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

Hello, everyone. Welcome to today's session, 3-step strategy to generative AI and trusted data in customer service. 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. Please note the slides will advance automatically throughout the presentation. And you can enlarge them by clicking the enlarge size button located in the top right-hand corner of your slide widget. 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 slide. 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 our presentation. And with that, I am turning things over to Allegra to get us started.

Great. Thank you. And hi, everyone. Thank you for joining this webinar on a 3-step strategy to generative AI and trusted data in customer service. Now I recognize there's a lot of buzz around this topic. And our goal today is to leave you with practical advice through a playbook on how you could be thinking about incorporating generative AI into your strategy to change the way you connect with your customers. Now before we get started, I want to first give a quick reminder that Salesforce is a publicly traded company. And you should always base your purchasing decisions based on products and services that are currently available. As a quick intro. My name is Allegra, and I'm on the product marketing team here at Salesforce. I've spent my career in the data and AI space, so really thrilled to be here today talking about 2 of my favorite topics: data and AI. I am joined today by 2 of my colleagues. Noopur is a product leader here at Salesforce with deep experience on all things data, AI and service. And Tim is a distinguished solution engineer who knows our product probably better than anyone. We are so lucky to have both of these thought leaders with us today to share more around our product direction. Now we have a jam-packed agenda for today. I'm going to start by introducing the 3-step playbook when it comes to AI and data in service. I'm then going to hand it over to Noopur to share our latest product innovation. And Tim is going to bring this to life with a demo. We then have what I'm going to call a real talk panel discussion with Noopur and Tim, where we're going to get into some of those hard-hitting questions. And finally, we're going to wrap it up with some closing thoughts before we close out the webinar. [Operator Instructions] There are a lot of you on the line today, so we'll do our best to answer all your questions. And we will follow up, after the webinar, to any questions, so -- that we do not get to, so please do continue to ask them. Now to start, I wanted to start this webinar a little bit differently than every other webinar about AI right now because, when it comes to AI, we really should be starting the conversation around data. And if you take one thing away from today, it's that data completely underpins your AI strategy. The key to great AI is great data. And over the past several months, we keep hearing about AI, right? Generative AI just seems to be dominating the tech headlines. It is arguably one of the coolest, most revolutionary technologies of our time. And I know generative AI might still seem like the buzz of the hour, but there's widespread consensus that we're actually just at the start of this AI revolution. Generative AI really unleashes a magnitude of far-reaching opportunities to disrupt every industry. And I might be a little biased, but we think that there's the most opportunity in service. And it turns out that a lot of you listening agree, with more than 60% of you service leaders believing that this technology has the power to better serve our customers, which is really exciting, but as exciting as this technology is and as much as CEOs clearly want to embrace it, what we're finding is that there's a trust gap with this powerful technology, right? 59% of customers do not trust companies with their data. And that's a challenge because data is the foundation of AI, so what we're seeing is that AI inherently introduces risks and concerns as what to -- will happen to all that data, right? You've likely heard of hallucinations, toxicity, bias, which is why we really wanted to focus this webinar today on the whole story of AI, which has to include data, has to do -- has to include trust. And that's why we wanted to give you all a 3-step playbook. Now this is the 3-step strategy that we recommend to really help set you up for success so you can create GPT-powered experiences that you and your customer love. First, you need to build your source of truth. Remember that we keep talking about the importance of data. This is step one. Now we recognize that this step is hard because getting your data house in order [ can take time ], but the good news is that you can start small. We actually just recommend starting with that updated knowledge base. You can then longer term start thinking about harmonizing and connecting your data. Then you need to set up your digital channels, right, because your customers are going to want to interact with GPT using their preferred channels. That could be messaging to self-service, so you need those channels set up to meet customers where they are. And then from there you can launch your AI journey to really set the foundation for GPT. So this means creating AI models and testing new use cases for your AI. And actually, in a recent survey with a lot of service leaders like yourself, you're telling us that you really want to use AI to create and personalize service communications as well as better automate customer service [ comms ]. So data, channels, AI. Now you're really ready to create a GPT-powered experience, but I really want to highlight that word trust, as that was the biggest concern we just discussed. So how can you actually do this all safely? And at Salesforce, we really believe in having a human in the loop giving feedback, reviewing, improving that output of GPT-powered content before it gets deployed. This is how you're going to create that scalable experience that every customer can trust. And now Noopur is going to go deeper into this 3-step playbook through a product lens, so let's hand it over to our product leader.

