Platform Keynote: How To Build and Govern Custom AI Agents and Apps | Dreamforce 2024 | Salesforce

Please welcome executive vice president and general manager salesforce platform, alice steinglass.

Wow.

It's amazing to look back at how our platform and really the entire industry has changed over the last 25 years. I just dropped my kid off at college a few weeks ago and he's going to major in computer science and I look at the classes that he's going to take and how different those are from the classes that I took in the 90s. Now some things are the same. We're both making cheesy video games. Mine was a version of Snake that had ASCII art. It was pretty cool. Despite the ASCII art, he his has like amazing sounds and 3D graphics and he has AI helping him build it and he's calling it Escape the Duck. I don't know what that means, I don't know what that is. I don't know how to Escape the Duck. But I do know AI and I wish that I had AI to help me build. It's clear that what we build and how we build it is, is changing dramatically in this world of AI and we couldn't do it without you. So I want to start by saying thank you. I want to say thank you for joining us over the last 25 years and I want to say thank you for being here today at the platform keynote. I appreciate all of you in the room and I want to know who's in the room. So we have over 21 million trailblazers, I think some of them in the room, some of us joining online. Do we have anybody here who works in it? Can you clap if you work in it? Okay, I got a bunch of them out here. How about admins? The admins are always pretty loud. Yes, I got my admins over here. Developers. Any developers in the room. Sweet. Together you have built 9.5 million apps on the platform. That's incredible. And you've automated even more. We're running over 300 billion automations every month. That's 350 million just during this keynote and that's doubling year over year.

Now I know we also have some partners in the room.

I saw some of you earlier. I don't know where you're sitting. Can you clap if you're a partner?

Thank you.

You have built over 10,000 businesses around the app exchange program. What all of these numbers tell me is that this is the platform for it and together we're going to take it into the future. Now the apps you build today look something like this. They're useful. You wouldn't have Built so many of them, but they require the user to do a lot of clicking to figure out what's going on. They're putting a lot of burden on them to find information. Imagine if they looked like this. We have agents who can work with you conversationally to build, to create, to take action with your users. It's not just what you see on the outside, it's AI behind the scenes, powering automations, powering the components that are on that page. AI is going to fundamentally change what we are able to build, but we know it can be challenging. It might be the number one priority for your CEO, but 88% of IT leaders feel they can't meet the AI demands safely. And this list of challenges, disconnected data safety concerns, tech debt, it's familiar. It's the same list of concerns that you've had for years and that we've been putting on these slides for years, but they're compounded when it comes to AI. So that is why we are making foundational changes in the Salesforce platform, so that you don't have to. We're going to make this easy. So you saw Mark, he was showing us, this is his slide. We're going all the way to this world of agents. And then he put up this slide which is our agent force, what AI was meant to be. But because this is the platform keynote, I want to dig in on that bottom layer there. So let's expand that out. Let's talk about what's in the platform that's going to make this possible for you. So it starts with Hyperforce. This is our next generation infrastructure built for the public cloud. It allows us to have greater security, availability, compliance across global regions. Above Hyperforce, we have our data layer. This is not just your CRM data and it isn't just all of those custom objects that you've built on the platform. Last year we talked about how Data Cloud and Mulesoft were going to allow us to bring in data from across our enterprise. Data that was trapped in snowflake or databricks and use it with zero copy as part of our solutions as native objects in the Salesforce platform. This year, talking about unstructured data, how you can bring in a pile of contracts and PDFs and recordings. Yes, I got a yes here, thank you, Steven. That you can bring in and use them natively as part of the platform. Above the data layer, we have our AI layer. You've heard a lot about this. I'm not going to go into it in depth, but one thing I Do want to point out is that the AI layer similar to the data layer is open. So just the way I can bring in my data from another system, I want to be able to use models like Anthropic or OpenAI. But also the model that you've trained, you should be able to pull that into the platform. And above the AI layer we have trust. Now security is built in to the Salesforce platform and we also have platform native solutions that allow you to do things like event monitoring or advanced encryption, backup or archive that can increase your security posture. Finally, we have my favorite layer, the tools. These are the low code tools and the DevOps tools that are going to let you build on the platform, build your own agents, build actions for those agents, build those automations and build those apps and all of this, everything you see here is built in to the Salesforce platform. So when you build on Salesforce, you get all of it.

Now.

