Salesforce, Inc. (CRM) Earnings Call Transcript & Summary

Everyone, welcome to today's session, transform Agentforce's potential by surfacing hidden unstructured data, and thank you all so much for joining us today. My name is Ariana, and I'm on the marketing team here at Salesforce. And before we begin, I'd like to cover a few quick notes with you about our webinar platform. Today's webinar will be available on demand after we wrap up, and it will be accessible through the URL that you are on right now as well as being e-mailed out to you tomorrow. Please note the slides will advance automatically throughout our presentation today, and you can enlarge the slides or any other widget on your screen by dragging the bottom right-hand corner. Should you need technical assistance, click on the Help widget located on the bottom left corner of your console. We've also added some additional resources, which are available through the resources window to the right of the slides. There, you can find some additional related content. [Operator Instructions] We'll do our best to answer as many questions as we can at the end of the presentation. And with that out of the way, I am turning things over to Parth to get us started.

Thank you, Ariana. Hi, everyone. My name is Parth Shah, and I'm part of the Data Cloud team as a product marketer. I have an amazing and really knowledgeable co-presenter with me, Vandana. Vandana, do you want to quickly introduce yourself?

Thank you, Parth. I'm Vandana Nayak. I work with our customers on their data and AI strategy based in Dallas. Looking forward to this webinar.