Thanks, Allegra. All right, now you've seen the 3 steps. Let's deep dive into the innovations that are going to enable you to realize this strategy. We live in exciting times, right? Generative AI is going to fundamentally change the way we do business. However, as Allegra just pointed out, right, these features are as good as the data they are rooted in, so say hello to Data Cloud. You probably heard of customer data platform or CDP that originally was used to power better marketing engagement experiences, but now with Data Cloud, your real-time single source of truth, you can focus on personalizing engagement across [ the customer ] journey, of course including customer service with that unified profile for every customer. Let's see how this magic comes to fruition. We live in a world where the inputs are many and disparate. And the outcomes can surface across a multitude of channels based on the customer's channel of choice. As shown on the left of the slide, Data Cloud has built-in connectors that bring in data from a variety of different sources, third parties, mobile, web, APIs, even legacy data from MuleSoft and historical data from proprietary data lakes in real time. I know you're thinking, "So what? I can build connectors to store data wherever I want," but hear me out. Data Cloud is -- not only brings all that data together, but it also harmonizes all of this data, billions of profile into unified profiles that are your real-time customer truth, into a single customer graph. This then becomes the customer truth that evolves as your data evolves. Now this real-time data can then be activated across your Customer 360, across all of your Salesforce apps, Sales Cloud, Marketing Cloud, Commerce Cloud and, of course, service. And remember that's step two. This data can now be used for proactive personalized experiences across all digital channels. Now when a customer reaches out, say, over a chat session, an agent can see the customer's unified profile that spans across the customer's relationship with the entire brand versus just that service relationship. And companies can also build out calculated insights like inclination to attrit, propensity to buy, so the agent is even more empowered to provide a personalized interaction. Now tie this real-time single source of truth with generative AI, boom. And as Allegra mentioned before, I may be biased, but I believe that service is the most powerful professional use case for generative AI. And so we at Salesforce are absolutely excited to introduce to you Service GPT, where they focus on Salesforce's key values, trust and innovation. So how does this empower service teams? Well, of course, you get to do faster resolutions with the help of AI, but with Service GPT, there's always a human in loop. And so the human is working hand-in-hand, verifying, editing and modifying those responses back to Einstein and feeding it back into the model. And what this helps with service teams is they embrace the AI knowing that it's there to help them work faster, not necessarily take their jobs. You can also empower your teams by boosting CSAT by personalizing every customer experience in real time. This is where the Salesforce Data Cloud comes in, right? It brings in all of your data sources together [ through that ] unique identity that spans across your entire relationship. And you can also unlock great amounts of efficiency and cost savings by bringing all these customer touch points together like self-service, contact center, field center -- field service all onto one platform. And because Service GPT is part of Salesforce Customer 360, service is connected to sales, marketing, commerce, IT, all of your customer data, allowing you to create seamless experiences across that customer journey, all of that powered by AI. And as you can see on this slide, we have several exciting features in pilot; and they're happening now, and many more to come within the next few months. Yes, you heard that right, not the next few years, not even the next few releases. We're talking about the next few months. That's the pace we're all moving at. Now let's deep dive into one of these features. Let's take a look at service reply. When a customer is chatting with a service agent, Service GPT will take that message and then use a prompt template to create a service response to that customer, but now if the only thing we were using it to do was just the customer message, the response wouldn't be any better than ChatGPT can do right now. There will be no personalization. And this is why we've taken it a step further in the process where we ground the prompt with the CRM data. And then we add a bit of flair to it with the company's language or procedures while masking personal information, so then when you do get to step three and you send that prompt to the large language model with all that relevant data, we now receive accurate service reply that is personalized. And that can also be used by the agent as is or can be modified. And that's the important part, right? The human is still in the loop. Pretty awesome, right, but let's dig into that question we've had at the back of our mind about trust. And so I want to focus on that middle step we talked about, grounding the data, which is built into the Einstein GPT trust layer; and that is the key differentiator. Now at a high level, the Einstein GPT trust layer is basically bridging the gap between your data and the large language model. It allows teams to use generative AI capabilities while still maintaining control over data, security and privacy. We use a technique called dynamic grounding to use customer context from your CRM to enhance the prompts given to the AI models. And so with the trust layer, we can remove personally identifiable information or PII from the prompts using masking tools so sensitive data is not processed by the AI models. The generated results from the models are also checked, harmful or biased content, and are obviously audited for compliance. What is very, very important to note is that your data is never stored outside of Salesforce. Once an external model processes your prompt, both the prompt and the generated output are forgotten. They are not stored or kept for any reason, whether it's monitoring or training. Your data belongs to you and that's the end of it. Now finally let's take a little bit deeper look into how Service GPT as a whole works. The first step is the one-standard gateway for calling large language models. Now that single gateway gives us a standard set of integrations to integrate with these models whether they're from a partner or a model that we've developed at Salesforce, so the GPT trust layer that I just talked to you about, that is a signal central point of governance and middleware services for tasks like prompt engineering. The next component to highlight here is the Customer 360. Now with Data Cloud and the Customer 360, we have access to relevant data to build prompts that are tailored to your use case. Our ability to ground the prompts in 360 data is critical to protecting trust, to mitigating the risk of model hallucination and delivering relevant personalized output for users. And in the end, this stack allows us to do things like generate service replies like the example we just looked at, write work summaries based on conversation data, draft knowledge articles or surface relevant and accurate answers. Well, enough of me talking about it already. Let's see this in action. And so for that, over to the [ legendary ] Tim Casey for our demo.