Let me be clear here. When I say build on the Salesforce platform, I don't just mean build on sales or service, I mean build for every department across your organization. You can build apps for hr, for finance, for marketing, for operations, and you can build apps for yourselves for your own IT team members. To get you started with that, I am pleased to announce the GA of AgentForce for developers. A round of applause. If you didn't get a chance to try it during the beta, you should check it out. It does inline autocomplete, it does test case generation if you don't like writing test cases. And it can explain that code that was written 10 years ago that has no comments. All part of agent force for developers. Now I'm sure all of this has got you inspired. You want to be building, you want to think about what you want to build. So to inspire you, we thought we would build an app live here on stage. And to do that we figured we would make up a fictional production house called Coral Studios and we're going to build a recruiting app for them to be able to manage contestants for their TV show. Which of course means we need a TV show. So anybody here like basketball? A cat likes basketball. So I've got a basketball for me. I'm a Bulls fan because I lived in Chicago in the 90s, so I have to be a Bulls fan. Any Bulls fans out here? No. No? Okay, so anybody here like reality tv? I feel like there was less applause for that. You can admit it. I like reality tv. So we're going to do basketball, reality TV show, hooping. Come on. There we go. Hooping with the Stars and this show isn't real, but all of the code we're going to show you today and all of these steps are out absolutely real. We've got real live code demos that we're going to run here. We're going to show you how to pull in that unstructured data to use it as part of your app. We're going to build an agent with low code. It's not for sales, not for service, it's for recruiting for my basketball reality tv. And then we're going to make it very real with security and governance. So to get us started, I'd like to welcome to the court our number one playmaker and an amazing architect, Leo Tran.

Thanks, Alice. Hooping with the Stars sounds so fun. You know the star I want to see on that show? You've seen him all over the Bay Area. His name starts with a C, ends in a Y. You can guess it's Cody. Of course he is 7ft tall. I looked it up and he might not have the biggest wingspan, but he is unstoppable in the post. Well, leaving that daydream aside for a moment, I have been an architect and engineer of the Salesforce platform for over 16 years and that video we saw at the beginning was an amazing reminder of platform's continual transformation. Agent Force is the next big leap for building apps with AI and agents. But for AI and agents to really work with you, it needs to know about you and what's unique about your organization. And so we have some work to do. And that work starts with data. AI and apps are better with activated data data they can discover and understand. Today, many of you are using Data Cloud to connect and activate your data across your data stores, your data warehouses and third party systems. This data though, is structured data. It's usually tabular, comes from digitized forms, has great metadata that says these bits and bytes represent an email address versus a corporate address. But what about unstructured data? This highly valuable data is locked in PDFs, documents, slide decks, audio and video files. It's highly valuable data that's hard to discover and even harder to quickly understand. So we've built a few things to help you. First, we've expanded data Cloud to work with unstructured data. So you can also activate it and integrate it into your apps and your users flow of work. How does AI work with unstructured data? Let's talk about Retriever Builder and rag or retrieval Augmented generation. RAG is a really big idea for generative AI. What is it let's have AI retrieve or look up the information it needs to augment and improve its generated responses. Simple enough to understand, but the hard part, how do you actually find that relevant specific piece of data, especially if it's unstructured? That's why we built the low code Retriever Builder. Retriever Builder walks you through how to find a specific and relevant piece of unstructured data. Then takes care of the heavy lifting of query crafting with Data Cloud. Once you've built a retriever, you can use it over and over across the platform. So with Retriever Builder and Rag now AI can work with the latest information about your organization. You don't have to deal with the cost and frustration trying to do something like build a custom model of unstructured data that's immediately out of date as soon as it's done. Now C is believing. So let's see how a developer can use unstructured data and Retriever builder to improve their apps. Let's meet the developers of Coral Studios, the hit maker of reality shows like Hooping with the Stars. So please welcome first our six time all star app builder, Brian Shellack, Senior director of Platform Product Marketing and straight from Paris, our own three and D specialist, Lead specialist SE Arnau Barton.

Leo, we're thrilled to be here. Put us in coach.

Passing the ball to you, Ryan. Let's load up the apps and Coral Studios.

