CPO Rising Series: Salesforce President & CPO on AI's Transformative Power in Product Leadership

Thank you, Renee. It's an honor and thanks for all you're doing to promote great product leaders around the world.

Yes. A little bit about myself. I have a real passion for building products and I've worked in enterprise software for over three decades. I work as the president and chief product officer at Salesforce, the second largest software company in the world. We're the leading provider of what is known as CRM software, but also playing a lot of other spaces with Slack, Tableau, and Mulesoft. And we are super excited about the move to agentic AI and generative AI, which we'll talk about. I have a long history in the software business. I'm an engineer by training. I went to graduate school and out of graduate school I worked at this little software company called Oracle in the 80s. There I met somebody that I currently work with, Marc Benioff, who went on to found Salesforce. I met Tom Siebel, who was a great mentor to me, and he and I left Oracle to found Siebel Systems, which was a pioneer in CRM before Salesforce. And I've been fortunate to be involved in three or four very successful software companies. So it's a great team sport. I've made a ton of friends in this industry and my passion is having amazing teams building amazing products. So that's a little bit about me.

And those are the only three CEOs I've worked for. So it's a sample set. And then I founded my own company, Velocity, that Salesforce acquired in 2020. So that's how I got to Salesforce. So yeah, I've had great mentorship.

Sure. Yeah, I think some of the fundamentals of building great software or great products, if you zoom even further out, are still intact. You know, it really requires great taste. And so it's not just about implementing a set of features, but there's sort of a zen or an art form to it as well as the science of building products. And I believe both are equally important. There's the left brain and the right brain part of delivering amazing customer experiences with award-winning products. And so those things remain constant. Some things go in and out of fashion, sometimes things remain in fashion forever. And so those are sort of timeless in my view. But what's changed is the rate of velocity and the acceleration in our industry. So we've been tripling down on AI in the last decade and particularly in the last couple years at Salesforce, first with predictive AI, now with generative AI. And literally the AI industry is changing every day. And so this is like a global AI race and the best teams in the world, the best technology companies are all competing at a global level and the stakes are super high. And so that really changes the way that you build software. So Salesforce, 25 years ago, we pioneered three releases a year, which was sort of revolutionary. Unlike the on-premise companies like the SAPs and Oracles of the world, you had to do a giant IT upgrade. Now our customers, three times a year, literally 200,000 plus of our customers and tens of millions of people just get the new release just like that. It's just magic in what we now call the cloud, and that's sort of become a standard. But we're going even faster and faster with our AI and data products and increasingly all of our products where we're doing weekly and monthly releases, and we literally build and rebuild the software every single day. And so the cadence and the velocity and the agility has changed dramatically. And I think in the product designing and building area, it's really gone from the classic terms of waterfall to agile. And agile is everything. It's the only way any successful software company builds software now. And the theory behind that, of course, as you all know, is with waterfall you might have a one-year release cycle and you get some customer needs, and by the time you get to that one-year release, the software is already out of date or the requirements have already changed. And so now there's the term of failing fast where you do things very quickly and you immediately get feedback from customers and you're constantly revising the target, and you maybe are never done revising that target. The software is never done. It's sort of more liquid software is how I think about it. And you're constantly talking to customers, constantly interacting. And the faster those feedback loops, generally the better the output. And so it's changed quite a bit, but I would also say that the fundamentals are still in place too. You still have to do the basics to really build award-winning products.

Yeah, it's a great question. I had a mentor a long time ago tell me early in my career when I started out as a product manager that great product managers have great taste. And the way I explain that is if you go to the men's tie department to go buy a new necktie, 90% of them are horrible colors with horrible patterns. And it's kind of amazing that they're all so ugly. And yet people buy those or they wouldn't have them at the store. And so a lot of people just generally don't have great taste, which sounds like an elitist thing but it isn't. It's just, I think it's just true if you really look at it. And great product managers can not only prioritize features, which listening to customers and prioritizing what they want is definitely a key success factor, one of the most important things that great product people do, but you also have to step back from it. And just because a customer asks for something specific doesn't necessarily mean that's exactly the right way to deliver on that. And so the taste and therefore the art of it, I think, is super important. And two topics that I think help deliver on what I call software artistry are design thinking, where you're really looking at the design of the software and the user experience much more than just the features themselves. And then there's a relatively new area that I call SaaS 3.0. Most of the leading SaaS software companies, and we're certainly all in on this, engage in product growth where literally the product is so simple you can just click a button and just blow it up, and you get immediate feedback on what people are using, and people can dial up or dial down what they're using as they use the software, and you're really trying to optimize the experience of software as a service. The software part is still important, but the service part is that entire what some people call the whole product experience. And that is super important. And that helps you deliver on what I call software artistry. So that it really is memorable, it really is unique, it really is differentiated from your competitors. And the best companies in the world really focus on those things: software artistry, product growth, and really thinking about the whole product experience.

