Good morning. Thank you for joining us. I'm really excited for today's session, our Q3 deeper look into our latest product strategy and innovation. As you heard me mention on Wednesday, Salesforce is in the midst of transforming ourselves and helping our customers transform into agentic enterprises. Today you're going to hear about that journey from our leaders. We have a wide swath of talent here from across the organization, focusing really in on data and Agentforce deployment motions. Quickly, first, some of our comments today may contain forward-looking statements that are subject to risks, uncertainties, and assumptions, which could change. Should any of these risks materialize, or should our assumptions prove to be incorrect, actual company results or outcomes could differ materially from these forward-looking statements. A description of these risks, uncertainties, and assumptions, and other factors that could affect our financial results or outcomes is included in our SEC filings, including our most recent report on Forms 10-K, 10-Q, and any other SEC filings. Except as required by law, we do not undertake any responsibility to update these forward-looking statements. With that, I'm going to hand the call over for a brief round of introductions with the group of talent that we have here that I'm very pleased to introduce. First, let me introduce Madhav. All right. Thanks, Mike. Happy to be here. My name is Madhav Thattai. I've been at Salesforce for five years, and I am the COO for our Agentforce Product Organization. Ravee? Hey everyone, good morning. My name is Ravee, and I lead Engineering for Salesforce Data Cloud, EVP of Engineering. I've been in Salesforce for six years. I've been in the data industry for a lot longer than that at Microsoft and other places. Happy to meet you all today. Gavin Patterson, I'm the Chief Digital Officer. The youngin in this crowd, as far as tenure goes, I've been at Salesforce about six months. Immediately prior to that, SiriusX M, and prior to that, Disney, where I was the Chief Technology Officer and led the development of Disney+ amongst other things. With that, I'll hand it over to Andy. Thanks, Joe. My name is Andy White, and I have the privilege of leading our team's efforts of supporting primarily our sales and our customer success and professional services teams as they support our customers. All the technology that we use, we try to be the best example of what Salesforce looks like and drive Salesforce on Salesforce and Customer Zero, which you're going to hear more about today. Anna, one of my partners. Thank you, Andy. Hi, my name is Anna Lee. I'm the COO for Customer Success. I have the privilege of serving our customers every day. In this context, I also get to be a customer. Back to Mike. Great. Just prior to handing it to Madhav here, the context I gave the group and for this audience to understand is we know very much top of mind for many of you out there is what the adoption curve, what the growth curve looks like for Agentforce and Data Cloud, what we're seeing inside customers, some of the challenges, some of the wins that we've had, and what the expectations are moving forward as we help customers kind of through this journey. That's the backdrop by which we set up this call and that we're looking to explore today. Of course, we're going to end the session with Q&A. We hope to have, call it, 25-q 30 minutes for Q&A once the presentations are over. As you're listening to the presentations, if there's questions that come into mind or if there are things you maybe heard on Wednesday relative to the dialogue today that you've got questions about, we'll take those questions once the presentations are over, and we'll try and get through as many as we can. With that, I'm going to hand it over to Madhav. Thank you, Mike. We've been on this journey with Agentforce. It seems remarcable to think that the product only launched nine months ago. We're very, very excited for all the incredible innovation we're going to be bringing to Dreamforce, which I'm going to touch on here in a minute. We really wanted to share with all of you how we think about this business. Agentforce and Data Cloud are consumption businesses where we want to really drive value for customers and ensure they're getting through their implementation cycle, and they are learning this technology, these products along the way. We really center the performance of this business and how we manage things on this funnel. If you look at the funnel, you'll notice we specifically emphasize the funnel in an inverted shape because what matters the most for us is that our customers are being successful. This daily kind of obsession with customer success, including, of course, our remarcable colleagues at Salesforce as Customer Zero, really drives our energy, our activity across our product team, our engineering team, our customer success team, and our sales team. If you look at the bottom of the funnel, that's really where we think about our sales process. How do we manage our pipe? How do we manage our closed deals? As you all know, we now have both your traditional employee-based SKUs, where we've launched now the Agentforce Edition, and our consumption-based SKUs. We started with our conversation SKU, and then we launched our Flex Credit SKU, which we talked about in earnings as well. That's doing really well. That's the sales process. We then get into the critical phase of implementation. Here our customers are really thinking about what do they want to use these agents for? How do they think about the use cases? How do they think about building them? How do they think about testing them and then getting them out to their deployment? That critical implementation phase is where we, our partner community, our professional services, our customer success teams are very involved in helping a customer be successful. At the end, you reach the consistent usage phase. What we see here, and you can see some of the incredible customer logos, I'm going to share some use cases here in a minute. What we see here is customers moving both depth as well as breadth. In a single use case, let me try to make the use case more sophisticated. I'm also going to start to expand across a multitude of use cases, and we'll show you some examples here in a minute. I want to ground you all on this because this is a fundamental way in which we think this business can be successful and how we manage it. As we start going through this funnel, we really manage it with a variety of programs. Go to the next slide. Across the funnel, we think about, of course, our critical sales programs where we're focusing on industry, really driving a lot of flexibility from a pricing perspective. In the implementation phase, over the last year, we have been very deep with our customers to help them think about their agentic transformation. This is not just a technology initiative. This is a customer thinking about what value they are trying to create for their business, what are the kinds of use cases that are going to drive the biggest impact. This strategy, this change management, and of course, the technology that comes along with it is something that we go very deep with our customers on. We began working with them with our forward deployed engineering team more than a year ago. We have a set of customers that we deeply focus with that are trying truly cutting-edge use cases. We call these our momentum and our hyper-focused customers. Of course, we're really investing in the community. We recently launched our agent plays or training that anyone can take online to start to really get comfortable with Agentforce and all of the incredible platform capabilities we built. We're enabling, driving our partners, and of course, working with our professional services team. These programs are really catered at ensuring we are driving customer success as customers think about this journey. However, they're also really critical because we have the privilege to work with well over 12,000 customers now that are on this journey, and we learn from them every single day. That learning directly drives our priorities from a product perspective. In the