Salesforce CEO Marc Benioff and the AI ‘Fantasy Land’

0:03If you've heard of Mark Benioff, it might be because he spent more than a billion dollars building the tallest skyscraper in San Francisco, now the world headquarters for Salesforce, the 25-year-old enterprise software company he leads. Or maybe it's because he spent$190 million of his own money to buy Time magazine. But if his latest big bet pays off, someday he might be known for something totally different, namely replacing thousands and eventually millions of white-collar workers with AI agents. Benioff is one of the most outspoken voices in tech. He doesn't keep his opinions to himself. And as he tells it, AI agents are the future.

0:42They never get tired, never have to take a break, and they're potentially much more autonomous than today's limited AI chatbots. But, and this is important, he doesn't buy into some of what he calls the false prophecies around AI right now. Number one, everyone is not going to need their own nuclear power plant to run their data centers. This whole energy fantasy. And number two, that everybody has to start DIYing it and building it themselves. Let's just take these down one by one. He has some harsh words for his competitors at Microsoft and others racing to hop on the AI train. And some advice for the millions of us who are along for the ride.

1:22And I think we all got drunk on the chat GPT Kool-Aid. Oh my gosh, cancer is cured. It's like, guys, we've got to keep these things in perspective. From the Wall Street Journal, I'm Tim Higgins. And I'm Christopher Mims. This is Bold Names, where you'll hear from the leaders of the bold-named companies featured in the pages of the Wall Street Journal. Today we ask, if Mark Benioff thinks the hype around AI is overblown, why did he almost change the name of his company to reflect his new focus on AI?

1:58Before we get started, we reached out to Microsoft about Mark Benioff's claims. The company declined to comment for this podcast.

2:09All right, Mark, it wasn't that long ago that folks were saying AI might be in some way the end of Salesforce. And now you are, of course, very excited about AI. I heard that you were thinking of changing the name of the company to Agent Force. For the uninitiated, what's an AI agent? Why are you so excited about it? Okay, well, number one, let's take it from the tippy top. Salesforce is the largest supplier of enterprise AI in the world today. We'll do more than 2 trillion transactions this week of our AI platform, which we call Einstein. We've had that around for about a decade. It is incredible.

2:48It works across every one of our customer touch points and across our platform. And it's done predictive, and now it does generative. And because we have such awesome AI, we've also done a huge amount of AI research and contributions to the AI community. So this kind of false narrative, one of many false narratives that I've heard, that Salesforce somehow is not a leader in AI or hasn't been a major leader in AI over a decade is a false narrative. And you've said that AI being delivered by your competitor, Microsoft, you've compared it to Clippy. And for those who don't remember, of course, Clippy was that inane helper that would pop up in Microsoft Word and offer unhelpful advice.

3:33And it's the butt of many jokes about the late 90s. Anyway, that's a bold claim, considering that Microsoft is pouring tens of billions into Copilot. When you compare Copilot to Clippy, what do you mean? You're right. Microsoft Copilot is the new Microsoft Clippy. And the reality is that Clippy came out because they're like, hey, we have an agent-based future, which we've all agreed since we watched the movies, Space Odyssey or Minority Report or Terminator. We know what agents can do. So we have to get super clear right now. What are we really trying to do? Microsoft has deceived our customers by telling them two things.

4:11One, that the future of their interaction with AI is CoPilot and they need CoPilot now. It'll transform their business. And two, they need to build their own models and train their own models. And when you put these two things together, you're going to transform your entire enterprise. Well, guess what? Where is the list of customers who have done that? And I already know you don't have a list of 10 customers in front of you who have transformed their business with Microsoft CoPilot and driving their own models. And this is a fantasy land. And I can walk you through piece by piece right now why that is technically false and why customers do have an opportunity right now to transform themselves, to change, to evolve, to go to a whole new level with AI.

4:58It is a wow moment for everybody and customers that I bring into and say, hey, look, let me build you an agent in a few minutes with your existing metadata and data and business process. And then they say back to me, oh, my gosh, we have been hypnotized into thinking that we need to build and rebuild these models and DIY our AI. But on their last earnings call, Microsoft CEO Satya Nadella said, we have never seen adoption of an Office suite product be as fast as it has been for Copilot. He said there's been 60 % year over year growth. He has said that corporate customers are coming back and buying more licenses, not fewer.

