EP 127: Marc Benioff (CEO, Salesforce): the Future of Digital Labor & the Agentic Era

0:05Welcome to the Logan Bartlett Show. On this episode, what you're going to hear is a conversation I have with co-founder and CEO of Salesforce, Mark Benioff. Now, we go deep into a number of different things related to artificial intelligence, including Mark responding to Satya Nadella's recent comments calling a number of SaaS applications simply CRUD databases that aren't going to survive the future of AI. We also talk about the trillion-dollar opportunity that Mark sees for digital agents, as well as what the impact of AI is going to be for different revenue, including what he thinks the split is going to look like for agentic revenue versus more traditional software revenue that we see.

0:44I'm joined in this episode by my partner at Redpoint, Jacob Efron. Jacob leads a number of our artificial intelligence investments here at Redpoint and also has his own podcast called Unsupervised Learning, where you'll go deep in AI with people like employees from OpenAI, including Noam Brown and Bob McGrew on recent episodes. If you enjoy going deep into artificial intelligence, I highly recommend subscribing to unsupervised learning. Now, without further ado, you'll hear Mark Benioff. Well, Mark, thank you for joining. Appreciate having you on. It's great to be with you, always. So one question, I guess, we're going to talk about the commercial applications of AI and how Salesforce is using it externally to your end customers.

1:22I'm curious, I heard you say something on a podcast recently about maybe not hiring more engineers this year, keeping that slattish and maybe customer support going down over time. Can you talk about how it's actually impacted the business of Salesforce? Oh, yeah. Let's just talk about that because I just think it's such an exciting moment. For the last 25 years at Salesforce, we're really all about helping our customers to connect with their customers in new ways. We automate all these customer touch points. We're helping them manage and share information and manage data. We're a huge database of 230 petabytes of information.

1:58That's awesome. And all of a sudden, and that's a huge TAM, by the way, all of its own. I know you guys are venture capitalists, so you like to talk about TAMs. So, you know, the TAM world is, you know, you look at like, you know, just like, well, just take one, like call it Slack, the Slack world that we're in. Hopefully you guys are on Slack. You know, it's like a hundred billion TAM, the collaboration world. Then you've got, you look at the sales and service and marketing, like kind of the traditional CRM world. Maybe that's a multi-hundred billion TAM. The analytics world, we own Tableau. That's like a hundred billion TAM.

2:29But we're kind of in the hundreds of billions of dollars, TAM. But I think we're in a new world. And this new world is the world that, you know, none of us have been in before, but we're all about to go into, which is the world of digital labor. And this idea of digital labor, this is a trillion-dollar TAM, a multi-trillion-dollar TAM, this idea that we're not just supplying information management tools to our customers anymore to help them manage and share their information, But we're now providing digital workers, you know, digital labor. And that is like the huge awakening that this is a new, exciting world that is what's really transforming.

3:12And I think as we've moved into these different levels of AI from kind of the predictive world where, you know, we kind of started 10 years ago with Einstein, which was this world of, you know, it's about machine intelligence, machine learning, predictive intelligence. agents. And then we moved a couple of years ago, two or three years ago into generative and where generative AI was really cool. And we're all on there and doing our search and saying, Hey, what, I'm looking at buying this new camera and which, which one, Oh, the Sony camera could do that. Oh, I had no idea, you know, and then listen, it's like, how is it doing this?

3:44And then now we're though moving into the agentic world and the agentic world is this exciting world. And then we're moving into the robotic world. And I'm sure you guys are making investments in that category, you know, in the agentic category. And you probably were in the degenerative category. And I know you were in the predictive category. And when you look at these four areas and how it's played out, it's awesome. But this third chapter has started and we're all in it. And customers are really excited. I mean, we've signed thousands of customers now on our AgentForce platform, and I've never seen anything go faster.

4:22Yeah, no, I'm curious. I feel like a question a lot of folks are asking is just the timeline toward, you know, some of the sygentic revenue and kind of that larger TAM that you talk about. I'm curious, as you look, as you think about like a few years out, you know, two, three years from now, like, what do you think the relative split ends up being of kind of this digital labor revenue versus, you know, different types of traditional software revenue? And then, you know, even looking at a decade from now, where do you think that goes? Wow. Well, those are like three or four questions. So let's try to take it apart one by one.

