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

Please welcome to the stage, EVP, Global Investor Relations, Mark Murphy.

Thank you. Welcome, everyone. Great to be here. Look at the energy in this room. Love it. Love it. As mentioned, I am Mark Murphy, EVP of Global Investor Relations with Salesforce. We have just a fantastic Investor Day lined up for you. I want to start by thanking everyone in this room. We are so grateful to you for your time and attention. We hope you're going to make the most of this unbelievable event by getting out there and engaging with customers and partners, ask them anything you want, ask them what's changing, ask them how this transformation is occurring for them. This is at least my 20th Dreamforce. If you can believe that, I've kind of lost count, but it's at least 20. And it's just unbelievable to be here and to be on this side of the room for the first time. It means so much to me. And as I reflect back on having been here through 20 Dreamforces, what is actually amazing is what has not changed at all. The location hasn't changed. We're still here in San Francisco and Moscone Center. The keynote is in the same room. The founders haven't changed. We have Marc, we have Parker sitting right over here. They're going to be on stage later today. The energy level hasn't changed. At the beginning of the keynote, the Hawaiian ceremony hasn't changed. I think a lot of the jokes that get made have not changed somehow in 20 years. What has changed is the technology, and the technology has changed quite a bit over the summer. There's a new ability to unlock trapped value. That has legged up in a pretty exponential way in the last couple of months. And so because of that, we're going to have a totally different structure to the Analyst Day. It's going to be unlike other Analyst Days that you have seen. We are going to be opening up laptops, and we're just going to be showing you live this kind of transformation that is happening. If we're going to be trying to show you what's possible, we're also going to have many more customers on stage. We're going to have Siemens. We're going to have Adecco, and we're actually going to have Anthropic on stage toward the end of this event. One more thing that has not changed in the last 20 years is the legal disclaimer. And so I'm not going to read it to you, but we're going to make some forward-looking statements that are subject to change. And we would strongly encourage you to refer to our most recent SEC filings, including 10-Qs and 10-Ks. So this agenda is going to be fast and efficient. We're going to really fly through this. We're going to start with Patrick and Rohan. They're going to go a lot deeper on the technology. We are then going to move to the go-to-market execution engine. We're going to have Miguel and Alexa for that section. And then batting cleanup, we're going to have Robin with the financial framework for this Agentic opportunity. After that, Marc is going to join us, and Marc will be on stage, and we will have a Q&A with the full leadership team for as long as they want to go at the end. And after that, we're going to have an investor reception and demos. And that's actually going to be right outside of this room. So with that, I'm going to hand it over to Patrick, who is President of Applications and Marketing. Patrick, welcome.