middle, for example, you see the things that customers want to really start to emphasize as they start to expand these use cases. How do they make their data readiness easier? How do they bring more determinism and control to these agents where you take advantage of all the goodness of the large language models, but you are also able to execute on structured business process and logic? Customers that get further in this journey have new questions. How do I test these agents at scale? How do I observe? How do I know what my ROI is, what my analytics are? This is just a wealth of information that we work with our customers on that directly goes into the product. We're really, really excited about bringing a lot of this innovation to Dreamforce, which you'll all see very soon. Before I wrap up and head to Ravee, I just want to show a few customer use cases of some incredible successes our customers have seen as they've been on this agent journey. I'll touch on a couple of these. Indeed, of course, is an incredible customer of ours. What's so inspiring about Indeed and their agentic transformation is that they've got a laser-focused North Star metric. They want to improve the time to which someone gets recruited by 50%. That is an inspiring goal that will have a lot of impact on their constituents. Indeed is really working on a multitude of use cases with us. They've got use cases that face their candidates, which is the most important and critical experience that they've now expanded their footprint on significantly. They've got a whole host of business processes, verification, ensuring that you actually have the right employer matching, and internal use cases that are driving productivity for their employees. Indeed, very, very inspiring use case. We'd also love to touch on Engine. This is a travel company based in EMEA. They are performing cancellation processes, reservation processes, all using Agentforce. This isn't just answering questions and simple FAQ. This is really going into core business logic at their company. They're now projecting a pretty significant 15% improvement in their handle time, which, of course, is going to drive a lot of savings for them. Finally, I'll touch on DirecTV. DirecTV is really going very deep on use cases that face their employees. How do they help their employees resolve billing resolution issues much faster? They're one of our most significant consumers now of our Flex Credit SKU, really starting to drive this across their employee base to see productivity. Just wanted to give you a sense of a lot of different use cases that customers are doing right now. With that, I'll hand it over to Ravee to talk a little bit about Data Cloud. Awesome. As is evident from what Madhav Tatay is saying, all our customers are looking at agentic enterprise through the lens of the various use cases that they're looking at that spans many different dimensions. As you all know, we have been on this journey of understanding customer data and intent and bringing all of the insights from across the enterprise and projecting it through a customer 360 across the space. We seem to be in the right place at the right time in the context of actually what the need is, not just about data technologies. It's about deriving the context so that our agents can be more meaningful, more relevant, more timely across the space, all with a consistent platform for governance. This is what is resonating very well with our customers. As an example, Wyndham, they have been on this journey of trying to understand their customers more deeply. Customer-facing, but how their company runs, how they think about their franchisees, and so on and so forth. This pillar of data has really enabled us to unlock the entirety of the enterprise with the right context. If you look further, how are we really accomplishing that? If you go to the next slide, what we have been doing across the space is to—can you please advance the slide, please? What we are seeing across the space is to really think about not just tapping structured data, which has usually been the realm of how I thought about it in the traditional CRM, but really unstructured data and how do we really bring together both the productivity, employee-curated content from Microsoft's ecosystem or Google's ecosystem or Slack, and putting that in context and being able to use that across sales, service, or across the space as it relates to something like tech and IT as we are entering ITSM as well. You would see a plethora of these advancements come to Marcet, whether we are ingesting or zero-copying or searching across the enterprise to bring the right context has really been our focus. Now, moving on to the next slide, while you have all heard the financial numbers in the calls, not just this quarter, but from before as well, what's really, really interesting is the usage and adoption. Across the board, whether it is FedEx or Indeed or Wyndham, what we are seeing again is the phenomenal adoption of not just ingest capabilities, but zero-copy has taken off quite a bit with 30% of our traffic right now as it relates to data coming through external sources with data zero-copy. Now, more important is also equally the fact that we are activating a lot of interesting use cases. Customers often start with one. They come back and refuel the tank, as Miguel said in the earnings call as well. Just this quarter, Q2, we had 40% of our growth, and you see we come from expansion deals where customers are seeing a lot more value and being able to not worry about how to expand from Marceting into sales very easily, very quickly because their platform is robust and set right. The best example to really articulate all of that is our Customer Zero. We at Salesforce are doing phenomenal work. Joe, maybe you can help the team understand how we are unlocking. Thanks, Ravee. As mentioned, I'm the Chief Digital Officer, and the IT function reports into my organization. That includes the technology that we build and use ourselves. It also includes all the third-party technologies. I think it makes me a good proxy for a lot of the customers that we are selling to who are the CIOs, CXOs inside or outside the company, I should say. The challenge that I think that we have with this, but it's also the opportunity, is the agentic enterprise or what we like to call the lean agentic enterprise is really a new concept. Much like when Salesforce started and we weren't just selling a different version of CRM, we were selling SaaS, which was new at the time too. We had to educate the Marcet, and we had to use ourselves as an example of what was possible in this new model. My team is really charged with using technology to help assist us to be the lean agentic enterprise. I think there are all sorts of aspects of that. I'm not going to drain the slide, but I think when I really focused on it is there's no template for it. We have to be ruthlessly focused on data and prioritization and making sure that what we're doing, we're constantly measuring efficacy on. We can go to the next slide. This is also one of the things that we have from a strategic standpoint, core principles. You can see that focus on quality. A lot of times when people want to get into agentics, they want to do a bunch of things simultaneously. The fact is, focusing on the critical use cases that you think the agent can be successful of is like a big part of how you get started and a big part of how we get started or got started. We're going to look at some of those examples a little bit later. Measurement is really important. A lot of times what people wind up with is if you're going to implement an agent, you come to realize that you don't actually have great instrumentation around the humans. If you're going to start to try to figure out, is this agent capable of doing this task as good or better, or can it offload, can it augment? You really have to have comprehensive measurement of the entire process. That's obviously one of the things that's a core capability of Salesforce. Even ourselves as Customer Zero are finding that we have to put instrumentation, more checkpoints at different spots inside of workflow so that we could actually