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

Excellent. Thank you, everyone, for joining us this afternoon. My name is Keith Weiss. I run the U.S. software equity research franchise here at Morgan Stanley. It looks like [indiscernible]. Very pleased to have with us this afternoon from Salesforce, both Robin Washington, Chief Operating and Financial Officer; and Joe Inzerillo, President of Enterprise and AI Technology. So Joe and Robin, thank you so much for joining us.

Thank you so much for being and Keith, we had to be sure that we gave you one of our latest. We hear there's a big announcement forthcoming, but also just for right. You have to be sure to show it to everyone. I knew it.

I never got to join in the club in San Francisco because I wear suits all the time. So now I can finally fit in [indiscernible].

I got to show everybody. It's all about [indiscernible].

Very nice. I like it. Thank you.

So you guys recently closed your FY '26 and a transformational year for Salesforce when it comes to sort of product strategy, which Joe is going to talk to us a lot about, but also the sort of operating model within Salesforce and pricing models. So maybe to start out, Robin, could you tell us a little bit about some of the key accomplishments from a financial perspective and how this has set you up for FY '27 in terms of what's been going on in Salesforce in the past year?

Yes. Well, thanks for that start. Yes, we had, as you said, a terrific year from an innovation as well as a financial standpoint. We announced our results last Wednesday, as you had in like record revenues, record quarter, record cash flows, pretty much taking you back to Investor Day, Keith, I think what's important to note, we laid out a framework over the next several years. We talked about that continued elevation of net new AOV being greater than AOV over time. And we saw those inflection points play out in Q3 and Q4. So it gave us even further conviction to what I said back then of the 12- to 18-month trajectory to return to organic double-digit growth. You add on to that Informatica, you're talking about '27, but we're even thinking broader, amazing integration already, great return. We were able to make that accretive fairly quickly here within 1 year and feel really good about how it's fitting in with our core data component of our platform and really helping our customers, particularly our customers. So very excited about that. In terms of other things that we've done, we're seeing our investments start to pay off. We invested a lot in AE capacity last year, and it's something we want to continue to do. We've also invested in FDEs, very focused on deployment of Agentforce. And as Joe will talk about, a way for us to continue to fine-tune our agents to make them easier to install, et cetera, and infrastructure. So all things that really set us up well and on the continuum to meet our objectives over time. There's also just our growth playbook that we've had in place in Q4, if you think about our premium SKUs, 300% quarter-on-quarter adoption of those premium SKUs, which really for us shows the value of our stack in general. So that's really important to us. We've also invested in industry playbooks and processes even with some of our new products like [indiscernible] that are really helping customers kind of figure out where do I start? How do I take and become an agentic enterprise and what are some real focused use cases that I can accelerate. So we feel really good about '27, particularly that second half acceleration that I mentioned and continuing to grow profitably. We have doubled down on investments in FY '27 to meet that FY '30 framework, ended the year at Rule of 44, feeling very good about that trajectory of Rule of 50. So prioritization on growth in revenue and agentics as well as profitable growth. And I think you'll get to the capital question at some point.

I hit on that. I want to bring Joe into the conversation. And you mentioned the stack. And I think that's important when we're talking about Salesforce because there's a lot of asset at play here. You guys have built out a very robust sort of underlying data platform. There's a lot of application capabilities on top of that. Now there's an agentic layer on top of that. Slack is part of the equation. And Joe, you've had a really remarkable career spanning 3 decades. You've had leadership roles at Disney Streaming, BAMTech, Chief Technology Officer at Sirius XM. And when I think about your career, it's building big systems. And this is a big system that you need to build out here at Salesforce. So can you talk to us about sort of your vision? When you come in and you take this role, what's the vision of what Salesforce could bring to the market in terms of enterprise AI and what that's going to mean for your customers?