And together, we're here to show you how you can transform Agentforce potential by surfacing unstructured data. Now before we get started, of course, just like with any other Salesforce presentation, thank you for being our customer, our partners and members of the community. We wouldn't be here without you helping us get to where we are today. So your success is a priority for us and your commitment of being a partner, driving success together in the base. And of course, forward-looking statements, please make sure to make all of your purchasing decisions based on the products that are currently available on the market, especially since we're going to be talking a bit about the road map in this presentation. Okay. So what exactly are we going to talk about today when it comes to unstructured data and Agentforce? A couple of different things. One, I'll walk you through what exactly is unstructured data? What is the value behind it? What does it really unlock? And then how does Data Cloud unlock it? And then once you unlock it, how do you bring that context from that data to Agentforce to make sure that Agentforce is more intelligent and accurate? We'll also show you a real-life service demo that reduces the average handle time when we use Agentforce. And then we'll talk about what's next on the road map, making it a lot easier for you to bring that unstructured data into Salesforce. And lastly, we'll wrap it up with how do you actually get started today using Agentforce and Data Cloud. And of course, as Ariana mentioned, we'll leave some time for questions. [Operator Instructions] Overall, we hope that today, you'll walk away with how to transform Agentforce potential with unstructured data. Okay. So let's get started. Over the year, I mean, we've heard it multiple times over the past year that future leans AI. And here at Salesforce, we believe that the future is AI and humans working together to drive success, which is why we introduced the Agentforce platform. Now whether it's humans gleaning insights to create those personalized experiences or Agentforce becoming the first line of response and customer support, each and any of those experiences are built on reliable, relevant and trusted data. We know that. Data foundation is what brings people and AI together. But did you know that data actually comes in different forms? Structured and unstructured. So structured data is what we've put to work for years, elevating customer experiences to new heights. This data is your customer information like name, address, phone number, e-mails, stuff that's stored in CRM, for example. It can also be financial data such as revenue, expenses, transaction data or data that exists in accounting systems. It can also be SKU numbers, prices, descriptions, data that exists in inventory management, sales orders, quantities sold, again, data that lives in ERP. In other words, all of this data is usually organized in tables. And because it's organized in tables, it makes it really easy for us to search, filter and use. It's very valuable. But we've only tapped into a fraction of the data that the customers have given us. And what I mean by that is, over 90% of the customer data is actually hidden in unstructured content, such as phone calls, chats, images, social media comments, PDF, case notes and on and on. This data is called unstructured data, and it holds the key for us to unlocking customer perceptions, opinions, tone, sentiment and a lot more. This is the data that we really need to deepen our understanding of the customer and for AI to truly craft those personalized and tailored experiences anywhere, anytime. Now the key here is this data has existed, but historically, it's been really hard to unlock and process. And it's because unstructured data comes in many forms and formats such as video, images, social media posts and a lot more, making it really difficult to standardize. And extracting consistent insights, let's say, from an image or from a chat, is going to be very, very, very different and challenging depending on the type of data that you have. And once you do extract the data, it can be costly to organize, store and centralize for easy access, again, adding to its complexity. And finally, when you have all of that data, it needs to be refreshed, it needs to be active and, most importantly, it needs to be connected to CRM to activate. This makes unstructured data very hard to use until now. Introducing Data Cloud, I think many of you have probably seen this slide. Data Cloud is the heart of Salesforce platform. It surfaces data to both humans and agents and everywhere in between from Salesforce Customer 360 applications, Agentforce, Flow and Analytics. It really creates a foundation for personalized customer experiences and real-time analytics, triggering data-driven actions and workflows and safely driving AI across all of the Salesforce apps. And unlike many of the other platforms, it has the capabilities built in to process both structured and unstructured data. Now you've probably seen how it processes structured data in a lot of the other presentations that we've done at Salesforce. But today's focus is all about unstructured data. So with Data Cloud, you can bring together and unpack insights from diverse knowledge types like PDF, text, audio, video, knowledge articles and a lot more in a few clicks. And you can also activate this unstructured data in AI, automation, analytics or any Salesforce application using no code or pro-code tools. And finally, you can bring all of that customer context to every AI experience like grounding AI with customer insights, preferences sentiment and a lot more that's hidden in unstructured data. In other words, turn unstructured data into meaningful insights using Data Cloud. Now many of you have also heard that we have recently closed our acquisition on Zoomin. Zoomin is a leading data management provider for unstructured data. And what we're hoping to do is, with Zoomin, we can bring together existing enterprise knowledge that exists virtually in any source of format such as SharePoint, knowledge articles and an confluence of webinars and YouTube and make it available in Data Cloud. This again really confirms and reaffirms our ability to seamlessly power Agentforce with unified knowledge, making every interaction more precise, more accurate and so on. So how does Data Cloud really unlock the power of unstructured data? This is a very busy slide, so I'm going to walk you through it. At a high level, Data Cloud can connect data from external sources, which is all the way on the left, then it can process it and make it available for you to activate anywhere within the Salesforce platform. So starting with the left, we have connectors that bring in structured and unstructured data. Now specifically for unstructured data, Data Cloud takes that data, let's say, if it exists in an image, it figures out the key text or objects that it needs to extract from that image, let's say, for example, a product description, then it converts it into a numerical representation that captures its meaning, then classifies it, organizes that data into different categories, stores it into a vector database and then indexes for fast retrieval. In other words, in simple terms, it really breaks down that image, understands the pieces hidden in that image. It sorts it, stores it and makes it ready for search. And then using no-code retrievers, you can now call or search that image data or that unstructured data in search experiences, analytics, automations or my favorite, Agentforce. And of course, all of this is native and built directly into Data Cloud. So we talked a lot about Data Cloud. I've talked a lot about unstructured data, what the value of it is. Vandana, can you actually show us how Data Cloud and unstructured data works with Agentforce?