5:44Those numbers have to mean something. Why is the clippy of AI so popular? Number one, there is a huge demand for AI products in the enterprise. But this idea that Microsoft has hypnotized the industry, that this is the panacea, this is the messiah of AI for the enterprise is a false prophecy. And I just get down to just show me the customers who are transforming themselves with this approach. And I have a better approach that works, that you can use today, and that customers are having an incredible impact with. And have you talked to Satya Nadella about this? You must move in the same circles.

6:24Of course, I talk to a lot of people and I tell them this, this idea that these AI priests and priestesses are out there telling the world things about AI that are not true is a huge disservice to these enterprise-y customers who can, number one, increase their margins, increase their revenues, augment their employees, improve their customer relationships. Yes, you can do all of these things with AI. But this other part that we are all living in minority report, no, we're not there yet. Maybe we'll be there one day. Terminator, maybe we'll be there one day. War Games, I hope we will never be there one day.

7:08To me, part of what's special about your agents is they're working within this scaffolding of how your software already works. How do you define agents at Salesforce? How are yours working in a way that can deliver a real result? Let's also extend the false prophecy one more second before we jump into that. Number one, everyone is not going to need their own nuclear power plant to run their data centers. This whole energy fantasy. And number two, that everybody has to start DIYing it and building it themselves. Let's just take these down one by one. And technically, let's just walk through exactly how we're going to do it.

7:44Yeah, let's start there. So how are your agents at Salesforce working in a way that can deliver a real result? All right. Well, let's just take it from the tippy top, which is, you know, we have a customer, Wiley. They're the textbook company. They're great. If your kids are in school, they're buying their textbooks from Wiley. So Wiley typically, who's been on our platform, by the way, for more than a decade, they're awesome. You know, they use our sales product and our service product and our marketing product and our commerce product and our analytics product, so forth, Slack. usually about this moment they're like you know business is scaling we need to go hire all these people just for the next 90 days it's our season let's roll and we said wiley what if your workforce had no limits you mean we don't have to hire all these people like once a year exactly and we're going to introduce you to not only your new sales force or your new service force your new marketing force, your new commerce force.

8:38We're going to introduce you to Wiley's new agent force. And they said, we don't believe it. And so to be clear, these AI agents are a drop-in replacement for the seasonal customer service agents that Wiley would normally have to hire during this peak of the textbook season. Sales agents, service agents. So service agents, basically handling, I got the wrong textbook. Please send me the new one. Sales agents. Hey, hey, we know you bought that book last year. Have you thought about buying the new one this year or marketing agent? Did you know we have a promotion happening at Wiley for these next generation books, but also build your own, roll your own.

9:17And you're not going to hire your own engineer to do it. They're not going to hire people, not programmers. They had to become, by the way, 10 years ago, AI experts when we entered predictive. Then they had to become data experts, and now they are going to become agent experts. And they will be agent blazers in the way that they are trailblazers today. I feel like we're talking to the Oracle here in the Matrix. You're telling us about what you say are false prophecies. What about nuclear? Some of your rivals are talking about how nuclear energy is going to be required to run all these server farms that power AI.

9:50And there's this kind of rush around the world to generate a lot more power for this stuff. And it sounds like you don't believe it. You're holding up a red pill. the truth. Here you go. Here you go. I'm ready. Here you go, baby. Okay. Listen, we all know where we're going. We saw the movies. We all know we want to have these incredible things. Okay. But what it's going to require is software engineering. And as somebody who's been a software engineer for 40 years or 45 years, I can just tell you that I love software and this is the next version of software. Let's not get away from that. Like even LLMs, we all know what those are, large language models.

10:31We all remember the first time we're on chat GPT and said, actually, I just said, what is the best chicken soup recipe? And it's like, don't worry, I got your back, Mark. And it's like, here it is. Amazing software. But LLMs are not a data storage mechanism. We don't even know how the data is stored in the LLM after it's trained. Okay. But it is a cool intelligence engine that we can connect our data and database and metadata and different pieces to and then bring it all together in a really smart way. It doesn't abstract you from also managing your data and all the other safety, security, sharing model.