All right, thank you very much, Noopur. So hey, everybody. Tim Casey here with the Service Cloud solution engineering team. And today, I'm going to walk us through a quick demo; and show you all how customers can unify their data, create that unified view of their customers and ultimately deliver the service their customers love. So I'm going to tackle this in the order Allegra showed with us, so let's go ahead and start with step one. And I'm going to talk about building out that source of truth. So it's a true story. I just got back from a trip and I actually had a poor service experience. And when we contacted the hospitality company, it wasn't great, right? In my profile, it probably looked a lot like it's blank. It's empty. It's lacking context for personalized service, right? And that's a challenge. It's a challenge for many of the organizations I work with, right? Their data is incomplete. And it's because it's missing data from many other systems, right, whether that's engagement data, streaming data, mobile apps, et cetera. And that's where Data Cloud can help, right? The first step, like Noopur walked us through, is we connect all of this customer data that's scattered across all these other systems. By using these prebuilt connectors, we can connect to any system, any format and in any volume. So something I see all the time is where customers have data scattered across multiple Salesforce [ boards ]. We can use these prebuilt bundles to quickly harmonize that data. And if it needs to be transformed as it's coming in, we can use data prep. It's a point-and-click tool. Take this data that we're ingesting. Filter it. Combine it with data from other systems and then transform it into a brand-new view. And that's what we see here where we finally connected over 130 data streams. Now if we take a look at one example. And this is one of the many data streams I've connected. And it contains streaming data from the website, so now how can I take this data and turn it into a unified view? Thankfully, Data Cloud is going to help us map these incoming streams into one or more data models. Now these are special because they're -- be able to provide a shared definition of our data that we can use to resolve differences between these systems. The example we see here is called an individual. So this represents a person. And that's what's missing from my travel example earlier, right? So even though organizations may have dozens of systems that define individuals in slightly different ways, we map all of those systems out into this shared data model. And together, that's going to form a graph that helps us understand the shape of our data, which we can then use to create this unified view of our customers. The last step really is as we're ingesting and mapping all this data. A lot of times, customers end up with duplicates or conflicting records, and thankfully, Data Cloud can help with that too. So using identity resolution, I've set up matching and conflict resolution rules that help me turn 64 million records into less than half of that as unified records. And that's how I'm able to turn these unified streams into a unified view of our customers. So what does it actually mean for service customers? I'm going to run us through a couple of examples. And we're going to talk about connecting our digital channels and step two here. So what we see here is an agent logging in to the service console. And as new work items are coming into our contact center, our agent uses a single presence to handle incoming work. [ An omnichannel are ] always going to find the right agent at the time with respect to queue, skill, capacity and availability. So as this call is answered, this is where we see that unified profile come into play, but right away, as an agent, I have a visibility to who this customer is, who they represent to my organization; and can start to deliver personalized and relevant customer service. If I need to drill down deeper, that full C360 profile surface automatically. And now as I have this unified profile, again, I know who this caller is. I have that full picture, like Noopur talked about, across all clouds, across all touch points. And most importantly, I can dive in and start to offer up some automation and intelligence, making sure we hit our SLAs, correctly classifying this interaction based on historical data and again making sure I'm proving that intelligent customer service. As the conversation continues to progress, I can start to leverage things like sentiment or keyword and dynamically surface next best action requests for my agent in real time, so when my customer here says she wants to make an appointment, that recommendation surfaced automatically. And as an agent, I am guided step by step by not only the proper phrasing but what exactly I need to do to get this customer set up with an appointment. And this is just the beginning, so Service GPT is going to introduce even more opportunities to delight your customers with generative AI for CRM. So I'm switching over to an example using chat. I want to talk about how we can further enhance the experience for both agents and customers when we finally launch that AI journey with step three. So what we see here is, as chats are coming in, Service GPT is able to generate responses before I can even think of an answer. It's relevant and it's contextual because Service GPT is using data from all clouds unified with Data Cloud. Now as an agent, I always have the ability to review what's been generated before sharing with my customer. And this is what I mean when I talk about leveraging that "human in the loop" framework by using your own data by running dynamic grounding. Again all this is happening in the background, but that's how we end up with generated content that's going to bring down handle time and ensure your customer's data is protected. Now this one looks pretty good. It's on brand. We go ahead and send that reply. And what we see is, as more questions start to come in, again Service GPT is going to continue to combine external data with your own CRM data, provide a relevant and personalized response. So again, rather than looking through knowledge articles, reading them, extracting the relevant information and writing a response into the chat, Service GPT is able to automate all of these steps, automatically suggests the right response, takes the burden off the support team and not only help deliver that exceptional customer service your customers expect but allows you to continue to scale as your organization continues to grow. Now as the issue is resolved and I'm ready to wrap up, Service GPT is able to automatically generate a natural language summary. I don't need to spend any time documenting the conversation. Generative AI is doing all the work for me. And I can even take that summary and turn it into a knowledge article and share that out with the team, and this is huge, right? So capturing that institutional knowledge in the moment, making sure it's available to the whole team, again continuing to help you all scale as your businesses continue to grow. And that's just it. So when we talk about AI plus data, plus CRM, organizations are empowered to accomplish more and scale the service that customers love. It's all built and maintained via no-code, low-code tools. And most importantly, it's built to make your organizations more efficient, more productive as you all continue to scale and grow as well. So that's it for now. Back to you, Allegra.