Thank you, Parth, for the wonderful overview. Imagine a customer calling your support line and the agent, be it human agent or AI agent, not just answering their questions, but anticipating their next need. That's the magic of adding context with Salesforce Data Cloud. Let's take a closer look at how Data Cloud adds context to Agentforce. So let's start by addressing one of the biggest challenges businesses face today when implementing AI solutions. The idea of training a large language model, or LLM, on your own business data, especially when you have dynamic data. On the surface, it may seem like the most logical approach for creating personalized and intelligent AI systems. But the reality is far more complex. First, training an LLM on your business data is expensive. The cost associated with building, maintaining and scaling this infrastructure are prohibitive for most organizations. And second, it's time consuming. Training these models to understand your specific data can take months, if not longer, slowing down your time to value. And it's not sustainable. Every update, every change to your business data requires retraining or fine-tuning the model, further increasing cost and complexity. The result? A disconnected approach where your data, AI and the channels fail to seamlessly work together. This is the challenge we are solving with Salesforce Data Cloud. Instead of training the model itself, we provide a framework that seamlessly retrieves and augments your business data in real time, making it ready for AI applications and customer interactions without the need for costly and complex retraining. Let's see how this approach bridges the gap between your data, AI and the channels you use to interact with your customers. So here, this slide introduces one of the most powerful concepts in AI today: retrieval augmented generation, or RAG. Let's break it down further. It all starts with the data. In any business, data is everywhere, be it e-mails, call transcripts, CRM systems and beyond. The challenge lies in making this data actionable, especially unstructured data like Parth mentioned below. It could be the customer feedback or support tickets. Then comes AI. The generative AI models are capable of doing a lot of amazing things, but they are only as good as the data they are grounded in. Without the context, their responses can lack relevance or accuracy, which limits their effectiveness. Ultimately, these insights need to reach your customer-facing applications, be it a chatbot, a customer portal or your support agents. So how does RAG solve this? RAG acts as a bridge, bridging the relative context to AI. Basically, it brings that context to AI, which is really, really powerful. With RAG, we are able to work with what's known as the coffers of information. This could be your PDF, documents, spreadsheets, videos, audios. It pulls all the relevant information from your data sources and enriches the AI prompts. This enables AI to provide that response that is not only accurate, but also deeply personalized. This is done via augmented prompt that you see on the screen there, where the input to the model isn't just users' query, but it's [ hydrated ] with relevant insights from your data. For example, if a customer asks about their account status, the AI doesn't just generate a generic response, it tailors the response based on the CRM data, past interactions and current interactions, real-time engagement data and any related updates. In essence, RAG ensures that your AI doesn't work in isolation, but is deeply connected to the heartbeat of your business, which is your data. This makes every interaction more intelligent, relevant and impactful. For example, let's say a customer asks the AI agent, am I eligible for a credit card upgrade? Without RAG, what would happen? The AI might just form an incorrect or probably hallucinated answer as the AI model is not trained on your data. But with RAG, the system retrieves the context, which is your prior e-mails, transaction history of the customer and any notes that may be in your CRM or other systems. It enriches the prompt. So the AI can respond with accurate information that, okay, the customer is eligible for upgrade and how to proceed as well, really detailed information. Let's take another example related to customer support. Imagine a call center agent using a tool powered by RAG. A customer asks, "Why is my bill higher than expected?" The system can quickly retrieve from billing records, usage logs, recent call notes and agent can see an enriched summary that shows that, okay, higher bill was due to increased data usage during their recent trip. They also noticed that they are eligible for a discounted data plan upgrade. Isn't that transformative? Isn't that what you really need for customer service agents and AI agents? So how do we make it all possible? But before that, let's talk about what it takes to build retrieval augmented generation, or RAG, on your own? and how does Salesforce Data Cloud make it effortless? What are some of the do-it-yourself RAG challenges? First, a disconnected stack. You are dealing with multiple external connectors, data stores, databases, data lakes. Each component, vector databases, search, indexing, retrieval requires custom integration and it needs complex code. Developers must write and maintain complex code to handle embeddings, similarity searches, query chain, ranking logic. Even a simple query involves multiple steps and often require specialized knowledge of libraries and the frameworks, what they need to use. As you can imagine, it's time consuming and error prone. Beyond writing code, you are constantly managing infrastructure, troubleshooting, optimizing. This can basically delay your delivery and makes scaling extremely useless. So what does Salesforce Data Cloud approach do? A fully managed no-code solution that simplifies everything. We provide easy-to-use interface. As you see on the right there, with just a few clicks, you can configure search parameters like selecting fields