 Welcome back everyone to SuperCloud 4. Focus on generative AI. I'm John Furrier, host of theCUBE. We're here for our in-studio live performance and some guests can't make it. Our next guest couldn't make it into the studio live. So she's coming in remote, Alastair Glass, executive vice president and general manager of the Salesforce platform. Welcome to theCUBE, SuperCloud 4. Thanks for coming in remote. We appreciate you coming in. Thanks John, I'm excited to be here. You're new to Salesforce, but you've had many stints in the industry cycles where you've seen a lot of change. Microsoft for many, many, almost over a decade at Microsoft, code.org, compelling work there, changing the landscape of education, opening up opportunities. AI is this generational shift and it's kind of fun to talk about it from a society standpoint, but from a business standpoint, SuperCloud and the next generation cloud technologies are changing how business is going to be done. So it's a super exciting topic, but with that comes some disruptive enablement, if you will. Certainly AI has dropped into the landscape. Everyone's talking about it at the app layer, but there's a lot of stuff going under the hood, databases, how code's being generated. This is a big topic. What's your current situation there at Salesforce? Take a minute to explain what you do there and what's your take on how AI is changing business? Well, okay, I guess I'll start with what I do. I run the platform team. And our platform is what enables our customers to build and deploy customized, personalized experiences, almost every company has something unique about their workflows, about their processes, about their data, and the Salesforce platform enables companies to customize for their businesses with no code, with low code, with pro code tools. We also support a vibrant ecosystem. We have over 7,000 apps on AppExchange, but I think what's really exciting about it right now is that it's not just no code, low code, pro code. We're talking about how you can use AI to customize how your company works, how the workflows happen, how the line of business happens. You know, John, I have two kids. I have a 15 and a 17 year old and I told them you are not allowed to use AI to help you with your papers in class, but we're deciding how business works at our companies. And we can figure out how does AI become a part of our workflows. And I think that our companies have the opportunity to do this right now and our platform is what enables them to do that. So I'm excited to be here. I'm excited to be working on the Salesforce platform and enabling companies to really connect with their customers in new ways. Well, great to hear the story. So what's the answer when they say, you know, what? I can't use the AI? What do they say? They lean into it or they like saying, okay, sure, yeah, right. Or what's the answer? Exactly, exactly. Yeah, and you get the mom, but... That's awesome. Well, I think it's going to be a great conversation because humans plus AI is more creative. There's a lot of opportunities with AI and this is what everyone's trying to figure out. A lot of experimentation. Clearly the enthusiasm is at an all-time high. What are some of the things people are doing? What do you see? Because there's a lot of people leaning in that run some experiments. There's some production discussions going on. That's the big topic here in SuperCloud 4 is some workloads going into production, but still it's early days where people are looking at formulating use cases, identifying low hanging fruit projects. What do you see right now in terms of adoption? Where are they putting the toe in the water? Are they jumping in the deep end? What do you see? So I think first of all, it's worth noting that AI itself is actually not new. Salesforce has been doing AI since 2014. We currently have over a trillion predictions every single week, but this is all the predictive AI piece. And now what's happening with this generative AI is we're looking at how we can use AI in new ways. What we see is that AI is now a CEO level priority for every one of our customers. Everybody is trying to figure out how do they use this to really accelerate what they can do. And I think there's a few differences that we see between what you could do with the predictive AI and what you can do with the generative AI. One is it's more accessible and approachable. We were always using it for everyone in the business, but you would need someone in IT, someone to help you set it up, someone to think about it. And now with the democratization of this, what we see is that all of the teams, the non-technical teams, sales, customer service, marketing, and commerce, they're all playing around with these tools. They're playing around with these tools directly and they're looking at how they can use it to accelerate their business. And in terms of where we see it, you ask where across the business do you see it? Everywhere, right? Sales teams are looking at how they can auto generate sales tasks like composing emails, service teams are talking about how do they auto generate knowledge articles, working with marketing teams to dynamically generate personalized content, work with commerce teams to figure out how do you personalize product recommendations and generate product descriptions, and IT, I love IT. The IT teams are using it to figure out how are they going to write code? What does it look like both in no code, low code and pro code to use generative AI to accelerate my development? So I don't think there's a part of the business John that isn't using AI right now. It's interesting. I mean, it's every aspect up and down the stack and across all industries are impacted. I love the aspect of loving the IT side because a lot of platform engineering use cases are coming out. And everyone's all gaga over the app side, which I get, but there's real value on the tech stack right now. What do you see there as opportunities? Obviously, automation has been talked about for awhile. We've had machine learning in the past, but now you got generative AI. How do you see that integrating in where data now is a valuable aspect of it? And by the way, the word walled garden is coming back. It's okay to be a walled garden if you're interfacing in. We're hearing that loud and clear here today that walled garden is not a bad word if the data is clean and seeing a lot of distributed data sets. What is going on? How do you see the platform engineering evolving? It's going to be data engineering. What's your vision? Oh, well, you're asking a platform for, I like, you're like, hey, not many people want to talk about IT, but of course you're asking the platform person. Yes, I want to talk about IT. I love that you want to go here. So I think there are multiple pieces here. The first thing that when we talk to IT teams, the first thing you need to get right is trust. And when you look at rolling out this kind of technology across your