weave humans and agents working together in an observable and then continuously improving type of way. We can go to the next slide. One of the things that's really important here is the fact that agents require the fact that people constantly have to attend to them. When I was working at Disney, the Marvel movie guys had a great saying, which was they never actually finished the movie. They just shipped it. I think that's sort of the case with agents as well. We get agents to a point of efficacy where we find them that they're impacting the business or delivering value, but it's a continuous improvement. We get feedback from our customers that use it, internal customers and external customers, every single day. It's incumbent upon us to take that feedback to continue to improve the agent. That measurement I was speaking about, where the baseline is of what the current human performance of that particular task is, is one thing. We're already starting to see where we're getting better than the humans at some of these very bespoke tasks that we're having agents do. When the agent gets better, the only way that we can figure out what the real headroom is, is to constantly improve, test, and measure. That's a big part of the process for developing the lean agentic enterprise. We go to the next slide. This is a dashboard. The data is not real because obviously this would be Marcet-moving data. This is a mockup that we use from a sales standpoint to show people without showing the actual data. I would say it is directionally exactly what's happening right now and a dashboard that we built to talk about our SDR agent. Marc mentioned this a little bit yesterday or Wednesday on the earnings side of it, where the SDR agent or the sales development agent is part of our sales agent. What it does is it essentially cultivates leads. Those leads that it's cultivating, this one in particular, are leads that previously we had scored so low in propensity that they really got automated follow-up from the company. Humans were not reaching out to them just from a scale standpoint. We just couldn't possibly afford to have as many humans talking to all these people there because the hit rate was so low because the propensity was so low. What we now have is a sales agent that could autonomously work those leads. From what I'd like to call the sawdust on the floor, we were able to pick this up and turn those leads into actual pipeline. Right now, our sales agent has done over $1.5 million in pipeline that's been created from leads that were essentially previously just thrown into the automated bopper that are now being worked by an AI using Agentforce and our Sales Cloud and Data Cloud behind the scenes to mine the data that's required to provide this function. We're now actually seeing real results on it. I think this is the point of really finding a use case, focusing on it, iterating on it, and then continuing to push the envelope and see how far you could get with it. If we go to the next slide, there's a lot of these focused on these hero agents, and you can see some of the data here. It's not just these outside-facing things. It's an employee-facing thing. It's outside-facing agents. We're going to talk in a second about help. These are all different functions that we think agentics is going to play a part of. Part of the ultimate evolution in our vision for this is humans and agents working together. The ways in which they work together are not agents taking over the complete job of a human. It's processes or particular aspects of a job that the agents are uniquely well-suited for that we can put them in. We can continue to tune. We can continue to improve. The sum is greater than the parts. Humans and agents working together to deliver outcomes. If we go to the next slide, this is one of the things I think is super exciting. The help use cases, both in my time in Salesforce and prior to coming to Salesforce using agentics as far as help goes, the help and support use cases are the ones that are actually very obviously very well-suited to agents, partially because the work that's done by humans is very regimented and very instrumented. A lot of companies, including us, would outsource these to various different call centers, etc. Those call centers have a high rotation of people. The folks coming in and out average tenure somewhere between 12 months and 18 months. You had to build a training curricula, i.e., you had data in a pretty good spot. You had to put instrumentation. You want to find out which of the reps is doing their job. You were measuring outcomes. Those are the key ingredients for an agent, a human agent. It's also the key ingredients for an automated agent. These drop-in support use cases, which are not exactly drop-in, still require work and refinement, have really been delivering results for us. If we look at the next page, we talk about a deflection of 77% of our cases, which I think is really important. We could go to the next slide as well. One of the things, though, is coming back to what I've been talking about this entire section, which is this constant improvement cycle where you measure, you understand, you refine, and you improve, and then you keep coming back and forth between that. As evidence of that, while I run the IT function, which reports into my organization, I actually work for our Chief Product Officer, Steve Fisher. We are part of the product development roadmap. Part of the job of Customer Zero is we take the challenges on head-on, helping refine the products so our customers don't have to go through that same pain. We can show some of the ways that we've then taken that mature product and enabled it to actually continue to improve the business and continue to improve what we're doing. With that, I'm going to turn it over to Anna to talk a little bit more about the use case. Thanks, Joe. I appreciate you going through that. As I kind of mentioned, it's our privilege to serve our customers. One of the fun things has been to be a customer in this space. Just to kind of add to this slide, there's been certainly lots of iterations and lots of learnings. From a perspective of someone delivering service, anyone who's in the service industry will know that it's not just about answering customer questions, helping them in the moments that matter, but also in how we make our customers feel in those moments. That's been a key learning for us. That wasn't an area of focus for us in the beginning, where we were focused on the data and hydrating it and making sure it's all, you know, the sources are good and it's clean and the quality of that. We focused on making sure the agents can answer questions accurately. What we really kind of learned in a lot of our testing, and I'll kind of share with you an example. It was Christmas Eve. I went to our agent and just asked kind of a plain question. It's Christmas Eve, and I have a lot of questions. I'm really concerned. What should I do? Our agents kind of came back candidly, kind of cold and unsympathetic and not very empathetic, and really at the level of care that we wouldn't really accept for ourselves. That's been really a key learning for us that this is really the moment where we say, this is why our agent is not bots, right? They're not bots. We want our agents to serve our customers in a way that humans would deliver that. That's really the power combination of being smart, right? The big brain, and then also doing that with a heart of service. It's been a pleasure for us to really have gone through this journey. As Madhav has said, in some ways, it's hard to believe that it's been less than a year that we've been in this journey. We launched this at the very beginning of October. We're coming up on our one-year anniversary. In some ways, it's been phenomenal to see how Agentforce can really scale for an enterprise like Salesforce. I'll pass it back to Mike for Q&A. Great. Thank you. Thank you, team, for the presentation. We're going to move to Q&A now. We have two ways that you can submit a question. One, you could submit via the chat window on the webcast, or you can raise your hand, and then we'll call on it. To get the Q&A moving as