has helped them with this mission. With Einstein Forecasting, we have predictive, collaborative and actionable forecasts right inside of Salesforce. So for all of our customers who are already forecasting inside of Salesforce, this is a no brainer. So here, this looks kind of like your typical forecast, but it's, it's a lot better than your standard spreadsheet. So here we have Einstein's predictions. He's telling us when we're going to hit our quota. And we can even drill down into this information by team, by individual rep, by geo, whatever lens that we want to look into it. And what's even better is we are now shipping our beloved pipeline waterfall chart with every forecast, because we know that behind every forecast is a pipeline. And customers want to understand, well, which deals are going to close, which deals are going to get pushed out. Where am I sandbagging? And so now you have those two views side by side, your forecast and your pipeline. And you can understand how they're influencing one another. But we're Salesforce and we always like to outdo ourselves. So of course, we made this to run on mobile devices. So now every sales leader, not just those at U.S. bank, can access their forecast on their mobile phone from anywhere on the go and have that complete, transparent view of their business. With that, back to you, John.

Right.

Great job, Allie. How about a huge round of applause for Allie and the demo team? All right, now we have the privilege to hear directly from U.S. bank. And so I'd like everyone to give a warm Dreamforce welcome to Bill Hoffman, Chief Analytics Officer and head of CRM US Bank. Bill, come on up.

Thanks, John. Appreciate it.

Thank you again. Thank you for your partnership coming. It's been great to get to know the team over the last couple months. So in the, in the video, we heard a lot about One US Bank. As the Chief Analytics Officer and head of CRM, can you sort of explain what that means to you?

Sure. Before I do that, though, I just want to say a deep grateful thanks on behalf of 70,000 US bankers to you, John, the entire Einstein team, and to the best salesforce client service team in the business. Brad, Jerry, Tom, I really appreciate you guys and the partnership. So really there's. There's two answers to that question. There's a customer facing question and an employee one. So on the customer side, we want to make sure that we put the customer at the center of everything we do. Customers don't look at us as, you know, I'm a. I'm a mortgage bank customer of U.S. bank, or I'm a small business customer of U.S. bank or a credit card customer of U.S. bank. They're a U.S. bank customer. So we need to have a unified view and then create a unified customer experience. And really the way to do that is to enable our employees. So we're very focused on leveraging AI and ML in service of our frontline employee. And that's a big. That's a big differentiator for us, we think. So The A and AI should be augmented, not artificial. There's nothing artificial about a high quality personalized relationship that a financial advisor or commercial banker can build with a customer. And using AI and ML to take some of the minutiae out of that. Right. So who do I call? When do I call them? What should we talk about so that you can enable the human being to do what they do best, which is be a human being and build a deeper relationship. So that's what we're doing with the One US bank strategy.

Great. So can you shed a little bit more light on how you're using Einstein specifically to sort of deliver on this vision?

Sure. So when you think about a personalized experience, and all of us have those with the best companies and brands we relate with, really the first step is to recognize a customer. Right. The second is to know them. The third is to understand, and then most importantly is to anticipate. So we're using Einstein and the partnership to help us get ahead and understand what that customer is likely going to need next and actually make recommendations for them before they even might think about them. So getting ahead of that and a couple of use cases as was mentioned. So one of them, the lead scoring use case. So finding within our 16.9 million retail customers, who are the customers that will resonate with our wealth management value proposition. We've had a 2.34x increase in conversion from our previous approach which was modeling. So leveraging Einstein's already created that with our, with our advisors, which is a huge proof point. And then also being able to identify customers that are existing wealth customers that might attrite and being able to help the financial advisor know first off who those folks are and secondly what to talk to them about so that we can reduce the likelihood of attrition. So those are a couple ways our frontline employees are leveraging Einstein in service of a better customer relationship.

That's great. It's awesome to hear. So Bill, you're clearly an Einstein trailblazer. You've already accomplished a lot. You've got a great vision for where you're going. But a lot of the folks in this room are probably just getting started. So can you give them some tips and tricks and sort of advice on how to get started and, and with Einstein and with AI, That's a great question.