Great. Thank you so much, Tim, for that demo. I just really love it because it brings to life the power of the 3-step strategy we just talked about, right, data, channels, AI. Now what I want to do next is just have a real chat with Noopur and Tim. So I feel like, in conversations with so many of you, you want to dig under the hood a bit more and have some real honest questions about like how does this all actually work, so let's start with a question I hear all the time.

There's so much noise about AI and generative AI solutions right now in the market. What makes Salesforce different? And Tim, I feel like this is a good one for you just because you spend so much time with our customers.

Yes, definitely. Thank you, Allegra. I think the first thing I always call out is I think it's really important to point out that Salesforce has been in the AI game for over 10 years now with Einstein, right? We've released over 60 features in the Einstein brand. We're churning out over 1 trillion predictions per week. We're already very much an expert in this space. And our approach to AI, it's really grounded in 4 pillars. So trust, relevance, security and ecosystem; and I'm going to start with trust. We're not only investing here, but we really have a track record of doing so. And Allegra and Noopur both talked about this as we were going through the beginning part of the presentation. We've had this ethical AI practice as part of our [ office of -- ethical we've ] used for the past 6 years. We're -- built out this "human in the loop" framework, again, so customers can verify everything that's being created or generated before it goes out to their customers. And then we train that model over time so it continues to become smarter, more accurate and more relevant to your company data. That leads me right into my next one. We talked a lot about the [ important ] of relevance. So AI is only going to be as good as the data that you use. And Service GPT isn't just going to use external data which can be, of course, unreliable. It can be inaccurate. It uses your own CRM data, carefully vetted, and trusted external sources so that you can be more confident. I think that leads right into where I'm going with, security. So we've built out this very cool proprietary framework that Noopur touched a bit on earlier, where you can use a variety of different LLM models without sacrificing security, right? Well, Service GPT is going to maintain secure data access. It's going to protect your PII data, and it's purpose built for what we're doing with the Customer 360. And that leads right into my last piece. So my final differentiator, I think, is ecosystem. So we have this really amazing research team internally that's been building out generative AI tech for years, but we're also partnering formally in this space to make sure that we're bringing our customers the freedom of choice. And to me, when you look at the whole picture, all of these factors really differentiate Salesforce and have us very well positioned to continue to innovate in this space.

Thanks, Tim. I think what you said in the beginning is actually something I especially want to highlight. I actually think a lot of people don't know that Salesforce has been investing in AI for years. We were the first to launch that intelligent CRM over 10 years ago. And to your point: We're doing billions of predictions a day, so we have and will continue to be a partner in this AI space. And I just really love these pillars and diving deeper there. Okay, another question I'm hearing about is how to make this real, right; and if we have any customers that are actually already using this 3-step strategy already, thinking about data and channels and AI, so Noopur, I'm curious. Any customer examples come to mind for you?