All right. Thanks, everybody. Does anybody want to take any bets on whether this agenda is fast and efficient, as Mark put it? Because I'll be happy to take a couple of bets. I think I know where. But that's okay. Okay. So I'm psyched to be here with you all today. Don't look at me like that, Miguel. And I -- we're going to walk you through a whole bunch of stuff about where we're going. We're going to kind of break down what you saw in the keynote a little bit more today. I have been getting a lot of questions from yesterday on how did we do that. There's -- it's funny, there's like 2 sides of belief. There's the side that kind of has an inkling and kind of wants to know more detail and then there's the other side that's like that couldn't have been real. And it totally was. So we're going to break that down and show you how all of that works. Rohan and I, the way to kind of think about these 2 sections in the agenda here is I'm going to start by walking you through how I think our apps are going to change and how they need to become agent-ready or a more technical term agent legible, meaning every app needs to be usable by agents in a composable environment like Cowork like you saw earlier. And then Rohan is going to come up and start walking you through how we are in the midst of transforming into a much bigger data company and how we'll leverage that data in an enterprise harness for the future. So that's the course of the next 30 minutes or 40 minutes or so. Now what I want to do is something just mildly annoying for a moment is I want to hop out of the slides, and I want to go to the demo screen just for a second, if we could. Thank you, team, back there. Okay. So I mentioned -- actually, how many of you saw the keynote yesterday? Oh, most of you. Okay. Great. I thought it'd be about 1/3 of you. Good job, everybody. So I mentioned in the keynote yesterday that you can go from nothing to something pretty usable in about 6 to 8 minutes. And then after that, you might have a few days of iteration as you build and you refine and you kind of design the user interface that you want. So I want to try to bring that to life a little bit here. So I've got a prompt up here, and I worded it kind of in a funny way, but I'm saying build a one-shot artifact and artifact is actually a capability within Cowork, which you'll see in a moment. And I wanted to list my open Salesforce opportunities. I wanted to use the headless toolkit to do it. And then I wanted to build a model where Claude drafts a task and posts that over to Slack. And then I say no questions. One shot at the front and no questions at the back are me trying to get Claude not to ask me any damn questions and just build the thing so that I don't have to keep coming back in and hit Enter while this is going. And what I'm going to do is this is like the cooking show. If you've ever seen a cooking show, where they pull the turkey out of the oven, even though it's been in there all day, is we're going to hit Enter, and you're going to see this start, and it's going to start working on this application, this opportunity application that I just asked. It's loading the tools. And all in all, this is probably going to take about 6 minutes end to end to come back with a completely working application. And this is really -- this is no longer vibe coding, right? This is we're taking the acceleration that these coding agents have delivered to developers, and we're putting it in the hands of knowledge workers. That's really what Cowork is. There's -- you're never going to see any code here. You're not going to see any of that background. Coworker just takes care of all of it for us. In fact, we can kind of expand and see what's going on. Now this is going to take maybe 8 minutes or so. And so we're just going to let that run, and we're going to go back into our slides. And then when we come out of our slides, God willing, it will have produced something mildly useful, and we'll take a look at it. Okay. We can go back to the slides. Okay. So I think you heard yesterday that we really believe that there is a big interface revolution going on. Another way that I like to say this is that there is a user behavior or a user experience revolution going on, which is that most people in the world in software, at least, are starting to figure out or starting to sense that there is a better way to use software, right? The way that we've used software for 50 years is we have to know what we want to do, and we have to know then how to use the software itself, and we have to click around and hunt and peck. The analogy that I like to use is that if you're a digital photographer, like a wedding photographer, part of your job is being -- having a good eye and knowing how to use your camera, but like most of your job is just knowing how to use Photoshop, right? Like it's a big complicated application. Salesforce is the same way. It's a big application. It has a ton of surface area, and your ability to extract value out of it is limited to the human's ability of what they know about how to use the application. And what's happening now is we are taking an agent and putting it between the software and the human. And we're just taking intent from the human. It's giving it to the agent, to the knowledge worker agent. In this case, what you'll see today, that knowledge worker agent is Claude. Claude is looking across all of the tools and capabilities, which have been connected inside of Cowork, and it is composing a response based on all of those tools. So you are getting a tremendous amount of additional value out of not just Salesforce but any other platforms that you plug into it. The other interesting thing that's happening there is it's kind of aggregating software into one environment because no longer do I have to do a discrete task in Salesforce and then a discrete task in SAP and a discrete task in Workday, I can -- if I have a task that requires all 3 of those systems, if they're all wired up in the same way that we just showed you yesterday like we can do with Salesforce, if they're all in that Cowork environment, it's now all aggregated and that agent can go off and kind of work and operate across those systems in ways that, frankly, we've been trying to build integrations for, for the last 50 years. And now we have this entirely new way to kind of drive these integrations. So it's a big, big revolution. And of course, we've been working towards this. Marc talked through this yesterday, really trying to simplify our platform to make it ready for this user behavior change that we see happening. We've really focused on our deterministic systems, of course, our data layer, which Rohan will go into; our apps layer, where we have a lot of the semantics and business processes that are codified into Salesforce. And then we added our Agentic layer and then now the new AI Force layer on top. And that's really where all of this starts to come together. So AI Force, you can think about it as 3 kind of new surfaces. One is Claude Force, the second is Slack, the third inside of Lightning is Agentforce Coworker. We will do more of these in the future. There is no shortage of knowledge worker apps that are beginning to emerge where you would want to be able to use Salesforce in the way that I'm about to show you. And then, of course, all of that comes together on this platform with Data 360, Customer 360, Agentforce and AI Force. When we click into Claude Force just for a moment, and I'll show you this in the app. Basically, we came off of Headless at our developer conference back in March. We launched Headless. We kind of -- we sensed the user behavior change that was happening. And we were like, we better get MCP servers out because our customers are trying to do this themselves, and it's a little bit dangerous. So let's get some first-party MCP servers out. That was, I think, very successful and a little surprising. I don't think people expected that posture from a software company to actually endorse that pattern. We got out there with it quickly, and then we started seeing people pick up and use it very quickly. And then on top of that, we started seeing them run into problems that we needed to fix. And Claude Force is really the extension of that and AI Force in general. It is us productizing the capability of those MCP servers into something that can get implemented and deployed across your organization way faster than every individual in the room having to connect their own MCP servers, which is just really not very practical. And then, of course, we'll talk more about this later across the sections, but then we wanted to package all of that up into one relatively easy to buy platform addition. So we have our new AI Force Max edition at $550 per user per month, and it includes everything that you're about to see today, including Slack and Slackbot, which you'll see glimpses of for me today, but not a ton. And despite this new addition, for the time being, Claudeforce is in open beta. So we're trying to maximize test area of customers that can touch it right now. One day soon, that will move into a GA, and you'll have to buy it to keep it turned on. But right now, we really want people to touch it. Okay. So I'm going to go back to the demo, and I'm going to break down a little bit what all that looks like. And then I actually have a customer -- a little bit of a surprise for you in a moment. We're going to bring a customer up, and we're going to see their live Claudeforce environment against their live production data as well. So let's go back to the demo. And you'll see -- there we go, I have to learn how to scroll. So it's still -- it did ask me a question. It's Claude managed, just constantly asking me a question. I'm just going to say, roll up command. I'm not really reading what it's asking me, so God knows what we're going to get. But I do have some confidence that it will be pretty good. Okay. So we're going to let that keep running, and I can kind of go into into the rest of the demo. So you saw this big beautiful dashboard yesterday, right, and all this whiz-bang stuff happening on the screen. But that's not -- that's a capability within Cowork. It's an incredibly exciting one, but it's also not the way most people start using Cowork. The way most people start using Cowork is they start on the chat screen. And so that's exactly what I'll do. I'm just going to create a new chat here, and they just start asking questions. Before anybody figures out that they can build applications, this is the way people usually start. And so what would a typical Salesforce question look like? Well, it might be, tell me about my pipeline, say, open pipeline. So a fairly typical question. Now what's going to go on in the background here as this thinks, and you'll see the work going on here is, first, it looks at the question and it's deriving the intent from the question. So just from the word pipeline, it is looking then at its tools that have been connected, and it is trying to connect a tool to the intent, okay? And in this case, it's going to connect Salesforce to that intent. It's going to say this is a question that Salesforce can likely answer, and then it's going to go ask Salesforce that question through the MCP server, and then it's going to bring back an answer. That is what is happening behind the screens. And not only did it bring back an answer, but you'll notice that it brought it back as a dynamically generated kind of mini application right here in the channel. And that is all made possible through a new protocol within MCP called MCP apps. Sorry to get mildly technical for a second. You'll also hear us refer to it as HXL, which is just our internal name for it. And all it means is when we ask that question, Salesforce on the other side, it says, not only do I know the answer to that question, but let me give it back to you with some additional detail about what an application could look like on the other side. And then Claude says, thank you very much for that information, and it draws the application right here in the chat. And all of that took place everything that I just said took place in -- well, way faster than I was even able to tell you what was going on, right? So it came back with our answer. And what's cool about these little mini applications is they're actually somewhat interactive. I can scroll over them. These aren't just like images that come back. I can interact with these things as well. In fact, I can even design buttons in. And all of this kind of sits on the Salesforce side. We send these instructions over to Claude and then Claude puts it on the screen. So I've got my North Star app here. This is an opportunity. This is actually the one that we were looking at yesterday at $4.8 million opportunity. And let's say, I wanted to ask a follow-up question. So tell me about the open activities. And effectively, you're going to see the same kind of thing happen. It's going to go to Salesforce or first it's to read the intent. It's going to realize that this is a Salesforce question. It's going to look up the skills and the tools to see if it knows how to go get it. And then eventually, it's going to come back with an answer. And ideally, that answer is also going to be some sort of dynamically generated interface, which I expect that it will be as soon as you see Headless toolkit, yes, there we go. So these are all of the activities that are logged inside of Salesforce, just brought back immediately to me and even a little activity chart, kind of a heat map where you can see when the activities were happening. Now these seem like relatively simple queries and they kind of are, but I want you to think through how much kind of internal human clicking is going on inside of Salesforce to do this the old way, right? You have to know how to get to opportunities, you have to load all of your opportunities, you have to click into every single one. You have to look at the entire opportunity screen for every single opportunity. You have to commit to memory, what's going on with those opportunities, then you have to synthesize that as a human being, and then that tells you what you should go work on or you can just ask Claude. And that's effectively what we've done here with Claudeforce. Now let me show you a couple of other things. One thing that's very cool is, if there are certain things that you do every single day or every few hours or whatever it is, you can schedule these. So that's what I have. I have a daily briefing skill here. So here's the daily briefing. So this tells me what I should be worried about every single morning. And I don't even have to come in and run this every single morning. I can just tell Claude to schedule it. It saves it as a daily briefing schedule, which you see right up here. And in a funny way, what you've done there is you've basically created -- just by asking, you've created an agent. You have created an agent that every morning is going to go out. It's going to inspect Salesforce. It's going to build your daily briefing, and it's going to bring it back and drop it right here in Claude. And of course, that goes to your phone as well. You get a notification every morning. So it's an incredibly powerful capability. I can also show you -- so here is the command center that we looked at yesterday, see if the sound is actually working. It is, which is unfortunate for Parker because that means Parker is going to come up. There he is. And we're just going to minimize Parker because I think he's heard enough of that. But this is a very fancy version of that prompt that I started with, right? Remember that first prompt where I was like build me an application. That's where this started. And then it just grew and grew and grew in capability as I asked it for more and more, which I think somewhere -- yes, right here, you can kind of see all of those prompts going through, and there's actually more than this behind the scenes. In fact, I can even show you. So remember a few moments ago, I told you that the kind of cool thing about this is the aggregation of software. This isn't just Salesforce, it's Salesforce and anybody else who builds these kinds of MCP connectors. Well, one of the companies that's really out in front on that is this company -- I don't know where I put it, here it is, is this company called ElevenLabs. And everything you see here on this screen, as I scroll through it, is if I know how to scroll, I actually don't even know how to scroll on this, I guess, we'll go this way. There we go. Everything you see here, I did not create. This is ElevenLabs internal kind of capability for creating voice models and for creating 3D avatar models. I didn't touch this. I didn't create any of this. All I did was create an ElevenLabs account, hook up the MCP server. And then when I was building my command center, I said, here is a graphic of Parker in the silly lightning costume. And then I gave it an audio file from a keynote that Parker has done in the past. Those were the only 2 inputs. And the Claude connected to the ElevenLabs MCP server, did every single thing else. Really powerful. And actually, there was a funny moment in here, if I can find it. It might be here. I forget, yes. I look at all the silliness as it generates this all out. I think it's here, filtered. Yes. So what happened is ElevenLabs didn't like this graphic because it thought that it demonstrated violence because of the lightning bowl, and it actually stopped me and prevented me from going further. And then we just had to change the initial graphic to make it look less like a real spear, and that's how we got through this. These AIs are constantly stopping you to make you do things. But anyway, none of this was -- 0% of this was manually created. This was all just from those series of prompts and the capabilities all hooked up. Okay. So I'm going to wrap up here quickly by going back to -- if I can find it, where is this? Probably in here somewhere. Is it this one? No? Stephen, help me out here. The very top, this one? Keep going. Here, it raised its hand. Okay, that's a good sign. That means it's about done. Okay. So it's ready to publish the artifact. The artifact is the application that I asked it to create at the beginning. So if I accept this, boom, there is my pipeline, and I should be able to expand this and maybe shrink this a little bit. So there is my full roll-up of all of my opportunities across my team. Super simple, just from that one silly prompt at the beginning. And I can make this better and better. I can make it look like whatever I want. For example, there was -- there's mine, which looked super fancy and had all the whiz-bang stuff, but what if I wanted it to look like a newspaper, all you'd have to do is ask in the instructions and you get a totally different view of all of your Salesforce data. What if you wanted to build something that looks more like a deck. So this is Salesforce information kind of in slide format. So you could actually build a full presentation. If you had to go do a Board presentation, everything is just there, all from a couple of sets of instructions. And then finally, what makes this all work? Well, it's pretty straightforward. It is this new plug-in behind the scenes where we've taken the many, many MCP servers that make Salesforce work and a handful of skills. Right now, most of those skills are focused on sales questions and salespeople, but we'll be expanding across our entire platform. And then things just kind of start to work because you have this powerful agent sitting between the human beings. So the key here is like this is not just about getting answers to your questions. This is about building full form applications and dashboards that you need to go get your business done. In fact, there was a thing that happened this morning. I'm not going to bring it up, but there was a -- I am going to kind of bring it up. There was a thing that happened this morning that I woke up and the very first thing that I said is I need a command center, so I can see the surface area of the problem that's going on, and I just jumped into Claudeforce and built it out and then got my team rolling on it. So we're using it in real life. This is rolled out to Miguel and Alexa's entire organization. It's rolled out to the Anthropic selling organization. We have about 40 customers in the beta -- well, in the pilot. As of yesterday, when we put the beta link up, I actually haven't checked to see how many customers have signed up for the beta, but it's probably several hundred or 1,000. And with that, I would like to bring up one of those customers. We're going to bring up Rick Janssen, who is the CIO of Siemens. You heard a little bit from Siemens, and we'll invite Rick up, and he's going to show us how he is using Claudeforce. We didn't test his laptop before that. So we'll see.

Hi everybody. Thanks for having me.

Thank you. I appreciate it.

So let's see whether this is working without any [indiscernible].

So this is going to be fun. We turned Claudeforce on for Siemens, what, 1 week ago, Frederik, 2 weeks ago.

There we go. Very good.

When do we turn this on for you?

A week ago?

A week ago.

A week ago, roughly.

Yes. And you've built the full sexy command center with a 3D avatar of yourself by now as well, right?