Yeah, I ask questions in the interview process. I've been doing this my whole career that sort of try to narrow in on that. So I ask people, one of my favorite questions is, tell me about a software product that you like and then tell me about one you don't like and explain to me why you don't like it. And that's how I got one of my first job offers. I didn't go to the company, a big successful software company. But it was back in the 80s when I was coming out of grad school. I could name the company. It was Microsoft trying to hire me a couple times, and I explained to them all the problems with Microsoft Excel, and then they hired me because they realized that I was a power user and I could tell them all the flaws. A lot of design flaws, gets to software artistry, where they had implemented a lot of the features but they certainly weren't fun or pleasurable to use.

Yep, that makes perfect sense.

Yeah, it's a great question. Let me just step back and say my personal belief is AI changes everything. Literally the world as we know it will never be the same again. And it's not going to happen in 10 years, it's happening now. And in two to three years, the world as we know it is going to be completely different. And the example I would give is, Renee, if you told me 25 years ago that I could use a really cool mobile app on my phone and that mobile app would disintermediate the taxi cab industry, I would have laughed, but that's exactly what Uber did. And now if you go on the streets of San Francisco, there's a company called Waymo that's disintermediating Uber. They're doing this completely in an autonomous fashion with autonomous driving. It's safe. If you haven't tried it, I recommend that you do, because it's sort of fanning out now to other parts of California and ultimately I think around the world. And this is the future. AI literally changes everything. And Uber's announced that they're going to do this too. So it's not that Waymo is necessarily one, but I like to use two examples that are Salesforce customers: Uber and Amazon. They changed the game in transportation and in e-commerce and retailing, and both of those are Salesforce customers. And so they used our software and other technologies to completely change the game. So AI changes everything, and we recognized that 10 years ago at Salesforce. So we went all in. And I think we were the first SaaS software company to go all in on AI. And we do today two trillion AI transactions every week. Two trillion. And those numbers are way up year over year. So I think it's the high watermark in the SaaS software industry. And a couple years ago, even though we had built our own large language models ourselves, and the entire software industry and the entire tech industry woke up when ChatGPT was introduced. We knew what foundation models were, we knew what large language models were. We have an AI research team run by a Stanford professor that was building his own models, but no one really knew that you could do something quite so compelling as ChatGPT. And so we decided to not double down but triple down on generative AI, and we put generative AI in all of our products. And then it's recently culminated in our most successful product launch ever, called Agentforce. Now in a couple of minutes, you could come in with your laptop and log in with your Salesforce credentials and you can turn on AI agents that work and do real meaningful things for your company. And companies like OpenTable, Vivint in the home security space, Saks Fifth Avenue, we demoed how they're using Agentforce at Dreamforce. Or Matthew McConaughey was recently in our Super Bowl ads with Heathrow Airport using Agentforce to have a better experience while you're in the terminals. And so we think that AI, as I mentioned, literally changes everything. And we made a hard pivot. We were already working on AI, but to really go long on this latest set of generative AI and foundation model capabilities, which we think is a big deal.

