Salesforce, Inc. ($CRM)

Good morning, and thank you for joining us. I'm Valmik Desai, Vice President of Investor Relations. This session marks the fifth in our series of quarterly post-earnings webinars aimed at providing you all with a deep dive on our latest product innovation, super excited for today's session. We will deep dive on Headless 360 and Slackbot. And as you heard on our earnings call earlier this week for Mark and the team, we're incredibly excited about the opportunity ahead with Headless 360 bringing together humans, agents, headless platforms. So customers can really use Salesforce with any coding agent across any surface. We're also seeing incredibly strong momentum with Slack and with Slackbot, which we're super excited about, personally my favorite product right now to use, I wake up to it every day. It's rapidly helping our customers become agentic enterprises. Some of our comments today may contain forward-looking statements that are subject to risks, uncertainties and assumptions, which could change should any of these risks materialize or should our assumptions prove to be incorrect, actual company results or outcomes could differ materially from these forward-looking statements. A description of these risks, uncertainties and assumptions and other factors that could affect our financial results or outcomes is included in our SEC filings, including our most recent report on Forms 10-K, 10-Q and any other SEC filings. Except as required by law, we do not undertake any responsibility to update these forward-looking statements. All right. With that out of the way, I'm super excited to have Joe Inzerillo, here with us. He leads our enterprise and AI technology business. also has the coolest background in the business. So thank you for joining us, Joe. We also have Rob Simon, who runs Slack here. We're going to start with the brief presentation, with Joe kicking us off and then we'll jump straight into your questions. Now I know this is definitely not a shy group, but please do submit your questions in the chat. And with that, over to you, Joe.

Great. Thanks, Val. So back at TDX, which seems like a lifetime ago, but it was only, what, 6, 8 weeks ago, we announced Headless 360 and the response has just been unprecedented. People are super excited about it. But there's been a lot of questions about what it is. And so I thought we'd spend a little bit of time talking about that. So we go on to the next slide. The thing about the whole concept of Headless 360 is really leveraging the stuff that we already have. So if you think about things like our data layer, the layer of context, the layer of work, where all of our apps at the genic layer between Agentforce, Slack, Tableau that sits at that level. And we have an engagement layer up top, where you could see our new Agentforce Coworker, Slackbot, some of our customer apps, custom apps and things like that, that sit up there. But Headless 360 is really meant to be that glue that gives the customers and actually ourselves internally huge flexibility in how we want to represent all of that goodness that sits in the stack below it. And so previously, at Salesforce, we and our Salesforce admins, we're really talking to the computer to provide instructions in advance to create interface for humans to use. We're now in a world with agentics where we could actually have the agents create the UI on the fly. And Headless 360 is the way that the agents know how to use the Salesforce platform and leverage that 27 years' worth of goodness and history that's baked into it. And so it's really about liberating folks and the way that they want to do it, but liberating them through agentics. And so that's the part that's so exciting is that unlock -- to get the wheel spending to get these new types of applications that people have never thought to build working and when it works, it's like magic. So we're going to dive into that a little bit. So generally, in the Salesforce ecosystem, you have a user who is asking for something directly. And so they may be talking, let's say, Claude. And the agent knows how to then go and hit MCP end points inside the Salesforce ecosystem to ask questions. And it knows what questions to ask because these aren't just APIs, they are MCPs. So they describe semantically what they're capable of doing. And that allows humans and agents working together to drive more value. And obviously, this works all over the place. It works in Slackbot under the hood. That's the way Slackbot's working right now today. It works in Codex. It works in Claude, works in all the major AI platforms. but it really is this work where you want to work with the tool you want to use, but still leverage the capabilities of the Salesforce platform. And what we've seen is when people adopt this pattern, they consume more of the Salesforce platform than they ever have because they're liberated to do so in the way that they want to do it and they can use the agent to really double down on the capability set. Go to the next slide. So I think one of the things that I love about this is just seeing the utilization. So you could see [indiscernible] I think we skipped 1 there. I like the 1 forward. Here we go. You can see Saster on the next slide here. And I love that example with them because they actually put out a tweet that essentially said, I'm paraphrasing, they've reduce their Salesforce licenses, but they've actually been paying more for Salesforce and it's worth every penny because they were able to use some of the early capabilities of the Salesforce MCP, which is really where Salesforce Headless is going to actually build their own custom interfaces. And that allowed them to just leverage more of the platform than you've ever been able to do before. And that's the kind of creativity that we've seen happening in the community and Headless 360 is really speaking to that. And so the Headless brings the humans and the agents together in Slack. And so we use this internally when we're building Salesforce experiences in Slack, they come preintegrated. So if you think about, for example, our Slack CRM is really Salesforce power sitting behind the scenes, the trust layer, the application layer, the context, everything that you've always loved on this. And at the same time, it's actually in Slack. Looking native. Nobody would ever know that Salesforce is behind the scenes because the interface is the Headless -- the Headless interface is Slack. And so no better way to actually show you that. So I'm going to bring on Rob, who's actually going to take you through a couple of slides and then show you what this looks like in practice. So with that, I'll hand it over to Rob.