Give me the rock. Come on, let's do it. All right, so Ellis said Coral Studios is building this custom recruiting app to manage the casting process for Hooping with the Stars. I would totally watch the show and look, they've got a lot of great information. They're already relating in here. You've got some data cloud related lists coming in around past seasons you have your applicants, but really important for the casting managers are these applications. Now they get a thousand of these applications every single day and they need to be able to go through these and understand the qualities and characteristics of any one of these given applicants. And the nice thing is you've got a lot of good data here within the contact record. So we've got a bunch of information that we've also related in from data cloud and prior casting calls they've been on, applications they've submitted and a bunch of rich detail. But you know something that's really interesting here is how can we help these talent managers get through all this data really quickly and that's where you can bring AI and data together and put this data to work. Right, so let's go ahead and build a prompt template that can help us with that. And I've got a pretty good prompt template right here. I kind of like it. Here's what I got. I've got a record snapshot which is grabbing all of the related information on the application object. It's pulling that in. And I've got a flow that I brought into this as well. So I'm grabbing related information to that applicant. But you know, Leo, what I'd really love to get at are those videos because every prospective Hooper is submitting one of these audition videos. Takes a long time to get through them.

And it makes sense. Coral Studios make TV shows. They want to see what their prospects would look like on TV in video. But these audition videos contain so much useful information that would be great for AI to use. Ryan, you got any ideas on how we can have AI use these videos?

Can I get to my beta bag of tricks? Can we see some beta products over here? Some new stuff?

New stuff.

Okay.

Okay.

Well, I'm going to S3 a logical place to go because this is where we have all of these audition files that people are submitting every single day. In these you can see the personal characteristics, the little quirks that make them a great app kit. It's how you know they're here for the right reasons. If you know, you know, that's Bachelor Nation reference. And look. So here's what we're going to do. We're going to bring this into data cloud and, and we're going to do that through a new feature called our audio video transcription service. And this is going to enable us to bring this unstructured data to Life, even those MP4s and use them across the platform. So I've ingested some of this information already. I built an unstructured data model object and a search index on top of it. And I can already query this right here within the query editor. And I could write some Apex to get at it as well. But I mean, Leo, we're all about low code, right? Let's talk about some other ways we could use this. Yes.

This is going to be the cool part. You've heard about rag. Let's see. RAG in practice on the platform. Ryan, how are we going to have AI retrieve these audition videos?

We're going to use a very well named product, Leo. We're going to use retrievers. Retrievers are brand new within the platform and retrievers enable you to create this artifact with low code that lets you pull relevant data chunks from any of these unstructured data sources, MP4, S, M4, as PDFs, all of them work. Here's what we're going to do. We're going to build a retriever here together and we're going to use another well named product retriever builder within Retriever Builder. I'm going to go grab the data model object. We have those addition files that we had in our S3 bucket, and I'm going to grab the search index that was automatically made from it. From that point on, I can filter, for example, to choose only certain pieces of information from that unstructured data. But I'm going to pull it all in over here and then identify what attributes I want to pull from this. I want some direct attributes and that can include things like what exact file are we pulling from. Another one that I want to pull in might be a related attribute. This could be something like within any given file, what was the chunk sequence? Which is basically a way of saying within this video that someone sent, when did they talk about when they were a youth basketball coach or when did they share that they were at Duke University for four years playing ball? Now we can put all this to work and how are we going to ball out with this thing? We're going to go back to our prompt template. So back in our prompt template over here, we're going to actually insert this retriever in using clicks. So over in Einstein Search, I can choose the retriever that we just built. Once I bring this in over here, you can see that I can provide a simple natural age instruction on what I'm going to grab from any given candidate. I just defined it as something as simple as for this given candidate, do they have basketball experience? Now I have something I can use across the platform. Leo.

Wow, we have seen a lot. So I want to take a moment and, and go over the big picture architecture of what you all just saw. The problem statement we want better prospect summaries that make use of the audition videos. So what do we do? We leverage Data Cloud, which now supports unstructured data, to process those audition video files. Then we used Retriever Builder to build a custom retriever to find that specific prospect's auditions video when we need it. Now we're able to augment our prompt. We can have that summary prompt, use that retriever and include that specific prospect's audition video. And the outcome of this is richer, more complete summaries about each and every prospect. There is a lot of technology going on here. It's hard stuff. But what you're seeing here is the platform's approach to simplifying to help you build better apps with AI and data. Ryan, let's see the outcome of all this. Great work.

I love it. Let's do it. Let's go back to that application record page. And over here I've got my handy dandy magic pencil. We don't know actually what to call this, but this is where we can actually activate this field generation prompt template. What this is going to do, you heard from Leo, it's going to retrieve the relevant chunks of information from those MP4s we brought in. It's going to use them to augment that prompt. And the last thing, the G here, it's going to generate this great rich summary of this particular app kit. And now we have something that can scale across the entire candidate recruitment process, helping all of the talent managers and cast managers understand at a glance if someone is a good fit for one of these shows. Wow.