or defining the context to enrich those AI prompts. No coding required, similar to Salesforce platform, you are very familiar with. This makes it accessible for nontechnical teams while saving developers countless hours and we provide seamless integration. Data Cloud integrates directly with Salesforce CRM and other external systems as well with over 200 connectors that we provide out of the box, even with external data sources. You don't need to stitch together all these disparate tools and it's built for scale. Unlike do-it-yourself approaches, Data Cloud RAG is designed to scale with your business, whether you are handling thousands or millions of queries. Let's compare these options for a real-world impact. Imagine you are building a product recommendation system. You have to code vector database, searches, rank the results, write API codes, all while hoping you don't miss any relevant data. With the Data Cloud RAG, you simply define the input like find products with input context and it does the heavy lifting, retrieval, augmenting, sending enhanced responses to your AI model. So on this slide, we'll try to capture the essence of how Salesforce Data Cloud RAG process, adds context to the Agentforce and transforming it into a very powerful and insightful tool for you. So the process begins when the user interacts with the Agentforce or an AI Assistant, let's imagine a scenario where a customer asks Agentforce, "Am I eligible to upgrade my credit card?" On the surface, it may seem like a very straight query, but answering effectively requires AI to retrieve the data from multiple systems, interpret the context of the question, too, and generate a response that feels very natural, accurate and relevant. So the first step of RAG is basically getting the relevant customer data, the data retrieval. Salesforce Data Cloud can fetch data dynamically from structured and unstructured sources. Unstructured sources could be your past e-mails, customer feedback you have received or a recent call history. And then the semantic search, it uses AI to interpret the meaning behind the customer queries rather than just looking at the keywords that we have done historically. For example, when a customer asks, "Can I upgrade my credit card," the system searches all relevant records with contextual understanding. And then we also support hybrid search, which basically combines semantic search with the traditional keyword search for maximum accuracy. The system uses embeddings to represent customer data as vectors, which is 0s and 1s, and then allowing it to compare the semantic meaning of the query with stored information. What I mean by that is, say, there was a previous e-mail that mentioned about card eligibility. The system recognizes it as relevant even if the exact phrase isn't the same as the customer asked. And then once the relevant information is retrieved, the context is added and enriched the AI prompt. And here is how it does that. It does it through structured data grabbing, ensures that the information like CRM records or recent call history or transaction lots are included. And then unstructured data processing is also provided. And the Trust Layer ensures that only accurate, secure and compliant data is used. This protects against hallucinations or inaccurate responses. Data Cloud also applies Zero Copy architecture, which means that data is accessed without duplicating it, ensuring both speed and compliance. The augmentation happens seamlessly in real time. For example, like the augmented prompt includes statements like, based on prior e-mails, the customer is eligible for a card upgrade. This takes time and ensures that the AI response is accurate and actionable. The enriched prompt is then ready to be sent to large language model. The LLM then generates a response based on the detailed context, built-in and custom prompts. Salesforce allows for pre-configured prompts or you can also create custom ones tailored for your specific use cases, which provides flexibility, and the result is output that feels human-like and tailored for specific customer query. And before sending the enriched prompt, the Trust Layer ensures that the sensitive data is masked and as needed, it maintains compliance and security standards. And then moving on, you get the response and the response, as you see, is very accurate, relevant. And we also ensure proper data governance is in place and accuracy checks are in place as well, and we ensure the output aligns with the business standard. Why does it all matter? Speed. First, the entire process, retrieving the data, augmenting the prompt and then generating the response all happens in milliseconds, allowing real-time customer interactions, very crucial; accuracy, by grounding the responses in real time, the AI avoids hallucinations, irrelevant or incorrect answers, and scales them. This is another key aspect as well. Whether you are handling 10 or 10 million queries, this approach scales effortlessly. In summary, Salesforce Data Cloud RAG turns AI from generic agents into a trusted adviser, delivering accurate, conceptual and impactful response every time. This is how we bridge the gap between data, AI and actionable insights that I mentioned earlier. So we spoke about how RAG works. In this slide, we will show what are the different ways to ground your AI prompts. It doesn't matter if the data is in CRM or external system or unstructured document. Agentforce can access it in multiple ways, ensuring the comprehensive, accurate response every time. So breaking down the method. First one is CRM merge field, which is dynamically referencing CRM records, replacing the actual data at run time. And when customers ask about their order status, Agentforce can pull the latest order details from the CRM using merge field, ensuring, again, making sure that the response is always up to date, and this ensures you are leveraging the freshest data from your systems. And then grounded flows. This incorporates complex logic by dynamically pulling in the data through