Thank you, Joe. And as you heard on the earnings call, obviously, we're having a ton of traction and success with Slack in the market right now. We're very excited to meet this moment because we think it truly is where AI is going to work moving forward. So if you go to the next slide, 1 of the things we're seeing is, obviously, there's a ton of money that is being poured into AI as our customers and a myriad of companies around the world these days, but there is a missing piece. And outside of agentic coding, you obviously like a true translation of that spend directly to employee productivity. So if you go forward, we think what's missing is something that actually connects all your people, you look at a lot of the traditional communication vehicles, whether they be e-mail or tax or whatever it might be, they're siloed in their nature. And you look at all of the agents that are out there today, like typically, they are a single player, they are deployed into a web-based interface or embedded in a SaaS op, they don't necessarily allow people to talk to those agents and then share those results back with other people. And then every platform ultimately isn't necessarily talking to each other. So what we think needs to exist, it's something that connects all of these things, but does so very importantly, into a multiplayer run time. So if you go to the next slide, we call that a work operating system for this AI age. And if you go to the next slide, that is how we are seeing people start to use Slack within our customer base, where all of their employees, their partners, their contractors, even their customers are working with them in Slack. They've deployed all of their agents from Agentforce and other third parties like [indiscernible], Claude or OpenAI directly into Slack. And then they have all of their other platforms connected into Slack as well, which can then be used by their agents and by their people. If you go to the next slide, [indiscernible] actually play out at some of the biggest companies in the world. But what's really exciting is the most innovative AI companies in the world have gone all in on Slack. And they not only use Slack to run their business, but they're actually building their products to be used in Slack because it is that multiplayer AI run time. And a couple of great examples are actually Anthropic and Shopify. So if you go to the next Slide, you may have seen on Lenny's podcast on Cat here, who's the Head of Product for Cowork and Cloud code at Anthropic said, Slack is basically the core operating system of their company. What's really interesting about Entropic is every single one of their employees has a public thoughts channel. We do this within Slack. We've seen Versal do it as well. A number of other players do it, where every single employee actually just think out loud in Slack. And the default is work happens in public. And so if you think about what that actually does is that raises the collective knowledge of your employees, but it also raises the collective knowledge of the agents that you deploy into Slack. And if you go to the next slide, you'll see at Shopify, what Toby says here, Toby and Shopify built an agent called River. So they want all of their agent coding actually be done in public in a multiplayer fashion. So what they've done is they've actually deployed it into Slack. And you can only actually use it in public. And what's fascinating about this is what Toby says here, "So the risk of today is that AI does the work, and we as humans don't learn from it as a group." And I think that's the downfall of kind of the single player scenario. But when people work together with their agents in public, the opposite really happens, which is the best pattern spread and the best knowledge spreads. And that's what we're actually seeing in Slack with third-party agents and with Slackbot. And so without further ado, I'm going to show you Slackbot. So I'm actually going to walk you through a series of things that Slackbot can do. And let me share here. Can somebody give me a verbal that you can see.

Yes, looks right.

Cool. All right. So this is Slack. For those of you that don't use Slack, I'll do a little bit of orientation on the left-hand side here is our side bar in the middle of what I have open is a channel, and this is a Salesforce account channel. So you see at the top part [indiscernible] account. This obviously has a message stream on how account managers keep the entire company aligned around that customer success. But you can see within here, I can see the details of that account from Salesforce right here in Slack. So first great example, I think, of Headless is this is all dynamically rendered interfaces based upon the metadata that exists in Salesforce. But what I'm going to do is open my personal agent, which is Slackbot on the right-hand side here. So if you look over here in this flexion, what I'm going to do is I'm actually a lot of salespeople are out on the road, and they may take chicken scratch nodes, either hand written or in their note-taking tool of choice and may eventually need to put those into CRM. So what I'm going to do is I'm going to take the chicken scratch notes that I wrote on a cocktail napkin and about this contract, Kevin Marshall, and I'm going to ask Slackbot just log a call, for that contact with for the B2B commerce opportunity. And so what's interesting here is you're going to see Slackbot. One could actually read this image, which I think is pretty straightforward these days. Two, has the context of the account because it knows that from the screen that we have the account channel open. So it knows what account it's related to in Salesforce. And three, it's actually going to Headless interrogate Salesforce, find that contact, take these notes from this image and actually write it into a task associated with that contact, that account and that opportunity in Salesforce. So we always put the human in the loop. So in this case, this is giving me a notification that it's going to create a record in Salesforce. I'm going to go ahead and say, "Great, go ahead and do that." So I think that's a first quick example of using Salesforce Headless. Again, think of Slackbot as kind of the first best customer for Headless and using Salesforce Headless to meet users where they are and provide an exceptional and conversational experience. Now what I'm going to do is I'm actually going to ask Slackbot what sales plays and customer stories would be relevant to leverage for this particular opportunity? Now what you're going to see here which is fascinating, is another example of Headless. So this is actually fairly specific and strategic questions being asked here that requires some domain context specific to the sales organization, their content, their coffee, et cetera. So if you look down here, you can see Slackbot's actually asking the sales agent. So this is Slackbot calling and orchestrating directly an Agentforce agent. And it's doing so through an MCP tool call. I could open this and I can see what is asking the sales agent. You can see it's given this specific prompt that is given to the sales agent to go off and find those sales plays and customer stories. And again, it's doing it within the context of this particular account, the opportunity that I mentioned and that contact. And so here, it's coming back with customer stories. It's denoted those that are shareable with Kevin externally. Those that are internal that I can't share with the customer. And so this is extremely helpful and saves a ton of time for our sales rep. So now what I'm going to do is I'm actually going to ask it to translate all of these insights and these customer stories into a Word doc that I can share with the customer. So what's really important is all the Slack users typically have their calendar, their productivity suite, whether that's Microsoft or Google already connected into Slack. And Slackbot because of that can then act in those systems on that user's behalf. So one, this has created or log the call with Salesforce for me; two, it's found relevant customer stories; three, it's actually going to write a document. I can go off and share with my customer, but it's going to do it. in the productivity tool of my particular choice. So my company, in this case, uses Microsoft OneDrive and Word. So this is actually going to go create a Word document. So it's -- I'll actually in fact, that this is real, it's going to -- you can see it said that it said the Word doc, but it didn't actually send me the link. So here is that Word document. So when I open that, I can see what it has generated. All right, now I want to keep the team going back to the multiplayer angle, I want to keep the team up-to-date what's going on. everything here has been kind of a single player. It's been me working with Slackbot. So now what I'm going to say is, has like, well, go ahead and drop the message to the account team that summarizes the customer stories that I mentioned and our next steps for acne. So when I do that, Sockbot is going to compose that here, but can actually send a message on my behalf back into this channel, and that is going to take what has been a purely single player experience and make it multiplayer, which is immensely powerful. And I think one of the things that we were able to do in Slack that you can't do it in a single player environment. Now I'm going to hop over and I'm going to show a few more things that we're very excited about that are a little bit more involved. And so the first thing I'm going to do -- sorry, I actually wasn't ready for this, I apologize. Let me open up like Google Drive real quick. Can you give me just a second. I'm going to open.