I actually had my team build a little avatar view, which I wanted to show up here, but...

You know what, I kind of deserve it actually, if you did that. Why don't you show us where you are?

Yes. Okay. So I mean, first of all, to start with, we are really excited. And I would fully agree from a Siemens perspective to what you have been calling the interface revolution. We believe that this is really the necessary step. And it's great to see that Salesforce is taking that, some might say, disruptive step to open up. I think the architectural readiness with Data 360, Headless 360 is there, and now introducing the partnership with Anthropic is just a great additional step. And I mean, as you can see on my machine, I'm running Claude Desktop. And to also quickly share here in the settings, we are running Salesforce for sales, the plug-in, right, what you just shared. I also have the team provide me 2 skills. 1 is the Siemens report, so just a layout and how we would like to create reports. And then the other 1 is the Salesforce data masking. We are, of course, looking at productive data here. So we want some data orchestration for the report we are generating in a second tab.

We were coming up with this on Monday night. We were sitting in the W with his laptop thinking about what to prompt. And he's like, but the problem is it's going to send production data. Miguel was trying to figure this out as well for his demo. And I said, no problem, just tell Claude before it outputs the answer to obfuscate your data, and it will do it. And it does. So we'll take a look.

So, and I mean, what I prepared for the demo is basically a prompt where I want Claudeforce to list me the top 10 Dutch accounts and their opportunities. And I want it generated as a report based on the skill I was just sharing in the skills section. And I think let me now hit return. And of course, we're going to have to bridge a little bit of time here just starting to think and execute.

And the fun thing about these demos is the AI works for a little while. And so you have to riff in between. The voice demos are also challenging because you don't mute in time, then they don't, it's like a whole new world of figuring out how to do these demos. How is that? Terrific? Did I go long enough?

No.

No.

No. No. Maybe we can quickly discuss where we could take it next, right?

Well, before, actually, maybe before where we go next, so this is, you said, a little bold for Salesforce to do, but maybe equally as bold for you, right? I mean this is an AI that now, click again, I told you that AI is constantly asking you. This is an AI that is like looking at and reading your Salesforce data and in some cases, maybe even writing your Salesforce data, maybe a scary thing, but you jumped into it. Why?

A, because we believe, as I just said, the revolution has just started and maybe you're delivering the blueprint for what many other companies would follow. You maybe saw our CEO also on stage at the keynote. And I think for him as well, it triggered somewhat, some kind of process that he says we need to maybe open up in the same way. I mean you're the #1 in CRM. So you have all the data about the customer and about how you are serving the customer. We have all the data about the product. So you have the customer truth, we have the product truth.

Absolutely.

So basically, similar path. Another thing is that, I mean, this is what I would call the Claudeforce version 1, which we're now launching. It takes away the need for many of our sellers, people in service and marketing, to really go through the already established interfaces, which we know from the past. But it would most likely start with reading data, combining data.

So you're starting with read... and synthesis...

Correct. But of course, the real magic would happen as soon as we would be able to also write in a meaningful way back into the systems because that is then doing exactly what you also were pitching. It takes away the elements where software has the human do all the work, and that is, I think, the next level of productivity, which we, of course, also want to achieve.

Absolutely.

And there's maybe even cool things we can do together.

We should have started this 1 before, just like I did the other prompts, but it's rocking and rolling here. This stuff is going to get...

Still going...

Still going. This stuff is going to get way faster over time. Really, what's happening in the background, it's not latency within the platforms. What it is, is overthinking is what it is. It is over, right now, the AI is over reasoning over the tools that it has available to it. And so all of that can be tuned as we move further in. And this is exactly why we want to get this out to as many customers as possible so that we can identify where you have what should be a simple question being over reasoned on as you come back in. Another way to solve that also would be to change the model and go to something like Sonnet, which will do much less reasoning and thinking, and it will just kind of immediately, the bigger the model you use, the more thinking it's going to do, which, by the way, is also more tokens. So that's relevant as well. Let's see, can you, let's see if we can expand, gives you a little expand on that. It might not. Yes. So this is what it's doing. You can kind of see all of the thinking. Actually, this is what's called the chain of thought. You're actually seeing step by step what the AI did there. So I got the Dutch accounts, now I need to. This is like literally, it is writing down its thinking as it goes, and this is incredibly useful information for us from a tuning and performance perspective as well if we have the telemetry to look at it.

So it should be ready in 1 second, but I think it's also worthwhile mentioning the additional connectors, which we would be able to bring in overtime. Yes. So connecting to the Microsoft Office environment, having all the Office 365 topics like conversations with customers, e-mails, documents, Exchange tasks noted down, documents are shared in SharePoint, et cetera, bringing that in as additional context, of course, something which would also be a very meaningful addition. I think the teams are working on making that already happen. And then the SAP connection is also something which is kind of powerful because we do billing, cash collection, everything on the SAP side. So having the end-to-end view on what's really happening in the different customer accounts.

How much of that from the Salesforce side, we've kind of solved a lot of that for you. You said teams working on it. For the other platforms, maybe not so much yet right? You guys are wiring it up yourself.

Right. We are wiring up it ourselves, but I would expect that you're going to have connectors, which are maybe then double tested or double checked with the different other tech partners to really have it rock solid and then out of the box basically.

Yes. Yes. I got you. So, this is what I meant when I said fast and efficient isn't going to work out. We're still rocking and rolling here.

I did somehow prepared for that scenario and now...

Yes...

I've created... perfect. And I maybe just open it here in Google Chrome so that everybody can see it. I want a nice extension, but I think here we go. And those are now the top 10 Dutch accounts, right? Obviously, it is, as I said before. But you would find it by value. In the Siemens report, as we define it in the skill upfront, you could drill down into the opportunities. So you see where your pipeline sits, what the status of the different deals is, and which stage they are, whether they are closed. And then also what patterns maybe stand out, win rate, cancellations. And then you could take it, of course, and dive deeper and double-click on the different elements to better understand what action you might want to take or to identify what maybe other people did to fix situations like this in other customer scenarios.

You'll notice it did something which I didn't expect, which is it actually went and got your real logo, and it probably even went to your website and stole some of your CSS to kind of make it roughly look a little...

That's the skill I showed...

You built in the background.

Yes, exactly because then we know that the system would always provide it in that kind of style and layout.

Incredible. Well, where are you going next with this?

As I said, I think we want to make sure that Claudeforce is really used across the organization. We have now the pilot going on with 25 users. The first initial feedback is excitement. People love it. And so I think we want to sit down and need to figure out what we are agreeing in terms of the pricing model behind it. But ideally, we can bring it out to our sellers ASAP and then innovate from there because I think that is what now the market is expecting. And I think we can speed up the transformation of our company with your support.

You have 25 now. How many do you think it could be?

18,000.

There you go on the go. 18,000. Amazing. Amazing. All right. Well, Frederik, really appreciate it. Thanks so much. This is super exciting. If you have any problems, then you will, you'll run into a few. Yes, please go to me, and we'll get it picked up, and we'll chase those 18,000 users.

Perfect. Thank you.

Thank you so much. All right. So I think from here, we're going to shoot it over to my colleague, Rohan, and we're going to start talking about trusted context and data. And Rohan, over to you. Thanks.