11:14But you're not planning on buying any nuclear power plants. We look at our CapEx on our income statement. You don't see it scaling crazy. And we're number one. Now, why is that? Because our platform operates in a highly efficient way. And look, this is an exciting moment. But in 2008, when Steve Jobs called me and said, Mark, FedEx and you the iPhone, you're going to love this. That was also very exciting. And I felt the same way. Like, I can't believe what's happening. And now I go, I can't believe what's happening. And in the same way with my customers who I love, like, oh, I'm going to transform your businesses with a phone.

11:51Now I'm saying, oh, I'm going to transform your businesses with AI. I'll give you an example of AI, power of AI, Disney. So here's Disney. They use us on every customer touchpoint. You go to the Disney store, they've got Disney Plus, the call center, Disney guides. If you go to the park, you know, with your kids and they're walking, getting you through the lines. And if you look down and look at your Disney guide's phone, you'll see Salesforce and Slack on the phone. And here's the funny thing, like, because I go to Disneyland about four times a year. And I'm on my way to Rise of the Resistance.

12:25It's a great ride, but very technical. And it breaks a lot. It just does. It's just the nature of the ride. and all of a sudden the guide is going to turn to me and say oh mark rise is down but we're going to bring you to another ride where do you want to go instead the agent is going to call him on the phone and say hey i know you're on the tour of mark rise is down i already looked at mark's ride history i know every ride he's ever been on in the last 20 years and by the way i just scanned across the whole park to look at flow control at every ride, I recommend you take him to Toontown. And I'm on my way to Toontown.

13:07The AI is able to do things that we cannot do as humans. We've just heard how Mark Benioff is reshaping the company he co-founded. But does that mean he agrees with the idea that all companies have to be led by their founders to succeed? Here is advice for how every one of us can operate in founder mode. That's after the break.

13:41In the Valley, there's been a lot of talk about the power of founders. There's something called founder mode, which is essentially this idea that founders have more power or more effective in bringing about change or running their companies. I got to imagine you really have some thoughts on founder mode. Are you an advocate of this idea that really the founder is the one that can affect change? I don't really buy into the whole founder mode thing, by the way. So I think the founder mode, I give it an F also. But I think my F is that it's a fail. I think every executive should be in founder mode.

14:16And the founder mode thing almost emasculates normal executives from feeling that they can't be in founder mode. But listen, at the end of the day, unleash the founder within. All of us have that energy inside us. Maybe I should write a book. I don't know. I think part of this kind of conversation, this debate is around strong leadership has been some of CEOs struggling or voicing concern about return to the office. We've seen Amazon pulling everyone back five days a week. Where is your thinking these days on work from home? It's such a great question. And I'll tell you that, you know, first of all, you're talking to someone who's always worked from home.

14:52And, you know, because you're seeing me on video, not just audio, I am at home. You see my dogs walking around. And I have that whole ADHD thing going on. So I need like an environment with very low stimulus. And I'm just more effective from home. I just always have been. And we've grown Salesforce to 38 billion and 75 ,000 employees. And I've done it from home. and more than 25 % of all the Salesforce employees have always been, always been remote. And a few years ago, there was this pandemic and in the pandemic, we all had to go home. And then we were all at home and all of a sudden we all discovered this thing called Zoom.

15:32And then we bought our microphones from Yeti and we're all sitting there talking to each other and going, oh, this is actually pretty good. So, you know, we can be effective from home and we can be effective in the office. I think that for some executives, especially the new ones who are coming in, onboarding, kids right out of school who are meeting people, collaboration, you know, the whole idea of the social environment in the office. I think being in the office is extremely important. And I think that when I was in that zone for my own career, that swivel chair experience where I'm in my chair and then I'm looking at the person next to me and they're talking to me about what to do and how to do it, that's really important.