our audience gathers the questions, I'm going to pose a question that we get a lot from the investor base in everyday calls and really focused on Agentforce adoption. I'm going to ask Madhav and Joe to chime in on this one as both spend the entire audience here, the entire group here spends a lot of time with customers, but Joe and Madhav in particular from their angle. What are some of both the opportunities as well as the challenges we run into when we start getting into customers and start to walk through the process that both of you laid out as customers think about the use cases, but more importantly, as they go from pilot to pilot to production? As we make those jumps and then think about the ramp in production, what are some of the hurdles or some of the challenges that we're working through with customers to help them get over that hump? Maybe I'll start with Madhav. Yeah, great. I think there's three major buckets of things that we work with customers on. Number one, the data layer matters a lot. Without the right data, as we know, without the right structured data, without the right unstructured data, giving the right context to the agent, ensuring that the agent has access to the right data at the right time as it's executing on these things is really important. Remember, our customers are not just answering questions with these agents. They are executing on workflow. They're executing on logic, and the type of data they have access to really matters. That is a really important thing. What we advise customers to do here, though, is you don't want to take the stance of a massive data re-engineering project without kind of an end in mind. The way we work with customers is think about what the use case is, what's the outcome you're trying to drive, what's the relevant data for that particular outcome, and let's really optimize for that. The data is the first thing. The second thing that's really important and has been a huge lesson for us, our forward deployed engineers have spent more time on this, I would say, than anything else over the last nine months. That is bringing consistency and control to the agent. Why is this important? If you recall, we lived in a world, as Anna said, of bots. Bots were really difficult because setting up a bot required an incredibly complex array of choices. They were very rigid. You had to maintain them. You had to change them. Just a difficult thing for people to really do at scale in a significant way. LLMs unlock this incredible ability to now communicate with this technology in natural language. I can give this technology instructions like I speak to someone, and it's able to execute on those instructions. That's phenomenal, right? That really expands the remit and the democratization of the people that can build these agents. However, you still need control for structured process. Our big insight really here is how do we ensure we bring these two things together: the flexibility, the freedom, the natural interaction layer that you get with the large language models, but the traditional Salesforce strength of process logic that we can then integrate into these agents. We're going to have some really exciting things to talk about on this front at Dreamforce, but we've started to bring some of these capabilities in for our customers so they can retain context, so they can understand what the next step is, so the agent can perform consistently. That's number two that I think is really, really important. The third one is the interface layer itself. At the end of the day, as Anna said, you are serving customers with this experience. The customer experience has to feel empathetic. It has to feel rich. Customers' expectations certainly have been driven significantly by phenomenal consumer experiences that are out there. We want their experiences with companies to feel exactly the same. Whether it's on the voice channel, whether it's on the text channel, how do you make sure that you're creating these rich experiences where customers are, whether they are employees internally working in Salesforce or working in Slack, or it's external customers that are living on a website, living in an app? We spend a lot of time really helping customers think through what's the right user experience you want to create in this agentic world so you're creating the best experience for your cohort. Would love Joe to add with all his experience on actually bringing this technology and creating these agentic enterprises. Thanks, Madhav. I obviously agree with everything that you said. I think it's certainly crucial. I'll take a slightly different lens on it, which is, over my career, I've implemented a bunch of different technologies from a bunch of different vendors. There are two things that are really just radically different about agents, and it's just the new normal. The first of it is it used to be that when you would try to do a pilot, most of the work was getting the pilot to work. Then transitioning to production was work, but it was actually like just chopping wood at that point. You kind of knew what you were doing. What we're actually seeing with agentics and with Agentforce is actually getting the pilot to work is actually not as hard as that barrier was. Getting to production a lot of times is much, much harder. It's not because the technology is harder. What it comes down to is the second point or second lens I'd like to put onto it is that while these large language models are incredibly powerful, they're also inherently non-deterministic. If you expect it to be right 100 out of 100, then it's going to be very, very difficult to sort of refine that. In a pilot, you might say, oh, OK, it's almost there. That looks pretty good. Let's go ahead and try to take this into production. I think that it's just the nature of the technology that is somewhat non-deterministic. That's where we've put a lot of effort into Agentforce is to make it more deterministic, like Madhav's saying. In some cases, make it explicitly deterministic inside of the Agentforce technology wrapper. It's still one of those process things that I think people just have to adapt to and the industry has to adapt to is that it's different than deploying procedural code. You can't just make a bunch of unit tests. You can't just do a bunch of things to get to production on these things. It requires a different set of tooling. I also think that's a massive opportunity that we're here to meet the challenge with, which is providing that next generation of tooling to people that actually gives you the same sort of compensating controls, but in a very different way. I think we're just, from an industry standpoint, not there yet on the maturity curve to really see the hockey stick yet. Some people are seeing the hockey stick. We're seeing the hockey stick internally on certain things. Once people understand that these are a new set of tools, a new set of processes that need to get it done, and again, it's stuff that we're hopefully pioneering, we're trying to pioneer inside of Customer Zero. When we provide those things to our customers, we're going to see them. I think it was Ravee that mentioned, you know, the 40% of business that we saw was increased utilization of things like Data Cloud. I think those are the green shoots that you see, that people that actually have got it locked in, that have figured out how to do this formula, they're doubling down and tripling down and quadrupling down. I think that bodes well for the fact that we are both on the right track and we're really starting to see the very beginnings that a hockey stick left, you know, leap off. It'll take a little bit of time, just as all industry transitions take time. I'm very excited about where we're at, and I'm very excited that the learnings that we have are improving our tooling to a point where we're just seeing customers be able to act, outside customers being able to activate faster. Great. Thank you, Joe. Thanks, Madhav. We've got a question that was submitted, and this is a good one. I'm going to paraphrase it a little bit because it comes up a lot in conversations, both with investors as well as with customers. Ravee, I'm going to turn