Salesforce flows that you're familiar with. For example, for a customer eligibility query, a flow can run logic to check transaction history, account type and even customer business rules before returning the response. This makes it possible to integrate decision making logic into data retrieval process. Again, no code needed here as well. And moving on the semantic search, uses AI to retrieve structured and unstructured data from vector stores that we spoke about earlier. For a customer query like, what is the latest update on my case, semantic search scans case records, e-mails, call transcript to retrieve that relevant context. Again, this ensures that even complex unstructured data is included in the conversation. That way, it's looking into that 90% of data that Parth mentioned. And the next method is about integration. This makes external call-outs to fetch data from third-party systems like an ERP transportation system or a shipping system very easy. If a customer asks for their loyalty points balance, Agentforce can make a real-time call-out to the loyalty system. Or if a customer asks for shipping status, it can make the call-out placed to the system like the UPS or FedEx to get that status. This allows you to unify your customer interactions even when the data resides outside Salesforce. And then the last one is about Data Graph. This is tapping into real-time engagement data from Data Cloud, including the behavior insights and interactions, such as what items a person is looking at, what plans they're looking at on their website, or are they looking to buy something, adding something to the cart. Basically, all looking at the recently browsed history, Agentforce can use all this data to recommend similar or complementary items. This turns raw engagement data into actionable insights, helping you improve your business. Together, these methods ensure that Agentforce is not limited to a single data type or data store. It brings together a rich blend of structured, unstructured, real-time and external data, all seamlessly integrated into the conversation. So with this retrieval methods, Data Cloud ensures that every conversation is grounded, up to date, and Salesforce transforms AI from a reactive tool into a proactive assistant and truly [ intelligent ]. So now we have understanding of how Data Cloud covers agent intelligence through RAG. Let's see it in action with a quick demo. And in this demo, we'll see how service teams can augment their workforce with an AI service agent, which we'll provide out of the box, that can tap into unstructured data like knowledge articles to handle customer questions. Let's dive in. So meet Sam. She's a member of the support team at Cumulus Bank. Sam has just opened up her computer and has received a question from a key contact at one of her accounts. That's Omega Inc. She has reached out asking what is the minimum size for loan equipment financing options. So without Data Cloud, without RAG, what happens? So Sam is still somewhat new to Cumulus Bank and has not had a lot of direct experience with financing for equipment. Typically, we all know, like she would ask a colleague or she would look through, search through all the different documents, knowledge sources. Some customers we talk to have over hundreds of knowledge, different knowledge sources, where they have to look up. Or if they can't find it there, ask another teammate or increases the average handle time. This whole process could take at least more than an hour to help answer just one question. So with the Data Cloud and RAG, Sam can take this entirely different approach that you see there at the bottom. Basically, she can ask the AI service agent for help and quickly get answered all in real time. So let's open up Salesforce application and see how it really works. So what you are seeing here is what Sam sees when she logs into the application. Using the chat here, she can ask simple question to the agent, as you see again, she's doing all this in her regular application. She doesn't have to go to another application or change screen. She's looking at Omega Inc. And here, she can just ask the same question a customer asked her, do we offer equipment financing? And what is the minimum loan size? And resend that to agent. And what agent does is, after receiving the question, within a couple of seconds, it comes back with the details. As you see here, yes, Cumulus Bank offers equipment financing. The minimum loan size for equipment is $25,000. As you see, it's a very detailed answer she's able to get without having to dig through all the different documents or without going to go ahead and ask her colleagues or other team members. And as we see here, this is one screen where she can see everything, including the account transactions, how Omega Inc. is doing and also recommendations as you are seeing here. Let's see how it all works. So here, I have opened up Agent Builder. So Agent Builder, as you see, it has topics. It has different topics and each topic has instructions as well as actions. So these topics help guide the agent to answer specific questions or take specific actions. And at the topic level here, you can -- for product questions, you can see the different instructions we have provided, basically before retrieving any information from a document, ask the user for specific keywords, sections or topics they are looking for, all tailored to make the responses very accurate, that way less room for hallucination. And even in the topic, when you look at the actions here, you see 2 actions. One is around product feature brochure and then financial transaction summary. And the output that is generated is this is -- what you see is the input, and here is the output with all the details that is generated by the agent. As you see, everything is done in one screen, easy to do all through setups using clicks and not code. Double clicking on the prompts here. What you see is the prompt. And in this prompt, we have the retriever that is defined. Again, even for the prompt, you