Someone at home is pausing every single frame and analyzing what Rob is doing.

Exactly I'm going to give me just 1 second while I stop sharing here, and I'm going to pull up the spreadsheet real quick that I need for this demo. So if you give me a hot second.

I think this also emphasizes our commitment to transparency and showing it actually happening as it happens, that Salesforce.

Absolutely. So let me share again here. All right. Can you see now?

Yes.

Cool. So what I'm going to do is I'm going to go ahead and get this problem going because this one is actually going to write some code. So over here on the right, I'm going to basically ask Slackbot to use the S&B Q2 forecast tracker spreadsheet and build the 6-page interactive report for me. And it's a pretty detailed set of instructions. So I'm going to kick that off. This actually actually take a few minutes because 1 of the things that's interesting about it when I show you the spreadsheet that it's going to find and read, it's actually very computational and data intensive. And so it figures out that it needs to use Python and it's actually going to write Python and execute that Python to do the analysis. But I'm going to hop over and actually show you that particular spreadsheet. So -- but I think the first thing to tell you is that I didn't give it a link to the spreadsheet. I just said, "Hey, can you look at that Q2 forecasts that are in a spreadsheet." And it goes off into this case, Google Drive and it finds it. So what I'm going to do is hop over to Chrome, and I'm going to show you the spreadsheet. So this is kind of dummy data, but it's an actual spreadsheet used by a sales team here at Salesforce. And so I'm going to click through really quick and show you like just the degree of complexity that's in the spreadsheet, the number of tabs, the amount of data, this is just not something that a human being can like genuinely look at and be able to make the most sense of. But Slackbot can do that for me. So it's off executing against that right now. What I'm going to do is I'm going to go pull up a prior version that I executed and show this to you. So when I hop up here, you'll see what it comes back with. So it came back with, hey, I've actually created that for you. But when I open up the details of what it's done, let me actually just show it too, which is in is a better example. This has actually created a dynamic user interface on the fly that didn't exist before. And this is, I think, the power of Slackbot, but also the power of the Salesforce Headless 360. So this is query Salesforce. It's red that spreadsheet. It's joint the information between the spreadsheet and Salesforce. It's taken my prompt and then dynamically generated this interface. So I can see a risk matrix of my deals where I have gaps to commit. You can see where we have velocity and coverage issues. I can do somewhat if in scenario modeling, if I wanted to, to see what might happen and overall health signals. This is something that probably would have taken a human being a while to take care of before that we're now able to do in a really quick fashion. So I'm going to hop back over. I'm going to close out of this, and I'm going to show you a different prompt. So when I hop back in here, you'll see the second prompt I gave it is actually saying, okay, we have a QBR later this week. I want you to leverage the presentation design skill and put together an executive summary PowerPoint presentation of the team's performance. So based upon this dashboard that's happening here and what -- this is important to highlight, this actually leverages our skills framework, which is new. We just turned on for customers this week. So Slackbot skills are a way for end users to actually define kind of micro agents or tests or processes, if you will and then share them with their team. We've got a series that have been built by us. We've got those that have been built internally at Salesforce. We have some that are recommended based on social proof in sharing. But these are really, really powerful. And so when I hop in back in here, you will see that what Slackbot used is that presentation design skill, which I'll open up here. And this is through natural language gives us some instructions on how to actually build a presentation, and it spit out this PowerPoint presentation, which I'll open here. So as I scroll through this PowerPoint presentation, you'll see this is a translation of that spreadsheet, what's going on in Salesforce in that dashboard that was created into something that's consumable in a meeting that we can share. So I think this is, again, something that people spend a ton of time actually going through the production of that we can get human beings out of the production work and back into that kind of creative and critical thinking. So I will stop -- actually, one last thing I'm going to show you on the multiplayer element. I want to show you a real channel in Salesforce called [indiscernible] Slackbot. This gives you an idea of the kind of social nature and multiplayer nature of AI and Slack. This is not a channel we created. The user has created it themselves. It's called How I Slackbot. We now have 4,600 employees that have opted into this. And every single day, these employees are sharing skills that they have built. And any employee can hop in here and add those skills to their Slackbot and releverage them. And it's just become this incredibly vibrant AI community. It is ultimately up to the AI influence of the Salesforce employee base as a whole. And with that, I will stop.

Awesome. Thank you, Rob. Thank you, Joe. That was awesome. We have a ton of questions coming in. I'm going to start with one here from Gregg Moskowitz at Mizuho. It's focused on kind of the opportunity that we have with Headless. And he says, "while Salesforce, Agentforce disclosures have been very encouraging from an adoption standpoint, it hasn't yet been visible from a revenue standpoint. Given that Headless 360 opens up, the Salesforce platform to external AI agents and coding tools via MCP. Do you think this could be the mechanism that puts Salesforce more directly in the token path?" So maybe, Joe, you can start with that one.