All right. Well, it's fantastic to be here and be with all of you. I, Rohan Kumar, I joined Salesforce 3 months ago, and it's just been a lot of fun. As Patrick started off, there's 2 big focus areas. Obviously, how do we get all the value that we have to be agent readable. That's a huge focus with headless and the work we are doing there. And the other big piece that I'll get into is where we are transforming to become a data company. And I'll talk about why is that and the importance of why we need to go there. And it's interesting, you look at like a lot of the value that AI is accruing, there's 2 very important things for these agents to be very successful. It's the model itself, which has the intelligence. It knows a lot about the world. The challenge is the model doesn't know anything about the business, right? When you look at customers, products, employees and how the interactions happen, the history and memory of how your business functions, a lot of that, the model doesn't understand. So by itself, no matter how intelligent the model is, it can't reliably reason and act on behalf of your company. And that's really where the data and the enterprise context to like that understanding of your business becomes very important. The challenge is building the enterprise context is very hard. There's, you look at like the data is in silos. A lot of times, the decisions that get made, they get made in meetings, there's no structure we are capturing all of that. And without having that comprehensive understanding, this is where agents hallucinate, right? When you look at how do you build trust inside agents to do complex stuff, getting that enterprise context right becomes very critical. And what I'll walk you through is how do you go create that? And it's, creation is 1 thing, but this is the most important IP of a company. And with these agents, you got to make sure that you secure it and then you have to manage it at scale when you have like a huge digital workforce that's running in your enterprise. So the first step in terms of creating the enterprise context is you have to get your data ready for AI. And what I mean by that is today, if you look at any reasonably complex enterprise, they have many SaaS applications, which have very high-value data. They essentially have data lakes, data warehouses, storage systems, where a lot of high-value data is getting stored. And it's actually really hard to even discover this, right? And so we have this product suite called Informatica that essentially helps data leaders in an enterprise discover all their data assets, clean them, connect them, figure out the semantic meaning and make that, take that first step of having the data ready for AI. So before the agents really understand your business, it's important that our customers understand the data assets in the first place. So that's step number 1. Now once your data is ready, it's ready, the next step essentially is to create what's called the enterprise context. And there's a difference. Data being ready means you can have like a bunch of structured data in your SQL tables that needs to be synthesized into figuring out what does revenue mean for the company, right? So that's a higher-level concept as an example, or what customer segmentation means for you. And to sort of create that context, we have this product called Data 360, which essentially not just relies on Salesforce data because there's obviously a lot of your enterprise data that's outside of Salesforce. And we've built this cool technology called Zero Copy. And what that does is it's able to reach into all the data systems that have been discovered by Informatica, remember, the first step. And then without physically copying the data, it looks at the metadata of each of these systems and it's able to synthesize the right context in a central location. So that's the second step. So you've made the data ready for AI with Informatica. And then Data 360 using the Zero Copy is able to get your Salesforce data and everything else into your context layer. Now once you have built out the enterprise context, the third step is essentially creating a semantic model. And semantic model essentially is your business glossary. When you think about what does customer churn mean, what does ARR mean, what does customer health mean? These are not just some static fields that you see inside a database. These are like business metrics, business relationships that have been created over decades, which are very important, like this sort of information for an agent plans are becoming very important. So Tableau is the tool that we have, the product suite that we have that helps you create these semantic models. And it's interesting, these manifest as dashboards today. If you look at Tableau is the BI tool, it creates dashboards. And of course, dashboards is being commoditized right now, like every AI tool helps you create dashboards. But the value of Tableau is never the dashboard itself. It is the semantic model layer that gets manifested at the dashboard. So in fact, if you think about, if creations of the dashboards become very common, you can do that in any AI tool, the value of your semantic models essentially has really gone up because if your semantic model is not right, then all those dashboards are actually giving you false information. So that's really where Tableau comes in. And now you have all this trusted context plus semantic model that you've created, it's available to all your agents. I mean this is great because it makes the agents a lot more accurate because they're working on a deep understanding of the enterprise. And then it also makes them more efficient. So they're not burning a lot of tokens. I think the point that Patrick was making where if you don't have the right data organized, then the agent is spending a lot of time trying to get to the right information, which is where they're spending a lot of cycles and a lot of tokens. So it becomes very expensive. So creating this enterprise context early on lands up becoming very important. The challenge, of course, is once you have all this information, if you put yourself in the shoes of the security leader, it's extremely risky because if these agents that are using this context start leaking it, deleting it, that can become a huge problem. So this is where we introduced this product called Salesforce Guardian. It's the evolution of what we have with trusted services in Shield. In Salesforce, we basically had capabilities like encryption of your data, data masking, et cetera, to protect your assets, and that's evolving into 2 additional capabilities. 1 is called agent identity, which is the embodiment of how the agent runs. For very complex stuff, you don't want the agents to run on behalf of the user because that gives them a lot of privilege. You want them to have identity which can be controlled. So that's 1. And the second thing is actually protecting this enterprise context itself, like capabilities that can actually classify your data, mark things which are highly confidential and have specific data policies that actually protect those. So that's how you secure AI. And finally, again, as this digital workforce starts getting built out, you can imagine you can start with like tens of agents, hundreds and then get into thousands of agents across your teams, platforms, applications, et cetera. And that becomes a nightmare for the CIO, for the IT leader to go manage. And this is really where for several years, IT leaders have relied on the product called MuleSoft to manage their APIs, which connect their business. And that's what we are evolving into this product called Agent Fabric. It's, think of it as a single registry that helps you manage all your agents, not the ones that have been built on Agentforce, but they're built on any provider. We basically are able to scan; monitor them and do things like FinOps to really look at what value you're getting of the agents versus what you're spending on them. All right. Let me actually jump to the demo and actually show you specifically Agent Fabric and Salesforce Guardian because there's 2 very important, it's not very obvious. The sort of the boring tasks that need to be done in enterprise, but it's extremely important that you get this right. So this is basically the homepage of Agent Fabric. And as you can see, we essentially tells you there are 726 agents that are running. And what you see are all the scanners. So these are basically tools that are running continuously in your enterprise to discover new agents as they're coming up, right? So, and if I go take a look at the Agents tab, let's go to the list view. You see on the provider side, agents from Azure, AWS, Google, Agentforce, so all the providers. And if I want to go deeper into any specific agent, let me just pick 1 of them. It's the recruiting agent. This was shown in the demo yesterday, right? As an admin, I get a bird's eye view of all the skills that the agent has. Like this is important because you really, it helps you understand the logic that the agent is using to come up with decisions. As an admin, I can go deeper, look at lineage. Here's where you see the integration across products. So this is the grounding data, the enterprise context that this agent is using. And all this information has come from Informatica. Remember, Informatica helps you discover all your data assets, and that has been integrated into Agent Fabric. If I'm an admin, I can actually map these data sources onto the agent, and I can look and say, hey, is it hallucinating? I can come and debug things over here. The other important thing is you want to monitor these agents as an admin like what's the average latency? If these agents are front ending your customer experience, then you really want to know how long does it take for the agent to respond, what's the error rate, how many policies have been violated. And then what's the total number of requests? Like how much of volume is being sent to the agent because that will be directly related to your cost, right? So here, as I see the agents, the specific agent is seeing a spike in the number of requests. So let me go dig into the cost management piece of it. And here's where Agent Fabric does a really good job of enabling the IT leaders to manage their budget really well. So you can think about creating sections of your budget, if you will, for a certain class of agents, and that's the maximum amount that they're about to spend monthly, and then you can manage things like that. Another interesting thing that Agent Fabric does is that behind the scenes, it actually observes the work that the agent is doing to come up with recommendations in terms of like what changes can be made to make this agent a lot more optimized in terms of token usage, right? So that's directly impacting the cost. So in this case, I can go to the agent I was looking at. There's all these optimization opportunities. And obviously, you have to be careful that these don't change behavior of the agent. But assuming that's true, that it doesn't, then I just go apply and that's it. Basically, that's very simple. So you can manage and govern the agents, you can look at their performance and the value creation that's happening and manage their cost. It's going to be very, very important as the volume of agents go. So that's the IT leaders' part of it. Now let's take a look at Salesforce Guardian. You'll see like the way we are trying to create this is you have this notion of a security score, which is overall, like what does the security health look like, how many assets that you have across your enterprise contact center I mean monitored here, you see it's about 94% coverage. But the 2 very important things when it comes to agents are essentially agent identity. So as the agents do their work, there we actually observe every action that they take. I imagine the scale at which these agents are running. And based on the actions the agents are taking, you determine whether the agent carries risk or not. For example there, the agent was supposed to do X and it's doing A, B, C. Well, that's, they're deviating from what they're supposed to do. So we capture those observability logs and then analyze that and determine the risk of the agent. So you see in the dashboard over here of all the monitored agents with different levels of risk, 2 of them have been called out as high risk. That's something for the security to take action on. The other important part is the data assets, which is your enterprise context was built out of a certain volume of data, how much of that data has been classified and understood, right? Because if that percentage is low, then you land, there's a lot of risk that these agents might be touching data that you don't want them to touch. So anyway, that's just a bird's eye view of how Guardian works. So Agent Fabric for the IT leaders, Guardian for the security leaders. So beyond just selling you the agents, it's this whole manageability piece, the security piece that lands up becoming very important. All right. So we can go back to the slides. Yes. So all the stuff that I just showed you is things that we have today, right, which all these products, there's a method to the madness. It's not just a random suite of products that we have. The journey that I walked you through is exactly what an enterprise would do. Now the interesting thing now is I'm going to pivot a little bit into the future of this Data 360 layer that you saw in the stack, how that evolves into what we're going to call the enterprise AI harness and what's our perspective over there. It starts with the frame essentially is every role in the enterprise is evolving. I mean we've sort of spoken from the perspective of the business users. But like I said, the IT leaders are going from just managing devices and apps to sort of managing these agents. You take a look at the data leaders, they're not just managing data warehouses and lakes, creation of that enterprise context that makes the agents accurate and efficient is going to be a very important part of their job. This is things like that, that most of these roles, things are going to change. So what is an enterprise harness? This term has sort of become like 1 of these platform terms, right, which is extremely confusing. But fundamentally, this is the link between what your AI models, the intelligence that you have and then the understanding of your business. The harness is essentially what brings it together for the business outcomes that you desire. Now in Salesforce, like this is a stat that we showed, the bottom layer Data 360 essentially is going to evolve into this Salesforce's version of the enterprise AI harness. And it has 6 capabilities. I won't sort of go into the depth of each of these and an AI control plane that brings it all together. But fundamentally, today, there is a lot of discussions around harness to protect a single agent, right? So where does the agent run, the run time of that, the evals, which basically tell you the agent is efficient or not, right? You test; you make changes to the agent. And then depending on what responses it's giving, you manage that against your evals, that's, think of that as your test set. And that's, those are the only 2 things that enterprises are focused on. We believe there's a much larger play over here, right? Starting with trusted models, which is we wanted to give you a choice in the models that you pick, right? It's not, you don't need the most extensible model for everything that you do. As a platform, we need to figure out based on the outcome that the customer is seeking based on price performance, what's the best model to pick. Trusted context is everything that I spoke about, which is synthesizing your business understanding, your semantic models, all of that sort of stuff. Agency and actions is how you, when a customer describes the outcome that they are seeking, the agent and the model they create a plan. The sequence of steps that need to be taken to execute that decision-making comes in the agency framework. And then the actions, the trusted actions is how you execute each of those steps. And every step that gets executed has to be governed just to make sure that the agent is not doing anything that it's not supposed to do. And then trusted governance is essentially getting all your data policies right because that's going to influence how your trusted context gets created and security is about agent identity and data protection, right? It's a very comprehensive view that at Salesforce, we are taking in terms of harness definition, not just protecting the agent, but if you can wrap your arms around the harness, you should be able to bring that with the headless APIs to the AI of your choice. This is really where we are sort of going towards the future. And an example of the model choice is what we announced in the keynote yesterday. Koa, it's actually very exciting. It's the very first CRM reasoning model from Salesforce. We basically post-trained the open-weight Nemotron from NVIDIA. And like 27 years of deep product understanding and appreciation of the CRM workflow, more than 10 years of data and AI research, it's sort of been baked in. So there's a lot of value, very excited. And this model has been designed for long-running agents. And what I mean by that is agents that can actually create a chain of thought which has multiple steps and make decisions at each step. So you can imagine a complex sales opportunity where what's the next best step to take to go make that happen or a very complicated service case that an agent is dealing with to get the customer to a better state. So complex situations like that, Koa can greatly help. Very, very excited about it. And then finally, evolving Agent Fabric into this AI control plane, which essentially helps you discover all your agents, govern them, connect multiple agents through standard protocols and eventually do your FinOps, the cost control, all of sort of coming together. That's a big part. Now 1 important thing is it's a very open and composable system, while Salesforce is going to have a very comprehensive answer for each of these challenges because we believe without having a complete solution like this, production deployment of agents is just not going to be successful, right? That's why we are taking a very holistic approach here, but it's open and composed. We realize that like our customers may have a different security vendor, a different data vendor or a different vendor for the AI control plane, all that's fine because through our headless API, we are able to integrate with each of these partners. So that's a big design point from the very beginning, saying we have a complete solution, but you don't have to use everything from Salesforce. And the thing that I'll finally say is while this is, we are sort of evolving this into the harness, the approach we are taking in terms of building this product is very incremental. So if you are a Tableau customer, if you're a Data 360 customer, if you're a core platform customer of any of our clouds, Sales, Service, et cetera, MuleSoft, you already have taken steps towards actually getting into this harness because we're breaking down the capabilities of all these products and bringing it together in a very composable and a unified way. It's pretty exciting. I think this is something which is going to be very important for our customers to do their Agentic transformation right. And would love to hear your thoughts on this after the session. So thank you.