Yes. I mean, thanks for the question. I think you're right. On one hand, you can look at my career and say, like I built things of big scale. Disney+ was ginormous. The other stuff they've done has been pretty big as well. But the other way to look at it is I really spend most of my time in the direct-to-consumer market, not like a traditional enterprise technologist. But I think agentic is one of those things where it feels much more like direct-to-consumer. And it feels it on a technical basis, like these continuous improvement loops, how do you constantly refine things, how do you get them better? How do they learn, how do you learn? That's one aspect of it. But I also think it comes down to just the way we interact with technology. And so in an enterprise setting, you say, okay, well, here's a screen, and I'm going to try to optimize it. Now every salesperson needs to conform to that screen. But if you go to sales and Agentforce or in Slack, now all of a sudden, you're having a conversation and knows who you are and knows the questions that you ask, it's a very personal relationship, not unlike the customization that might be in a Disney+ or a Sirius XM streaming or Pandora, where it gets to know you, that's now going to become in the forefront, and we're seeing it now becoming in the forefront of how these technologies mesh together to drive better outcomes for our customers.

Got it. So I think the past two like investor debates behind -- and investors obviously have a lot of uncertainty about sort of what the future holds, particularly for the SaaS application layer. But you -- I mean, you came to Salesforce, you came into this big system, this big incumbent vendor with full knowledge of what was going on, right? Like the models were already in play. And you saw opportunity here at Salesforce. So how do you get comfortable with some of the -- like the investor concerns? Like number one is the DIY concern, right? Now with code generation tools, it's so much easier to develop software. Why does that not present a more of a threat to Salesforce and what you guys have built and erode some of the moats that people have traditionally thought about within software businesses?

Yes. Look, it's a great question. And I can see how people would think about it that way because they'd say, "Oh, well, look, you have these amazing tools." But I think about it more like a master carpenter, right? Like back in the day, when you had to cut with a hand saw, like you really had to be a good carpenter in all parts of it. Now you can cut on a table saw and it's probably going to be a pretty good cut. So the tools are raising the boats for everybody. So yes, the DIY market is getting more sophisticated, but so are we. And we have a backlog of data and features and things we've always wanted to deliver to our customers, but it sort of sits in a queue based upon what capacity you can afford for meaningful and disciplined growth. Now all of a sudden, we're seeing that good engineers and our teams are going 20x, and they can really sprint ahead of this. And so I think it's really in my career, I've done a lot of DIY, built a lot of systems from mostly whole cloth. But even I had the -- like why would I want to waste time trying to build something that I could buy that's fit for purpose. I'd rather spend my development money to use those tools to do the thing I'm trying to do, whether that's Disney+ or another direct-to-consumer product, that's where my value is. And so I just think that like the combination of us continuing to accelerate the rate at which we're delivering meaningful outcomes to our customers as well as them being able to then use those outcomes to really specific on what their core business is I just can't imagine there's a lot of people that are going to want to go backwards and say, like, well, let me build a better Salesforce. Like I don't understand where the ROI is in that.

Right. Right. And that would also be a defense against another investor concern of start-ups, right, of AI native start-ups being able to kind of move faster. But we think about start-ups, it's a dynamic between best-of-breed and suites. And if you guys can innovate faster, if you could close that feature functionality gap between sort of what a start-up and focused on a single technology versus what you could bring into a broader suite, it seems to tilt the balance in your favor of like let's put all this solution, let's garner these new capabilities from our incumbent vendor who's already automating a lot of our business process shortly.

I mean everything you just said. And then I'll add another one, which is if you think about social media. So if you set sort of TikTok aside, that's really a state-sponsored company. It's tough to argue that they competed on the even field. But look at the social media companies. They're all the same ones that were social media companies that were at the birth of it years ago. And the reason is it's not just everything that you just said, it's also the data. It's that deep semantic understanding of what your customers are doing and what the processes look like. In the case of social media, that sort of identity graph, the social graph that ties all these people together allows them the fuel to continue to innovate on the interfaces and things like that. And yes, start-ups come, some of them get acquired, some of them are inspiration, but they haven't really mounted a real threat against them. And I'm not saying that we think we're invulnerable. But at the same time, we have all of these assets, and we have 26 years of real data that tells us where people are having problems, where they want to go forward, how can we help automate those things with that tool set. So the data is just as important because it provides continuous inspiration for how we're going to try to solve problems and move the bar up for what our customers are able to achieve.