So when we think about it, the first thing is to make sure you've got the right question. As an analytics professional, you know, we're all about trying to get the right answers all the time, right, but sometimes, you know, we have the questions wrong. And to me, the question here is not a technology question, although you need to have great technology like Einstein. It's a change management question. So how do you get folks in the field to want to use Einstein? So we found, you know, some of it's school of hard knocks. Other key lessons learned on our journey, there's really five things. The first is have an agile approach. Don't be afraid to fail test and learn and then scale from there. So be flexible and agile in your approach. So that's one second, data quality matters, right? But don't you know, small, fast data is a lot better than clunky big data. So don't feel like you have to, you know, load your entire analytic engine or use every single parameter you have for your data. We actually use 40 to start with Einstein and then went to hundreds and then thousands. But thinking about starting with tens, not thousands of parameters in the data. Third, have a relentless focus on frontline user engagement. What are their needs what are their unmet needs? What's their feedback as you're rolling this out? How could you learn from them? What's in it for them? This is really about that frontline employee and how you can enable that group. The fourth is make sure you're bringing in all your stakeholders early. So thinking we're in a highly regulated industry, perhaps some of you are as well. Financial services, healthcare, making sure that you're bringing your risk management compliance partners in the mix. Anyone who could potentially, you know, support the work or be, be a detractor, make sure to bring them in early. So as our CEO says, you know, we're, we're operating down, straight down the middle of the fairway when it comes to, to fair data use. Right. So bringing our compliance partners in early was a critical piece. And then the fifth and last one is, you know, celebrate your success. There's no reason that each one of you can't be here doing the Einstein keynote next year in 2018. Why not? Like we wouldn't have thought this last year. So whatever your successes are, whether it's here, whether it's in any form you have at your company, you know, market the heck out of your, your learnings, your successes, and your company will continue to give you the resources to execute an AI and ML. And we've, we've just really loved the Einstein platform. And I think we're just getting started.

Absolutely. All righty, Great. Thank you again for your business. Thank you for your partnership. How about a big round of applause for Bill Hoffman? All right, now we're going to switch to the platform side of the story. And to do that, we're going to tell the story of College Ford, this amazing, amazing nonprofit in Austin, Texas, led by an amazing CEO. And to tell this story, I'd like to welcome up on stage, Marco Castellana, VP of Product Management for Einstein. Come on up. All righty.

Boom.

All right, well, thank you all. Einstein is great for large businesses like US bank, but it can also be very powerful for small businesses. And so today, as we give you a look at the incredible new capabilities that are my Einstein, we're going to do so from the perspective of not just a small business, but yes, a non profit College Forward. College Forward's mission is to get underprivileged students to and through college. And they do so by employing a network of coaches that work with the students and ensure that they are successful. Now, in the area of Texas that they serve, less than 10% of underprivileged students ever make it to college. But when they go through College Forward's program, they have a college enrollment rate of over 90%. 90%. So these coaches are working with these students and they log every interaction with every student in Salesforce. That gives College Forward a rich data set in Salesforce, and that's fertile ground for Einstein to do its work. Now, we have in the demo booth over here, a member of our product marketing team, Julian Sauvage. Julian, hello.

Hey, Marco. Hi, everybody.

Julian, if you don't mind, can you give us a look at what a coach for College Forward would see in Salesforce? Well, here we have it. Thank you. So, like many of you here, they have heavily customized their Salesforce instance. So along the top, we see various custom objects, students and schools and semesters, things that make sense to their business. Along the right hand side, you can see tasks that are due of the coach today and conversations and chatter with colleagues. And in the middle, we have Einstein Analytics. But right now, these analytics are looking backwards. They're looking at students who have enrolled and who have dropped out. And the question is, can we make this look forward? Can we make this into an early warning system that will tell us which students are going to drop out of the program before they do? So, Julian, is that something we can do with Einstein?