Thank you, Patrick and Rohan. The demos were sensational. Next up, we're going to have Miguel and Alexa. And I want to just jog your memory. 1 year ago, Miguel was on stage. Miguel was our Chief Revenue Officer. His remit has expanded. He has all of go-to-market motion now. That includes customer success and partnerships. And along with that, Alexa's role has been elevated. And how about a big round of applause for Alexa now, our new President and Chief Revenue Officer. And I couldn't think of a more critical time to have you on stage. So thank you.

Thank you, everyone. Rohan, Patrick, thank you for teeing up so nicely for us. Alexa and I are the lucky executives that get to take this amazing software stack to market. I was here a year ago, same exact location. I think it was October in front of many of you. And Robin and I, we were telling you how excited we were with the momentum that we were seeing in the business, the massive opportunity ahead of us with the Agentic Enterprise and how we were investing heavily and how we were investing heavily and wisely to capture the opportunity. The interesting thing is months later, we started hearing this chatter, this noise in the market, something called SaaSpocalypse or the false narrative that AI was killing SaaS. You know what, we are very competitive. We are like what the hell is going on. This is not what is really happening. I mean, luckily, the market are starting to come to its senses. And fortunately, I think by now, all of us, I think we believe that it's time to put this false narrative to bed because the reality of the business that we've been operating on for more than 4 quarters already, more than a year, is like the piloting phase of AI from 2 or 3 years ago came to an end. That's what we told you and people were going in production at scale, betting on the big software platforms like us. That's been the case throughout. So we call this new phase SaaSceleration. Some people call it Renaissance, my accent is not great, but you get my point, right? The message here is AI is not killing SaaS. In fact, AI is making software vendors like Salesforce more relevant than ever because AI needs the trusted context and the secure governed deterministic execution to drive value in the enterprise. I mean that is fundamental. And we have all the pieces of that software infrastructure that you saw earlier. I'll give you a small anecdote. It just happened to me in the last meeting before coming here, I was meeting with 1 of the largest PE firms in the world. Okay? There aren't that many in this category. And 1 of the leading partners, he was telling me that 300-plus large portfolio companies. 75% of them are on Salesforce. And he told me something that I wrote down a piece of paper that was pretty amazing. He said, Miguel, I just want you to know that the Salesforce platform, your business application is the most prolific, prolific of all the platforms. And he named SAP, he named ServiceNow, he named Oracle, he named Workday in terms of driving value with AI to our portfolio companies. And I'm like, you're going to come on stage with me later and say that, but he's conflicted because he's probably 1 of you, he may be here, I don't know. But the net-net is it's a really renaissance era for us. These are the 3 key messages that Alexa and I would like to share with you today. Number 1 is we see increasingly strong demand in this area. AI is accelerating SaaS, okay? And we are, we continue to invest aggressively, again, wisely also in the top AI markets to capture this opportunity, okay? Very similar message that Robin and I delivered last year. Second, this is a bit new because we were learning last year. We, and Alexa is going to go in detail on this point, the second point. We have now many different levers to monetize AI. We identify a bunch of drivers that are making our customers buy packaged solution with AI embedded. And we also have multiple commercial frameworks to monetize that opportunity. We are meeting customers where they are. Not every customer is in the AI transformation journey at the same place, but we have a pricing, a commercial model for each of them. And then finally, hopefully, I mean, I was very, I was worried that you guys came today for this event and then left and you were not at Monday, Tuesday, the keynote. I'm very glad that 2/3 of you were at the keynote. Hopefully, you've talked to a lot of customers. I talked to a lot of customers myself. At every meeting, I have to pinch my cheek. I'm like, 'Oh my god, this is so incredible what the customers are telling me.' The proof is in the pudding, and you've heard from many customers on the keynote, et cetera, you're going to hear from 2 more customers. We heard from Frederik, we're going to hear from the Adecco Group. We've heard the marketing story. We're going to hear the real story here by their technical team. And then also, we're going to hear from Anthropic, and it's super exciting when you hear from customers. Now on the first point, I'll cover the first point. I know that you like this slide on the left side from last year. So Robin and I, we were very confident about the business. We knew that things were going very well, okay? We had really strong and healthy succession of quarters, okay? And we told you last year that we were going to see an organic subscription support revenue reacceleration within 12 and 18 months, okay? Guess who said 12 and who said 18, okay? This is like a partnership. Well, it's been 12, 13 months, okay? So we have guided Q3 subscription support revenue. If your models work correctly, you'll see that there is the beginning of a reacceleration. Why? Because of what we told you last year is the chart on the left side, the magic word, net new AOV growth. Net AOV growth at that point when we were here, the lines were crossing. We knew they were crossing, okay? They did cross in Q3. So the net new AOV growth outpaced the AOV growth. Mathematically, when that happens, the AOV accelerates. AOV always grows, obviously, but it was decelerating for years. Why? Because there was a lot of, because of this. Because we have some years after COVID where the net new AOV was actually even negative growth, and it was pulling down the AOV growth. Now since Q3, the last 12 months, the last year, net new AOV growth has outpaced the AOV growth. Okay? And we feel very confident in this net new momentum going forward. Now look at the business metrics. I mean, most of them, I think all of them we've already shared at our earnings. cRPO, not a bad number, 14% we delivered in Q2. We guided 14% this quarter for Q3. That 14% for Q3 does not include the Contentful and Fin acquisition that we just closed last week -- this week. Pipeline at record levels. That's a little bit of a new info for you guys, but it's in the very high teens, very healthy coverage ratios. Coverage ratios are healthier than other years with increased commits, obviously. It's looking good across the board. Okay, across segments. Contract lengths are increasing. I mean we gave you this stat in Q2. We do 100,000 transactions per quarter, give or take. A lot of them are renewals. Many of them are new bookings. It doesn't matter whether we're talking renewals or new bookings. We give the average of everything. It doesn't matter the segments. We took every deal band, contracts were expanding in every deal band. That's good. That's good. Then you add the fact that seats, sales, seats and service seats. Remember, the world in SaaSpocalypse world, the world is coming to an end. There will not be any more sellers. There will be not more service agents. Guys, you have a seat-based model. Salesforce is going to cloud. We haven't seen still any decline on number of seats for sales or number of seats for services. And we track that obviously very closely because guess what, at some point, there might be. But we are not even worried if when that point arrives, which so far hasn't arrived, it's still growing. We have so many new commercial frameworks to monetize the AI opportunity that, we'll be okay. We'll be okay. But so far, it's increasing. And of course, Slack seats are flying through the roof, okay? It's the hottest product right now, I think, in enterprise software, definitely within Salesforce, but I think in enterprise software. And then the consumption, AWUs are exponentially growing pretty much every quarter, we do as many AWUs, Agent Work Units, which is the closest thing, it's not tokens, it's the closest thing to productive work delivered to the enterprise. We do more in 1 quarter than in the history until now. I mean obviously, mathematically, that will have to stop at some point. But it's a lot of consumption, a lot of customers, Agentforce customers consistently consuming. So, and then, by the way, I didn't mention something that I think it's going to be 1 in your slides, Robin, every single AI company, the same AI companies that were going to kill SaaS and replace all the CRM systems, all of them are going fully into Salesforce. In just 1 year, in this AI segment, we multiplied by, actually, it's 100%, nearly 6x fold the business that we do with them. Not bad. You take the top 9 AI companies in the world, the top 10, 9 are wall-to-wall Salesforce, okay? The 10th one, probably you can imagine who the crazy company is, it's an amazing company, by the way. They are actually not wall-to-wall, but all top 10 run on Slack and 9 out of 10 wall-to-wall Salesforce, okay? And they're multiplying by 6 the investment that they're doing in us on a year-on-year. These signals, these business signals are clearly not a SaaSpocalypse situation. So I'm not going to use the word SaaSpocalypse anymore. I hope you don't either. And the market will come to senses this little by little, and we are ready. Now because we see this momentum, because we are excited because we are talking to customers, because we're seeing the demos because we're seeing the new surfaces because we, I, Alexa, all our teams, we are now working. I mean I go to work, I think I'm going to a video game. I have my, I mean, if you guys see my surface where I work, you don't believe that this is a serious company. This is -- I'm having so much fun. I have a screen that looks like Patrick's demo. By the way, in your, when we finish this session in the break, 1 person in my team is going to show you my cockpit, which is a skin that I put on top of Claudeforce, okay? Because the good thing about Claudeforce is you can say a few things to Claudeforce and it builds a new skin on top of it. Okay? And until yesterday, I couldn't demo, definitely not to the finance community because particularly this is a public event, it is broadcasted. And you will see all my numbers, the numbers for Q3, which I'm sure you're super interested to know. And all of them are going to be on my laptop. Ning, who works in my team is going to show it to you, but we told something to Claude this morning. Create a toggle button on the top right that says demo mode. So when she clicks that, everything that is data in our metadata model, everything that is data gets blurred. But she's going to do from time to time like this so that you see that this is real data. So this is not, because I think, Alexa, your demo is going to be dummy data, okay? It's on the Slack, there's another surface. But it's the same system with dummy data because you didn't ask Slackbot to do that. But if you do it, so, but listen, it's fun. And if it is fun for us, it's fun for our customers. I mean, Rik, I mean, he's just playing with it. I mean I feel bad for him because he's just been playing with it for 2 days, and he was here in front of the world doing a demo. And this is very basic, what he just built. Imagine if you build that, you iterate 1 week and you have the most amazing dashboard in the world, driving value to you, sending you signals. Every time that you put a prompt, you are creating an agent that does things for you. So I don't log into Salesforce anymore. Parker, thank you very much for leading the way. I think it was here when you said I don't understand why people will log into Salesforce in the future. You were totally right. I was sitting there and I'm like, this guy is crazy, okay? He's my age basically, but white hair, I'm like this guy is becoming crazy. Why he's a visionary? And I pretty much 1 month later, I stopped logging on to Salesforce. And I use Salesforce 100x more, okay? Because of all this excitement, we are tripling down, investing in capacity, also under my new role, investing across the success area at these builders, I have 20,000 more people in our revenue because we're bringing the whole go-to-market together. And I gave all of them 1 focus. And you guys can imagine which focus I gave, which objective I gave to those new 20,000 people in customer success, professional services, partnerships, 1 metric. You can speak. Not you. Look at the chart, net new AOV. And why net new AOV and not customer success? Because net new AOV means it's the ultimate metric of customer success. If customers buy more and are happier and adopt more, which is net new AOV, okay, it's good for everyone. So anyways, things are happening. We are doubling the capacity. We've added 30% more AEs in the last 2 years. And at the end of this year, we're going to have double-digit growth, 10%, 12% growth, more AEs in the core countries than 1 year before. We're also adding 1,000 builders as we speak. So we are very confident in the momentum and the direction and the net new momentum going forward. And then we have, we are very confident about the software infrastructure. Now one; remember the words from the PE person. This PE person represents 180 companies. And these are not Mickey Mouse companies. They are pretty big companies because this is the biggest PE firm in the world, 1 of the biggest. We are the platform that drives more value to the end customers through AI because people want to consume AI through packaged applications. This is the trend. We take care of all the complexities. We take care like Claudeforce. Everybody can connect MCP servers and build their own MCP servers, but you want to do it securely. You want to do it with 0 data retention. You want to do it with all the skills that matter. So that's the power of our platform. We are very confident. And with that, I'm now going to hand over to Alexa. We already applauded her. I'm so proud to work with you. She's amazing. She's a force of nature. And it's my time, my time.