And I think, Keith, that's the key differentiation. Like I said, they're not -- customers just desire to see if I don't just want technology, I want things that ultimately improve my interaction with my customers, my bottom line, my productivity. You can't do that without the data, right? So they're looking for solutions. And we've built 26 years of being the trusted #1 AI CRM vendor as well as being very innovative. And I think you combine all that together, we have the solutions, we have the technology, but we have the trust and the data.

Right. And it's not just data, like we're thinking about data like the data sitting in the database, it's also the understanding of your customers, understanding their business problems because that's ultimately where the value is solving business problems. All right. The second sort of investor concern that I want to pass by you, Joe, was the idea of an AI user interface, right? And I think Claude with co-work really and how well that did tool use really spark this fear of that perhaps going forward, we're not going to go into Salesforce to understand what's going on in our customers and then go into maybe Workday to understand what's going on with HR. We're going to have this one universal AI user interface that is going to handle all of our requests, understand all of our systems, understand all of our data and abstracts sort of the user from the end systems and make maybe Salesforce a little bit more of a back-end transactional system. So one, how do you respond to that concern? And two, can Slack play a bit of that role for -- at least for a Salesforce customer against Salesforce type systems?

Yes. No, I think it's a great question. I'll take the second part first. I mean, Slack, it's not an if, it is. If you look at the major AI companies out there, they're using Slack, like it is the way that they get work done internally. So I think that Slack is a natural place for it because it's where people who use Slack get things done. So why would they not also be getting them done with agents, especially because those interfaces and those interactions tend to be fairly textual. It's like a collaboration tool that [indiscernible] could say like when Slack was founded before we acquired it, that they thought about agentics, but they thought about people, but it turns out that the agents and people want to work in a very similar way. But to your other point about the disruption side of it, I think the way I think about it is I'm old enough that I've lived through these technological revolutions. And back in the '90s, when you wanted to run a computer program, you went to a very specific computer and clicked a very specific binary and did a very specific thing. And then really, Mark and Salesforce were the ones who invented, oh, no, no, no, you could just do that SaaS, it could be in the cloud. And there was a whole bunch of like, oh, the people who had these binaries were going to be disintermediated them. And yes, well, Salesforce grew out of that with a new company, Oracle and a lot of these other companies are still around. They adapted to it. Same thing with mobile. Mobile, very similar things like, well, you need an app and then all of these companies have apps. AI is going to fundamentally change the way we interact with the computer. And that's cool. But it doesn't just change it in a super narrow way. It changes it in a general case way. And Slack back to the first answer, Slack is that organic way that people are de facto actually going to Slack because it's really well suited to do that. So yes, I mean, look, if we were a different company and we didn't have Slack, maybe I would be worried about it, but I actually think we're leading this transformation with Slack and people are coming to us, the AI folks, they're building things into it. And so there won't be a monogomy of interface. There's going to be a plurality of interface, but where is the gravity? And I think Slack is a great example of the human collaboration with agents is the gravity, and that's where we're pulling people towards.

Got it. Got it. Robin, I want to ask you about a different investor debate, but another investor concern, and that's the risk of seat-based models. The idea that we are automating and doing a digital labor replacement of the very units that you price on. So how do you think about that potential disruption risk? Is there a risk in terms of not being able to sort of make up for kind of the seats with added value that you're bringing with the more consumptive elements of what we're doing with agents?

Right. It's a fair question that we get asked a lot. And I will say, when we look at the core data, we're not seeing that. We're seeing the great adoption, the momentum metrics around our agentic products, but we haven't seen seats year-over-year quarter-on-quarter decline. And I think it goes back to the earlier conversation, Keith. As long as we're showing value of our platform and in our way, agentics make our core apps even more valuable, I think that's what's really critical. Now to sit and say that over time, are you not going to have some type of attrition of seats, that could very well happen. But we really see a hybrid model of seats as well as adoption of our agentic products driving consumption. And I can't tell you exactly how those curves are going to grow. But overall, we see overall continued value of our core apps and the system, the integration of them in the context is being really critical. It's like you can't have one without the other to our overall conversation.

I mean I totally agree with what Robin is saying. But I'll -- Robin and I are sort of here talking about the company, but we also have a relationship because in addition to overseeing Slack and Agentforce, I also essentially oversee our office as the CIO and all the things that we're using it. And so...

[indiscernible] CIO...