Sure. Well, our CEO Marc Benioff was all in on AI as I mentioned. We started 10 years ago. We acquired a couple different AI companies. We established our AI research lab that's run by the Stanford professor of machine learning and data science. And so we have a series of PhDs and experts in AI. So we were well equipped. We have one of the best AI teams in my opinion in the world. But when generative AI came out, we assembled a meeting of our top 100 product and engineering leaders and Marc beamed in as a special guest for a segment. And I sort of teed it up at the beginning, which was the context of AI changing everything potentially, even though this was brand new at the time. And the decision to be made, because I'm a big believer you want to frame meetings in what decision you're trying to make. I teed up and said, 'Listen, my personal opinion is we should make a hard pivot and go all in on this, but we're going to talk about this for two days with our top 100 people.' And at the end of that two-day meeting, we talked about all different aspects of it and how to do it, what to do. It was 100 to zero that we should do a hard pivot. Everybody was nodding their heads vigorously. And so we went all in, first building AI agents and what we call AI in flow of work, and then ultimately Agentforce, not only helping humans in the flow of work but now doing this with autonomous sales and autonomous service, agents that can take orders, agents that can manage inventory, agents that can onboard employees, agents that can do ITSM, agents can literally do everything. And now those agents run in Slack. So I literally, 90% of my work is automated by agents working in Slack every day. And it's a huge productivity boost for us. I think we're the only software company that I'm aware of that has AI agents running on our main website. So if you go to salesforce.com, you can try out Agentforce right there on the homepage. If you go to help.salesforce.com, you can try out Agentforce. And we've deflected about 90% of the questions people have. But to simply automate the answers to every single question that you could ask about Salesforce in every single language about every single product is hugely beneficial. And so our customers see that too. And so it's really, really exciting. So a lot of people are talking about agentic AI. We're using it ourselves and living it, and the experience has been transformative. But the pivotal meeting was that one where we decided to pivot. We moved a few thousand people from whatever they were doing before to just work solely on this. And then we have something we call our V2MOM, which sort of prioritizes all of our work for the entire company. And we put Agentforce at the top of that list. And now it's Agentforce and a supporting technology that is our data lake technology called Data Cloud, which is the fastest growing product ever built in Salesforce history. So it's like our version of Snowflake and Databricks built in. And this product line has grown twice as fast as Databricks and Snowflake did at that amount of time. It works with Snowflake and Databricks and Google BigQuery or Microsoft Fabric or Amazon Redshift, so it works with the data lakes and databases and apps that you have, but it powers our AI Agentforce platform because we believe, and I think it's proven to be true, that data powers the AI and the AI powers the customer experience of the future.

It's phenomenal. I mean, we are again the second largest software company in the world, and we just delivered a fourth quarter which was our first $10 billion plus revenue quarter. So we're now on a over $40 billion run rate and continuing to grow and expand. And for a company like us, we think it's not the AI doing the entire customer service process by itself. Our vision is it's humans and AI work together to serve customers. So our CEO Marc Benioff gave a great speech at Davos. It was a lunch for about 120 CEOs, and he made a joke but it's really telling as an answer to your question. He said, 'Congratulations to all of you, you're the very last generation of CEOs that will manage an entirely human workforce, because going forward starting today, it's going to be AI and humans working together to do things and to help your customers and deliver customer success.' And so we're doing that now. We have a great customer support team, and we have about almost 40 agents now live in production, different versions of Agentforce doing different things within Salesforce, on our website, within Slack, onboarding employees, approving, automating approvals across Coupa and Concur and other apps. And I have one universal approval agent which is really amazing versus having to log in to all these separate systems which I really don't want to do. And so I can literally just click a button in Slack and approve things and see the whole history there. So the productivity benefits are enormous. And it doesn't mean that the AI replaces people. I'm a big believer AI is not going to replace anyone's job. It's going to be a person using AI that might potentially take your job. So I think it's going from an optional skill set to, it's kind of cool to understand what AI can do, to quickly becoming mandatory that people use AI in their line of work, whatever function you're working on and whatever industry you're in.