Yes, absolutely, Marco. Well, it seems like you want to predict the dropout, right? For students, it seems to be like a pretty custom use case on custom data. I think this is a great fit for my Einstein, namely Einstein prediction builder, Point, click and predict. So I'll just go ahead and get started. What I really like about this product, Marco, is that it tries to a bunch of different models, including random forests.

Whoa, whoa, whoa, whoa, whoa, whoa. No. If I'm not mistaken, you have a master's degree in statistics, do you not?

I do.

And that makes you a data scientist, does it not?

I guess it does.

I'm sorry.

We designed this product for Salesforce administrators who don't necessarily have a background in data science. And so we're going to do something a little different today. Is there any trailblazer here in the audience that would be interested in going up to that demo booth with Julian and building a prediction right now? Raise your hands. Raise your hands. Anybody? Come on.

Anybody?

Yeah, yeah. All right. This lady right here, she rose her hand pretty quickly. Come on up. Come with me. That's right. Let's build a prediction in front of like 2,000 people. What could possibly go wrong? Come on, let's go up to the booth. All right.

No, Pressure.

Watch the bump here. Okay, before we get started here, can you tell us, what's your name?

Avalon.

All right, Avalon, what do you do?

I am a data manager for a local nonprofit called 10,000 Degrees.

Also for a nonprofit. Excellent. Yes. All right, well, I meet Julian Sauvage, and please build me a prediction.

All right, Avalon.

Hello.

Hey.

You can keep your mic if you want. All right, so I'm not gonna be clicking. This is Avalon's fingers. So let's start with a new prediction here. You have no PhD in data science.

You sure?

No, I do not.

Okay.

Not yet.

Not yet. All right, well, let's get started. We'll give that prediction a name. So let's call that dropout. Because we're predicting Dropout, right? Yep. And then go ahead and get started. And then there's four things you need to do. Do, Avalon. Four things, and that's it. So step one is select the object to predict. We're going to choose that custom object called Student. I think you need to scroll down a bit. Here we are. Select that and then hit next. Okay, step one. Step two is we're going to select the field to predict. So let's go ahead and select Dropped out here in the middle. Perfect. Go and hit next. Step three is just confirm the data we'll use to make our prediction. Seems like a pretty rich and good data set that we have here. So I'll go ahead and. And hit next. Okay. Fourth and last step, Avalon, is save that prediction to a new field that we're going to call risk of dropping out. You can select that. Hit next, Review your prediction settings. We're good with that. And bam. Just go ahead and build your prediction. And that's it. That's what it takes.

Four clicks.

Marco, four steps.

Good job. Thank you. Thank you, Avalon. That is incredible. You got to give it up to her standing up in front of the crowd like this. Well, Avalon has just built a custom prediction on a custom object. That is incredible. But now what? Okay, so, Julian, you built the prediction. What are we going to do with it? What does it even look like?

It's a great question. There's multiple ways you can visualize the prediction in your pages, Right? So I'll just go ahead and finish that. You can put that in your workflows with Process Builder. You can do a bunch of things. You can just have your prediction in your page. So I'll just go ahead, open my Lightning app Builder, drag and drop that newly created Einstein Precision Builder component. Save that, hit back et voila. This is your. In your student record.

All right, I see. Lightning component, huh? Now, wait a minute, wait a minute. I know that a lot of folks here are still using Classic for various reasons. Does this work in Classic too?

It does.