Yes. Look, thank you. It's a great question. Thanks for the question. Absolutely. I mean I think Headless 360 is what the market has been asking for us to do with our Agentic technology. They saw what the capabilities were with Agentforce and what it could do internally. And I think it's one of those things that have clicked for us and a click for the market at around the same time that like Well, wouldn't it be great if all the coding agents could take advantage of everything that Salesforce had to offer. And I think the Headless 360 strategy is really about empowering the ecosystem of agents. So there's no question that there's 2 things that we have real high confidence and we have data that gives us this high confidence, which is this is something that people want. And the people that use it consume more. So it gives them more Slackbot because it takes the things that have been happening in Salesforce historically at human scale and changes it into agent scale. And that agent scale is the humans and the agents working together but the agent scale is just much, much, much larger and freeing it up so that you can just bring the tool of your choice to bear on Salesforce. In some cases, that's the completely vertically integrated Salesforce stack. In other cases, it may be components like Cowork or some other LLM coding tool or operations tool that you could just basically attach to Salesforce and go for it. So I absolutely think that this is 1 of those things where every MCP interface that we've turned on at Salesforce immediately gets lit up with traffic. And we see that traffic just continuing to grow because the demand is there because what we have in the core platform is just so valuable.

Awesome. I do think I'll add 1 thing. The way we're looking at the long-term opportunity here is exactly as Joe described, which is we feel that more people are accessing the platform, the more they're able to actually surface those insights wherever they want to do that work is super valuable for us, right? And it's super valuable for our customers. So when we start to see more usage patterns at more kind of use cases that really start to proliferate, that's where we're going to work with our customers, work with our ecosystem and understand what is the right way to kind of capture value on both sides of the equation for customers. Salesforce. So certainly more to come there, but it's totally the right question to be focused on. Our next question here comes from Ran Yagnik, my good friend and our investor at [indiscernible] Capital Management. "Is there any initiative to have the UI of Slack itself from purely text and image-based presentation interaction to a more dynamic richer interface. For example, having something like Tableau more closely integrated into Slack, so you can see ad hoc visualization." And I think Rob showed some of this, but maybe, Rob, you can touch on the future road map here as well.

Absolutely. We did show a few things there, which was we showed specifically the on-demand creation of like a dashboard or a dynamic surface. But I think what's most exciting for Slackbot is 2 things. One is MCP and the MCP UI standard that's emerging. So basically, any MCP tool that Slackbot can call can actually generate MCPUI that can be then rerendered in Slackbot, which inherently makes it more dynamic. But also coming back to Headless 360 is the Salesforce Headless experience layer. So there's so much rich metadata in the Salesforce platform for Salesforce customers like they've expressed the way they work and do business. And in Tableau and the way that they've built their layouts in Lightning. And that could all be transposed basically and shown through MCP UI and Slackbots. So yes, Slack, think by just the nature of consuming MCP and MCPU with the Headless experience layer from Headless 360 is going to become much more rich visually.

Yes. And maybe just to add on, if people want to see a glimpse of what this looks like, would you turn that dial to maybe not 11 yet, but at least a good solid 8 is take a look at Slack CRM and when you look at, you are doing deep work that historically, you would have gone to a rich lightning interface to do, but you're now able to do it in Slack. But to Rob's point, more importantly, the agent is also able to do it with you in Slack and so you see that you could do the deep work on a very rich surface, but you could also have the agent to do that work on your behalf and you can collaborate multiplayer with the agent. And I think that's a big thing. Multiplayer is not just a human thing. It's the humans and the agents collaborating together in a multiplayer scenario.

So we're going to stay on the Slack and Slackbot subject here with a question from Keith Bachman from BMO. How does Slack compare and compete against Microsoft CoPilot and I think it means Slackbot there. It seems like a lot of overlap with Microsoft and yet Microsoft has underlying personal productivity tools. So maybe, Rob, you can start there.

Yes. I mean the way I think -- I'm not going to sit here and necessarily talk about or competitor's product or a partner's product. But what I would like to say is like what makes Slackbot Magical is the work that we have done to actually -- we call it being a good host within Slack. But the work that we've done to kind of tilt the umbrella and help users on board and help with the AI fluency. So one, I think that is something that you don't see in a lot of like the single player instances of AI tools that are out there today. Two, the other thing that I would say is like there's just a tremendous system of context in Slack. So if you look at the way people actually use Slack and like the bias towards working in public channels and having these longitudinal channels that the membership changes over time. There's just a significant amount of context where it can like immediately. So I think there's no cold start problem they can start writing like you. It can start understanding your company's objectives and priorities almost immediately. And then another like ease-of-use thing, frankly, is that it inherits all of the connections you already have into your slack, right? And so 70% of our customers are Microsoft shops. And so I can use Slackbot to book and beating through outlook now? I can use Slackbot to write a Word doc as we saw or create a PowerPoint presentation for me. So it's just like -- it's incredibly fast in the sense that it helps you get up to speed on AI. It helps you share AI with your teammates. And it actually helps you connect to other systems and operate in those systems. I think faster than anything else we've seen.

Yes, absolutely. And just maybe to add to what Rob is saying. I think when we think about the fact that we live in an incredibly heterogeneous environment right now. There are agents everywhere. There's other systems everywhere. I think the big thing is it's the fit and finish and the way that people use the tools that's super important. So I think the sort of software world was really fixated on features, what is the Harvey ball or checkbox chart look like is to foot features you have. And those are important, don't get me wrong. But the big thing about Slackbot is how organic it is and how you use it. You don't have to at-mention agents in the stream of consciousness that you're talking to Slackbot with it, knows how to use these tools and orchestrate them. And I think that's on e of the reasons as a company, we've really been talking about AW user or these agentic work units because that's really the output side of it. If people are driving these AWUs, we know they're actually getting work done with the tool. And we feel like, again, we need to make it heterogeneous, we need to work with everybody. We really value our partners, but we also believe that it's the best place to get worked on is in Slack.