Sure. Well, the proof's kind of in the pudding. And so when we launched Agentforce, we turned it on in their sandbox, not in production, because we don't think our customers would want that. But you could show up to an AI campground and log in, and we had a team of experts, and we turned on over 10,000 companies, I think it was 10,600 in three days at Dreamforce in San Francisco, which I believe is the largest onboarding of AI agents in human history. It really is mind-blowing. And then in our fourth quarter, we added another over 5,000 brand new Agentforce customers. And so the magic of Agentforce is it's built on the platform we've been building for the last 25 years. And so many people, when ChatGPT came out, said, 'Oh, what's the big deal? I'm just going to take ChatGPT and plug it into Salesforce or whatever other software I'm using.' But that's not really the answer. Our view is the LLM, the foundation model, is like the chip in the computer. That's like saying the microchip is going to automate all your work. Well, you couldn't have a personal computer without a microprocessor, but you need an operating system and you need a storage device and you need a whole set of hardware and software components so that you can use the computer to do really cool things. We think the same thing is true with LLMs. LLMs are becoming a commodity. There's kind of a leaderboard of this one's 94% accurate, this one's 95%. So what we've done is built an AI and data platform where you plug in the LLM of your choice, and it dramatically outperforms the LLMs by themselves in terms of accuracy, in terms of hallucination rates. It understands the context of what you're trying to do, which is clearly important, and it accesses not only the data but what we call the metadata and the semantics that we have, what's called the semantic layer, which is the fundamental meaning of what a user is trying to do, and it describes the information in the database and in the objects. And so there's a real reason why it outperforms an LLM by itself. It includes an LLM but it does a lot more than an LLM does. And that's how you can deliver what we call agentic reasoning. So we're talking right now and there's neurons in our brain and I'm listening to you and you're listening to me and we're sort of processing that data and understanding what the semantic meaning is of what we're talking about. And then we have a planning engine in our brain that sort of comes up with a plan of what you're going to say or what you're going to do or how you're going to motion with your hands and your expressions. Well, the AI can do the same thing now too. So we built something called the Atlas reasoning engine that reasons just like you and I do. And that's a sort of a higher level function for AI. It's not just I ask it a question and it gives me an answer. And so we think that's really where this all goes: AI in the flow of work is where it started, and then it goes to AI agents. And we think once you have AI agents that can reason, then it's very similar to the Waymo example. You're going to take the AI agents and plug them into physical devices: cars, industrial robots, robots in the home, set-top boxes in your house. And the AI is going to do, it's going to be like right out of the Avengers movie with Jarvis. It's going to do incredibly cool things. You're going to talk to it and it's going to make all of our lives better, we believe. And will there be misuses of it or problems along the way? Of course there will be. That was true with the internet, that was true with e-commerce, that was true with every prior technology boom. It can be used for good or bad. But that's why we have a judicial system and a government that make sure that people don't rob a bank using AI, just like they did for cyber terrorism. And so I think it's going to really change our lives in a good way and make things easier, to allow people to focus on higher level functions. And in the product domain, as you asked, it goes back to the mechanics of building product versus the artistry. The agents will be able to do a lot of the mundane work that's time consuming and needs to be done in a tedious but comprehensive way, where you're checking off the boxes, so that people can sit back and say, 'Hey, what is the software really doing?' And I can talk to people and say, 'Do they really like the software or do they love the software? And if they don't love the software, what would it take so that you would love the software?' And those customer communications that are person to person, I think, are super important.

I think that's a great point, Renee. It's literally not only based on our AI work in the last decade, but it's based on everything that Salesforce did when it literally pioneered SaaS 25 years ago. And that's part of what we pride ourselves at Salesforce: making complicated things simple. And if you think about software as a service, that was the original idea behind it. Before that, there was on-premise software and it was like an IT project to implement it, to deliver it. You had to set up the servers, you had to set up the networks, and then when you had to upgrade it, you had to do that yourself. So that was an IT project. And the idea that it could literally just be a utility in what we now call the cloud, well it wasn't even called the cloud back then, was kind of revolutionary. Now everybody does that and it's just sort of a no-brainer that that's way better. So we've done that with SaaS, and then with mobile and social, and then predictive AI, and now generative AI, and now with agents and Agentforce. And in all those generations, there's what I would call DIY it. So people used to build their own CRM systems. When I started, that was the main competitor in this space. Everybody built their own CRM system and that was crazy. It's like, would you build your own ERP system or would you just buy one? An accounting system costs a lot less money, you just implement it. So why would you build your own? It doesn't make any sense at all. Then the same became true for CRM. And then if you could just turn it on in the cloud, it's even easier and simpler and less expensive. But people keep trying to do that. And so with AI, there's what we're now calling DIY AI, where people are trying to take our software and take a foundation model and do what's called fine-tuning the model, and then build their own custom data lakes. And then they have to hire their own Stanford professor or Caltech or MIT professor to help them decipher all that. And our view of that is you don't need any of that. In the year 2025, the magic of how we could turn on 10,000 customers in minutes is because the whole stack already exists and you're already using it. If you were at Dreamforce, those are our 200,000 plus customers. Why would you rebuild and reinvent the entire wheel? Why don't you just turn generative AI and AI agents on right there? And it's trusted, it works in the flow of work, then you can leverage our AI and data capabilities. So it's faster, it's easier, it's way better, the accuracy and the end results are better, and oh by the way, it costs less money too. So it's sort of a no-brainer. So we believe that the DIY computer hobbyist view, like all the prior successive trends, is going to fall by the wayside and everybody's going to buy this. They're not going to build this.

