Marc Benioff on Agentic AI and Salesforce’s Next Chapter

In this podcast, Motley Fool host Dylan Lewis caught up with Salesforce co-founder and CEO Marc Benioff to discuss:

• Why Salesforce doesn’t plan to hire any software engineers this year.
• Lessons learned from major acquisitions, like Slack and Tableau.
• Where AI goes next.
• More.

To catch full episodes of all The Motley Fool’s free podcasts, check out our podcast center. To get started investing, check out our beginner’s guide to investing in stocks. When you’re ready to invest, check out this top 10 list of stocks to buy.

A full transcript follows the video.

This video was recorded on March 08, 2025

Marc Benioff: These large language models are mostly commodities right now and have been. I’ve been doing models for a long time in AI, and they’re constantly leapfrogging each other. We’re trying to leapfrog everyone ourselves. It’s a fun game that we play in the tech industry. But the large language model game has been won by nobody yet. This idea that the model that’s successful today is going to be the model that’s going to be successful tomorrow is not true.

Mary Long: I’m Mary Long and that’s Marc Benioff. He’s the co-founder and CEO of Salesforce. Salesforce helps other companies manage their relationships with customers through a number of different Cloud-based software offerings. Recently, Salesforce has added something called Agentforce to that suite of offerings. The idea behind Agentforce is that autonomous AI agents take action on behalf of customers and employees. Listen to any Salesforce earnings call, and you’ll hear Benioff talk a lot about this concept of Agentforce and agentic AI. On today’s show, my colleague Dylan Lewis catches up with Benioff. They have a conversation about this move from generative to agentic AI and what that transition means for companies and consumers and employees. They also talk about building versus buying and how Benioff feels about the acquisition of Slack and Tableau in hindsight. Plus, they talk about reckless predictions and where Benioff thinks the AI race is headed next.

Dylan Lewis: I’m excited to have you on because you are one of the most authoritative people on one of the biggest themes out there in the market right now. AI was the theme of 2024. It is continuing to be the theme, although it’s shifting a little bit. I think we’ve seen things move from a real focus on strictly generative AI to a little bit more of a focus on agentic AI, and that is where Salesforce is really living right now. Can you walk us through that transition and what you’re seeing in terms of development there?

Marc Benioff: Well, Dylan you already know, this is the most exciting time I’ve ever had in my career. This is also not our first AI rodeo. We’ve been doing AI for more than a decade at Salesforce, predictive AI, machine learning, machine intelligence. Our Einstein platform is probably the largest most important enterprise platform for AI in the world. We deliver about a trillion enterprise AI transactions every single week. The generative revolution has just been incredible. What we’ve seen happening with these large language models over the last two years is really astonishing.

But I think for a lot of our customers, they’ve been very disappointed with a lot of the solutions that have been given them or even shoved at them. Even Microsoft has really disappointed so many of our customers with what they call Copilot. Instead, we have a different approach, which is the next generation of our platform that we call Agentforce. The idea that you can really take and create something that we believe is going to be the largest and most important transformation of the entire tech industry, which is the digital labor revolution. The digital labor revolution is like a $3-12 trillion opportunity. It’s really different for people in my industry.

We’re mostly people who just help companies manage and share their information. We do that really well, customer information, all kinds of apps that we have to help companies manage their information. Now we’re really stamping out digital workers at scale and giving companies the ability to manage digital employees alongside their human employees. That is very different, very exciting and really the next level of our entire technology industry. This is a moment that none of us will ever forget.

Dylan Lewis: I want to dig into those digital employees and the idea of these digital agents. As you guys are rolling out Agentforce and having more conversations with your customers around it, what are you seeing in terms of adoption and what are you hearing back from your customers in terms of what’s useful and what’s not so useful?

Marc Benioff: We’ve never seen a product or technology take off at this level. We just reported our quarterly results. As you know, we just had our first $10 billion quarter. We’re the second largest software company in the world behind Microsoft, and the fastest-growing product that we have in our product line is Agentforce. Now, we probably did about 5,000 different types of Agentforce transactions in the quarter, 3,000 of which were paid transactions, and we did about nine digits in total contract value in Agentforce. Now, why it’s important is we’re already starting to see customers deploy it. It’s getting deployed at a much faster rate than we’ve seen other products. Even this morning I was buying some new shoes from Vans. I went to Vans.com on my phone. Here’s my phone right here as a point of evidence. [LAUGHTER] I could probably bring up the site. It’s probably still on there.