Great. We have another 1 here that kind of goes into both Slackbot and Headless 360 from Allan Verkhovski from BTIG. "If customers are spending more on Salesforce for leveraging your MCP, how does that conversation go in terms of how much of an uplift they're willing to pay for Headless 360? And what is your right to win? And why should people gravitate to Slackbot?"

Yes. Well, I think those are 2 separate questions. I'll take the first. Maybe I'll give Rob the Slackbot question. But I think as far as Headlands 360, we have not unveiled our total commercial plan for this. And I think the reason is because we really want to work with our customers and make sure that we're getting this right. Like this is an important adjunct to what we've been doing historically, and there's a lot of complexity to it. And so we don't want to do it too quickly and just kind of shock on something out there. We really want to talk to the customers to understand how they see the value, how we can help with it. But the thing that we believe is there is value. So when we talk about what it might charge or if a customer is going to pay us more or less, the whole point is if the customer is using the platform more, there's a way to come up with that's totally fair and beneficial to both parties to monetize that extra value that's created. And so I think sometimes in the customer service use case, for example, a lot of people talk about savings, but we're really fixated on growth. So when we think about things like qualified and sales agents and the things that we're putting out now agentically using the platform, it's increasing what a company can do from a revenue standpoint. It's increasing the stickiness of the interactions with that company, and we believe that, that value creation by using these new tools, there's going to be a fair way to split it, but we just don't want to be too overly prescriptive without really socializing it with like our CIO advisory boards and other folks that are out there to make sure that we're really hitting the mark and making sure that we stick the landing on, again, appropriately monetizing that value. But we're confident the value is there, and we're confident it will be very accretive. I don't know, Rob, if you want to take the Slackbot portion of that, though, because I think that's a great question as well.

Yes, absolutely. The first thing I would say is as far as right to win, the most important thing for us is to actually be the home of all AI for employees within enterprises that is used in a multiplayer way. So period. It's actually not Slackbot. The most important thing for us is that Salesforce, all of your other AI tools show up in Slack exceptionally and can be used in this multiplayer run time in a way that like makes a ton of sense for humans and agents to to collaborate. And then after that, I think Slackbot is our most important priority. And as far as our right to win, I'm not going to say it's our right to win necessarily a particular category of software, but we do typically have a right to win your use. And the way that we do that through Slack is obsessing over the user experience, and obsessing over how teams work together, right? And I think that's what you'll see with Slackbot that when you see internally at Salesforce, we've got 70,000 weekly active users of it, 93% week-over-week retention. People are just kind of, as Val said, kind of obsessed with it because it so naturally fits into their workflow. So there's 2 aspects of that. One is proximity. It's right there in Slack, we're already spending a couple of hours a day to work with your colleagues. Two is the context. So it knows what you're working on and it knows what's important for your company. And so I'd just reiterate, the most important thing for us is for the ecosystem to be successful in Slack and then making AI extremely easy to use and share through Slackbot.

Awesome. Great. We have another one here that's along the same thread, along the similar questions around headless economic impact. So Kirk Materne from Evercore asks, can you talk about Headless 360 from an economic perspective, it would seem there are different revenue and op margin implications for buying an agent versus building an agent inside Salesforce versus building an agent elsewhere and accessing a CRM agent via an MCP server, any way to think about the variability across these different scenarios. So maybe I'll start with just a high-level view, and then I would love on Rob to chime in on kind of their use cases specifically on a Headless use case that you're building on another platform as you heard us say, we're working through that. More to come there on the commercial model. I think where we start to see a differentiation specifically on comparing an Agentforce built agent, which is purpose-built understands the context understands what your task you're trying to accomplish and is deeply integrated in the full set of CRM data that we have, that drives a much faster kind of time to value proposition for customers. It's a higher accuracy outcome for a lot of these customers as well when they're able to use things like agent for script determinism in the flow of that workflow and that agent where they're actually able to not just leverage optimization of which model makes sense for the right step of this process or this task how do I make sure I'm actually kind of programming in the standard workflows, the standard if this is the topic that they're asking about, this is the way we should actually react here, and that actually helps you drive leverage on the cost side of how much you're pinging an LLM to get that answer done. So a lot of work our team has been doing from an R&D perspective. There's a lot of work that we even have done from an implementation and deployment perspective, Agentforce testing center. You heard from one of our customers at UCLA Health that talked a lot about the importance of having that testing center to have a high degree of confidence before they went live with a patient-facing agent, of course, in a world where their goal is to make the patient experience better, to triage faster on those health care requirements that are higher impact and higher priority versus ones that are lower priority that they can just surface with a quick answer, that takes a lot of confidence for a firm like that to actually go live with an agent facing their patients. So the ability to have that level of confidence, all the work we've done around the platform, we think there's a huge advantage there. But Joe, I'd love for you to come in and kind of chime with your perspective of what you've heard from customers as well.

Yes. I mean I talk to customers every day. And we have some customers that, again, love the vertically integrated stack. So for them, Agentforce is just natural. They have that skill set their trailblazers know how to use Salesforce. They know what they want to get out of the Salesforce. Agentforce unlocks these new potential applications of it. And so they want to do that. And I also talk to customers that say, "Hey, I have this investment in my coding agent or tools or whatever, and I would like to do it this way." I think Headless 360 speaks to both of them. And I think the way that I think about it is like cars, right? I'd say the vast majority of us buy a car and then we drive it around. And maybe we hang an air fresher, we certainly play our own music in the car. We do a bunch of things to make that car ourselves, but we don't physically change the car. There are people that mod cars, right? So they change the suspension, they switch out the rims, they do all those sorts of things, and that's the way they want to drive their car. I think with us, we give you a couple of different series of cars that are fit for purpose for those particular industries or size companies, small business, et cetera. So we give you a bunch of different choices in cars, and you can take the vertically integrated car and not touch it, just customize it in the ways that we talked about or you can go and full mod that car and you can give a component of it and say, well, look, I want the Salesforce engine transmission and tires, but I'm going to put my own chrome on it. I'm going to do this and that. The other thing to what Val was talking about is if you want to take that Salesforce version of it and you want to do really great things with it, it also comes with a pit crew, right, and a mechanic. And that's really testing center and all those sorts of things. And so when you think about it, we're really trying to provide choice, we would like to win more than our fair share of folks that want to use the vertically integrated stack because it just delivers so much value so much faster. But we also want a rich environment of people taking components of it that they think are going to make their particular application and my analogy, their car better for them. and we're trying to do both of those things. And Headless 360 is the mechanism by which we're actually able to do both of those at the same time without having randomly different development efforts to try to support that.