Yes. When she and I have questions like about, okay, should we invest in this? We're putting our business hats on as two executives that run a 75,000-person plus company, right? And we're looking for value. It's one of the reasons that we introduced this like agentic work unit as a measure because it really matters what the outcome is. Like you can spend 1 million tokens and that could be good or bad, depending on what you're doing with it. Either you use or what you're doing with it. And so when we try to think about investments, we're thinking about them the same way our customers are, and we're saying, okay, what is it doing? Like is it delivering the value? How do we get there? How do we measure that impact? nd I think the whole industry is evolving. It's super frenetic right now. Nobody has a playbook of exactly how this is going to work. But in a simplistic sense, if we deliver real value, like we're going to get compensated for that real value. And the higher value we can deliver because of the complexity of the task that the agents can sort of orchestrate across our stack and other stacks, that's going to bode well for what we can do from a pricing standpoint regardless of the vehicle.

And I think just to take an example like that, and Mark, our CEO, uses it a lot, is our helpdesk.com. We have been able to save on reactive call volume. Now we've been able to reallocate those resources to other areas of the business. It's adding incremental value. We're able to measure that and reinvest it or readopt it. And in some ways, it allows our customer service reps to interact more with the technology, the whole idea of humans and agents working together so they can be more focused on proactive value-add call. Our call volume is still going up because we're growing, but our ability to reallocate those resources and leverage agents really helps our productivity and allows us to rebalance. And so when I think about it as a user, the value of my users hasn't gone down. It's just allowed them to be more value additive, more focused on ensuring that they're meeting the needs of my customers, and it gives my customers 24/7 support for things that don't meet human engagement.

Okay. So maybe to kind of sum up this like investor concern and competitive dynamic and maybe shift the conversation more to a constructive of what Salesforce is going to bring to the marketplace or Salesforce's positioning it's unlikely that your customers are going to try to DIY their own solutions, right? That's what they look to you for. They're looking for solutions. You have strong competitive positioning against start-ups. But there is this new white space. This is added capability that large language models and generative AI bring into the overall system. And there's going to be a competition for who gets at that white space, who's able to create this further automation and further productivity for the end customer. So Joe, maybe you could talk to us about what sets up Salesforce well to win in that competition? What gives you guys the right to win in building out that additional capability, additional workflows against the front office and even broader into going into stuff like ITSM?

Yes. I mean I think when you think about it, the whole paradigm is sort of like upside down from where it was before. So you think about like the way in which you build code was like very like, okay, let's take this thing we want to do, break it down into steps, do all these steps very iteratively, figure out how we get there and all sorts of stuff like that. We're now starting to go the other way. We're now the imagination is sort of starting at like the user level where you can put tools in the hands of users and then observe how you can continue to make them faster and faster and faster. And for us, I can't understate how important it is. It's not just the models. The models themselves are incredible, miraculous, frustrating creatures that exist now, and they've completely shifted the paradigm and they keep making improvements, but they're not delivering like a results-based system that you can depend on for a business. You need that data. And it's not just the data like you were saying in a database that sort of sits there at rest. That data is now kinematic because of these models. It's always sort of being introspected and moved and things like that. So when you look at us, we start with sort of this data layer where we have all of this amalgamated knowledge, but we can represent it to the upper layers of the stack. And so you start working your way into the activation layer, the apps themselves, the facilities that the apps provide and then you have the agentics sort of orchestrating that whole thing and then back to Slack. You have this interface layer that then participates in the entire thing. And so does that necessarily assure that we're going to be successful? Of course, not. Does it sort of show that we have this incredible advantage in vertical integration where we have like a really strong foundation of how these components that need to all exist, need to all interoperate, need to all be observable can work together. And then because we've worked so much on the fit and finish of how they fit together, you also get into the situation where I can't emphasize enough these continuous improvement loops. So it's not just does it work today. It's like how does it get better tomorrow? How does it react to the change in human behavior? How does it react to more data becoming available. And that's how these things are just going to go. Like there is no going to be steady state, done, ship the software we're done. It's always going to be at this frenetic user level as opposed to the sort of architectural level. And I think we're really well positioned because both through organic build and kind of the DNA of who we are and acquisition, we filled out that layer in vertical integration. And we've seen with hyperscalers and things like that, how much that vertical integration is a huge asset in delivering solutions that actually work and drive value.