All right, so for those of you using the Classic interface, yes. Einstein works with Classic as well. So now we have a prediction. In this case, it's a prediction of a student dropping out. Now, you can imagine that many of you would build a similar type of prediction, maybe not of students dropping out, but of customers attritting or of employees leaving. Einstein Prediction Builder can handle lots of different use cases. And if we take a close look at it, we can see kind of the essence of what a prediction is. A prediction of this type is a probability. A probability of a future event happening. In this case, of a student dropping out. And like the other Einstein capabilities that Ali showed earlier, when Einstein makes a prediction, it doesn't just give you the prediction, it also gives you why. So it's got those predictive factors down there. Now, those predictive factors came from the data. So when Avalon went up there and she made those few clicks that she did, that kicked off this whole process where the Einstein engine went through. All the data that College Forward has, it looked at all the students, and for the few of them that dropped out, it found these patterns, patterns that pertain to dropping out, about attendance, about gpa, about. About living conditions, and many more. And it applies those patterns here to arrive at this risk score. But all of those factors there are coming from standard and custom fields in Salesforce. This is all structured data. But College Forward also has a lot of unstructured data. As I mentioned, these coaches, when they meet with these students, log every interaction in Salesforce. But they often do so in. In the form of free text, in the form of counselor notes. Now, we can use another part of my Einstein Einstein Language Services to extract useful information from this free text. Let's have a look at that. So, as you can see, Julian has already added a counselor note. Julian, would you mind submitting that? Now, this note indicated that the student was having trouble finding housing. Einstein Language Services can read that text and it can extract information from it. In this case, it did so and set this flag on the student's record, saying that she's now in a troubled living condition. The Einstein Prediction Builder prediction that Avalon previously had built discovered that that was one of the risk factors. And so, as a result, because that flag was set, the risk score has now risen. Now, it's one thing to look at a prediction and see it on the page. But what you really want to do is take action on that prediction. But the predictions made by Maya Einstein are eligible for all the rich workflow capabilities that are built into the Salesforce platform. And so here we've built a workflow with process builder that when a student is above a certain threshold of risk and when one of the factors is a troubled living condition, it will automatically create a case to assign a social worker to work with that student and attempt to find that student housing and get them back on track. So now that we have this prediction in the system, when a coach logs into Salesforce, they will no longer be looking backward, but looking forward. They can see at a glance all the students that they're currently working with and their risk scores so they can know immediately which students most require their attention. And now to tell us a little bit more about how Salesforce and Einstein are changing the way College Forward changes the world is none other than the CEO of College Forward, Austin Buchan. Austin.

How you doing, Marco?

Welcome to Dreamforce.

Thank you so much. Well, excited to be here.

Can you tell us a little bit more about the problem that College Forward is solving?

Absolutely. It's a daunting one, everyone. As you showed earlier, only 10% of our nation's low income students are going into and actually completing their college degree. This is compared to about 70% for their high income peers. And this is a number that has been with us since the 1960s and is really only getting worse. So every single year, we're seeing hundreds of thousands of students, students enroll and subsequently drop out before they complete their college degree. And that's exactly as doom and gloomy as that sounds. This is where College Forward steps in. In combination with these incredible young people that we employ through an AmeriCorps program, we're able to really prove that anybody with a desire and a motivation to go off to school with a little bit of support from someone who's been there themselves, maybe just recently graduated from college, can make this happen. And the wraparound support that one on one individualized coaching to demystify the process, to connect them with resources, is really a way for us to level the playing field at scale. And we've been doing this. So despite the odds, despite the cards that are stacked against so many of our students on their path to higher education, we see time and time again thousands of our students beating the odds, becoming the first person in their family to graduate from college and really start breaking that cycle of generational poverty.

Absolutely Wonderful. Thank you. Now, how did you come to use Salesforce?

Well, I think like a lot of small and mid sized companies in the world for profit or nonprofit, College Forward was really struggling. Data management was a key value of the organization we knew had a huge critical role in helping us evaluate our work, better understand what it was that we were doing well and what we weren't doing well. But we were doing this, I'm only slightly embarrassed to say, on a combination of 27 Excel spreadsheets, Google Docs, and a third party software that we basically bent until it broke. We were very fortunate. Unlike a lot of nonprofits, finding resources to invest in data platforms is really hard. But we received a large grant and I mean, it's no brainer why we picked Salesforce. It was customizable, it was in the cloud. We have 130 young people that, that worked for us who were coming straight out of college themselves. So the usability had to be there. And that's really what we did. We spent about six years customizing and converting the entire platform into a 360 student information system that allows us to do all the incredible work that we're now looking at today.