Yes. I mean I always love to talk about our story with Formula 1 that has been an amazing customer, also more so because I'm a huge Formula 1 fan. Max Verstappen is the GOAT. I'm not taking any questions around this statement later, so there is that, but just in terms of how Formula 1 has been an absolute trailblazer in this opportunity of taking the AI, data and service model and building that end-to-end story, right: So let's talk about Formula 1 has 500 million global fans, right, 1.5 billion TV viewers. And they've used this 3-step strategy to really attract new fans and ultimately bring in new revenue, right? They turned to Salesforce because they needed that platform that can look across that customer journey like we talked about earlier, so it's not just about the service experience. It's service, sales, marketing, right, like bringing that thread together. And then with Data Cloud, they've been able to have a complete view of their fans across every channel, all right? And when I say channel, it's whether they're actually at a Formula 1 event or they're visiting the website. Using that harmonized data based on location, content preference, favorite driver, they're able to have those real-time insights across service interactions, right? And by also using automation and AI, so they're layering that on. And they've been able to help their agents be more productive. So in fact, having built this end to end, they now have seen a 86% first contact resolution. And this is an absolutely powerful story of data, AI and service all together.

Yes. I always love the story. And I know you're a true fan here. And it's just such a great proof point with clear metrics. And it goes back to that 3-step strategy that you mentioned and we've been talking about so much around connecting data, engaging on channels, leveraging the full power of AI. Okay, I have one more question for the both of you because we talked so much about trust throughout the presentation. How does the trust layer really work that Noopur mentioned earlier? So Tim, maybe you could break this one down for us a bit further.

Yes. No, happy to, Allegra. Again this is a question I get from customers all of the time, right? So what does it really mean when we do talk about trust and generative AI? And there's a few things I want to call out on this slide, and I'm going to start with context. We see that on the bottom left there. And this is so crucial. And this is exactly what we're talking about when we talk about step one, right, building out the source of truth. This is how we're able to do dynamic grounding using your own private data, combining that with the best of external or internal LLM models and ultimately using that to generate content that's personalized. It's relevant, yes. And most importantly, it's secure, right? If you do have sensitive data, we're able to mask all of that PII with data masking before it even passes through our LLM gateway. It never crosses our Einstein trust layer. And that's not the only stuff we're taking, right? So I don't want to gloss over this. Again I think it's a huge differentiator, but we're using what we call 0 retention architecture with all of our LLM partners. What this means is that, even masked data, any data at all, that prompt is going to immediately disappear. It's never stored outside Salesforce or used for any sort of external model training for any reason, right? From there we're able to run that toxicity detection that Noopur talked about. We're going to ensure these responses align with our own ethical use of AI. We're going to run the prompt hydration. We're going to log an audit trail, capture that feedback right inside Salesforce before ultimately serving up these generations to our C360 apps like Salesforce Service Cloud. And that's really what we mean when we talk about the Einstein GPT trust layer.