4:54And then if I've missed one, let's go back and kind of close the door. So when we were getting ready to talk about this at Dreamforce, I thought the most important thing for us is to start talking about how it's humans with agents driving customer success together. And that, you know, at some point these agents are not like our agents are 85 to 95 percent accurate right now. And we could talk about that with you and what we're seeing. But at some point, the customer says, actually, I want to talk to a person. And bam, you're back in the customer support world with a human. And that whole screen in front of you is filled with the whole agenda conversation.

5:32And you see everything playing out dynamically from what happened in the agent. And now I'm seeing what's going on with the humans. So that is very powerful. and that is the deep integration between the two worlds is that we're not in a pure agent world. Maybe we, you know, I don't know if you're talking to her every night, you know, from the movie and on your way home, you're having a relationship with a digital agent, but I'm not, you know, but if you do go to help.salesforce.com, you'll see that I'm trying to be customer zero and I took our stuff. And we have 9 ,000 employees doing support at Salesforce.

6:14And we basically put an agentic layer on our whole system. And you log in and we ground you. That means connect into your Salesforce data so we know what products and all your cases and everything. And you're working with the agent. And why that's important is we do about 35 ,000 of those customer support inquiries every single week. And a month ago, 10 ,000 of the 35 ,000 have to go to humans. So they're on there working with the bot or something, trying to kind of work inside the electronic world. And then they eventually got ejected into the human world. Well, now only 5 ,000 are not resolved by the agent.

7:00So we've gone from 26 ,000 resolved to 31 ,000 are resolved by the agent. So there's still 5 ,000, but we just need less support people because of that. So that is a big thought where all of a sudden, wow, we're getting a lot more automation, a lot more capability, but there's still a deep integration required between the agentic layer and the software layer. And connecting humans and agents is extremely important to deliver customer success. So that's certainly how we see it right now. And you have these great case studies. I mean, similar to what you've done internally, I feel like you've published with like Disney and Wiley and all these folks, some really impressive outcomes of using AgentForce.

7:40We have signed now thousands of customers with a revenue relationship. That is incredible. And these deals are from very small deals to deals that are as big as eight digits just for agent force, in addition to all the other products that they're then, you know, buying as well. So that is very exciting. And, you know, we can see, you know, you know, when we do a deal with a customer, if we're talking about a large customer, like let's say a large, I'll give you an example, or telecom customer deal we recently did, it's a large deal, you know, it's a nine digit deal. And it's very significant, and it's multi-year and so forth and so on.

8:18And the total we call the total contract value or the TCV is this long, this big. And then in that deal, we saw like an eight digit Asian force deal. So that is very exciting. And it pulled also all the other products along with it because it makes all the other, we don't just have a service cloud. Like, you know, we are the number one service cloud in the world. If you look at the Gardner Magic Quadrant on service cloud, we're up into the right. Oh, by the way, that's all pre-agentic. Now we're the first agent-first service cloud in the world. We have a complete agentic platform around our service cloud.

8:56And for all those customers who have service cloud, who are about to buy service cloud, it's all going to be agent-first. So that is very cool. And it's not just the only agent-first service cloud. Our sales cloud is now agent-first, our marketing cloud, our commerce cloud. Even if you go to our website, you'll see that we took, you know, WordPress, and we have built an agentic layer on the front of it, ingested our website into our data cloud, and then built an agentic layer on the top of it. It's an experiment that we're running. But I think it's a very neat idea that agents can be, you know, front ends to websites as well.

9:33So we're in a new world. I don't think all of us understand where we are. It happened so fast. I don't think we had time to all really debate it and discuss it like we normally like, oh, hey, this is what's going to happen. Like, boom, it just happens. And so now we're all just trying to go as fast as we possibly can. I'm curious, like, as you think about that example with the telecom and the service cloud revenue and the agentic revenue, just from a proportionality standpoint, I guess, as we sort of internalize some of these use cases, does it end up being, you said, an eight-figure deal for the agentic?