can pick different model types. You can use the standard models or you can also use custom models, which is relevant, especially if you already have some models implemented. And we also provide the flexibility to select a model, different model for this particular prompt. And you have also options for different languages as well. And in fact here, you have preview languages available as well, and you can test this. As you see, very easy to set up, all done through clicks. And here, you have ways where you can define different resources. It could be using flows that I mentioned earlier. It could be Apex or it could be Einstein search that we have referenced here for financial brochures, retriever or it could be based on current user or it could be free text. So now we are looking at Data Cloud. How does the Data Cloud enable that? These are the different data sources. These are the unstructured Data Lake Objects that Cumulus Bank has created by ingesting all of the knowledge articles, which are originally existed as PDFs. And using Einstein Data Libraries or block storage like S3, the documents were ingested and 3 data streams were created automatically. The directory table and then the index and then the chunk -- vector embeddings. They are all created. Let's quickly take a look at what is going on, how is the vector database doing it? I'm here. This is the search type. We have defined it as hybrid search, which is both Symantec as well as keyword. And as you see here, chunking strategy, which is HTML, text, and PDF or it could be log, and we are using large embedding models here, and we have defined data sources. All this is done in Data Cloud all through clicks again. And we provide all this out of the box where it can easily set up, you can easily configure these indexes -- search indexes, retrievers without having to do heavy coding. And then where do you define the retrievers. This is where Einstein Studio comes in. Again, you can define the data model objects. This is from Data Cloud where the data -- how the data is modeled and then the search index that we saw earlier. And again, you have all details around what are the written fields, you can easily set this up here. And very important aspect. The other one is around governments and security. This is key and Data Cloud and Agentforce both need this, and we provide this. Data Cloud enhances governance, security, safe data sharing and Agentforce solutions by automating it. Here, as you can see, there's classifications that are on CCPA, GDPR, HIPAA and you can define all this out-of-the-box different classifications, policies, taxonomies, everything based on your data, you can tag them. You can -- there is tagging manager. Also, we have AI supported auto-tagging functionality, makes it all very easy from a data governance perspective. All this we are doing in Data Cloud as we see on the data streams, all your integration sources and then data lake objects where the data gets stored, how you model the data using data model, all this, again, you set up your identity resolution, data action, data governance, all of this set up -- provided out of the box, you can easily set it up. So let's take another example. So here, we ask another question, very similar to earlier question, okay, do we offer financing for custom equipment, the construction equipment? And again, it goes, gets the details, very easily gets the information that's very relevant with the right context that Sam can now use with -- to help the customer. With that -- so going back here, now that we have seen the demo in action, let's take a step back and explore how Salesforce Data Cloud and Agentforce power AI across the enterprise, transforming the sales, service, marketing teams operate. This just isn't about automating the task. It's about driving intelligence, personalization and efficiency at scale. So this is a sales use case, identified upsell and cross-sell opportunities for using sales agent. How does it work? Sales agent can ask Agentforce questions like, okay, can you identify my best upsell opportunity? Data Cloud retrieves customer purchase history, engagement data, preferences to surface those actionable recommendations. Let's take an example. Let's say a salesperson is targeting customer who recently upgraded service. Agentforce might suggest scheduling meeting with them to discuss complementary offerings like premium add-ons, increasing the likelihood of a successful upsell. Impact is the same, our sales teams to have hyper relevant conversations that drive revenue. Next example we have is for service agents. How does this work? Service agents use Agentforce to ask for SLA details or case-specific insights. Data Cloud retrieves the relevant data from CRM and service logs in real time. Example here we have is, if a customer asks about SLA coverage, Agentforce can quickly pull the contractual details and provide precise answer, avoiding delays and increasing customer satisfaction. And the next one is about campaign agent, personalizing every campaign in marketing. How does it work? Marketers can ask Agentforce to segment audiences, craft campaigns or optimize messaging, Data Cloud uses real time engagement and historical data to deliver precise targeting. Example is imagine a marketing agent ask Agentforce to create a campaign for high-value customers. They can use natural language processing to generate this as audience segment. And again, this supports high level of personalization, which can drive higher engagement rate and ultimately boost your campaign ROI. What ties it all together? Data Cloud's ability to unify data and provide that single source of truth, whether it's sales, service or marketing, Agentforce is constantly retrieving, augmenting and generating insights that drives these intelligent actions. This helps break down silos across the organization and ensures that every customer interaction is informed, personalized and impactful. With that, let me turn it back to Parth to cover the road map and how we can get started on this.