I’m getting ready and I wear really wide shoes. I just got on there and there’s five icons at the bottom of the Vans site now and the middle icon is the Agent icon. You click on Agent and you have Agentforce, and that is deeply connected through the whole company. I’m very interested in, hey, what are the wide shoes and which shoes are going to be the most comfortable for me? Also, hey, I need support on returning shoes that I just bought and I need help here, I need help there, and the Agent is able to work out and resolve all of these things for me. That is what is really cool and exciting. Hey, what can I say? This is happening faster than I expected. I didn’t know that was going to happen. The CEO of Vans, somebody I’ve known for a long time, used to be the CEO of a tech company. It’s a company called VF Corp, their amazing company.

I’m like, “Wow, this is unbelievable.” He said back to me, “Well, people are telling me this is the most exciting thing that’s ever happened on our site.” I’m like, “It’s hard to put it in context what is happening because instead of having to call some advance, someone to email someone advance, I’m able to work with a very intelligent digital worker who’s helping me to complete my transaction to get the customer service and support and all the things that I need.” That is very exciting. That is what I am motivated by every single day now. I’ve got so many customer stories and that one I just found randomly this morning when I got up. I’m like, let’s just keep going. It’s so cool.

Dylan Lewis: As someone who wears a 12 EE, I’m appreciative of everything that you’re doing through my shoe-buying experience.

Marc Benioff: You should go and try them. You’ll see they’ve got a really nice, cool, wide shoe on their site.

Dylan Lewis: Earlier you mentioned Microsoft, and I’m going to take the bait there. You’re not alone in pursuing this type of technology. They are also announcing that they have agents out there. They are trying to work that into a lot of the software they bring to their customers. When you think about these offerings, what do you see as the differentiators and what are you really competing on?

Marc Benioff: Well, I think there you know, obviously, we’ve transformed, changed the conversation, as you can see that before it was about a Copilot and they repackage and resell ChatGPT as their AI solution. They don’t really make one yet. They’re trying to make one. They’re basically an OpenAI reseller. I think the problem is that for a lot of the customers, they’ve had to DIY that into what is a complex enterprise architecture, and it has not exactly worked. You can talk to the customers because likely we have a stream of customer stories now. I think that the reason that’s happening is we’ve just extended our core platform. We’re not really asking customers to add anything new. We’re just asking them to turn it on.

I was with one of our largest customers an airline in Singapore, Singapore Air, we’ve all heard of it and the CEO Goh is amazing and they’ve been a customer for a long time, great relationship with a client. They’re like, “Yeah, we’re just turning this on. This is amazing. We now have the agentic layer around the entire airline. We can do sales, we can do service, we can do marketing, we can do all of these things.” That is really cool and really gives us this ability to augment their workforce, and they’re shocked at how fast they can install it and get it going because they’re just flipping a switch and turning it on. This is different than that Microsoft approach, the Copilot approach. First, Microsoft Copilot and we talk about that, customers realize it’s not just one Copilot. They have a dozen Copilots on all their different product lines, none of them are connected together. It’s not one source of data. It’s not this one piece of enterprise code. It’s a real hodgepodge approach. That’s why it’s failed. I think our approach is much simpler, easier and more successful. With real customer success because of the architecture. I think it’s the architecture that’s making it difference. That ability to deliver that, high-quality semantic layer on the data that’s really giving us this capability with AI.

Dylan Lewis: I think with artificial intelligence, one of the big things we’ve had to acknowledge a lot in the last couple of years is it is going to change the way that a lot of businesses work and the way that a lot of companies work. I saw that Salesforce isn’t planning on hiring engineers this year. How do you see agentic AI playing into the way that work gets done for you guys, but also for a lot of your customers?

Marc Benioff: Well, I have a lot of software engineers, probably more than 10,000, maybe close to 12,000 now, and they’re a huge part of our workforce. We have 75,000 employees, but I’ve noticed they are just so much more productive using these next generation of tools. The so many new next-generation productivity tools. In that same way, humans will be augmented. In the example of Gucci, we took our technology and we applied it to a call center in Florence, Italy. We augmented the call center, and it’s really incredible what happened. All of a sudden, we were like, OK, are they going to need less support agents? It turned out revenues went up 30% in the call center because the support agents were able to sell products, market products. They could understand products that before they did not have or understood. That was awesome and incredible that we could make their revenues go up and make them more profitable.