I didn't get walkup music. I was very; I was sitting in that seat wondering what it might be. Thank you, everybody, for having me here today, and thank you, Miguel, for the introduction. You guys don't know me yet, so maybe I'll share a quick personal tidbit. I grew up in a family where we had a motto that was per aspera ad astra. So through difficulty to the stars. And I think it was really appropriate for what we've experienced over the course of the past couple of quarters because difficulty, I think, sometimes is a necessary condition for success. And I think what it did for us, as painful as it might have been, was forced us to get really clear on what our customers needed from us in that and also what we do incredibly well. And I think at the intersection of those 2 points is we discovered that our customers had a huge amount of trust in us and our customers needed us more than ever to help them transform. So that's what we've been really focused on. We're sort of ignoring the noise and focusing on how do we go execute to drive against that customer expectation of transformation and trust. And so candidly, I'm even more excited to step into this new role in a period of, we won't use the old word anymore, but in the SaaSceleration or the Renaissance. But I think that difficulty has been a huge learning aid for us. So let me sort of help you understand a couple of things. 3 things on this chart, although it's going to be a pretty dense discussion of those 3 things that I want you to take away. The first is that we think we have an incredible opportunity to go monetize AI. The TAM is exploding. I think that the word we will not reference anymore did not incinerate opportunity for software, it expanded it. And so Robin will talk a little bit in her section about some of the hard numbers and how we see that spread out across industries and customers. I think 1 interesting way to think about it is the expansion of knowledge workers. There are, give or take, roughly a billion knowledge workers today. We have 50 million of those knowledge workers roughly on our platform. There is huge upside to go capture a really significant portion of the market. And that is before we ever think about an agent as a knowledge worker, which I think is going to geometrically expand that knowledge worker base. There is so much that we can go do here. So that's the first piece. I think the second piece is to think about how are we going to put our capacity. So Miguel talked about all of this capacity that we're investing into the market, double-digit growth from the builders. And you're going to see us, I think, do a couple of things. The first is recenter on the markets where we think AI is going to grow the fastest. And so I think top markets has always been pretty central to our strategy. I've been here for a decade, but we're going to double down on the 9 markets that we think are the biggest and are going to grow the fastest where we have infrastructure to go in and accelerate. It doesn't mean we're not going to invest in the rest of the world. It doesn't mean we're not going to look for opportunities to move really quickly and find upside potential that we haven't tackled yet, but you're going to see us concentrate our, put all the weight behind the arrow and go concentrate our effort and our energy. We're also going to be really thoughtful about what does the capacity mix need to look like. You saw the stat that we've scaled we'll exit this year scaling to 1,000 builders. And so as I sit with customers, and I spent, I probably met with, gosh, I don't know, 150, maybe getting close to 200 customers this week at Dreamforce, 75 CROs yesterday. The #1 theme coming out of -- I spent 2 hours with some of the top CROs in the world. The #1 theme coming out of that session was help me build this. So they saw Claudeforce, the demo that you saw Patrick show earlier. They also saw a version of how we run our business in Slack. There will be other examples of that on other surfaces they want and need us to help them build that. And so part of the capacity modeling that we're doing right now is thinking about, I don't want to show up to a customer or anyone on my team to show up to a customer with a PowerPoint. I want to show up and build a prototype; I want to push that into production as quickly as humanly possible. So that's sort of a thought in terms of how are we mapping our capacity to the demand function that we see. And then I think in the spirit of how we started, we're being pretty thoughtful about, okay, what are the demand drivers? What are the patterns that we see. And what I would tell you is this is indicative, right? And you guys are talking to customers, and I think you recognize that there's alternate patterns, but I think this represents probably the most common one. So let me sort of walk you through what I see more often than not as the way customers adopt the platform and how we're sort of thinking about driving that monetization motion. So the first, when I talked about knowledge workers, is we unlock new categories of knowledge workers. We unlock additional seats. Miguel talked about, we are really confident because I see seat expansion without even uttering the word AI, I see seat expansion as an opportunity. And I think a good example of that is Replit. Everyone in the room is familiar with Replit, 1 of the fastest-growing digital native or AI native companies. They scaled from $10 million in ARR to $0.5 billion in ARR. We scaled with them. Those are core sales seats that we landed into the account, and we're growing as they grow. The thing I think that's really exciting is it's not just sales, it's not just service, it's also Slack. You heard the story about Adecco on the main stage. We're going to tell you the real story today, so I'll skip that one for now. But I think the net message I will leave you with, and I think I heard this in the CRO Summit yesterday, too, people still need to build contact centers. They still need to build out pipeline management systems. They're still going to build marketing campaigns. So there is a big opportunity here for the core portfolio. I think though, and Replit is a great example of this, as they expanded their core sales and service seats, they started to think about agents. And so the most recent opportunity we had to engage with them, they expanded into agents, and the place that I typically see clients start is in this employee agent use case. And so they want to build out capabilities to make just like we think about productivity, their own employees way more productive. I was with Bouygues, one of the largest European telcos in Paris, probably back in March, and I did a ride along. This is really fun. I grew up in Canada. I have a grade school Canadian French. They did the entire meeting in French, which severely tested my capabilities. But we did a ride along in their contact center. They have 6,000 people sitting in a contact center just outside of Paris. And they have implemented Agentforce for internal use cases for those 6,000 contact center employees to help make them more productive. They wanted to stop swivel-chairing, and they wanted to be able to help automate the process of sending text and e-mails, searching knowledge articles, first and foremost, 600 knowledge articles, but also doing outreach to customers who are inbounding the contact center. So I think it's a really powerful example of how customers can start with their internal use cases, but they don't stop there. I think once they understand the power of an agentic capability, they start to think about, gosh, I have to get that in front of my customer. And CrowdStrike is another really good example. I've worked with CrowdStrike for 5, 6 years. They started their agentic journey as a Sales and Service Cloud customer. We implemented a whole bunch of agents, but the 2 biggest agents were an ask legal agent and they follow MEDDPICC, so they do sort of like a qualification process, and they wanted to help boost their sort of pipeline quality. So we launched those 2 internal agents for them. Fast forward, they now have multiple agents in production, but they now run an external agent, their service agent, and that is containing 83% of the inbound requests that come to their contact center. You can't actually file a support ticket without touching Agentforce when you talk to CrowdStrike. So I think that's a really powerful example of the pattern. And as customers do this, and you heard Rohan talk about it, invariably, what they recognize is they need to harness their data to make sure that they're high grading those agents effectively so that they can serve their clients and they can serve their employees. And so FedEx, another client that I spend a ton of time with, they produce an astounding amount, a petabyte of operational data a day. They have 600 different, actually more than 600 different data sources, and they're using Data 360 to kind of harness all that data and bring it together. So their 2 largest data sources or the data lake environments are Azure and Databricks, but we bring all that data across those 600 sources into internal agents that are helping them do a bunch of things. But 2 of the most powerful things they do is look at dormant accounts and look at quotes that have been abandoned. That process took them 3 or 4 hours, but the ROI has been like tremendous off the chart. So I think that's a great example of kind of how the discipline around agents ultimately drives the motion towards we need better data. And as they build out more and more agents, they got to think about governance. So ADT is another customer that I spend a lot of time with. I'm giving you a lot of bright and shiny ones. So let me give you 1 that was a little difficult. Maybe 1 year ago, July 4, actually, I got a call from Marc saying that he'd heard from Fawad, who's the COO at ADT and that he wasn't happy with the Salesforce implementation. And so we obviously parachuted the team in. I think Mark Sullivan got the same phone call. He's had a history with Fawad at State Farm. We had to do a bunch of work to rebuild their data model because what they wanted was to be able to go way faster. They were swivel chairing across a bunch of applications. They wanted to implement agentic technologies and the apps or the orgs that they built just weren't ready to go do that. So we suitcased our team in, great news. They've sunset a bunch of these legacy applications. Everyone is standardized now on Service Cloud. So again, an example of number 1. But they're starting to think about or they're working with us deeply to partner on, okay, as I build out all these agents, how do I do 2 things really that are really important to them? How do I govern them? How do I understand where my agents are? How do I register them and kind of manage their behavior? And how do I manage their costs? They're super focused on managing their OpEx responsibly. And so I think this is a huge unlock for them as they continue their agentic journey. And then finally, and this is a little bit of a departure. It is a super new use case. I'm being sort of transparent and probably leaning in a little bit to an opportunity that we see from a monetization perspective that's really, really new. I love it for a lot of reasons. A, we didn't see this kind of business opportunity at, call it, the beginning of the second quarter. So in the last 90 days, this has shown up. And what we're seeing now is Slack emerge as the agentic operating system for agents. And as people launch their agents into Slack, what they need is a super high latency performance pipeline to make sure that they are fueling their agents with Slack data. And so this is something that I am really excited about. We have probably, the 3 deals that we've closed, a bunch more in the pipeline. Think about digital natives as really being the target area for this. And the thing I love is that it breaks the association. Miguel talked about seats. I feel really comfortable about the seat expansion opportunity that we have, but this breaks that association. I monitor every digital native and how they use our Slack system today. And I think 1 of the really exciting things as you look at it is that the correlation between API calls and usage doesn't map to headcount. You can have a really small but fast-growing company that is using, that is building agents like crazy, and they don't need to have a lot of employees. So I think that's really exciting. And then finally, we think about more ways to buy. And so you might say, no, actually, you guys already have a lot of ways to buy. But I think this is really important for us, right? We need to meet our customers where they're at and make sure that they've got very flexible, very low friction ways to onboard onto the platform. If you look at Q2, 7 of our 10 largest deals were AELAs. And I think as we start to launch Salesforce Commit more fully across the team, that's going to be a major driver for revenue acceleration for us as well. On a final note, and Patrick, I think, showed you this slide, the premium upgrade motion, I think, is going to be a huge acceleration vector for us. So for those who aren't familiar, I'll just show you, you saw this before, so let me orient. We haven't introduced a new value edition SKU in over a decade. And so in the last, call it, 8 quarters, the team has been focused on how do we go take that base of core CRM users and move them up the value chain. For every 1% of the base we move, it's $100 million. But more importantly, I think it's an opportunity to push all of our customers to fully leverage the rich capabilities that we're offering in Max Edition, including headless. So this is something that every single seller, every single person in the company has been enabled on. They understand this is the call to action. I have not walked into a room at Dreamforce and not asked people to upgrade. So if you guys are excited about upgrading your companies, you can scan the QR code and learn more about Max. But I think this is going to be really powerful, and we've oriented our partners and our company around driving that upgrade cycle and making sure that we work with customers to make them successful on it. I think just a final note, and you heard me talk about the 8 quarters. In the 8 quarters would, frankly, I would tell you like partial emphasis on this, we've built out $1 billion of AOV. And when you look at the sort of mechanics of it, in the upgrade pattern, you see about a 60% to 80% uplift depending on the customer and which edition they're coming from. But when you look at the individual sort of account level data, what we see is a 1.4x expansion of the ARR. I think this is really powerful. A, we're growing, and that's kind of the key message. But when customers renew, you guys can do the math and kind of math, the ARR is slightly less than the upgrade fees. They're remixing the portfolio. They're realigning to where they need to go and kind of fuel their growth or their transformation efforts, or the products that they need to fuel their growth and transformation efforts with. And so not only does it lead to top line revenue expansion for us, but I think it also mitigates any fear of attrition in the future. So I think this is really powerful, and we are going to put the whole force of the company from marketing and campaigns through enablement, plays, programs, incentives and calls to actions in the field. This is the execution point. Everyone is going to be focused on monetizing the value edition. So I'll end there on the monetization note. I think there's been a lot of discussion around Claudeforce at this conference. And I think 1 of the things I'm hungry for is to show you a little bit more about how we're using our own tools, Slack included, powered by Anthropic, but Slack to drive our business. Claire is going to help me here. I was wondering if you're in the back actually building a demo toggle into the Slackforce demo. But before, while she's setting up the demo, a couple of things. So we're obviously very focused on driving growth, increasing our sellers, improving their productivity. 1 of the ways that we do that is through sort of revamping from the ground up our sales processes. Forecasting is 1 of the processes that we know people spend a ton of time on. So we have totally rebuilt that function across the go-to-market organization since the beginning of Q3. We rolled out, and let me know, Claire, when you're all set. I'll do the Patrick riff, see how long I can do this for. We rolled it out at the beginning of Q3. And this is what we use to run the business now. You're not going to see live data, as you heard Miguel talk about, you're going to see dummy data. But otherwise, this is exactly what we use every Friday when I gather my leaders together to run the business. So I think it's really powerful. How many people have seen Slack before? How many are familiar with it? Okay. All right. A few more things to me. You guys weren't going to have seen Slack, you guys, you know what you're about to look at. Are we ready to go? Okay. Do you guys mind switching over to the demo? Okay. Great. For those of you who haven't seen it, this is a surface in Slack. I think this is 1 of the most powerful -- I use Slack to do 9 million things. I use Slack to prepare for today. I use Slack to prepare for customer meetings. I use Slack to collaborate with my team. I use Slack to run my global forecast process. I'm going to point out a few things here that I think are really important. So the first thing that you're looking at is a dashboard. Everyone in this room is familiar, but there's way more to it than a dashboard. I look at this every morning when I start my day, and again, this is not real data. I look at this every morning when I start today, my day, to kind of understand, all right, what's happening in the business? What's changed overnight? What should I spend my time on. There is no more precious commodity than time. I want to know where to orient it. And so this does a couple of things. It tells me where we are at any given point in time, and it tells me who are the drivers behind it? Who are the people who most materially represent the opportunity that we have in any given quarter. You can see that here. It also calls out, look, where might you have some exposure in the business? Where do you need to go pay attention? This is the what to watch. Miguel talked about the fact that our pipeline was up this year. And so I think that -- if I looked at this and saw that my coverage was high, I think that's basically driven by momentum, but I'm still going to go double check that to make sure that we're operating with the same quality of pipe that we typically do. And so this kind of gives me a good orientation for what's going on, what do I need to go focus on. And then I might dive more deeply into the forecast by leader. Well, actually, I'll show you. So this is just, again, dummy data, so it doesn't necessarily show you a pattern. But I would look at this to say, okay, what's happening? Is there a trend that I need to understand here from a macro perspective? Is Europe doing really well and EMEA is slowing down and Asia Pac has a problem. So I can look at this kind of global view. But I tend to spend most of my time in the leader view. And so I might pick as an example, Lenore Lang. Lenore Lang succeeded me in my previous role, so she runs tech media, telco and our data foundations business, our largest operating unit by revenue and people. I might pick this business and say, okay, tell me about Lenore's business and what's happening. And so I can see here that Lenore is calling $172 million. Slackbot thinks that's high. Now in this case, again, not real data, I'd be like, look, I think Lenore is going to crush it this quarter. So -- but it will give me a view of what Slackbot thinks and why it thinks this is going to happen. If you sort of -- this is obviously instructive and helpful. But if you think about it, I have teams of people who spend a lot of time. Forecasting is really important. You depend on me to give you an accurate view of what we're going to do. My team depends on that, too. And so we spend a lot of time making sure that we deeply understand the numbers, but I think Slack can help us do that better and faster. And so I can actually dive into for each of my leaders and I'm going to stick with Lenore here [indiscernible]. So I'll tell you this is a real scenario. It's not -- it didn't happen with Lenore. It happened with somebody else, and it wasn't this deal that you see here, but it was a real deal that happened last week on the forecast call. And so I can go through and I can understand her right. This is Lenore's -- like the overview of Lenore's business, here are her top 3 deals. And I might see that she's got a big deal at a tech company that is at risk. And so we'll talk about that on the call, and she may give me some feedback about why that deal is struggling. And one of the things that I want the team to do is get Marc involved right now. There isn't a deal in tech that shouldn't have Marc connected to the CEO. And they are not eager to do that. They want to be perfect. They want everything buttoned down. I want them to do it fast. I want Marc to reach out to the CEO, so I know that if something is wrong -- when he executes that reach out, I'm either going to get good news or bad news, but I'm going to get it fast and I prefer it fast. And so I'm on this team's case right now to go and get that outreach. But I can also look at, hey, what's happening in this deal right now? Are they doing SICs at Dreamforce? What are the exact connects that they have happening? And then I can actually go into it and kind of say, "Hey, team, let's get this down or let's get this text message set up for Marc and let's have them reach out to the account really quickly." So that's just kind of a very quick run-through of how I'm using Slack powered by Claude, powered by Anthropic, excuse me, to run the business. And a final note, this is Miguel's favorite example, so I promised him I would do it. I will be in London later this month. And one of the things as we sort of make these trips into the market that we want to make sure we're doing is spend time with customers. So I can ask Slack, what are the top deals in London in Q3. And it will go through and it will search. And you'll notice it's going to come back. I didn't tell it how many, but it's going to come back to me with, I think, 7 deals when I checked it this morning. That's going to say, hey, here are the top 7 deals in London. And so further, I can say, look, I want you to send a Slack to every one of the opportunity owners for each of these big deals, copy their entire management team, tell them I'm going to be in town and ask them what I can do to help get this closed in Q3. And so you'll see just down at the bottom here, it comes back and says, okay, do you want me to send it as a DM? Do you want me to post it in the channel? Do you want me to curate it so that it's fit for purpose for that individual team? I say, yes, that's what I want you to do. So super quick example, but I think it's indicative of the transformation that we're driving as Customer Zero of this technology. And I think the opportunity that we have to go help all of our customers transform. So with that, I'm going to pass it back to Miguel, who's going to share 2 real-life examples of customers we're transforming right now.