Got it. So like we were talking about before, so Joe's bringing solutions to real customer problems. If you guys are bringing a solution to your customer, they're going to pay for it. They're going to value it in some way. And you guys have developed a whole menu of for Agentforce pricing. We have agentic enterprise license agreements. We have consumption-based pricing via per call. We have Flex Credits. We have seat-based SKUs like Agentforce One Edition. Why is this so important to have so many pricing options? Like why -- it creates a lot of confusion for us, you don't like confusion. Why is this so important to get adoption in the marketplace to have this menu of options?

Yes. It's confusion, but it's also agility. I think what we're finding, to Joe's point, this is not a static market. We're competing. Our customers are looking at options. They're trying to scale. And our job is to be sure that we've got solutions that meet customers where they are. Some want to pay per user, particularly they're looking at other options. Others want certainty. They want to understand what this means. They like the user model. And so the different menu of options that we have allows us to meet a customer wherever they are on that journey, Keith. And over time, the ELAs are great. If you want to go all in on us, you don't need to worry about whether this agent is going to hit up against this LLM too much or it's going to be. You can really decide what are the right use cases? Do you want customer facing? Do you want employees. So we believe that agility that we have, sorry for the confusion, really helps take off the table for the customer any concerns they might have around cost. And it allows us to kind of just double down with them relative to being their platform of choice.

Yes. I mean, to Robin's point, I'd also just add, living through the hyperscaler revolution was an example of like it was a very different model. People were used to this CapEx model and I do this, I get a data center and all that kind of stuff. And by the way, like I used to be really good at building data centers. I thought that was an awesome skill set. I haven't built a data center in 12 years. And so like that all of a sudden became less interesting to me. But I think when you think about that transition, we think about it as if it was like this square wave transition that just happened. And everybody was like, yes, well, of course, it's how you pay for a cloud. But the reality is it took a long time. In retrospect, it seems short, but it took a pretty long time, took a decade really to get that into the full mainstream. The same thing is happening as far as the pricing models go with agentics right now. The only difference is because agentics are moving so quickly from a disruption standpoint, that entire time line is compacted. And so like to your point of us changing models, we're trying to be reactive to the market, trying to be reactive to our customers, meet them where they are. If you actually slowed down time and kind of expanded that to a decade thing, it wouldn't seem as frenetic it's just because there's so much opportunity and so much disruption right now that it feels like they're stacking and it is, and it does confuse customers, but we think the inaction is much worse than that. So we really want to try to get some stability, really trying to mature it and get to a point where we're meeting everybody where they're at. But I think the time scale is what people don't really appreciate is the fact that this is all just happening really fast for the whole industry.

Right. And I think the receptivity that we've had gives us the ability to monetize in any framework that the customer wants over time, it's more predictable for us. It's again, back to that hybrid model. But our goal is to ensure that our overall platform is sticky, it's retained. And we think this all-in model really helps us with that in our regard. And to Joe's point, we're iterating with the customer, it's a very different selling model. It's not users and we go away. We're out there our forward deployed engineers working with them and fine-tuning. And these pricing options, including credits, give them a lot of different options and ways to absorb that based on the success that they see and the iterations that they go through to be successful.

Right. So I mean, what I hear from a lot of investors and then what they're looking to me for and what they're looking to you and the IR team probably even more so is clarity. They want absolute certainty of that $1 of seat revenue is going to turn into x amount of agent plus seat revenues. But listening to Joe about the -- how quickly this is evolving, how quickly the capabilities are evolving. Listening to you about how customers are still trying to figure out how they want to pay for it. It seems like maybe we're looking for something that would be too limiting, right? If you guys narrowed it down to one set of functionality and one pricing model, you're going to limit your opportunity. So with that being said, it's near term, maybe maximizing from stock price, long term limiting. What should we be looking to? Like what -- is it AWS? Is it the agentic workflow units? What should we be looking to? And what gives you guys confidence? Because you called for acceleration into the back half of FY '27. What are you looking to, to get that certainty to give that forecast? Because I know you're conservative. Like you're not going to tell us about acceleration until you feel really comfortable with that acceleration.