to you for this one. It's about the data estate that we see inside customers. Joe and Madhav just alluded to how critical it is in helping customers get their data situated and ready for use in agentic AI and leveraging the technology. Obviously, we've got a zero-copy partner network. A lot of the data estates inside our customers are super complicated. Whether it's us or a competitor that might be trying to integrate, can you talk a little bit about what you see inside the data estates, how we help customers, or what the challenges are you run into in helping customers get their data ready? How does that interact with the ecosystem? The Snowflakes, the Databricks of the world, etc., or Data Cloud, as it were. Can you help us walk through a little bit of the data estate dynamic that you see? Yeah, absolutely. I think this is an important understanding that we are going through. First of all, there are two inflection points. For most of the agents, we are seeing structured and unstructured data that has to manifest itself. Unstructured data, as you all know, is a brand new entire ecosystem of content that everybody is trying to process and grapple with. Just this quarter alone, we had about a 150% increase in the volume of activity we are doing on the unstructured content. Now, the challenge with unstructured content is the following. First and foremost, there is a lot of nuance here from the perspective of the type of content. As an example, we are working with a major medical device manufacturer, and they all deal with all their unstructured data in really flow charts of troubleshooting guides and so on. That is very different from a banking customer, a large one in India, where it is all about policy documents, which is all tabular and detailed. That is very different from what you might see in the context of a user manual from an auto manufacturer. These are all fundamentally different forms of data, different kinds of data that is unstructured. We really need a lot of important innovations to come through to really make all of them ready for a variety of different use cases that we want to light up, whether it is customer-facing use cases, employee-facing use cases, and so on. The other aspect that we are also learning is, while customers have a lot of important real estate in a Snowflake, a Databricks, a BigQuery, or Redshift, we are seeing a lot of them put to action in the form of the right semantics. Unless we add the right semantic model to it, eventually, the context for the agent requires all of the structured data and the unstructured data from the entire ubiquity of their landscape to come together. To give you an example, we have publicly talked about Fisher & Paykel. They are a very important customer of ours out of New Zealand. The key aspect there is how do we really think about bringing personalization and web and mobile events that they are having along with all the data that they have in their back office systems, along with all the advertising data that may come from the Google ecosystem. How do we really bring all of this diversity of data assets together? We see this partnership network growing rapidly. We have asked them to announce that now we have unlocked DB2 with IBM and Watson that went live last week. I think we made a public announcement about that, of how our zero-copy partner network now also extends into mainframe ecosystems. Our mental model is simple. We really feel context is important. Without the right context, agents are not going to be able to make the right decisions. In that, structured and unstructured needs to be woven together. We are going to continue to expand on this ecosystem through partnerships, both on the structured and the unstructured side. A lot of this is also algorithmic. As Madhav alluded earlier, we really need to arm the large language model-driven agents with the right information, with the right determinism, so that they are doing the right job as well. In many ways, the continuous growth that we are seeing in the data business is primarily fueled from the fact that people are realizing they can have many different use cases that they can light up once the data is ready to be able to take advantage of it in numerous dimensions, whether it is for analytics with Tableau Next or whether it is for transactional C360 use cases in a call center or an agentic use case, as the case may be. Thanks, Ravee. Now, I'm actually going to pull a thread on this question and ask Joe to chime in here. One of the common follow-on questions we get to Ravee, to what you just explained, is there's a notion of, I'll use the term super agent, that I think a lot of folks have a vision on. The underpinning of a quote unquote super agent would imply that there is a master data lake or data warehouse or what have you that cuts across the entire enterprise data estate. As we all know, that is a very complicated structure. I'm going to ask Joe to chime in a little bit from his lens on what he sees inside customers, building on what you just called out, where you have different data lakes or data warehouses that sit across various functions or vendors and how he thinks about that feeding the overall AI narrative. Yeah, I think it's a great question, great point. Like Ravee said, weaving the structured and unstructured is one of those things that we're trying to give as much capability to Data Cloud as possible to simplify weaving it together. It's also a little bit back to a change of mentality. Part of the change of mentality is we used to think that we had to get everything, all the data that we had, we had to get in this big tabular network with these tables and joins and all these kinds of SQL things. Certainly, that stuff's important. Agents are actually pretty good at looking at all that stuff and bringing it together in general, like agentic technology generally is. What they're not good at is two sets of very conflicting facts. When you have two different data sources that literally say the exact opposite thing, the agent struggles with that. That's where you get things that people are saying, oh, well, the agent's hallucinating. It's not actually hallucinating. It's actually just struggling like a human being would be if they looked up and got two different answers. I think when you think about the uber orchestration layer, the data hygiene becomes super important, but also the state becomes super important. In the same way that we don't, you know, not every person in a company has the exact same job, and we don't just do 1% of everybody's job, you have specialization of talent. The agents are ultimately going to be specialized and then orchestrated under something else, but they do need to share state about that customer. I like to call it omniscience, right? When multiple agents can share state about a given customer, a given record, a given company, whatever that agent's domain space is, when it can understand that state, then you can get to the point where you shouldn't be able to see the seams between the agents when you're starting to do handoff. There's no question that there will be orchestration agents. There'll be these uber agents that run on top of it. None of that's going to work if you don't do the wood that we're chopping right now on the back plane of making sure that the data is consolidated. What I'd love to do is maybe just throw it over to Andy, who's been doing a lot of this on Salesforce and was really leading the effort on help. I think that's a good example of where we had a bunch of disparate data sources. Andy, if you wouldn't mind telling about some of the work we had to do to get that into shape. A perfect example is if you think about all the platforms that we have across the company and how many different variations there could be of how to reset your password, right? It's a perfect example of confusing the agent and also the duplicate versions of that. Thinking not just about the help example, but our internal, we call our internal support team TechForce. How do you reset your password for your phone, for your Linux device, for your mobile device? We had different versions of those documents. It's exactly what you said, Joe. We