From the Salesforce executives. And now we're going to hear from the real heroes that are getting this incredible stack and driving enterprise value in their companies. So I'm going to start with Adecco, okay? And by the way, this has been a little bit impromptu because we saw Adecco many times. We talked about Adecco. The CEO was with us at the keynote, great marketing messages. And then right after the keynote, I met with the working team, and they went through all the progress. I typically meet with them every -- about 3, 4 months. And I'm like, wow, this is even better than I anticipated. And I had a good conversation with my friend for many years, Pierre Matuchet in the IT area, Senior Vice President of IT now Pierre join me. I think they're going to put a chart for us. Give a round of applause to Pierre Matuchet, Adecco Group. There. And honestly, the reason he was -- his eyes were like on fire when he was telling me the progress that he was driving with Salesforce and I'm like, you know what, what are you doing tomorrow night -- tomorrow afternoon, I'll say, tomorrow night also I invited him today to the concert. But what are you doing tomorrow afternoon because I'm meeting a few of my friends, and I would like you to be with me and tell your story. It's funny because Denis, the CEO, arrived at the meeting at the end, and I said, Denis, I need to get your approval to get Pierre to be with me at this meeting, and he said, yes, so here we are. So thank you so much. Let's take a seat. So Adecco, I don't know if you -- hopefully, you know Adecco, but it's one of the biggest, if not the biggest staffing global company in the world in HR services, about 34,000 people in 62 countries, about $25 billion of revenue. And we are so grateful that you're here, to your CEO, to the partnership that we've had over many years. And the first question, if you remember last year, I always ask the question similarly, what was our relationship between Agentforce? Sorry, before Agentforce.