Yeah, it is amazing. Now, I understand that you're not just a Salesforce user, you're also an App Exchange partner. So how do you anticipate that Einstein will change how you work with your partners?

Absolutely. I'm gonna steal John's phrase, we are scrappy. And part of our scrappiness is realizing that as quickly as we've grown from 30 students to 8,000 students to where we are today, it's only a fraction of the need that's out there. And we recognize this pretty early on. And in partnership with Salesforce about three and a half years ago, we took all these incredible customizations that we've made on the platform, packaged them up into the to this incredible new tool that we call Copilot, which has been on the App Exchange. We had one partner who's actually in the audience today. College possible in day one. Fast forward to today. We have 40 organizations collectively serving 240,000 students throughout the US across 20 states. And we had our first international partner in Rwanda doing this as well.

That is awesome.

So, you know, Einstein, it's really building off the vision of Marc Benioff of democratizing access to these sorts of tools. What we build at College 4 to support our staff and students doesn't stay with us. We can take these customizations. This incredible work that we've been building with you Guys put it into this managed package and really lift all boats with really just a few clicks.

That's right. Because Maya Einstein will be packaging too. Yeah. That's fantastic. Now I understand that we actually have here a graduate of one of your partner programs. Is that true?

That's right. One of our incredible copilot partners that we've been working with the past few years. It's this great organization called 10,000 Degrees doing very, very similar work to College Forward actually right here in the Bay area. And this is going to sound eerily familiar to you, but that young woman you brought up to do the demo, Avalon, is not just a trailblazer in Salesforce. She's a trailblazer for herself, for her family. And you'll hear a little bit more about her story. But I think you're going to really appreciate hearing this.

All right. Avalon, where did you go? Avalon, please come on up. What a coincidence. Avalon, please join us. Welcome, welcome back, welcome back to the stage again and thank you for building that prediction earlier.

Yeah.

Can you tell us your trailblazer story?

Yeah. So it's kind of a long story, but I like Austin mentioned. I am an alum of a a local nonprofit called 10,000 Degrees that's based in San Rafael, just a half hour away. And I am the first in my family to not only graduate high school, but also attend college and graduate college.

Congratulations.

Thank you.

And a lot of that comes from the support that I got from town thads of degrees. I was a high school student that didn't know any anything about the process of getting into college. They helped me figure out how to get in, how to pay for it and how to succeed in college. So I am endlessly thankful. Towards 10,000 degrees I went to Mills College, which is a local women's college across the bay. Women's colleges.

Yeah, Mills College.

And I studied, I did a lot of research and then that got me into data. And then after I graduated college, started working for 10,000 degrees, working directly with other first generation low income students and our peer to peer model. And I loved it. But I started doing more data centric work, getting back to what my experience is in after about a year and then a year and a half ago, we started working with Copilot and Salesforce and it just drastically increased our efficiency, our data integrity, our security and everything and making so that we can improve the work we do with students. And actually it's interesting the example that was put for the language interpreter, the language processor brought something up for me in that I also experienced housing instability when I was in college, and I was very much at risk of being homeless when I started college. And if we can bring this kind of work into with our advisors who work with students and they can have these tools to bring the finite resources that we have in college access programs to direct those resources to the right student at the right time so that they can succeed, we can do so much good in these communities.

Fantastic.

It's very exciting.

Well, thank you very much for being a trailblazer. Let's give it up for Avalon, Brave Trailblazer, and Austin Buchan.

Thank you, Marco. Thank you both.

Thank you, Avalon. And now, now to give you a look at the engine. Under the hood of Einstein is the designer herself, our senior data scientist, Shubha Navar. Shubha.