Give it up for Ryan and Arno. That is an amazing demo. As a platform architect, I think a lot about how to build an integrated platform that's continually transformational and expands what new and existing apps can do. And this demo is so great because it shows the beginning of what's possible. We showed you Data Cloud processing video files. But Data Cloud can process many types of unstructured data. It can process PDFs, documents, slide decks, audio files. We showed you retriever builder to make it easy to create a simple query. But we have a suite of low code tools that make it as easy as possible to build complicated retrievers and to work with unstructured data data. And because this is all deeply integrated into the platform, you can work with unstructured data and retrievers, not just in prompts. You can use it and work with it in flows, in Apex in our ui. That's the power of the Salesforce platform. It's integrated, it's easy, it's always transforming, but doesn't leave your apps behind. And so we have laid a great foundation for AI with data. So let's see what it looks like to build an AgentForce agent. To do that, I'd like to welcome up the Postman because he always delivers. Senior Director of Product Management, Carlos Lozano. Take it away, Carlos.

Thank you, Leo. Good afternoon, everyone. I've been chewing this language AI problem for the past 10 years and let me tell you, Agents is the most amazing technology I've had the pleasure to work with. And I'm humbled to be here today and show you how EGTforce works and it starts with a request. This can be a user request, a manual request, but it can also be a programmatic request. Now, at every request, what the reasoning engine is going to do is going to retrieve the relevant information from that, it's going to create a plan, plan of actions, and those actions are going to be the most relevant to fulfill the job at hand. And if at any moment it requires additional context and additional information, it will go ahead and collect it and it will do so autonomously. And that is the very different of these types of systems. And you can build these custom agents using Agent Builder. The beauty of this is that you have already started your agent force journey, because the assets, the building blocks that you're going to harness in Agent Builder are the Salesforce assets you already know are the Salesforce assets that you already use. These are lightning flows. These are Apex classes. These are the new kids on the block, the prompt templates. With that, we want to show you how we can put all of that into music with a live demo. Ryan, let's go ahead and let's take it off where we left off.

Let's do it. Let's go back over here to this prompt template, because this prompt template, which is helping us get at that unstructured data, this is also the beginning of one of our agents, one that is autonomous, it works around the clock, and it can take action on behalf of our users using this or any other artifact within our platform. Carlos.

And if we go to the site of Coral Studios now, notice that in the lower right side of the screen, we now have this custom agent, our casting agent, that's actually going to greet us and salute us to candidates like Sophia. And Sophia is going to be able to have a very fluid and dynamic conversation with this custom agent. This is not a traditional chatbot with rigid if then do that type of conversational flows. This is more dynamic, this is fluid, this feels much better. And this is great. This is an informational experience. But we want to do more. How are we going to take it to the next level? Ryan?

Well, Carlos, let's go over to Agent Build over here. And you're right, what we've created here, it's actually incredibly compelling. In just a couple of clicks, we can build a knowledge agent like the one that we just saw, and we can do that with all of our knowledge within Salesforce and Data Cloud, or we can choose specific subsets of it that we want this agent to pay attention to. But, Carlos, I think the real magic here is when we can enable this agent to not only inform and answer questions, but also take action on our behalf.

That is right. We want to give these agents the agency to act on the company's behalf. And to do that, the team player here is going to be topics. So let's start on the left side panel here for topics. And these are the topics that you're going to allow your agents to cover. These are essentially the jobs to be done that you want this system to take action with. And you can start with out of the box topics from the asset library. As you can see here, these are topics for verticalized use cases. And of course you can build it from the ground up. But today let's go ahead and double click on that general FAQ topic. And topics have two elements. The first element is the instructions that you're going to give. These are the rules that explain the agent how to leverage this topic and how to use the actions that encompass it. And we want to add an additional instruction here to cover an additional use case. And it's as easy as what we can read here on the screen. If the customer is talking, that wants to talk to a person, then provide the emails to be routed to that department. That is how Ryan just coded that very topic using the most universal coding language, which is human language. Now, this is all great if everything is going smoothly, the happy path. But what happens if customers are requesting things that the agent was not designed to do? Or even worse, if someone has a malicious intent? Well, this is baked in with guardrails from the ground up. And as you can see in this case, it actually triggered a reverse engineering topic and it graciously is letting the user know, hey, can't actually accomplish this request. Okay, now we want to enhance our agents and for that we want to leverage the existing assets that we have in the org. Let me tell you, Ryan, there's something that we would like to do and it's the ability to deploy the background check that we usually do in a different fashion. Are we able to actually bring it to a conversational experience?