Absolutely. Hazard of a job. But no, I mean, at the end of the day, it really comes down to customer success. And it's really about that partnership. We've used -- we've thrown out the AWU -- I shouldn't say thrown it out. It's a really good metric for AWUs, but we're looking at net new AOV. So we have a set of metrics that helps us really understand the direction of that customer journey. I think the other thing that is important, Keith, to your point, what's happening with the customer, and we also talk about the multiplier effect of being on multiple clouds, leveraging our agentics, et cetera. We're seeing real acceleration of ARR per customer as those customers go on this journey towards the agentic enterprise. So that's another comforting point to us. But you're right, we're definitely in a shift I think trust and customer success is absolutely the most important thing. And as long as we do those things, we're going to be okay and show that value and worth to our customers. We see it in the numbers. We see it in our pipeline. We're seeing it with our SDR agents, which are generating more leads. And to your point, I think customers are coming back saying, I don't want technology, I want solutions. And that's what we've been doing for 26 years now with agentics. I think that's the value that we see, and that's the direction of growth that we see our customers really leaning into.

Totally. And Robin sort of talks about the growth in ARR and like how we think about it. But I also sort of break it down into customers that I personally interacted with. And so like I was in APAC in November, I was talking to one of our customers I was just sort of starting their agentic journey in Japan. And that same customer about 2 weeks ago was in New York, and I got to see them again. And it's sort of like -- it's like your friends kids, right? You see a picture and I am like,"Oh, wow, they're big now. I had that same sort of impact with their agentics where they had sort of started and then you look at it and like, wow, and now they were coming in to talk to us again about what are the next 5 use cases we're going to do. And when you see that kind of like we did something, we work together, we partnered, they got results and now they want to expand that program. It's hard to like summarize those in a number in a spreadsheet, but you feel it when you see it happening. And I think that this particular customer was a great example of I got two snapshotted pictures several months apart, and I could feel the acceleration.

Yes. We see it internally. We call ourselves customer zero, and we started with everybody experimenting, 100-plus agents -- we've kind of got it down to 4 key categories: employee customer sales, back-office procurement, but we call them hero agents. And we're starting to see the acceleration. And we were just -- they are talking to employees. It's about -- it's given us permission to go faster because they now see the value add, and we now know where to invest and where to decelerate. And we see that same journey, as Joe said, with our customers.

And I think this is also the part that sometimes I see undervalued in the industry or underestimated in the industry is the fact that like these tools are due to everybody sort of almost instantaneously. And so when they come out, obviously, we have relationships with model folks, we get a little bit of a head start. But people are now starting to really know how to use them to deliver real results. And so when I talk to other CIOs, their second agent is a whole lot easier than the first. And the third is a lot easier than the second. And that's both because the products are maturing, but it's also because they know what life cycle, agentic life cycle works for them. They have the data. They know what they need to look at. They know how to make these things better. It's not like there's no appetite to take these things on. So really, what we're now into is the building of the confidence curve about the practitioners understanding that like, yes, I can sign up for a number that I'm trying to either grow or save. I know I can do that now because I've got this precedent of these last 3 things that I've done. And now I really feel like I have mastery of the technology despite the speed that it's going at. And that learning curve of the practitioners really starting to understand how to actually use these things to their advantage for a business, we see every day with ourselves and also when we talk to customers. And like I said, sometimes it's hard to see that read through into a spreadsheet in a very, very specific precise way like you all want to see, but you feel it when it's happening, and we definitely feel it.

So you guys take these innovations, you create solutions for your customers. You're starting to get real traction with these solutions, you're feeling from the marketplace and you come out with agentic work units to try to convey to us sort of that inflection that you're seeing in the business and the traction that you're seeing with these agents. Why agentic work units? Why come up with a new KPI? Why not tokens? Like everybody else in the industry is talking about tokens. Why do we have to come up with a new KPI for Salesforce in particular?