Really helpful to bring this to life: trust and better understand that trusted data piece. I know we did a webinar just a few months ago, and I feel like we're really figuring this all out together. It's so nice to see this come to life, and thank you, Tim, for getting into the details more on this. Now I could ask Noopur and Tim a million more questions, but I want to get to your questions on the line. And before doing so, I did want to cover some next steps and resources. So just once again, as a reminder: Please keep submitting questions at any time. We'll definitely try to cover as many as we can and get the answers out to any that we don't have time for today.

So now on to next steps and resources. I want to make sure that we're really leaving you with that applicable playbook. Remember this 3-step strategy, right? This came to life across our product innovation that Noopur walked through, our demo. It's about your data, your channels, your AI journey, all with trust in mind, so we hope you now have a playbook to use to really take advantage of generative AI. Now how can you take advantage of generative AI specifically using Service GPT? Well, listening to all of you, we wanted to take a moment just to share some of the use cases we're working on. So we do plan to continue to release use cases throughout the year, and I'm excited to share the ones that are just coming very soon. Service replies, the example that Noopur walked through in detail, auto generates personalized replies using trusted CRM data across any channel that you saw. Work summaries helps automate what we think of as a very time-consuming task but critically important, which is summarizing those conversations. So Service GPT writes these accurate summaries for you using case data and history. Knowledge articles are how you can help really institutionalize knowledge across the organization. And it's actually one of my favorite use cases because Service GPT will truly just write a draft knowledge article for you based on that closed case data, which helps save your team time and helps onboard agents faster. And I know there were some questions around case deflection: search answers. So as your agents or customers are searching for answers [ to a question ], Service GPT will actually surface a generated answer, grounded in your knowledge base, directly into the search page, which saves everyone time, so definitely I wanted to talk about that use case a bit as well. Now we're also extending this innovation to field service to help empower frontline workers. And so there's 3 features I wanted to call out that really is helping these frontline teams before their visits, during their visits and after. So starting with pre visit is the mobile work briefings so frontline workers have the right customer, asset and service history data for the job at hand. So this is information, right, like how equipment has been maintained and details of previous customer interactions. While on site, we have knowledge search, which provides quick searches for knowledge articles to assist any contractor employee, so really helping improve those first-time fix rates. And then for post visit, we have that post-work summarization which intelligently generate summaries that take into account quantitative asset data and shorten that visit duration, so these service reports include real-time data, future maintenance recommendations and even images taken on the job. As you can see, there are so many great use cases and we are just getting started. And it's truly clear that the use cases for generative AI are endless. As we wrap up the journey together today. There are a number of ways to continue your journey with us in the future. We just launched a new Trailhead, so you can start your training and learning now. We're hearing from a lot of you that this is exactly what you want to be doing, is learning and training, so wanted to make sure we flag that. We highly recommend connecting with your account executive today to learn more around Service GPT. And we also have a lot of events coming up. So we have upcoming readiness webinars that you can attend. If you're in London, you can catch us at London world tour later this week. And if you want to plan ahead for the AI event of the year: Registration for Dreamforce in September in San Francisco is officially open, so you can check our website for all the details there. Now we want to open it up to all of you to ask questions. And I've seen so many great ones come in. So a reminder to please continue to drop those in the chat. And I'm seeing a lot of questions around data, so Noopur and Tim, I want to bring you back.

So one question I've seen a lot is where is data being stored. Is it outside of Salesforce?