Carlos? We absolutely can. And so what we want to do over here is take this API. Actually, this set of APIs that we've already established. We want to turn these into actions and we can do that as part of the action creation process. But Carlos, walk us through how we create an action in order to create.

A custom agent action. It is as easy as creating a custom agent action right here. Clicking on new agent action, choosing the action type. What type of action is this going to be? The source type in this case, it's going to be an API coming from our Mulesoft API catalog. Now the system now is going to fetch everything that's available in our Mulesoft API catalog right here. Here we have already our background checks APIs and we're going to turn these into conversational experiences by doing the most important thing of this demo. If there's one thing to remember is that description is all you need. This is how we are telling the AI how it needs to use this action. So if you can describe it, AgentForce can actually do it. But Ryan, we should have our own Ryan here. If you can describe it, Agent Force will prescribe it. That's right. So check this out. Because Ryan has made this input required, if the information has not been provided, the agent will autonomously go ahead and collect it. When it has collected and has fulfilled the user request, it will move on to the next step. Ryan, we're ready. So we have our custom actions, we added our topic instructions. Let's go ahead and simulate how our users would interact with the system.

Let's do it. So you can see over here, I've brought in those three APIs, they're now registered as actions. Let's go over here and, and ask it something simple based on our instructions like can you. Let's get a background check status, give me the status of my recent background check so that I as an app kit can understand where are we within the part of the process. And the first thing it's going to correctly do is instruct me to provide its email. It's going to do this so it can begin an authentication process using the first of those three APIs. So I provide Safiya's email and what it's going to do over here is discern from my request the right topic to invoke, choose the right action and run that. And over here if I scroll down, we can see that it sent a one time passcode to my email. That can enable me to proceed forward with this. So let's go take this thing full circle.

Let's go full circle. Let's go back to the Coral Studio website. And now we have enhanced our agent and now our agent can give us informational use cases for hooping with the stars. But also we can run this background check right here. And then so when Ryan provides this beefed up utterance here, right here, the system is going to start reasoning. It's going to grab those inputs that email that one time password and here we go. Applicants like Sophia can even Ask for additional details. And Ryan has also added an instruction that if applicants have their background check validated, it will proactively suggest to schedule a casting interview. And this is something very interesting because this is the type of user experience that a traditional chatbot will choke on. But we can do it very naturally and very fluidly with AgentForce. Here we are now, the system is basically giving us the available scheduled times and we can go ahead and move on with the next steps. Wow. What, Ryan, what you just did here, you created a custom agent, you gave it a role, you gave it a purpose, you connected topics and actions to it, you gave it agency. Doing that, you hooked it up to data, and you decided in which surface you were going to deploy it. Now, this time around, we chose to deploy it in a surface, which is a visual surface, on the website, on that chat widget. But that is not the only type of surface where you can have agents. Agents can work in the background headlessly and can be triggered by data changes. For example, in the.

Org.

Thank you very much. For the next section on security and government, please welcome defensive player of the Year, SVP Josie Childs.

Thank you, Carlos. I am loving this Hooping with the Stars app, and in fact, it's bringing back some fond memories. I actually started my career working on the television show Survivor. That's right, that's me. And I'm going to let you in on a little secret. There are a lot of people out there that are trying to spoil a season of reality tv. In fact, for one of the seasons I worked on, the cast list leaked ahead of the season, which was a huge disappointment, and we definitely don't want that happening to Whoopin with the Stars. So now that we've built this incredible app, we have to ensure that we protect it. So today we're going to talk about how you can play both offense and and defense and govern and secure all of your data. As we've discussed, great AI apps require a lot of data. But as data grows, so does risk. So let's talk a little bit about risk. I'm sure everyone here has experienced an outage or a breach. Don't worry, I'm not going to ask.

You to raise your hand.

I'm not going to ask you to raise your hand. But whether in our personal lives or our professional lives, we've all been there. But it's not just outages and breaches that can be costly. New and evolving AI compliance is becoming more and more complex. Now, the EU is leading the way here, but Other countries, including the us are not far behind. And this is a place where you cannot afford to take a risk. Late adoption is not a strategy. So to combat these risks, you really need to do three things. First, you need to ensure that you are managing all of your customer data. And Unigine will implement strong security policies across your Org to have that really strong security posture. Second, you need to ensure that you're encrypting your most sensitive data and protecting it from breaches. And third, you need to ensure that you're tightening access controls. You need to ensure that people only have access to the data they need to do their jobs. Not everyone at Coral Studios needs access to Caitlin Clark's address. And this is important. In fact, it's so important, the companies are reporting a 15% increase in their security spend. And I'm telling you this today because either you're already in front of this or you can help your company get in front of these critical risks. And as your trusted partner, we're here to help. We're here to make this easy because we have a complete suite of tools to help govern your AI enterprise. And a strong data governance foundation requires all of the tools that you see here, including data accuracy to ensure that your data is up to date and correct, and data retention to ensure that you're retaining or deleting older, stale data. But to see this in action, I'm going to pass this ball back to Ryan. So Ryan, let's get back into that demo.