10:07Is that ballpark? ballpark, does that end up being like 10%, 20%, 30 %? We're just trying to quantify as venture capitalists how to even think about the incremental opportunity for some of these things. Every single customer is going to be different. And I think that it's impossible to say exactly what that's going to be. In that case, the customer was about a third agentic and two thirds traditional. I can't tell you yet exactly. I just don't have enough examples to be able to say, oh, this is exactly how it's going to play out. But in that case, that customer was about a third agentic. And that was a very exciting, you know, transaction and very exciting for us to take this very good size telecom company.

10:52Here is the platform to automate them. Let's get them to another level. We're able to paint a vision of what the future is for these customers. That is incredible. And also help them show they're going to lower their costs with the agentic layer, but radically increase their customer intimacy and their capabilities to be able to deliver value to those customers as well. So it's a tremendous use case and it's just an awesome opportunity. A CEO of this upstart business, Microsoft, Satya Nadella, said recently on a podcast, something to the effect, let me read it for you and I want to get your reaction to it.

11:26He said, business applications will collapse in the agent layer. They are essentially CRUD databases with essentially a bunch of business logic. And once the AR tier becomes the place all the logic is, they will start replacing the back end. I'm curious. It felt like that was maybe a veiled or direct shot at Salesforce in some way. So I'm curious your thoughts. It wasn't veiled. And I think that I had to watch it and then try to. I'm glad you wrote it down because I think a lot of people are trying to figure out what is he exactly saying. And this is what he's basically saying. Microsoft has disappointed everybody with how they've approached this kind of AI world.

12:06And today, when we look at CoPilot and what they've done, they've kind of repackaged OpenAI and dropped it into Excel or something to say, oh, yeah, now here, now you can try to do this, do something. But customers are not finding themselves transformed with this CoPilot technology. I mean, I've spoken to these customers. I mean, they barely use it, and that's only if they don't have a chat GPT license or something like that in front of them. So it's not at all what we're talking about. And I think that we're delivering an agentic platform in production today to our customers. And we're already the largest enterprise AI provider in the world.

12:51We'll deliver 2 trillion enterprise AI transactions this week to our customers. Now we're delivering not just predictive and generative, but also agentic AI. And we're delivering this at scale globally to our customers. And this is what customers really want. Look, Microsoft's a very good company. They're the number one software company in the world. We're the number two software company in the world in the enterprise. And they're very good fast followers. I'm sure they will try to copy our stuff like they usually do and try to move or copy our language and move towards us. But we're out there right now in production with thousands of customers, and they're just not at that level.

13:31And I think this co-pilot thing has been a huge disaster for them from a branding and kind of validation standpoint. Customers don't look at them and don't take them seriously in AI, nor should they, because they're not even making the AI themselves. One of the debates we've had internally on this is as we move to more of a UI-less world and we're interacting more and more with some chat-based system, does that remove some of the incumbency advantages that a system of record has? And it could be Salesforce. It could be Workday. It could be whatever. We can come down to the service now. I think that head on, because I think that if you kind of heard what he said and kind of what you're saying, which is we're moving into the environment of her.

14:18So if we haven't watched the movie, this is what you're saying, right? You're like, well, listen, I just talked to, hey, how's my sales forecast? And by the way, I'm trying to close a deal with these guys, you know, at their venture firm and want to make sure they're using our products. So let's make sure we follow up and get that on my calendar. And let's do this. And that's not how I'm, I don't know how you guys, how you operate every day, but I have this thing called a phone, which is this really cool invention that I will be on. Like when I'm done with the podcast, checking my text messages, my emails, my Slack messages, working on my forecast.

14:58You guys may not use it anymore because you're way out in advanced in venture capital world. But how I operate is like I have a phone, I have a computer, I do Zoom. I have a lot of things, these tools that I use, and I don't know what world that he's in where, you know, these SaaS applications are only CRUD databases. But the funny thing is, it's a mishmash of words because SaaS applications are CRUD databases. CRUD is create, read, update, delete. So yes, we are creating, reading, updating, deleting sales information, service information, and kind of the example of if you go to help.salesforce.com and hopefully your portfolio companies are all using our service cloud and our sales cloud, and they get on there and they are having a great experience in resolving 90 % of their things on the agentic layer, and then they go, no, I need to talk to a Salesforce employee now.