16:15But is it important for everybody and is appropriate for everybody? Maybe. But, wow, we have these great new tools that we can do so much from wherever we are. We can work from anywhere. That's been one of the great things. and the AI revolution in agents. These agents, by the way, they're also not in the office. They're working for both. It sounds like you have a nuanced kind of view on this. We're not going to see some kind of sudden emergence of a five-day mandate for everybody anytime soon in Salesforce. At the end of the day, you're either delivering the value or you're not. I don't really care, honestly, if somebody's at home or not.

16:56Are they getting their results or are they not? My guru in management is Hal Janine, who wrote the book Managing. He had a great phrase, managers must manage. You know, managers must manage. You're either getting the results or you're not getting the results. So that's just our new world. So let's not degrade everything we've invested in in the last 20 years. We've done amazing jobs with the technology, the computer, the Wi-Fi, the phone, the codecs, like the video codec that we're looking at, the quality of the video right now, it's amazing. That's the compression decompression and, you know, algorithms that are letting us to have this experience, you know, with all three of us at home and talking on this thing.

17:40We've invested massive amounts in technology. Let's really use it. We've just heard how AI agents are going to transform the world of work. When we come back, Mark Benioff talks about when he thinks human-level AI will arrive, and why he's backing a startup from one of the pioneers of modern AI, who is trying to create smart machines in a radical new way.

18:17So, there's one other area I want to ask you about, Agents. You own Time Magazine. which is a fact about you that I think a lot of folks don't realize. And when you're thinking about AI and agents, does that thinking ever intersect with how you think about the media and specifically how time or other outlets are preparing themselves for the future? Is the AI taken over? No. Has AI cured cancer? No. Is AI curing climate change? No. So we have to keep things in perspective here. We have incredible tools to augment our productivity, to augment our employees, to improve our margins, to improve our revenues, to make our companies fundamentally better, have higher fidelity relationships with our customers.

19:06But we are not at that moment that we've seen in these crazy movies. And maybe we will be one day, but that is not where we are today. We have to keep this temporally correct. And this idea, like I've heard, you don't need to manage systems of record. Oh, software writes itself now. You know, that technology has improved to the point where we don't even need enterprise software. It's like, oh, that is awesome. And what planet are you on? Because I'm on Earth. And I can tell you what the current technology can and cannot do. And it can do a lot of cool things, but it cannot do everything. So what is your prediction for when AGI is?

19:51This is kind of that godlike AI that can rival human intelligence. Do you see in the next 20 years? not coming at Christmas. And it's probably not coming Christmas next year. Look, I actually think we're hitting the upper limits of the LLMs right now. It's a very cool new model. By the way, if you look at the history of AI, it's a history of models. And we have the new model that we're hanging out with. But these models, what happens is the history of these models is they mature. And then we kind of get into an AI winner as we're ready for the next one. Like Fei-Fei Lee, who runs AI at Stanford.

20:25She just started this new company. I'm an investor now. And she believes that the next model will be multisensory. And this idea that there's going to have to be many, many different components of things, cameras and audio and all these things. And it's four dimensions. It's not just in the dimensions of our current physical world, but also time and the ability for it to span all these things. It's not an LOM. But she's going to have to take her engineers and they're writing their code and working on it and on and on and on. And maybe they'll release it at some point in the future. and at some point that may replace the LLM.

21:01And we just have to be careful how we think about these things. So we have to get back to reality. And I think we all got drunk on the chat GPT Kool-Aid where we're like, and what is that chicken soup recipe? And can you summarize these 10 articles? Oh my gosh, cancer is cured. It's like, guys, we've got to keep these things in perspective. It's a tool. And I hope that we're using it to improve humanity and making things better. Mark Benioff, thank you for coming on our show. It's great to talk to you always. Thank you for having me. I'm so grateful to you. Bye-bye now. Aloha. And that's bold names for this week.

21:45Michael LaValle and Jessica Fenton are our sound designers. Jessica also wrote our theme music. We got some help this week from Julie Chang, Catherine Millsop, Scott Salloway, and Falana Patterson. For even more, check out our columns on wsj.com. I'm Tim Higgins. And I'm Christopher Mims. Thanks for listening.

22:26the user survey so that the Prime Minister would like to CAT on its own personal Let's change now! The channel is available on this podcast channel for credit reform, post the podcast channel where it calls podcasts.