We’ve seen the same thing happen in a company in Miami called Lennar. They’re a huge home builder in the United States, maybe the largest or one of the very largest, and they have two great co-CEOs, Stuart and Jon, who I’ve been talking to because their team came to Dreamforce which is our annual Powwow Lapa Balza conference concert. We have like 40, 50,000 people come to San Francisco for that one. Their team came. They’ve been a customer for eight years. They do sales and service and marketing, they do all these amazing things. The team came back to Miami and they ran a hackathon to look at how could they use Agentforce inside Lennar. What they found was that there are five critical use cases. If they could do it, they could potentially generate hundreds of millions of dollars of more profit, huge amounts of revenue and basic things like being able to provide an agentic layer on all of their home sales that are 24/7, lead management, selling new products to existing customers like insurance or mortgages. You can not have to have a very big fantasy to see, wow, there’s a lot of things these agents could do because it’s just hard for them to hire more workers. They’re really extending their human workforce with these digital workers, and it’s a digital labor revolution.

For a Lennar or for any of these companies that I mentioned, it’s an extension. In some ways, it is a transformation. Another company in Florida that we work with very closely is Disney and for Disney, of course, it’s augmenting not just like the Disney Plus call center, but it’s also really providing a different interface for, first of all, the employees. They’re cast members. The products at Disney are so complex and so deep and they have so many that when cast members are trying to understand all the products and trying to communicate to the customer, they need tools to do that. That’s number 1. But the second thing is the ability to deliver an agent that can talk directly to the customer on a complex product configuration. Like for example, I’m coming to Disney World. I want to have a really incredible hotel room. I also want to have a restaurant reservation. I have an allergy. I need to make sure that the restaurant knows that when I get there. I also want to make sure I have all the accompaniments for the park when I get there, the right park pass, the right promotion, the right additional Disney Guide, all of the different products of Disney, they start to get bundled together into a SKU for this client who’s coming to the park on vacation with their kids. Wow.

They’re humans, they do have a hard time putting it all together. The agents can help and the agents can talk directly to the customer and then you can see this even on our own website any moment the customer, boom can move over to a human agent. Then a human agent gets a screen because we’re all operating in the same app, the same dataset. It’s that same Trinity. Then the human is looking at, oh, yeah, I can see, here’s the client, here’s the allergy, here’s the hotel, here’s this. This is not something that I’m just making up on the fly. This is something that we’ve built that I’ve seen getting deployed, that I’m excited about that this can be really is possible and maybe a year or two year ago, it was not possible.

We really have made huge strides. Whether it’s that hospitality or media company or financial services company or consumer product goods company or airline or a clothing company or tech company. Every single company can benefit from this technology, and that is why this is going so fast.

Dylan Lewis: You just talked about how this was something that you guys built. I know historically, Salesforce has been a mix of you guys home grow some stuff. You acquire companies that look interesting and make sense as part of your overall suite. In the AI zone, is that something where you guys want to focus more on home growing, or are you still looking out there and seeing what opportunities might be there in the market for you?

Marc Benioff: We believe that there is for a high growth, high-quality tech company, the perfect balance between organic and inorganic innovation. Even in the example of Agentforce, I was able to accelerate it because we already had the vision. We knew what we wanted to do, but we knew we needed to have an extended team. We wanted some technology that we did not have inside. Eight or nine years ago, 10 years ago, I bought a company. It was an amazing company that we integrated deeply into our product that became Einstein. The founders of the company left after about five or six years, and one of them started a company that we invested in. Then about a year ago now, I bought that company, and it was an aqua hire in one level.

That means just hiring the people. Another level, I like some of the technology they had, but I knew I could simply and quickly extend my platform with it to bring forth our AI in a new way, and that is Agentforce. Acquisitions are important. In other areas, we’ve entered new markets, and we have critical products like Slack. A lot of people use Slack. It’s a huge product. Incredible. He had a huge growth year last year, and one of the reasons why Slack is growing and more relevant than ever is it’s an incredible place where agents can live. In fact, I just retweeted, a developer came along this morning and said, Hey, I can see how Slack is going to become this place where all my agents are now operating, I’m communicating with them, and yet they’re all collaborating with all the other workers in our company. That’s very exciting for us. Slack was an incredible acquisition. Tableau is another one. Now, a lot of folks haven’t seen the new Tableau yet. It’s called Tableau Next. I have always loved Tableau. It came out of Stanford.