15:57And then boom, Boom. That person comes on the screen and then everything flips. And what happens for the Salesforce employee is a huge screen of all the information appears in front of them. And they're looking at the screen and they can see exactly what the agent has been doing with you and looking at the case information and looking at your company and all your agreements with us and all the products and every interaction that you've ever had with Salesforce is right there on the screen. That world is not the world that you are describing. You're describing the post-PC, post-phone, post-Zoom, post-Slack, post-email world.

16:37And I'm not in that world. I'm in the current reality. It's 2025. Here is my desk. I'm in the now. So I know what I could Oh, and by the way, the AI is not 100 % perfect. I don't know if you've tried it, but these things are good, but not great. And, you know, AGI is not here quite yet. I know you may have read some blogs this weekend, but I don't know what AGI they're using, but it's not the one I have, at least. So I am trying to make good of some very good AI that's a lot better than what we've had. But, you know, we can do a lot for customers, but we can do not everything, but we can do some things.

17:23I mean, I think what you're saying makes total sense. I don't know what I'm saying. I haven't really had the chance to respond to that, so I appreciate it. We're happy to provide the forum for it. It's a question that we've debated internally, even outside of that clip itself. We've, you know, we've thought about the UILA world. It's like this. Did you ever see the movie Space Odyssey, I think it was called? I did. And there's the moment in Space Odyssey where he says, open the door. And the computer says, no. Remember that moment? Yes. That isn't how I go into my house today. You know, maybe one day it will be that I'll be able to, you know, open my doors with my voice and everything.

18:08We're just not in that. Look, there's a spectrum of, you know, technology is a continuum. It's constantly getting lower cost and easier to use. And we can see where we are and where we're going. And it's great to have like a marker in the future. Yeah, we're all heading to her. But then to say that that's where we are now, I think it's a deception to customers and to people like you, good, hardworking investors who are trying to figure out what companies should they invest in to actually make money for your LPs in the next five years or the life of your fund. And that is not, I think, useful right now, which is that the reality is we do have a world where humans and agents are going to have to work together for probably the short term.

18:51And at some point, if we move into the computerless world, let me know, because that will be very exciting. But that's not where we are right now, is it? I appreciate you looking out for lowly venture capitalists, Mark. Someone's got to do it. all this fud and crud out there it's making it hard for us to do our job i'm an investor too and i you know we have a little thing time ventures you know we're an angel investor please include us on your deals and you know we're investing in a gentic robotic you know world you know regenerative biology you know these things are exciting but we have to kind of operate in the here and now otherwise we get too far out there oh it's probably not a great situation i mean it It makes total sense that the human support rep that, you know, is going to take the handoff from the AI agent certainly needs, you know, like something like Service Cloud to be able to navigate this.

19:43How do you think about like the, you know, as more and more just AI agents are doing some of this work and maybe, you know, at some point we get, you know, companies that outsource all of support to AI agents. What do those agents themselves need? I mean, do they, you know, does the tools that they need to like to log into or get access to or be able to, you know, get the information they need. Does that look like, you know, service cloud or is that something different? I'm sure it's something you thought about a lot. Well, I think that, you know, our fundamental architecture has had to change in the last couple of years and it's kind of happenstance how it's kind of started to line up with what we've had to deliver in the agentic world.

20:19So I think the big change that we've made is, and I think you understand how Salesforce has architected, which is our sales and service and marketing and commerce and even our analytics. It's all one platform now. It's one code base. It's one sharing model. It's one user model. And it's all amalgamated. So all the workflow and all of the capabilities where we had separate workflows because we acquired companies, we've really united them into one workflow that we have called Flow, one user interface, which we have called Lightning. and then you're able to basically bring in the functionality you need to have that application.

21:01And then we've also added a data cloud on top of that, which helps you amalgamate all of the data and federate that, which means connect off to the snowflakes and the data bricks of the world and do all of those things. And that idea of an application layer and a data layer are extremely important. And that's what we thought two years ago, pre-Agentic, because we didn't really think that we'd go into Agentic as fast as we have. And now that we're in the Agentic world and we're shipping an Agentic platform, on top of that is layer three. So that third Agentic layer is also part of a single code base.