Awesome. I think we answered Arjun's from William Blair's question already. So I'm going to skip ahead as he was asking about the vertically integrated stack and how we think about third-party agents versus Agentforce. James [indiscernible] at UBS is asking about data sharing. It's an area that's been an accelerant for Snowflake consumption, can CRM attack that data sharing use case between vendors, partners, customers to drive consumption and stickiness by enhancing the focus on Slack external partner connections or Slack dashboard sharing. And I know we have a massive Slack ecosystem. So maybe, Rob, you can chime in on what you're seeing there.

I think we have a tremendous opportunity to participate in that. And I think what's been really exciting, one, what's been really exciting is the adoption of the MCP standard. I think this 1 creating an interoperability between agents, but also identifying the access patterns to APIs. It has pushed us and made us change our APIs across Slack and Salesforce with these new genetic access patterns. But it's nice to have a standard around this, and I think participating in that is incredibly important, and it allows our agents to consume the tools, services and data from other systems and vice versa, and we want to play very nice in that. How it all ends up shaking out from a monetization or balance to trade perspective, I think, remains to be seen. But I think right now, we're obsessing over creating the things that actually help users and companies achieve what they want and figuring out the economics as we do that because the rate of change is just so fast.

I'd just add one thing. I mean specifically because the questioner asked about Snowflake and mentioned data, one of the things that we have is our data foundations layer, data cloud, Informatica, all the -- Tableau, all the stuff that sort of sits in that data foundation side of it, all of those are either are or will be available as Headless 360 as well. And so when you think about a lot of, let's say, Snowflake data or Databrick's data that's sort of landlocked, where it has part of the picture, but it doesn't have a full semantic understanding of the company, you can use data foundations to give it that semantic understanding that Symantec grounding, and then you could use that on our agents to make them better. They'll just naturally take advantage of it or you could use headless you can use data foundations headlessly to power other agentic experiences where Salesforce is just doing that data aggregation part of the semantic layer side of it. So I think what we really want to do is we want to have a set of LEGO blocks out there, where people can assemble them. But just like LEGO does, you also could buy the kit that tells you exactly how to build the spaceship that you want to build, but if you don't want to build that exact spaceship, you're free to customize it any way you want. But I think sometimes when people think about it is they don't think about all the things like Informatica that are now sort of either inherently or will be available Headless as part of that LEGO kit that allows you to just compose these things in ways that is vastly easier than trying to get everything consolidated in a single database.

Great. We have another question here from Kirk and Peter at Evercore. How are you managing compute costs in terms of customers' access to the massive Slack context library that continues to build on itself over time, meaning the context change or context library as more tokens on any query of Slackbot presumably going to be utilizing to fulfill this request. So Rob, maybe you can talk about that optimization path that we're going through.

Yes, absolutely. What [indiscernible] you may have heard it on the earnings call yesterday. We actually have -- so we launched our Slack MCP server 9 or 10 weeks ago, and we now have 1.2 million weekly active users of our MCP server that are making around 40 million weekly active tool calls that I think are pulling around 175 billion data tokens a week at this point. So it is a tremendous amount of volume. And so I think it speaks to the value to the second part of your question about the context that sits in Slack. I can't get into too many details, but we're working on a number of things to actually amenify the footprint of the infrastructure required to serve that while also still meeting the needs of the use cases. I think we're also doing as much as we can to I mean this gets into the technical details. We're doing as much as we can to preprocess and make as efficient as possible the consumption of those APIs. We've actually built, as I mentioned earlier, different versions of our APIs with partners for these new agentic access patterns because a lot of the APIs, frankly, that existed before for the traditional Slack app use cases aren't the best fit for these agentic access patterns that are much more chatty and iterative. And so we're actually building entirely new APIs that are agent focused. And I think over time, we'll be figuring out exactly how that fits into our plan structure and monetization structure with partners.

Yes. The only thing I'd add to it is, I think it's a great question. I think it's an astute question. But 1 of the things that perhaps we don't talk enough about is our transformation as an agentic enterprise using these coding tools. And so just to throws of love towards the Slack engineering team, it's incredible what the team has been able to do. And we now have groups of people that are 20x the productivity they were in the past. And so I think before these coding agents, there was like a real tension with like how much do you drive features versus how much do you drive efficiencies in cost structure and cost to serve. Because we have these coding tools because it's so entrenched and we're moving so much faster, we can do that optimization more contemporarily to the feature development that we've ever been able to do in the past because of the scale we're getting out of the coding tools. And so we, as a user of agentic coding tools have unlocked capabilities and velocities that like if you would have told us 2, 3 years ago that we would be here, I don't think anybody would have believed it, they would have thought of it as science fiction.

It's a great point there. And I think 1 request for Joe, I've requested access to Claude code personally, so I can start to agentify the IR process and help answer questions faster. So hopefully, that access gets approved soon. Our next question here is for -- from Omar Sheikh from Redburn. "Can you give us any examples of how early adopters are using Headless 360 right now?" So maybe, Joe, you can talk about some of the customer examples that are really exciting.