Josie, you said it well. Thinking about protection against business disruption, how we can ensure that our data is safeguarded. These are top of mind for everyone. So we need a full court press over here to ensure that this app is trusted. Now, it starts with how we build and because I'm not a terrible person, I'm building in a sandbox. What I'm going to do over here is I'm going to show you two things that exist now within sandboxes alongside all of your data. Your configurations, your metadata. The first I'm happy to share is that data cloud is in sandboxes. This is really exciting. I don't know how we actually tested this beforehand, but now we tested it within sandboxes and this is great. So as we stream in data, as we create calculated insights, we can do that in the complete context of our. Org. And, and I've got one more goodie for you. Agentforce also works in sandboxes. You might have already seen this in a pre release. Org. So get your agents going, build their actions in A safe and isolated environment. Now, part of how we ensure that we're trusted is we enable all of our team to build together within these sandboxes. But like my friend Arnaud, he's been trying to get some of the casting roster over here. He wants to know where the Celtics are going to be. I love you man, but I can't share this information with you. So how can we remove some of the information here or at least replace it so that when we're testing here, we're doing so with trust. And we do that through Data Mask, which has received a huge performance boost. It's 10x faster, it's got a slick new UI and it enables me to go into my sandbox and replace sensitive information with harmless jumbled placeholders like this. So that's how we can test with trust as we do this. But encryption is also key to our process. And I literally mean key, because platform encryption has been extended to data cloud as well. And that means that you can create and rotate on demand a customer managed key for data cloud. So you can have a similar security posture across all of your orgs and data cloud as you build. Another big thing here that we need to consider is understanding behavior. And we need both good and bad behavior. And for that we have event monitoring. And now with event log objects, which is brand new, we're able to see as a first class standard object all of these security and adoption events. When someone's report, when a report event is triggered, when a transaction security event is triggered, triggered, and we can report on them in places like CRM Analytics. Over here we can see, for example, all of this information around things like where we're having failed logins. We can also see some APM stats, some performance monitoring around where our application is scaling and where we need to do some more work with that information because it's all part of our platform. We can even create things like flows on top of these event log objects, enabling us to, for example, go and inform our entire infosec team via their Slack channel, let's say, of everyday security events that happen within transaction security events, all as part of the platform. Josie, this is pretty cool.

This is really cool because log data is now represented as a standard object. You have so much more control, but it's also so much faster. Instead of waiting 24 hours for log data to refresh, you're now going to be able to see that data in a matter of minutes. Okay, so now that we've protected our company, Coral Studios, let's talk a little bit about data management. So we're going to dive back into our demo and we're going to start here in Salesforce Backup. So let's talk about data resilience. You need to ensure that you have data resilience. So you're going to start in Salesforce Backup and we're going to go ahead and we're going to create a backup and this backup is just going to be running in the background. So that way if an integration goes wrong or someone accidentally deletes some data, we can retain or restore that data with just a matter of clicks. But what about the data that you do want to delete? Deleting data is not just about compliance. You also don't want to ensure that you don't have old or stale data running in your AI processes or syncing with data cloud. So here we are in Privacy center and we're going to go ahead and we're going to create a privacy policy. We're going to say let's look at applications that we received seven or more years ago and let's go ahead and let's just, with a couple of clicks, delete that data. So now we have less sensitive data sitting in our org, but we also have ensured that the data that we do have is up to date and correct. Now, what about the data that maybe I'm not ready to delete yet, maybe I want it for reporting purposes later on, but I don't necessarily need it running in my day to day org. Well, taking care of that kind of data is going to speed up your apps, but it's also going to help you save on security costs. So we are going to move over or sorry, save on storage costs so you can use those dollars for bigger production values next year. So we are going to go ahead and now here we are in the new Salesforce Archive. So welcome to Salesforce Archive and we're going to go ahead and we're going to create an archive policy. So we're going to say for shows that were canceled two or more years ago, let's take that data and we're just going to move it into a separate data store. So it's still within Salesforce, but it's just not running in our active org. You can think of this kind of like retiring a Jersey. We can still see the data and soon we'll be able to restore that data if we need it, but it's not running in our active org so we can focus on what really matters today. Okay, so that was super easy. How can you do this in your orgs today? Well, today we showed you Data cloud and sandboxes and Data Mask to help you test securely. We showed you database encryption and event log objects to help you increase your security posture across your company. And we showed you data management through Privacy Center, Salesforce Backup and Salesforce Archive. And we have invested a lot in this space and this is a lot of innovation, but it doesn't stop here. Earlier this month, Salesforce signed a definitive agreement to acquire OWN company, a leading provider of data protection and and data management solutions. Now we remain two separate companies, but once this exciting acquisition closes, OWN's capabilities will complement the existing products that we have, like Salesforce Backup Shield and Privacy and Data Mask. So stay in your seats because we still have a lot of exciting announcements, including our giveaway and Alice has a fantastic customer interview. So, Alice, back to you.