confused it. Connected with that, the other thing that we learned big time at the beginning, and you and Anna spoke about this some with the slide, I'm going to always butcher it because I think of it as the heart and the head, but I think we used service and something else. What's the proper phrase that was on the slides? It's the heart and the brain. We'll accept that, Andy. OK. This whole idea, we dumped all of this information on our new latest hire, the help.com service agent, and we didn't train it at all on the art of service. None. That's where, as mentioned earlier, it was lacking empathy when Anna used it right in the middle of the holidays. It's never how we would onboard a human. That's an example of where we learned we didn't have enough of the right kinds of data, which is what our insights about how after we went live, how our customers were using the agent, we were able to see new data sources we needed to apply based off of the questions they were asking. We had way too much of other data, and we had to spend a lot of time cleaning our data repositories. One thing we always encourage our customers to think about is the team that's onboarding the agent. This agent's going to be a part of how would you onboard a human? We just did this with the SDR agent. Joe, you showed some bogus data, but talked about some real results. We gave that agent the heart of a seller, which is what we learned from our experience on help.com. It shows persistence, and it's hungry, and it's going after the sale. It still has empathy. Those are some of the things we've learned and some of the pieces of too much data and not enough of the right data. Thanks for asking. That's actually a good segue to the next question. It's actually going to go to Anna and Andy here. When you think about our own journey that we've been on with customer support, a lot of times, if I weave it against a conversation, and I'm paraphrasing the question that was submitted, if I weave it against what adoption curves look like for our customers, can you give a little bit of insight into the journey that we went on from initial pilot through getting through to full production? I think the detail would be super helpful for this audience to understand. At each kind of gate, if you will, that we jump through as you ramped across channels, as you looked at success factors, can you walk us through what that journey looks like? I think it's very appropriate and similar to a lot of what our customers go through, similar to the conversation that we've been having here. Maybe Anna will do. Yeah, I'll start. Thanks, Mike. That's a really great question. It's a conversation that we have quite often with our customers on what to expect. Naturally, when our customers are implementing Agentforce, there's this expectation of here's my case volume, and I expect when Agentforce is working, that that case volume comes down. The big learnings for us, I would say, in the first three months have been that we actually really didn't see that. We actually saw an increase in case volume. What we've learned were really two things. One is our customers have multiple channels to contact us, right? They can create a case via web, they can chat with us, they can call us. There are multiple ways to go do this. When we approach this, obviously, we want to make sure the blast radius is really contained so we can implement, we can learn, and then we can fix what we've learned. When you do that, what happens is our customers, what we find is our customers are still skeptical. We find that our customers are going to other channels. That's been sort of like one learning. The other learning has been really this positive experience that our customers are engaging. As Agentforce is proving to be helpful, our customers didn't necessarily, you know, our customers who engage ask more questions, maybe more questions they would have asked to a human. I think really a couple of things, what we've learned is we really didn't see this sort of like deep drop-off in case volume as we had expected, which would be sort of intuitive. There are some behavioral transitions here for our customers. Also, you know, we've talked a lot about this, right? Agents are not bots. When we went down this journey, there's really still some skepticism. Certainly, we didn't do ourselves any favors when we didn't really train our agent to be human. That was kind of a really big learning. We saw really, I would say, the unlock after about six months and nine months. As you saw earlier, Joe presented this, is that we're at a million and a half customer requests with amazing resolution and satisfaction. I would say the first six months of learning is steep. After that, once you sort of win over sort of our customer trust and confidence, we start to really see that drop off. Anything else you want to add to that? Andy's been my partner in crime for the last 10 months here. Yeah, I think there was the other aspect of that, Anna, wasn't just like thinking about all the things you talked about. It was also different ways of working for us because we had to start thinking and working differently across our customer support team and the IT organization of how we went after this problem and breaking down barriers that existed before that were now no longer relevant from a technology or organizational perspective. We've continued to see that now of thinking things differently. This used to happen over here in this part of the organization, but it's really not relevant to our customers. Trying to think about what is the customer journey and how are we meeting them where they're at took us some time too because it's really just different. All of a sudden, we're like, wait a minute. We could solve this problem with Agentforce. It was something we could never do before, surfacing up more information or accessing a third-party system or answering a customer's question about the renewal, which is not necessarily something that customers had asked before. To your point, the barrier had been lowered of what customers were asking. We saw a variety of new things that we don't see when they're engaging with human support engineers. Tons of learning and also learning where we needed to pay attention. Joe talked before about instrumentation and really baselining what does good look like that our human counterparts are doing. We had to implement that. Your team was actually better than a lot because you've already done so much baseline and metricing. There's been new insights that we've learned from that as well on the journey. Mike, I really think depending on the level of complexity, there is a six-month baseline that you're putting in place that you then have the opportunity to scale. There's the other aspect on customer-facing and internal-facing where you're changing behavior. One key thing that Anna's team has done is taking away other avenues of doing things. I think that's so important, whether that's external customers or internal. I don't think agents are necessarily always best when they're additive. You should really be removing complexity and removing optionality out of the system. The example I would give you about this is a lot of our customers have bookMarced a case submission form. We have, over time, removed the ability for different types of customers to access that form directly because Agentforce can solve their problem that would otherwise be routed to a customer support engineer. Part of this is really talking about changing behavior. Like Anna said, showing, no, we have a better experience that we can provide for you. Don't submit that case. That's some of the things that come to mind. Perfect. Thanks, yeah. I love the example you left there with Andy, because I talk about it with investors all the time. The moment of truth, I can go, if I go back six months, I can remember like it was yesterday. The moment of truth was when we decided to remove the button, the Contact Us button on the website. There was a lot of nervous energy in the room when we decided to do that. We're getting some more questions in here now. We're going to turn a little bit more towards some external customer examples. I'll give this one