Before Agentforce.

Yes.

We were using your CRM for 15, 10 years. It was good. Nothing spectacular. We have a relationship.

Okay, how do you like the [ interview ].

Not a strategic relationship, not a strategic partnership. We were a bit stuck from our side with 42 instances of Salesforce, so not easy to manage, maintenance cost and so on. We need to build some complex mechanisms to have a view of everything. And we get more and more requests from global customers like Amazon, like Siemens to have the same process everywhere in the world to be able to give them a global reporting. And we were in Dreamforce and working with your team, we have decided to implement the [ Data Cloud ] and this was really for us the foundation of the change because in less than 3 months, we have been able to put all our data in 1 layer and to be able to run reports.

So that was more or less, what, 3, 4 years ago?

2.5.

2.5. So they were stalled a little bit of a boring usage of Salesforce. They were a pretty nice-sized company. They were -- sorry, a customer. They were in the 8-digit business. It was good, 42 instances of Salesforce, but it was a mess. It was a little bit of a mess. And we were not growing with them. In fact, we were having conversations of even reducing the fear attrition that some customers want to use less. That's where we were. And then you implemented Data Cloud and you killed it. It was...

Data Cloud 3 months bring us new vision of our business, capacity to implement new processes to better serve our largest customer, which represents 70% of our turnover. So very important moment.

Okay. So then that sort of -- the way I looked at it is that really appeals to you. You have now a global view of all your orgs. It sort of stopped the bleeding, okay? But then you came to the next Dreamforce and then you saw Agentforce. And I really would love for you to tell this audience with the same passion that you told me yesterday, what are you doing with Agentforce? What agents are you deploying, the volumes -- the volumes that you are deploying them? And most importantly, what is the business value that you are driving?

Okay. I may be a bit wrong. But I will do it in English. Even if Alexa asked me to do it in French, I will keep English. We have started in U.K. in May '25, 6 weeks to develop our first agent. It was a prescreening agent. Today, we have 7 agents running in 12 countries, representing 60% of our top line. In U.K., our first agent was what we call a screening agent. It means we need to prequalify candidates before a real interview with our recruiters. Year-to-date, we had 2.7 million conversation with these agents worldwide. And for us, the key business impact is a reduction of time to present the candidate to our customer. We reduced it by 40% by the usage of an agent. And we were with Parker a few months ago in a branch in the suburb of Paris, and Parker kindly visited the branch and had a discussion with our recruiters. And one of our lady as a recruiter talked to Parker. This agent has changed my life because at 10 to 6 in the evening before leaving the office, I launched all my agents. They do the job during the night. And when I come back in the morning, instead of having to call 200 candidates, the agent has already prequalified 20 candidates. And one of the key advantage of this agent is that it is running 7 days a week, 24 hours a day. When our branches are closed, we are still doing business, prequalifying candidates.

In fact, a big percentage of that work for the agent is in after hours, right? Okay. So that's the prescreening agent.

We have also the interviewer agent, which makes real interview with voice.

You have an agent, an AI agent that interviews candidate?

Yes. Including voice. We are using Voiceforce, last name of it. 20% of our calls are currently done by voice, and we do 20,000 interviews per week with our interviewer agents. And this has a direct impact in terms of business. We increased what we call in the hiring area fill rate. It means our capacity to answer to the orders of our customer by 10%, okay, by using this...

You placed 10% more candidates?

For the orders when we put in...

With the combined human and agentic workforce that you have put [indiscernible].