I'll go ahead and jump in here, Allegra, because this is one I get pretty much every time. It's the first question I get when I'm talking out to customers. And I want to reiterate and be super clear here that, when I talk about like the Einstein GPT trust later, what that means is data that's being used in your prompts, they're -- it's never stored outside Salesforce, right? And that's what I mean when I talk about 0 retention architecture. Your data, your private data is never stored outside Salesforce for any reason or any amount of time.

Thank you for making that super clear. I saw a lot of questions coming around that. Some other questions have been around public data, and I think this one is interesting. So when public data is used, will it pollute your internal data store?

So I can talk to that a little bit. This is the beauty of the solution. And think about what Tim just said, right? Like we don't store your data outside of Salesforce, but also, public data is only used to generate responses at run time. And so we don't actually pollute your internal data. It goes back to that we ground the output in the context of the CRM data you have, but we don't like store data outside or pollute your data with public data.

Helpful. A lot of questions got answered with that answer. Okay, this is interesting: This is around training AI models. Is customer data used to train AI models? Tim, I think this is one that I know we've talked about. Can you go a little deeper here?

Yes. I'll jump in here. This is probably the second most common question I get, and it's a good follow-up to what Noopur just talked about. And I want to be clear here as well. Like customer data or your own data, it's used to ground prompts, right? It's used to create relevant outputs, but it only ever trains Salesforce-hosted model. And what I mean by that is we have contractual agreements in place with all of our LLM partners that your data is never going to be used to train their models or any sort of public model.

Awesome. Thank you for that. I'm glad that we're getting into some of these FAQs. A lot of you are asking actually a lot of the same questions. There's a bucket I'm kind of seeing around data quality and data cleanliness and I think this is a good question. How important is data quality and data cleanliness? Noopur, I feel like this is maybe a good one for you. Is that cool?

Yes, yes, absolutely. I mean I'm so excited that we are talking so much about data in this webinar that is focused on data, AI and service because really, again going back to where we started and what I said during my presentation, the AI is only as good as the data it's rooted in. So data quality is extremely important, right? And by now many of us have used ChatGPT, personal or professional use cases. And oftentimes, I found those responses to be super generalized or sometimes even just plain wrong, right? And so this is where Sales Cloud's, Data Cloud and Service GPT responses come into play. Just reiterating everything we've said here is that the way this Service GPT has been built is keeping that in mind, that data quality and cleanliness is extremely key and important because this is the foundation on which AI is built. And so we have to have trusted data, clean data and data that we can be convinced is the right data for the right AI outcome. So yes, it's exciting that everyone wants to talk about it, but that is definitely a key tenet when we talk about the trust layer and how we process the data so you can apply AI to it.

Very helpful. All right, we probably have time for just one more question, just wanting to be respectful of everyone's times. Like I said, if we didn't get to your question, we will be following up with you. I love this question because I think it's a tough one: Why would a customer choose Service GPT versus building your own LLM?

Let me go ahead. And I'll jump in on this one too, Allegra. I've definitely heard this one before. I actually heard this from some of our own account teams as well, right? And I've definitely talked to some customers, some ambitious customers, I'll say, who have kicked the tires on like building something custom, right, but it always comes down to like who maintains that, right? There's a lot, I think, of inherent risk when you talk about doing something custom and potentially exposing customer data not only through like the accuracy or like the model hallucination that Noopur talked about, but we've all seen there's been a few negative incidents in the news. Or organizations have just done just that, so I think there's a lot to be said around partnering with a trusted generative AI partner like Salesforce who has been doing this for a long time, already has all of the necessary safeguards and agreements in place that are going to protect your own customer data, right? From there, Service GPT is going to be able to ground AI using your private data, activate it in the flow of work that your support teams are already doing and then make sure we keep all that data safely and securely right inside Salesforce.

Love that answer. And I think that's a great way to close. We're right at that 40-minute mark. Now that's all the time we have for today. As mentioned, if we did not get to your question, we will be in touch to follow up. We really appreciate you spending time with us today and look forward to partnering with all of you on your AI journey. A big thank you to Noopur and Tim for joining me today. Remember great AI starts with great data. So enjoy the rest of your day. Thanks all.