Yeah. So we built an application on the Salesforce platform. We connected it to unstructured data, we made an agent for it, and we did security and governance. But I'm guessing you don't just want to see it with a fictional Coral Studios. We want to see it with a real media company. So please join me in welcoming to the stage Garrett Boss. He's the EVP for enterprise and data platforms at Fox.

Hi, Ellis.

Thank you, Garrett. So I had the chance to go out to LA and visit your studios and meet with your team, and I wanted you here because I was so impressed with what you've built on the platform. Fox has over 75 applications running on the platform, covering everything from Fox Entertainment to Fox Sports. And you had an opportunity that I think many of us can only dream of a few years ago, where you got to rethink your entire platform strategy and you moved all in on the Salesforce platform. So can you tell us why you chose to choose Salesforce here?

Yes, of course. And thank you, Alice, for this opportunity and really appreciate the partnership with Salesforce. But Fox as a company started five years ago after Disney acquired 21st Century Fox. And as part of that process, they acquired all of our technology. So we had to stand up everything from new media and broadcast operations, all the way through new business applications. And we had a very short period to do it because we had only a TSA period for which we could use those, the older apps that went back to Disney. So we partnered with Salesforce and they helped extraordinarily in the process. And let me tell you how. So, first of all, it was a force multiplier in all the things that you Just presented a lot of the key functionality that is back office within an application around user management, administrative modules, all of these things you guys handled. We could then focus with our conversations on the business and building out the business requirements that they needed. And that really drove a ton of value and made us move quickly. Secondly, we had a single Org strategy. We deployed all these apps in a common org and we had a style guide for all of the apps, common data model. So as users navigated from one app to the next, there was a familiarity there. Right. And they were able to. The change management process as we were rolling out all these apps was amplified. And then lastly, when we partnered with you guys in 2019, we really wanted to partner with a platform that was open that we could move data in and out of. And we made that decision because you guys are an open platform. And shortly thereafter, you guys acquired Mulesoft and accelerated that process. We were a big Slack customer, and Slack joined the family, right? And we were able to serve up micro moments to our customers, approvals and notifications, and that's just expanding, which is very exciting. And then lastly, we're Tableau customers. And so Tableau joined the family, right? And we were able to then have all of our instrumentation and analytics be part of the flow of work within the application.

Okay, I want to hear more about some of those things you just talked about. So I was watching this weekend, my Seahawks, my hometown Seahawks beat the Patriots in overtime on Sunday. On Fox.

On Fox.

On Fox.

On Fox. And as I was watching, because I'd been down to your studios, I was thinking some of what I'm looking at is being run on the Salesforce platform. Do you want to share that with our audience here? Tell them about some of the specific things that you're building and using on the platform?

Absolutely.

Love to.

By the way, congratulations to your Seahawks. First place so far in the early start of the NFL season. But so we've built out apps across ad sales, production, marketing, legal. One great app that I'd like to highlight is within our sports production group. As you mentioned, we have over 1800 studio shows that we produce every year. Over a thousand events at stadiums across the country. NFL, Major League Baseball, college basketball, etc. It's a huge logistics challenge to get all the crew there, all the equipment, all the transmissions you guys set up for this event. Probably we do a ton of events every year. Right. And we use the Salesforce platform, an application called Playbook, to plan all of these events. And it has a mobile app that's part of it, that the teams on site can collaborate on what's going on within. Within the event. And it's a. It's a game changer for us. And it's a. It's a. It's a huge win. But it's not just sports production. We, like I mentioned, use it across ad sales, across many different groups, across Fox the platform.