Yes. I mean I think when I look at somebody like Williams-Sonoma as an example, that has aspects of it. that are delivered using essentially the Headless 360 APIs and MCPs, even though we didn't really call it that when we demoed it a couple of months before, we actually came out with the name. The underlying technology was the same. And so when you see these like super rich experiences like the shopping experience, William Sonoma is just really incredible where it understands what you bought from them. You can ask for recipes. You can see these rich care cells of things rendered into the Agentic side of it. We're also seeing like really incredible things with like [indiscernible] where we could actually drive the website itself, where the agent is interacting with us. And the acquisition we made recently, Qualified, I think, is a great example. In the past, if we bought a company like Qualified, we would be thinking about, okay, well, how do we do with these integrations into our platform. But because the Headless 360, we're actually able to just integrate with them the way that anybody else can integrate with the Salesforce platform. And so Piper, our agent there is now able to do more things with Salesforce than they could pre-acquisition because obviously, we're putting a lot of focus on that. But ultimately, the Headless 360 platform, other people could do that, too. So I think when I look at these genetic experiences, we're just scratching the surface. I mean some of them are really step functions better than what's been in the market before. A couple of examples that I gave that I personally like, but they're still so early innings as to what's possible when you start thinking about customers' agents talking to our agents when you think about the whole website essentially being totally dynamically generated by the agent. All of these things are avenues that are going to start coming up that we can take advantage of. And so I think those are just a couple of examples, but it's just so early innings with us about how good you could make it. And that's actually the exciting part of where we're at right now is just seeing people take advantage of things in ways that like we wouldn't have predicted, but we're thrilled to see happen in production.

Our next question here comes from Matt VanVliet at Cantor Fitzgerald and focus a bit on gross margin. So as both Agentforce and Data 360, MCP adoption grows, the strategy around gross margin as monetization takes different paths and usage of AWS becomes the near-term goal. So I can start with this one. I think with a portfolio of products like we have at Salesforce, of course, we have different kind of gross margin profiles across our core business and that's something that we manage really well. We have best-in-class subscription and support gross margins in the industry. We're going to continue to manage that well, right? With new products, you always expect kind of ramp cycle, a curve of efficiency that will take some time to get through, but we're managing to kind of keeping best-in-class software subscription and support gross margins over the long term. Now what does that mean? I think, first and foremost, customer success is super important, make sure we're optimizing for usage, getting Headless in the hands of more customers, making sure they're understanding how can I use this in my business, that we think actually will expand usage across the entire portfolio, which is super, super exciting. There's a lot of things from a monetization standpoint that you've heard us talk about over the last few quarters. Agentforce One edition, Agentforce for apps, more bundled SKUs, ALAs, driving more of this embedded Agentic use cases within the purchasing vehicles that we have is another way for us to capture more value. In the quarter, Agentforce 1 addition and Agentforce for apps actually grew 60% bookings growth year-over-year in Q1. And the ARPU uplift we're seeing there for the premium tier additions is really meaningful, 60%, 70%, 80% of ARPU uplift that we're able to achieve. So there's a lot of ways for us to go out there and kind of capture the value from what we're delivering to our customers, but we also manage a portfolio of gross margins and have committed to kind of maintaining that best-in-class level that we're at today.

Yes, Val -- I mean I think you nailed it. The only thing I would really add to that is, I think right at the beginning of your response, you embedded perhaps the most important part for this to understand is is the fact that it's growth in usage, right, growth and delivering value. And so there will be instances where customers existing licenses like some of what Val talked about, will be all they're paying us. And yes, their utilization will go up. We also believe we can drive efficiencies back to my last answer. But we're also seeing that we have these consumption models. We have those things where customers when they succeed, we succeed from a customer outcome standpoint, and that growth in it sort of inherently brings the gross margin with it because we're highly aligned to our customers. And they are paying us more, but they're happy to pay us more because they're making more money materially. And that's exactly where we want to make sure that when we think about pricing models, we get this right because we want to be equitable to everybody. But I think Val all said it very well.

Awesome. The next question here is from Terry Tillman at Truist. And as you think about Headless 360 longer term, is the bigger opportunity expanding the number of users who interact with Salesforce and increasing the depth and frequency of platform usage or is it opening up new agent-driven use cases that were not practical in the traditional UI-based model?

Yes, I think it's actually both. I think both of those things, it's a good question. But I think both of those are going to happen at the same time. I mean when you think about it -- when you take a step back and you say, okay, we look at our admins, we look at the folks that were like building these interfaces, like the UI was empowering, but it was also constraining, right? Every admin out there, every UI person, every trailblazer had a backlog of stuff that they wanted to get into the interface that they -- that somebody believe maybe they are a business user or a stakeholder that submitted it was going to drive incremental value to them. But there was a finite limit to how much of that you can do. Now in this world with the agentic coding tools and with Headless 360 and the sort of femoral business driven, let me talk to the agent and tell it what I need. Our belief is that people will have more meaningful conversations with the Salesforce platform. And when they have more meaningful conversations, they're going to have more of them. And there's no question that agents are assistive in that. agents are driving it and human beings are also driving it. So I think we see it happening in all vectors. There's definitely a lot of folks that in a company that didn't necessarily interact with Salesforce Director definitely did not interact with Salesforce Lightning that now, I think, are going to drive increased usage because they're going to talk to an agent that's going to be talking to Headless 360 in the Salesforce platform. And I think Slack is just the greatest example of that. When people use our products in Slack, they use more of them. And that is because of the fact that they're just so much more dynamic and so much more powerful when we can create this level of both control and just assistiveness to the individual users of it and not constrain it by how much the IT department could actually fulfill those requests. And so as a technologist, I look after the office of our CIO and the internal Salesforce systems, I'm thrilled with this thing and how it's working because it gives me the opportunity when I wear that had to think about how do I better serve my business users. And part of the way I better serve them is I teach you how to fish and then I give them the fishing rods. And I think that's what this technology is going to do as it expands to is that this concept of -- it's not even distributed IT, it's AI fluency lifting in the entire company, and therefore, the tooling goes much more horizontal I think that's going to drive utilization for short.