Hey, Marco. Thank you. So. Thank you, Marco and Austin, and I think that is an incredible story, and you guys deserve another huge round of applause. So I find it incredibly exciting that a nonprofit with a fraction of the resources of a Fortune 500 company can use AI to really transform the lives of underprivileged children. I've been a data scientist for a decade now, and I know that it takes a lot of work to bring this kind of intelligence into your business apps. You need to hire a team of data scientists and data engineers to extract your data, understand your business processes and your use cases, accurately model your predictive problem, and then surface the predictions right back in the business context. And it's the need for this kind of really specialized skills that puts machine learning out of reach for a majority of businesses today. And that's a problem we wanted to solve with Einstein. We wanted to make machine learning accessible to everyone. Now there aren't enough data scientists in the world to. To solve all the millions of use cases out there. So we had to do something special. We had to build AI that builds AI. What does that mean? What does it mean to build AI that builds AI? Well, it turns out that when a data scientist works on a problem, there's usually a very definite set of steps that he takes. And this includes things like extracting important signals from the data, transforming those signals, trying out many different machine learning algorithms and so forth. And it's a very iterative process, and it can take you several weeks to a month to build a really good predictive model. So instead of doing all of that, we decided to teach Einstein to do that instead, so that Einstein automatically builds the best predictive model for your data and your use case. So with that, let's take a look concretely, at CollegeForward, behind the scenes, and see how it builds and how Einstein builds these predictions. So now we all know that predictions are only as good as the data that you feed them. So the very first thing that Einstein needs to do is find out all of the College Forward data that could be relevant for making these predictions. Now, we haven't told Einstein anything specific about College Forward, what it is, what it does, what its business processes are, but Einstein does have access to the Salesforce metadata, and so it uses that metadata to identify all of the relevant information that CollegeForward has. So it finds the student object, of course, which is information about every single student. But it also identifies that the high school object is really important because it contains really relevant information, like whether or not the student's high school has a college prep program. And it finds that the task object contains information about each and every interaction between a student and his counselors. So it automatically identifies all of these relevant objects, and it joins them all together. And now that we've identified all of these relevant objects, there could be thousands of fields across these objects, and each of those fields could be sliced and diced in many different ways to generate signals that could be used for predicting. Einstein automatically does that for you. So it automatically identifies the number of times a student missed his interactions. His appointments with his counselor is a very important signal. And it identifies the number of times. It identifies the segments of family income that are correlated with dropping out. And so it does all of this. It identifies all of these important signals from these thousands of fields while throwing out all of this furious noise. And then we come to my favorite part, which is actually building the predictive models. Now, there are dozens of machine learning algorithms that Einstein could try out, and each of those algorithms comes with dozens of different knobs that could all be tuned to varying degrees. And so there's literally millions of combinations that Einstein could try. And finding the best combination can make all the difference between identifying that student who needs help and just letting him slip through the cracks. And so Einstein intelligently navigates this space of possible combinations to automatically find the best predictive model for your data in your use case. Now, one would think that the story ends there, but these predictions are really useless unless they are delivered in the right hands, where a guidance counselor can take preemptive action on them. And so this is exactly why we build Einstein into the Salesforce platform, so that you could get these predictions where you need them to when you need them. So that is all the. That.

That.

So that is all the things that Einstein actually does. And I find it incredible that we took all of that complexity and boiled it down to just a few clicks. And so I'd like to pause and just appreciate all of the incredible hard work that has gone into building this. And I'm so proud of our team. They're sitting right here. So now think about your business and your use cases and what's most important for your business and how you could use the predictive power of my Einstein to really be an AI trailblazer for your company. And so with that, that's all the product that we've been building, but we've also been doing some pretty amazing research. And so I wanted to call up our chief scientist, Richard Soter.

Take a number.

I was Lightning.

Record the bundle.

Thunder. Thunder. Thunder. Thunder.

Thunder. Thunder.

Thunder.