21:38So the Agentic layer, the data cloud, and every customer touchpoint at Salesforce is one code base and one piece of code and one data model. And why that is important is because the AI needs data to make the right decision. So it needs to have the seamless understanding of the data and the metadata. You know, the data is my phone number. The metadata is it's a phone number. So that idea that the metadata, the data, the apps, the workflow, the data cloud and the agentic layer or one code base is kind of Salesforce's unfair advantage because it makes our AI more accurate. And that's how we look at it.

22:20And I think that's why customers are, you know, going so fast with our platform. You probably saw we deployed because we kind of released it nascently into 135 ,000 customers. And now they just have to turn it on and then they can turn on and build within the platform their agents. So we should get to our goal of a billion agents. That is very exciting for us. And I think this is an exciting moment. And I do not think that this is just a CRUD database. I think it's a misunderstanding of the vast power of the platform of Salesforce and what you can do with it. And don't ask me, the customers because they're the ones who are going to really be able to deliver that, you know, value.

23:06I guess I'm curious, like, as you kind of zoom into the future a little bit, you know, it seems like today, you know, there's a bunch of enterprises that are adopting these support agents and they're talking directly with customers, you know, in the future, not to beat the her analogy to death, but I imagine, you know, consumers may have their own AI agents that are talking back and interacting with that. And, you know, where does this all go? And what does that mean for like the agents that we actually build? I don't know. No, I mean, it's probably a really good lunch at the Village Pub, you know, where we can, like, sit and everybody's talking about this conversation, really.

23:36Like, everybody, I think, all of a sudden has gotten thrust into this, you know, in a way. Because you can see, like, whoa, this is going faster than we all anticipated. And I think that's the cool part about our industry right now. And this is a moment where we are all, like, sitting there going, when are we all just going to be in the her world? And I think it's a totally fair question. And then I think it's a continuum. So where are we? I think it's a good point. Like, where are we in this moment? Because I don't think we can exactly predict when it will happen. Maybe there's people out there who can say, well, yes, in 2029, all phones will shut off and we will just talk into the microphone.

24:19But watch this. Hey, what is my schedule today? oh nothing happened by the way so oh what is logan's schedule today siri what is jacob's schedule today well given that we're vcs a blank calendar might be the right answer there anyway so i'm clear if that was tweeting uh you know podcasting everything vcs do so that is the amazing thing which is like it's not quite there yet and i think that that is the turing test for her. We just did it. One of the things I've heard you've been thinking about is just global labor shortages and some of the economic outlook or implications for this AI agentic layer that we're moving into.

25:04Can you maybe elaborate on some of your thoughts around that? I think that this is a big moment for our customers to be able to transform themselves to become agent first. And it's kind of a leads into your next world, which is we are going through a global labor shortage where we see all these declining birth rates. We understand that it is harder to hire, especially people here right in the United States in sales and service. You can see it. And this is going to give us the ability to do more. And as an example of that, look at engineering. I think in engineering this year at Salesforce, we're seriously debating maybe we aren't going to hire anybody this year because we have seen such incredible productivity gains because of the agents that work side by side with our engineers and making them more productive.

25:48And we can all agree that software engineering has become a lot more productive in the last two years with basically these new models. And this idea that there is somebody who's kind of along the ride with us and kind of having this incredible experience, you know, helped making our engineering teams even more successful. That has not been true in our sales and service and marketing teams. I mean, maybe it isn't content. I think that's one of the reasons that Adobe is struggling right now is because you can see what's happened in the content world. Like I watched that Jensen Wong keynote last night in Las Vegas, which was epic.

26:27And that keynote, almost every example is a content example except for the robots and the driverless cars. and like the examples were awesome on all the content that was being built but in all those cases i would have been using adobe products before to generate a lot of those movies or to do a lot of those images and then i'm like whoa this is like really powerful that there is just a huge ecosystem of companies you know that are doing these amazing products that are competing against Adobe in incredible ways, but even just the generic models are able to do incredible. I'm sure we've all had the experience like, hey, draw me this, and that's so good, I can now just put this on a t-shirt.