to Madhav here, and others can chime in. Can you give us, Joe put up a good slide earlier that looked at different use cases across the enterprise and how we think about agentic enterprise. Can you walk through an example or two of customers that we're seeing today that we're actively working with that maybe started small and are starting to expand and what that looks like? Yeah, absolutely. I touched on a couple before, but they're worth kind of re-emphasizing. The pattern that we see with customers is very common. By the way, this was true for us at Salesforce as well. Joe and his team have done a great job of really focusing our energy around canonical use cases we want to prove out while we run a lot of horizontal experiments trying a lot of things. You saw the slide of things that we've done. Two customers really come to mind that I think have really exemplified this. One is Indeed. As I said, they have such a clear North Star KPI, and any experience that they're doing really ties to that particular KPI. This is not just agents for the sake of agents. This is really very specific. They started out with, I would say, a pretty complex use case: make their actual candidate process better. They've got millions of candidates now. Our agent now is in front of a very large percentage of those candidates, so interacting with them all the time. Started out with, hey, simple questions. These are the companies I'm interested in. What can I look at? I need to schedule something. Basic business process flow in the candidate experience. That's kind of what they started with. They said, hey, if we're going to go do this, now we want to also make the humans better. The actual humans in their operations teams, as they're interacting with candidates, how do we make them more productive by giving them better information about all of this rich interaction that the candidate just had with the agent? Let's actually make the humans better up to speed on when they are now going to interact eventually with that candidate further down the road. How do we give them the best data, the best preparation so they're really tailoring and personalizing that experience for each candidate? The overall candidate experience gets better. That's an internal use case that faces their employees. Now they're experimenting with saying, hey, very often in our workflow, many departments in our company are involved. We already collaborate and coordinate on Slack. How about we ensure that the agents are surfaced on Slack as well so we can actually make sure we're getting all that contextual information? We can make the agent more productive, make the swarm effectively more productive with these agents helping us. That's a good example of a customer that you can just see. You start with a customer-facing experience. You then think about how do I now, as I'm handing off to humans, make the humans more productive. You think about, in a collaborative, orchestrated system with both humans and agents, how do you make them productive? That's a really good one. The second one that I love is Williams-Sonoma. Williams-Sonoma has an incredible bar for their customer experience. When you think about buying something from Williams-Sonoma as a company and all the brands that are underneath them, they really have the customer experience in mind at every point of the journey. Their journey has really been, as I said earlier, both kind of vertical and horizontal. What does vertical mean? Vertical means let's start with simple business process workflow. Where's my order? How do I get an order update? How do I cancel something? These are simple tasks that the agent can start to perform, and you have reliability and consistency in the agent performance. Within that use case, they started adding more things. Oh, can I get product recommendations? Can I think about what a customer might want next? Can I start to tie it into the Marceting and the sales journeys in some ways? You build depth in that agent experience. At the same time, they also have a lot of different businesses and a lot of different properties with fairly unique needs. They've now taken this agentic experience horizontal across multiple different departments in the company that all have certain different versions of it, but all tied to the single vision for what the customer experience could be. You see customers with that pattern as well make a single agent more complex or start to make the experience more horizontal across different sub-businesses, different properties. Great. Thanks, Madhav. This next question is, I really like this next question from Hannah. Joe, I'm going to ask you to answer it, but I won't read the whole thing. The question basically revolves around inside our customers. Marc mentioned on the call, you know, overestimation one year, underestimate five years. I think AI kind of squarely falls into that camp. Can you talk a little bit about embedded in the behaviors and what Madhav was just referring to, what you're seeing in customers and what you feel like are the, let's call them the major milestones, even though it will vary by customer, but the milestones that you expect to see and what you're hearing from customers over the next, call it what we're seeing now versus what you expect to happen 12 months from now, etc., that will help us increase velocity, if you will, of deployment, Agentforce and Data Cloud deployment? Yeah, I think it's a great question. I'll give an analogy that I think is helpful. It's sort of like self-driving cars. The first time, like, I had a Tesla Model S, and I remember I had the hardware for it, and then I got in the beta, and then the update came. The first time that I was driving the car, it's like I went around a curve on the expressway. I remember my hands were just air gapping, but huddling over that wheel, just like, is this thing actually going to work? Is it really going to make the turn? Oh, I got to make the turn. Now I think about it, I get on the expressway and like putting on auto steer to get on the expressway and drive for a while, I don't even think about it. It just happens. It's like one of those things that I've just become accustomed to. When it does something weird, you're surprised. You're like, why did it do that weird thing? I think that's a good proxy for agents, whereas initially, people are very concerned. Is this agent going to say something weird? Is it going to really mess up my customer support CSAT? Is it going to do these types of things? I think Andy's timeline is right. We're looking to compress it as much as possible. Maybe for most people, it would have been 12 months. Now it's six months. If it's six months for us, can we make it four months or three months for our customer? I think that's the type of thing that we're constantly working on. After that, customers get more confidence in what it's doing. They just start to layer on more and more and more complexity to it. It just becomes one of those things that's just part of the fabric of it. Slack was mentioned earlier. I think Slack is just, I really do think to some degree in this space especially, it's one of Salesforce's secret weapons because when you really are used to interacting with Slack, agentics, the sort of way in which you parcel small amounts of information back and forth and get answers and things like that is, I don't know exactly what agents are great at. What we see internally, for example, is when people use agentics, use an agentic agent, when they use it on a much more robust interface like one of our Lightning experiences in a cloud, and they use it in Slack, the sum of that is much larger than a person that uses each one of those modalities independently. I think what it shows is, again, the sum is greater than the parts when we get there, where people are going to start, they're going to be in this soft simmer, and then the boil comes, and then they really start to double down. You really start to see this hockey stick escalation in usage. That's sort of what we're seeing. It took us, you know, Andy, keep me honest on this, but it took us something like probably seven months to get to a million conversations and help. It feels like