Thank you, Vandana. Always love to see a live demo. I don't know if most of you caught it, but the first time when she talked to Einstein, the response that he got also had citations. That's one of the other advantages of using RAG, exactly where the data is coming from, it's not just hallucinations. Before we go on to the next two things, I want to make sure that we leave time for questions. So if you haven't got a chance, please ask any questions in the Q&A box. So what's next? Walking through what's coming up to make it really easy for you to unlock the unstructured data. Really, the road map slide breaks down into 3 different categories. One, we're making it a lot easier for you to bring in more types of data and more formats. You may have heard that we recently announced MuleSoft Direct for Data Cloud, which brings in the ability to bring unstructured data from Google Drive, Confluence, SharePoint, Sitemap. We also added in audio video support into Data Cloud. The second category really is how do we make it easy for you to activate all of that data, right? And so next year, we're going to add -- or this year, we also added no-code search retrievers to help you bring it into flows as well as prompt builders. And then we're also going to add in more activations around sales Einstein conversation insights powered by Data Cloud. And the last category is really about accuracy. Because when you do ask that prompt to Agentforce, behind the scenes, you want to make sure that it's retrieving the right data. And then when it does get to the results, it is pretty accurate. And so for that, we're adding in better search and RAG accuracy capabilities coming soon. And of course, more native support for Data Cloud on some of the features that Vandana just talked to at a high level. Now we realize that structured data is just one part of the equation of the data that you bring in. And so we're also making improvements to our connectivity in general. We're adding in more connectors through a Zero Copy Partner Network. We have over 200 new Data Cloud connectors available, and we're building out our partnerships with the major data lake and data warehouses that you may be familiar with. So whether you're looking to ingest data directly from Salesforce or unstructured data or an external applications, we have connectivity connectors available for you to bring that data into Data Cloud. And last but not least, this is quite important. Vandana had mentioned it towards the end about data security and governance, right? We want to make sure that, of course, Agentforce, all the data doesn't leak externally, but that you also stay compliant with all the data that you bring into and store within Data Cloud. And so there's a lot of new features coming out. The couple that I want to highlight here is what Vandana has mentioned or served before, AI-based tagging and classification. So when you do bring that data in, you can automatically classify and tag it into the Metadata Framework with AI. This is already in beta. You can also start making policies around the data that you have. So that way, it's not just users, but also the access that AI has or different departments have. So we're introducing policy-based governance, which is also beta this month. And then a couple of other features, you can now bring your own managed encryption key to encrypt all your data. And then you can also privately share -- privately and securely share all of your data with private and public cloud networks with private connect for data home, which is GA now. A lot of robust features coming soon to enhance data security and governance. Okay. So I'm going to take us home. I promise we'll leave some questions -- time for questions at the end. So last couple of slides. How do we get started today and what have we learnt today? As a quick recap, what are we here today, Data Cloud is really the heartbeat of the Salesforce platform, whether it's structured or unstructured data. No matter where it lives, you can bring that data into Data Cloud and turn it into native objects and fields. And because it's deeply integrated into the Salesforce platform, where data is now fully harmonized with the broader Salesforce metadata framework. And you can also use the other Salesforce platform tools like Flow and Lightning so that you, along with Agentforce, can easily activate your data. So you can power those insights, automation analytics or even any of the personalized experiences that you want to build out. In other words, this is really how you can transform Agentforce potential with Data Cloud. So how do you get started? Our recommendation is to start small. Start with those simple use cases going left to right on the slide, start with a simple QA agent that searches through documents and knowledge articles. This is available out of the box is how you build out those turnkey Agentforce. And then once you have familiarity with it, then you can start an app automation to it and then you can start building in those proactive agents. So that way, there's any sort of a trigger, they can initiate a conversation with the customer. And then, of course, last but not least, all which to the right is the fully custom agent that you can build on whatever use case that you want. But the key here is with Data Cloud and Agentforce, start simple. Start simple with simple use cases and build up as you get more familiar with our stack. So that being said, we covered a lot today. Hopefully, this gives you a good preview of how to transform AgentForce with unstructured data and data cloud. So how do you get started today, first and foremost, join our Datablazer community. This is where we're going to be sharing updates. This is where you can also talk to other like-minded folks that want to use Data Cloud. So join the community, we have a trail head that helps you get started with Data Cloud. You can also check out some of the use cases that customers have used Data Cloud for. And if you want to get started today, we have a starter bundle available for you. No need to take pictures or write-down these links. They're all going to be available on this platform as well as e-mail out to. And so with that, I thank you for being with us on this journey together, and of course, taking the time to listen about Data Cloud and RAG. So with that, let us open it up to questions.