I'll give you a really perfect example of it is one of the things that we've done with Agentforce in the last -- we announced sort of a Dreamforce last year, tail end of last year was agent script. And this notion of like the real challenge that a lot of people have when they go down the agentic path is you want a rich understanding of what the human is trying to do. But often, you want very prescriptive outcomes. So if you're filling somebody's prescription, you don't want the engine to guess, right? The models themselves are stochastic, right? They're probabilistic. So like you want that to be very precise. And it's really tough to do that within an LLM by itself. So part of agent script is this deterministic side of it that we have with our new planning agent, agentic adaptive reasoning, we call it. And when you look at it, you start to say, well, if somebody goes to do something and the token use goes down, that would be bad, except if they start to use things that actually use this agentic reasoning engine as opposed to the LLM itself, the token count goes down, but the actual efficacy goes way up. And so we really felt like tokens are a story, like you do need to look at tokens. But at the same time, you really need to figure out like how do you get work done. And the work unit felt like a very pursuable thing where you can say like we actually did something that helped further the cause of whatever this customer was trying to do. And so that's why we thought that like never fall in love with these things. They come and go. But we felt like we needed something that wasn't just tokens because it wasn't really describing the phenomenon. And ultimately, you want that mix of like efficiency, efficacy and cost to all be concatenated into one statistic. And right now, I just think the market is so immature that we don't have ability to quite do that yet. But AWUs, we think, is a big step in that direction.

And the benefits to that to us are it helps our customers better understand the ROI. It also helps us better manage one of your other questions is what's going to happen with gross margins, right? So the efficiencies of that, that engineering feed internally for us helps our gross margins as well. And again, I can measure Joe's success, our success in customer zero by looking at our work units. What are we actually getting done. So it's a great metric for a number of different reasons.

Right. Salesforce isn't just a wrapper around tokens. You guys are adding a lot of value that enables you to get leverage against tokens.

And that value is not only with the agentics, it's also with the core.

Exactly.

So that, again, it's this hybrid model of users and agents working together with context of data. That's the value proposition that we want to get.

And back to the... Absolutely. And you just touched on this. I think it's worth like just really trying to put a fine point on. Again, when we talk about like context and things like that, we're sort of talking about it in a past tense in the sense of like this is the thing that the agent needs to make its decision. That's true. But when it makes its decision, the thing that happen becomes context as well for the next call the agent makes, either to that same person, very personal context or sort of thematically across the organization or when we look at it from building the technology stack itself, it winds up being the more that you could do these, the more you get acceleration of people using it, the more exhaust you have from these transactions, which then make the transactions better, which makes people want to use it more, et cetera, et cetera. And you see that flywheel effect starting in a lot of things. We saw it internally. We see it with a lot of our customers at this point in time. But that's where it's really going to start to accelerate because each one of these things just reinforces the previous action and then predicts the future action better.

Got it. So I want to wrap this all up with the capital allocation question. You guys have talked about the Trinity of capital allocation. You started paying a dividend. You raised that 6% this year. You're going to have strategic focused M&A that is more shareholder-friendly as well as share repurchases. And you guys returned more than $14 billion to shareholders in FY '26?

[indiscernible] of our free cash flow last year.Yes.

You guys put out a very big number in terms of share authorization, $50 billion at that time, it was 27% of market cap. Does that signal a little bit more of a weighting on the share repurchases given where the share price is, given your excitement in the business versus maybe the other 2 parts of the Trinity?

Yes. I would say the way we step -- I mean, look, I said it on the call, if you look at what we see as the value of our products, our innovation in our company, there's a big dislocation. And so we see no better investment right now than Salesforce. That being said, we're going to do it in a balanced, disciplined fashion that doesn't preclude us from thinking about smart M&A. Last year, we acquired 10 companies, including Informatica. To your point, we're doing it in a disciplined fashion. You should consider dividends our floor. And we're generating a ton of free cash flow and expect to continue to do. So yes, they're all important. We see dislocation with us. We're going to double down and be a little bit more aggressive than we have in the past. But all, to your point, in a Trinity fashion that makes the most sense to drive long-term shareholder value. And at the end of the day, it's growth in that top line, the reacceleration of double-digit growth. We know that's how we're valued. We're not going to do anything around share repurchases that precludes us from doing that, but we see a really opportunistic opportunity to take out a portion of our market cap, and that's what we're going to focus on.

Amazing. Super exciting time at Salesforce. Joe, Robin, thank you so much [indiscernible].

Thank you.

Thanks.