So when we're all watching the super bowl, we'll now know how it's. How it's all being planned in the back end. I had another question for you. I know that you're a SHIELD customer. Can you tell me about why you're a SHIELD customer and how you use that?

Yeah. I think everybody who works in technology has security top of mind at all times. Right. No matter what you do, you have to do it in a secure. A robustly secure way. Right. And SHIELD really allows us to do that and move fast at the same time. We're able to build our applications, deliver value to our customers, and do it in a secure way, whether it be encrypting data at rest or in transit, whether it be protecting our PII information. Some of the features that you guys demoed in terms of masking data, et cetera, these are all great features of it and it's really critical to our strategy.

Yeah, I know you take security very seriously, so it's great to be able to partner on that. I have one more question. You all know where this is going. We've been talking about Agent Force and talking about AI. How do you think about Agent Force at your organization, Garrett?

First of all, we're very excited about it. I think we all recognize how transformative large language models and GPTs are and the new agent functionality within this, the platform. I think what makes Salesforce a differentiator here is that it's in the flow of work a lot of times. I mean, we all see the productivity enhancements that can occur with these tools, but if it doesn't have the right context and it's not in the flow of work, it's not as good as it can be. Right. And you guys have done a tremendous job building it into the application workflow so that we can present those agents and the LLM capabilities to the user with the right context during their flow of work. Example we were talking about with Playbook, if I'm scheduling the next game in Seattle, I might want to figure out what were the games that were over cost overruns or games that weren't cost overruns. You could ask these questions of an LLM in the flow work, and it has all the context and data from previous previously scheduled events.

That sounds like a great. A great use case and it, you know, means more things in Seattle. I'm all good with that.

Yeah.

Yeah.

Okay.

So before he leaves the stage, Garrett wanted to bring a little of the Hollywood magic to Dreamforce. So he has a special giveaway. Garrett, you want to share what you're giving away here?

Absolutely. Yeah. So one of you out there underneath your chair has a golden ticket for a trip to Los Angeles and a VIP tour of the Fox studio lot. So.

So let's take a look. I was out at the lots. It's going to be an amazing tour. Do we have any. Do we have a winner? Can you raise your hand if you want? I don't know where it is. They didn't tell me. So somebody here? Any winners?

Any?

Oh, I got it right over here. Congratulations. Thank you so much. Garrett. Thanks for the giveaway and thank you for being a Salesforce customer and partner for all these years. You're. It's amazing what you're building on the platform, and I really appreciate you joining us today.

Thank you for the partnership and huge shout out to the Fox team who makes it all happen. So appreciate it.

Thank you. Thank you.

If you didn't win a tour to the Fox lot, I've got more surprises in store. We have platform login credits, which we're just announcing and just launching. This is a great way for you to allow users who don't have a seat license to the platform, but maybe you're building an app that you want them to use. You want to be able to get started in a consumptive way. You can now use platform login credits to be able to do that and. Okay, there it goes. It's loading the animation. I want to get you all started building agents. Not just sales agents or service agents. I want you to build legal agents, procurement agents, finance agents, football agents, playbook agents. Whatever the problem is that you're solving at your organization, we are going to enable you to do it with agent fours and with the Agent Force Partner Network. The Agent Force Partner Network can get you started fast with our partners who have already built Copilot, sorry, Agent Actions and agents on the network. These are partners like Google and IBM, AWS Workday, who can get you started fast with agents. They're all available on AppExchange, so you can check that out. And I know we've gone through a lot here, so if you forgot any of this or you want to see these demos later. If you take out your phone, I've got our takeaway guide here. It has all the videos, all the takeaways, all the actions. So you can bring it home with you and leave your phones out because I have some recommendations for some fantastic things to do while you're here. Tomorrow at 9am we have the admin key. Minote.

Yes.

Yeah, I knew there were some admins over here. At 2:30pm we have the developer keynote. You're definitely going to want to see that. I showed you Agent Forest for Developers. We're going to see it in action at the developer keynote. You should also check out the Agent Forest Launchpad where you can build your own agent in Trailblazer Forest. We have a number of booths on security and privacy data compliance protection. So if you want to get hands on and and learn more about these capabilities, you can do it in Trailblazer Forest. And last, check out AppExchange Landing where you can learn about the App Exchange ecosystem. And with that, I want to say thank you. Thank you for joining us today and I hope you all have a wonderful dreamforce. Thank you.