They did this amazing work in business intelligence and analytics, always wanted to buy the company, and finally had the chance to do that. The product is awesome, but we’re moving into the agentic world. I have rewritten it. I rewrote it over the last basically a year. In that case, I don’t need to acquire it already had it. But now Tableau has a deeply integrated agentic layer, data Cloud, and app layer, just like we’ve been talking about when tableau users see this coming up in the new conference, they are going to honestly flip out. They call themselves the DataFam. To see the next generation of Tableau that it could have another decade or two decades of growth ahead of it with a new architecture is really exciting, not just for me, but I think for the entire community. I’m really excited about that, the new Slack, the new Salesforce. I think refreshing our whole product line with this trinity approach, rewriting everything into this architecture. This is transforming our company and our customers, and we’re going to see a lot of new exciting deployments this year and refreshed deployments of what was already deployed, getting extended.

Dylan Lewis: I’m glad you brought up Slack because I was going to have to ask you about it. It’s the biggest acquisition, and Tableau also something that has a lot of brand recognition. We’ve used at the fool, and I think a lot of people who follow the software space know well. Now that you’ve had a little time with both those, I’m curious, how would you grade those acquisitions with the benefit of some perspective? What’s the work to be done there that you still see?

Marc Benioff: Well, number 1 with Slack, I can give you a lot of class on acquisitions. I’ve done more than 60 acquisitions. It’s 25 years of Salesforce, more than 60 acquisitions, and so there’s a lot of lessons, but I think the Slack, number 1, when we bought the company, great brand, great product, great team. I think that the team that we acquired had a slightly different vision of the future than we did and why we bought it. They were struggling, competing against Microsoft. They didn’t understand how to differentiate themselves. There were issues, which is why we were able to buy it. Otherwise, we wouldn’t have been able to buy it because it was incredible. I think they were on their way from one billion, on their way to two billion. Now the thing is on its way 2 billion to 3 billion, and there aren’t that many $3 billion products in enterprise software, Slack is one. These products have grown, they’ve been refreshed, rebuilt.

But sometimes you have to rebuild the team. They have to realize they’re part now of a different company Salesforce. It’s a different culture. We have a different vision, we have a different architecture, and they’re going to benefit from being part of our architecture because they want to be part of our DataCloud. They want to be part of Agentforce, and that’s what gives our customers this incredible value because now, let’s take any one of those stories that I just went through, but let’s say you are in Singapore Airlines and you do have Salesforce as your platform and you have these agents and you have this agentic layer and you are doing sales and service and marketing and commerce and all these things. Now you also have Slack, your employees are collaborating and that’s all working together, and that is what is cool, and that’s why I love the new Slack, what we’ve done with it. We’re showing that off today a trail head DX, for example, customers are loving it, tweeting about it, we’re showing off some of the work. This is what gets us excited is to innovate, to create, to conceptualize the future.

What I love is you can take some of these brands that they may not have a clear trajectory without us, but then you bring them into our architecture, and it transforms them and gives them not only an expanded customer base to sell to because all of our salespeople now are selling Slack, for example. We have a huge Salesforce selling Slack where before Slack only had a couple of hundred people selling Slack. Hey, if you’re going to compete with Microsoft, you need a company like Salesforce who, A knows how to market and position against Microsoft but B can sell directly to customers. I was just talking with one of our largest customers yesterday, the CEO of a large pharmaceutical company, a great company. I use their products myself. I wear contacts. I use our contacts. He’s a huge Slack user, running his company. I was telling him how excited I was about the new Slack and what I thought was going to happen and what he could do now that he could not do before. I was really happy about it.

Dylan Lewis: As we wrap up here, we’ve spent a lot of time talking about AI, generative AI, agentic AI, I’m going to ask you to peer into your tech crystal ball a little bit. Do you have a reckless prediction on the direction or the development of AI, whether it’s in how customers will use it, how businesses will use it over the next five years?

Marc Benioff: Well, I think we see a few exciting things happening in the industry, and I think people need to reframe how they’re thinking about this. Number 1, you see these companies putting hundreds of billions of dollars into these data centers, not just one. You, Dylan, could probably list five or 10 of them on your own, without even notes in front of you. I could, too, and I am driving my cost down. By moving my platform to work on as many of these as I can because I can take advantage of a lower-cost infrastructure. I want to be a software hyperscalar. I don’t want to be a hardware hyperscalar where we have to build these data centers.