So at this Dreamforce, we want to focus mostly on the things that are now running in production that are impacting our customers. But rest assured, we're not going to stop there. You saw a couple of the products and features that Mark and we just touched upon in these keynotes, and several of those were just research projects last year and now they're impacting sales, service and marketing features and products. But there's so much more. We've already had some first versions of natural language understanding with bots and trying to classify your intent. We can find your company logos now on social media. We can identify scenes or count your products to some degree. But I want to show you now another long term vision that we have in Salesforce Research and that is to empower all of you with your own personal AI assistant and an AI assistant that understands all your data that is in the CRM. An AI assistant that you can just naturally talk to after you come out of a sales meeting and update all your fields in natural language. Or an AI assistant that can read all your knowledge base articles and allows your customers to ask questions about your business in much more natural ways by just typing them and without you having to teach all the potential questions that might come up. So in order to do that and to build this kind of meta AI assistant engine for all of you, we need to work on a couple of different things. So those are not going to be. This AI assistant is not going to come out next quarter, but we have a couple of projects that are laying the groundwork for that kind of engine. And I'm happy to tell you that we actually now have internally one of the most accurate speech recognition systems in the world. And we also acknowledge that for a Variety of all these AI features in natural language understanding. As we're growing internationally and as our customers also grow internationally, we need to do better translation. We want to be able to translate quickly documentation into different languages. And we now have one of the most accurate and fastest machine translation systems in the world. I want to show you a couple of demos too to see how this works. So let me start with a question answering model. So here you see a somewhat complex question. Why was AC Electricity gaining popularity? And as the algorithm reads the question, it is trying in different parts of this longer document. Imagine this was your knowledge base article. It's looking and trying to pay attention to different parts of that input and it's actually able to correctly identify the advantage in long distance high voltage transmission. And now of course, that is just one example. But imagine it could read through all your knowledge base articles all the complex service cases you've had before, and now you can very quickly deflect even more complex questions about your products. Another example is in visual question answering. Here you see a simple image and you have the question, how many men are on top of the truck? And you see in the beginning it's still focusing on the person who is just getting into the truck, but eventually correctly converges on just the two people on top. Now imagine this isn't a truck, but this is your supermarket or your store. You want to identify how many people with a handbag are standing in front of this aisle. So that is another example where we can now actually very accurately answer counting questions. And not just about the objects that you've already seen before, but counting objects that you've never seen before in a certain constellation, in a certain location and so on, but not all the time. Do you know exactly the question that you actually have? Sometimes you may have a very complex service case where already 50 or 100 emails went back and forth and now it gets escalated to somebody else and you need to summarize essentially what was going on. So we've also worked on summarization. Here you see the outputs of an algorithm as it's picking and selecting different words from a longer input document in order to summarize it more succinctly. And this is the first time a deep learning algorithm is able to generate long coherent sequences of text in the history of deep learning. And of course this is again a first version, but let's face it, with all these amazing features and just this keynote, we could all benefit from a little summarization. So with that, I'll give it back.

To John all righty. Great job, Richard. Boom. Great job, Richard. Great job, Chuba. So right about now you might be saying, as Richard said, we had a lot of AI, we've got a lot of Einstein. Where do I even start? Well, the place to start is obviously trailhead. So what's great about Trailhead is we have, of course, dozens of dedicated trails for Einstein. But it's not just about examples and best practices. You can spin up a trial of my Einstein right there in trailhead. Now, my question to all of you after all of these presentations is what are you going to predict? How are you going to use my Einstein to transform your business? And so here at Dreamforce, today is AI Day. We have dozens of sessions for Einstein. You'll see Einstein and all the product keynotes. We have a special call out for the fireside chat that Richard will be leading this afternoon at 3:30 over Moscone west on how AI is transforming the future of business. Also over in Moscow west, we have the Einstein Observatory, where our experts are there waiting to help you out and get you started with Einstein. And most importantly, fire up the Dreamforce app and take the survey. Tell us what you liked, what you didn't like, what was missing, what you'd like to see, what are your use cases? That's how we build great product here at Salesforce. And with that, have a great day and a great dream force.