the last half a million has gone like that. It's like three months, and now it's like, you know, we've added 50% to that number. I think that's because for two reasons. Customers have gotten more comfortable with it. Our customers have gotten more comfortable with it. It's also one of those things where we've gotten more comfortable with it. Things like removing the submit form entry is a good example. We removed it because we genuinely think it's better at this point. We think it's got a higher probability of helping a customer. I think everybody's going to go along their march, depending on the sort of data fluency and the data work that customers have already done to the data lake and aggregating their data and things like that. It could be shorter or longer. That's why Data Cloud and what Ravee talked about with zero-copy is so important. For customers that have been spending time for the last five, six years building a data lake and getting the data in one spot or in a federated array of spots, congratulations. You were right. You were 100% directionally reliant about what you had to do with your data strategy. Now you actually have something to do with it beyond just minute for insights. You can actually put it to work and action that data in a real way that's impacting the business day in and day out. I think depending on where you are on your journey, you're going to see more or less acceleration from the companies, depending on where they come from. There's zero question in my mind that everybody that we talk about that really gets that use case gets it nailed, it's just they're doubling down because they get confidence in it. It's just like the self-driving car. You get confidence in it, you use it more. Right. Thanks, Joe. This next one, Ravee, I'm going to throw this one to you. Marc on the call on Wednesday referenced our data platform, is the way you referred to it, talking about Data Cloud, MuleSoft, and now, of course, Informatica coming into the fold, hopefully shortly. Can you talk a little bit about how you guys collectively in the data organization think about the collection of those assets, where they overlap, where they reinforce each other, especially as we fit Informatica into the family? Yeah, absolutely. I think we recognize the plethora of complexity that exists. To give you an example, we have customers who have on-prem data. They do definitely have some cloud warehouses in the mix. They also have applications too, like a backend ERP system. Of course, they also have Salesforce. They want to do analytics use case. They want to do transactional use case. They want to do agentic use case. If you really see, there are different tools that we need to bring to bear so that we get the line of sight from each one of them. As an example, with Informatica, we believe strongly that we will have a much easier line of sight to all the on-prem assets and infrastructure that matters the most, particularly in the back office systems. Similarly, MuleSoft, as an example, is a great asset when we see lots of customers integrate with their existing application ecosystem that they might have, whether it's SAP or something else that is relevant to them. Now, with the investments that we have made deeply within Data Cloud itself, we are now able to advance quite a bit in terms of having 270+ native connectors to technologies that might be available with the line of sight directly on the website, sorry, on the Internet, on any of the hyperscalers. They could be in Azure or Google or AWS ecosystem. The confluence of all of this is really to understand the customer data, understand the semantics. We believe strongly in data fluidity. It's not going to be one solution, as Joe alluded to, even in Series XM. There are so much differences that exist, same in Salesforce ourselves. We think that the more we do to provide bridges and make the data fluid, the better off we are in being able to actually curate the understanding, drive the right semantics, and then be able to activate it with the right context in different places. Another dimension to this is governance and security. I think this is another paramount important aspect that we think is going to become more at play as you have a world of multiple agents, both within the enterprise and across the enterprise that needs to collaborate. How do we really provide the right level of ground-level security and access control so that, just like humans, you really need to guard the information these agents are going to consume? How do we really build that logical layer across the stack is also equally important. We see all these assets coming to bear in the context of the maturity of the data platform, as Marc alluded to, in the form of having good governance, security, catalog, lineage, all of the conversations that we have had with being able to reach into the on-prem and on the cloud across the cloud vendors and applications. In many ways, the big shift here is not so much a data center of gravity where the data needs to be in one place. It's more about fluidity and how do we still give all the semantics that are required for all the mature features that are required for agents. Great. Thank you, Ravee. We've got time to shoehorn one more question in here because I like it, and we get it a lot from investors. I'm going to hand it to Madhav. How do we think about, from a product offering standpoint, vertical-specific use cases? We've obviously led that charge on the SaaS side of the business with our industry solutions. It comes up a lot in terms of how do we help our healthcare or financial service customers, et cetera, ramp fast on Agentforce? Maybe you can talk about that. Yeah, I would love to. We've got very industry-specific business processes and logic in Salesforce for a long time. We invested in this in our industry apps. Ravee just said something that's really important. Our strategy ultimately is about using that data fluidity to drive action gravity. That's really where Salesforce shines, is that customers are able to execute on their work on Salesforce products. I think there are no better indicators of those products than our industry applications. These are applications built with very specific industry ontology in healthcare, in financial services, where the business process is aligned to what the business process is in that particular industry. Now you imagine in an agent world, you've got this kind of connected data, you've got business process and flow for those specific industries, and now you're surfacing up all of that to the agent. Our belief is that that's going to be among the most powerful use cases. Our industry teams have taken this one step further and have now come out in the last month with 200+ templates that customers can get started with in any industry. Oh, you have an agent that's specific to billing? Great. You can build that. You have an agent specific to a health record update? Fantastic. You can actually go do that right now. To help customers get started on that journey, these templates are very helpful, very useful. Our customers have built a lot of incredible logic on these applications. I mean, Banco was one of the customers that I skipped over in the slide. That's incredible. They're a customer that uses our applications, but have now scaled to a significant extent in a customer-facing scenario because they're able to leverage all this business process and logic in the way they've implemented their technology. We think the verticals are really important. We actually think that the agentic use cases, especially at the value layer, are really tied to those vertical outcomes. A significant area of investment for the product organization. Great. Thanks, Madhav. I want to thank all the leaders for joining me today. I've really enjoyed the conversation. Thank you all for tuning in. As always, we'd love your feedback. We'd also love to hear your ideas on future sessions that you'd like to hear about. We'll continue to do these as long as our investor base and our analyst base sees value in them. Please let the ideas fly and give us any feedback or further questions you might have. Thank you, everyone, for joining.