I want to have a portable platform that works across all of these data centers. That’s a huge priority for me, and I think that’s a tremendous prediction for the future, which is everything is going to be a lot lower cost to deploy. That’s super powerful that you have companies putting so much of their capital, and I don’t want to impact my balance sheet and income statement by doing this, so I’ll take advantage of the investments that these companies are making and partner with them. That’s number 1. Number 2 is you can really see, I’ve said this quite a few times, and I think initially people didn’t feel it was true. I think now they get it. These large language models are mostly commodities right now and have been. I’ve been doing models for a long time in AI, and they’re constantly leapfrogging each other. We’re trying to leapfrog everyone ourselves. It’s a fun game that we play in the tech industry. But the large language model game has been won by nobody yet, and this idea that the model that’s successful today is going to be the model that’s going to be successful tomorrow is not true. You can see, for example, I’m getting ready to delete Open AI and ChatGPT off my phone which is something that I would never have even thought of six months ago. But now all I’m using is this Grok app, and it’s fun.

It’s got this crazy edgy voice capability. Elon Musk has transposed his personality into the way he’s divided up the voice platform. I don’t know if you’ve tried that Grok-3. It’s crazy. Everyone’s got a crazy story. I did a demo yesterday for a friend of mine. They couldn’t believe it. Like this unhinged mode and storyteller mode and all kinds of other crazy modes. I’m like, Well, this is just better right now and more accurate and more fun than the ChatGPT app. Maybe I’m going to delete that off my phone. Why am I paying $20 a month for that? I’d rather be paying for this. I probably have six or seven of these things on my phone, and I’m getting ready to pair them down. Then, is Grok going to be the best app a year from now? I don’t know.

Maybe I’ll end up putting the ChatGPT back on. Maybe I’ll be using you.com as a great one. Maybe I’ll be using Perplexity. I have no idea what’s coming. I couldn’t have told you this is what was going to happen a year ago. That’s a commodity part of the infrastructure. The LLM, it’s like disk drives. Disk drives and RAM and the computer. These are all commodity infrastructure components that are being made by large tech teams competing with each other, and that’s how we drive our costs down. That is something that I think is very important that, yes, you have a new commodity part of the infrastructure, the large language model. When I look at those things, I think the third piece is everyone saying, Hey, AGI is coming, but nobody exactly can define it. Nobody can describe it. No one we can seem these movies, our futurist Peter Schwartz was a key writer on Minority Report and war games and Deep Impact. It’s like, let’s say AGI is coming and that we’re all going to be talking to how and saying, Hey, I’ll open the doors and, no, I don’t want to open the doors. But the doors, that’s not how my doors are working right now. Maybe that’s coming but it’s not here yet. I’m like, I’m willing to be optimistic on AGI, but I also realized like, what does AGI really mean? I think that for those LLMs, they’re great, but we know they don’t have 100% accuracy either.

They’re like, we have the highest accuracy, and the highest we’ve seen with some of our customers is in the 90s. We’re in the mid-80s ourselves right now on our own deployment. We’re going to try to get into the 90s. But we’re not going to get to 100 because the way that the thing is architected with these words are all hooked up with each other, and then they turn into these spider, neural network words, and these ideas aren’t going to give us 100% accuracy. New models will have to come with new AI ideas, and nobody’s going to want an AGI that’s only 50% accurate. That’s going to cause chaos. Maybe AGI is just going to mean 100% accuracy, and when will we get there? I don’t know. There’s some fantasy of what the next generation is going to be. Maybe it’s going to be a multi-sensory model, a world model. It’s a three-dimensional rendering model. But those are the three things I would say. One, much lower cost infrastructure that we’re all going to be able to ride on, two commodity LLMs and three, yes, these future states that are very exciting. I hope that we get there. The future is definitely coming, and we’ve seen the movies, and we all want that to happen. It’s going to be exciting.

Dylan Lewis: Well, Mark, if your game in a year, we’ll have you back on to see which phone app you are using and also the progress that you’re making on Agentforce.

Marc Benioff: It’s great to see you, Dylan.

Mary Long: As always, people on the program may have interest in the stocks they talk about, and the Motley Fool may have formal recommendations for or against, so don’t buy or sell stocks based solely on what you hear. All personal finance content follows Motley Fool editorial standards and are not approved by advertisers. The Motley Fool only picks products that it would personally recommend to friends like you. For Dylan Lewis and Marc Benioff, I’m Mary Long. Thanks for listening. We’ll see you on Monday, Fools.