Just a second while I pull up the questions. Okay. There's a couple of questions coming in. The first one I want to address is, will a recording of this webinar will be available? Yes, of course. You will get an e-mail with recording of this webinar in a few hours, I think by tomorrow at the latest with all of the resource links as well as the slides. There is a question about how is Data Cloud's ability to process unstructured data different than other platforms? Vandana, do you want to take this one? The question is, how is our ability to process unstructured data different than others? I can also take a stab at it, if you want.

Yes, go ahead, Parth.

Yes. So I prefer to add anything I've missed. I think, it's everything that Vandana had mentioned. I think the first thing is you're getting a fully managed tech stack designed to really unlock process and most importantly, activate that unstructured data seamlessly, right? We have built-in tools like chunking to break up the data into smaller pieces, metadata extraction, vectorization of the storage. You can index it for search, you have hybrid search capabilities available and then no-code retrievers. All of this is, again, fully managed and kind of built in, so you can bring all that data right in the flow of work. Vandana, there's also a question about -- oh, this is a good one. I already have a vector database. Can I use it? Is there still value in data cloud vector database? Do you want to take this one?

Yes. So currently, Salesforce does not have native support for external vector databases, including built-in user interface for that seamless integration. While customers can use Apex to perform callouts to external databases, this approach does require additional customization. However, it's important to keep in mind that full RAG system involves more than vector database. As we showed, Salesforce offers that extensive value beyond simply providing the vector database. It integrates various tools such as hybrid search, data retrievers, AI capabilities through platform like Data Cloud. This capability ensure that Salesforce AI goes beyond just the vector database, enabling more of that configurable, easy to built, quick time-to-market and supporting more of that context.

Thank you for that. There's a question about, was all the stuff in the demo today GA?

Yes, everything we showed in the demo is either beta or GA. There are things around policies. I showed only the things that is GA, but as Parth shared, there are more features coming from government's perspective soon in the 254 release, which is a [ spread ] release.

There's also a question about Sandboxes. Is unstructured data supported in Data Cloud sandboxes?

Yes. It is supported as of early November 2024. So we do support. The merge back retrievers from sandbox to higher augs is planned for early December.

Okay. I think this is a good one. The DMOs on data cloud or structured data, how do you map unstructured data to the DMOs?

So we do provide the functionality where, as we showed, there is like new UDMO, which is for unstructured DMO, Data Cloud provides that through UI, you can map the data.