Awesome. Our next question here is on AgenticWork units from Jackson Ader at KeyBanc. "Which tasks or outcomes consume more or less AWU use. Do AWs track closely to token consumption? Or are there some outcomes that are really valuable and high in AWS, but don't consume a ton of tokens?"

The answer is that AWUs are very token variable, I would say, in some cases, they consume a lot of AWUs. -- the particular task -- or a lot of tokens, the particular task that all puts an AWU may be very reasoning intensive. So it's sort of thinking about things and taking a bunch of stuff into consideration. And therefore, the token count is quite high. But I'll give you an example of why we use AWU as opposed to exclusively token count is if you look at agent script inside of Agentforce. It allows you to have a high degree of determinism. Well, the way we're able to ensure that is we're not using LLM to fulfill those requests. There's an agent -- agnetic LLM upfront that's trying to understand your request. But when you get into that sub agent, that is deterministic, we're running other types of models and other types of procedural code on the back end to guarantee that it will be a deterministic outcome but it's actually not agentic in that sense. The front end of it's agentic, but the fulfillment end of it is not. And so that would be an example of that's actually a very low token use case, but still very high value. And I think that that's 1 of the reasons we think AWU, at least at this point, are really good measure because it's like what work did you get done with sort of agentics in the loop. And the deterministic nondeterministic deep reasoning, wide reasoning, all of those things have very, very different token profiles, but we believe they all deliver kind of similar amounts of value. So that's the reason we added up the way that we do.

Yes. I totally agree, Joe. The one thing I'd add just for some color on what you saw in our Q1 AWU print, Agentforce service, the service use cases are still representing the majority of AWU consumption but let me call out some of the fastest growing areas within that AWU number you saw 1.6 billion in the quarter. That was growing really well quarter-over-quarter. We actually saw a real valuable increase in kind of the sales use case. We launched the sales agents I want to say 3 quarters ago now, and that, I think, grew above 200% quarter-over-quarter in AWU usage. We had certain industries that we've been hyper focused on from getting these out-of-the-box agents, more use cases, right, retail consumer goods, high tech, public sector that represented a real large amount of AWU consumption in the quarter and there's a way for us to really look deep into the AWU usage across these different vectors to understand what's working, what's not, where can we go faster, and that's a huge operational advantage for us. And the last thing I'll mention, Slackbot absolute amazing AWU usage growth quarter-over-quarter, more than 300% quarter-over-quarter growth in Slack AUs in just a short period since launch. So a lot of good things happening there, a lot of ways for us to really monitor optimize, make sure that these AWUs are actually converting real kind of token kind of raw intelligence into work enterprise outcomes. And that's what we're super excited about there. So the last question here that we have for the webinar today, and thank you to Joe and Rob for hosting. As AI agents increasingly outnumber human users, how does the Slack and Headless 360 product road map evolves? So maybe give us a little bit of a sneak peek, whatever you're comfortable sharing at this point on what can we expect to see out of these products in the future?

Yes. Well, I'll start. Maybe I'll throw it over to Rob for the Slack perspective on it. But I think what I think about agent 360, I'll just go a little deeper than what I was talking about before, there are parts of me that really believe that of the capacity of my IT organization is going to be building things that power agentic use cases for business users. So they're very much in the situation where their customer used to be the business owner directly, now it's the business owner's agent or their coding agent or the Co-work agent. And so when we really start to think about that, that's, I think, when you get massive scale because you also get massive customization, right? We in the abstract, we always wanted to have the most personalized version of Salesforce that people can have so that it felt natural, they used it, they loved it. It brought them joy, same thing with Slack, but there were limitations based upon how much code you could write, how much you could do for customization. Those limits have now been reset in a very, very major way. And so I think thinking about as an internal IT provider, internal technology providers thinking about the use case that you're no longer just providing technology for humans, you're actually providing technology for your digital workforce as well to get jobs done. I think this is just a super provocative concept. And I think we're going to get there a lot faster than people think. We think Headless 360 is a pivotal piece of that because that's the things -- the type of things that these agentic digital labor agents are going to have to use, they need that kind of foundation, and we're happy to kind of lead the way by showing people, yes, this is the way that deep work is going to be done in a hybrid workforce.

Just to build on what Joe said from a Slack perspective, I'd say we're very excited and we're tracking towards. I think you'll see as many agents in Slack as users as you will, humans in Slack as users. And I think that's, again, going back to our priorities, that's where it's exciting to try and nail the problem of becoming that multiplayer AI home for the ecosystem or multiplayer home for the AI ecosystem. And then I think to another point Joe made earlier, I think what you'll see moving forward is there was actually a mention of this on Lending's podcast by the but the CEO of EVRY, the other day. But I think you'll see Slackbot become a super agent for employees in Slack, and that will ultimately be the primary thing that they talk to that ends up interfacing with and surfacing other agents from -- certainly from Agentforce, but from third parties as well. So excited to see proliferation of agents and Slackbot help with the wayfinding with those agents.

Awesome. Well, that wraps up this webinar. Thank you all for joining so much. We look forward to seeing you over the next few weeks. One quick ask, if you have feedback on what you'd like to hear next quarter, what topic, what product, an area of the business that you'd like to deep dive on please send that over to myself, to Alex Chan, Alex Kingary. We also have a new member on the team that you'll be seeing out at the conferences over the next few weeks, Lauren O'Brien that we're really excited to get out there. So look forward to seeing you on the road, give us feedback, and thank you so much for joining.