Sure. Yeah, early in my career when I started out as a product manager, because I was an engineer by training, I assumed that great product management was all about just spending time with the engineers. So you might spend 80 or 90% of your time with the engineers and maybe a little bit of your time with customers. It turns out that's not the right formula. I sort of learned this by doing that the best product people spend more time with customers than with the engineers. And there's a reason for that, because you get your best ideas from your customers. So if 19 out of 20 of your customers say, 'Yeah, we really need this one thing,' or maybe 20 out of 20, it's probably right. And then it's hugely powerful because you go back to the engineering team with, 'Listen, I just talked to 20 customers. Here's what they all said.' So there's no debate, there's no subjectivity to that. This is what they said. And sometimes you can even record that or bring the engineers along with you so they can hear it for themselves, because hearing is believing. But I think that's super powerful. So I think the best product people spend maybe 60% of their time with customers and 40% of their time with the engineering team. The product and engineering team working together and being a thousand percent aligned is a critical success factor, no question about it. But you have to bring in that customer input. And even if you do surveys or even if you have advisory boards, nothing replaces that personal interaction with those customers. So on our executive team at Salesforce, I do more customer meetings than just about anybody. And I do that for a reason because I want to hear what the customers are thinking. I want to hear what they like and don't like about our products and how we can improve and be better. And I want to hear about what their thinking is about where they're spending their time and where they're spending their money and what's important to them as a company and what's their IT architecture of the future and how can we be their single most trusted partner. So we have five core values at Salesforce. Number one is trust, so that you can trust the people and our software as a service, particularly the second S, the service part of that, that it's up 24/7, it works to as many nines as possible, that we protect the data religiously, and you can also trust the people. But then one of our next core values is customer success and innovation. And those go hand in hand in my opinion. As you work with customers and you ensure that they're successful, then that fires and empowers the innovation because you get all this amazing customer feedback. So the best product people spend a lot of time with customers.

Sure. You can. Those other modes of collecting information are not bad. There's nothing wrong with them. They're necessary but insufficient in my mind. And so part of it is, if I go back to my example of 19 out of 20 or 20 out of 20 of the customers want this one enhancement or new feature or new product, it's probably a good idea. But often customers jump to the specific implementation, like we want a button right here that does a certain thing. And while the idea is right, sometimes the implementation of what they ask for isn't right. Because these are sometimes business people or IT people who aren't necessarily trained in software artistry and how to deliver something that's really fun and engaging to use. And what they're suggesting breaks what we call our design standards, which are very important, that there's a consistency and a coherence to how all the software works together. And so they're not doing this every day like we are. And so I think what you want to do is get the input so you really understand the what, but then have the discretion to determine the how. And sometimes when you get this from your sales team, the raw feedback is very interesting from your sales team, but they're also not trained in building software or designing software or delivering software either. So they sometimes miss it. And so they sort of get it partly right but it gets garbled a little bit. The old game of telephone where things start out, you go around the circle and by the end it's something totally different. And then you're not working on the right things anymore, which is a huge waste of time. I've never worked in a software company that has too many engineers with not enough to do. And so prioritizing what the engineering team works on is super important. I always tell my product teams, think of a visual of taking $100 bills and putting them in a trash can and just incinerating them. When you're working on the wrong things, you're just wasting time and money. It's just a complete waste. You want to be very focused on every ounce of our creativity and our smartest people and our engineering team is working on the right things. So that's why you have to talk to customers. It's really important that you really understand exactly what they're saying. And there's real nuance sometimes in how people say things, and how the requirement they're asking for works around the world, or maybe it might work this way in this industry but that way in another industry. These industry-specific requirements really matter, and the compliance is sometimes different in different locales. So you got to really understand that. So we spend a lot of time, but we do everything you said before. We have focus groups, we do surveys, we have in-app learning and in-app suggestions and all that. And that's all good, but you got to validate it face to face.

Exactly right.

Sure.

A great product leader is a great listener.

You know, I love Cursor, which is a new AI tool that literally writes the code for you. So you tell it what you want to build and it builds it. It's literally amazing. That is something that we're going long on. We've built tools like that at Salesforce, but we use third-party tools too. But that's a really great example of where this whole industry is going.

That's a good one. When we write our V2MOMs, we outline all of our priorities as I mentioned, and we constantly check that with an agent and we ask it, 'What do you think of this V2MOM and these set of product priorities? Do you have any suggestions on how to make it better?' And amazingly, the AI does. And so we use the AI in our flow of work to refine our plans for every single release and our vision for the future.

You know, I love American history and historical fiction. So that's one topic. But the space that really got me interested in technology is science fiction. So I love science fiction and there's some amazing science fiction out now that's about us meeting people, making first contact in the near future. So there's a great book called The Hail Mary Project.

It is an amazing book. It's a good book. Super cool. And I love the way off in the far future too, like Star Trek and Star Wars, which is what I grew up with. But it's cool to hear that, I think with people like Elon saying we're going to land on Mars, which I think is going to happen in our lifetimes, the near future of these things happening is what I really love. That's a new genre where they're literally talking about us landing on Mars in five years and us having first contact soon after. So that's why I love that book.

Renee, it was a real pleasure. Thank you for having me.