organization, the questions about hallucination, toxicity, where is the data being stored? How do I do PII? Hopefully my aunt doesn't watch this, but I just used AI to write an email to my aunt. And that's great. I can do that, but I don't want everybody at my company to just randomly use AI to start writing emails, sending PII into the AI. It's something that you need to do in a trusted way. So the first step I think for any IT department, for any builder is having that trust layer. So what we've been doing at Salesforce is starting there with the Einstein trust layer. And what that is is it's a secure AI architecture that's natively built into the Salesforce platform and is designed for enterprise security standards. What does that mean? It means you need to be looking at how do you integrate and ground the AI with trusted data? How do you mask the PII? We have zero data retention and PII protection built into that trust layer, toxicity awareness, compliance ready for AI monitoring to see what happens. So there's more there we can get into it, but I think that's the first step. And I think that's, before you get into step two, you need to have that trust layer, that foundation built in. And then on top of that, I think there's a really interesting space for the technical teams, the developers, the admins, at Salesforce, we have trailblazers who are thinking about how do I enable the workflows for my company? And when I say enable the workflow, what we see is, we did survey 45% of in our research, we'd use Generative AI more if it was integrated into this technology they already use, and that makes sense. If you're using it, you wanna see it built into the technology. So the question is how do you do that? You work with, let's say, a line of business owner and they say, hey, what I need to do, let's email's a good simple one. I wanna send emails. How do I work with that line of business owner to help them generate an email as part of the flow of their work? So it's just right there, it's a button, it's, or I'm talking to my co-pilot right there, when I'm ready to send the email, I have the choices of the kinds of emails I wanna send, an opportunity, an outreach, a, you know, trying to get you to upgrade, whatever that email is, these are the types of emails my company sends. And then how do I customize and personalize that? I think in the old world, maybe I thought, hey, I just have to write a lot of code for that. But there's a new world that we can use where as a developer, what I'm thinking about, I'm thinking about the job to be done, right? How do I send an opportunity email? And I don't want every single salesperson to have to figure out how to do that. By the time I put in all of the information, I might need to know, hey, John, what level of membership do you have, right? That might be important. Or, you know, more information from you about, you know, what did you do last week and what have you been looking at? By the time I type all that in, it hasn't saved me any time to send the email. So what we want to do as a IT person or as a developer is figure out how do I get all that right data and connect it into the prompt in a programmatic way and then be able to activate that in the course of work so that it's just in line. And so what we're doing at Salesforce is we're building those tools. We're building the prompt builder tool so that you can create that connection as a developer, right? You're connecting that data into that scenario, activating it as part of the course of work so that you're enabling your team to do it and then being able to tell what worked, being able to monitor that and look at it and say, hey, is that actually helping sell things, right? Did John take me up on that offer? And being able to go back and say, hey, no, I want to improve this or change this. So the way we think about developing software, I think is changing. And then the other thing that's changing is we're giving people a co-pilot to do it, right? And are you in mute? I can't hear you. I'm listening. Okay, great. I don't want to interrupt the flow. Yeah, yeah. So we're giving people a co-pilot to do it and we can talk more about that and what it means to be working with support from a co-pilot. Let's get a couple of things. I love the trust angle. Trust is huge. You mentioned activation, active. Love that word. And one topic that's come up a lot here is if you get the governance and compliance right on the front end, the data scales. And you got the picks and shovels coming in to help the developers with co-pilot and tools. This is an important kind of nuance. It's kind of an IT thing, but if you get the data guardrails right on governance, you can scale the developer and the user experience of the tooling side. So how did people do that? Because this is the number one question. The word guardrails is used a lot. You know, I see that in, well, we've got to put guardrails to protect everyone. But legit guardrails can be things like we see in DevSecOps. Shifting left is guardrails for developers to code security in the CICD pipeline. So that's cool. Data, are we there yet with data? So what's the guardrails and governance principles that are needed to ensure responsible use of generative AI, trustful use of generative AI, and scalable? That's a great question. I think it starts, one of the reasons, I only joined Salesforce six months ago. And one of the reasons I joined Salesforce, one of the things I loved about it is at Salesforce, our data is not your product. Our customers own their own data. They've complete control over how it is used. We're acting as a trusted custodian to help companies manage their data. And we also have to give them the tools so that they can have that same promise with their customers, right? So they can guard and protect their customer's data as they're using it. So I think it starts with a set of data governance. It starts with the tools, it starts with protections that we were talking about. You need to understand your data. You need to have it tagged correctly. You need to understand the PII for your data. You need to write security permissions in place, right? And so a lot of this base infrastructure of just having the right set of enterprise data governance in place for your data is critical and it's a starting point for using it with the AI. And then you need the tools on top of that that allow you to use the AI while respecting the data that's underlined it. So that's things like transparency and accountability. So being able to explain how these generated AI systems are working, being able to say what's happening, having security and privacy in place. So you're protecting your users' data. The way AI works with data is different from some of the ways we've worked with data in the past. If I put a bunch of data in a database, I know exactly where that data is. I know if I put John, I put your name in the database, like I can search it, I can find it, I know exactly where it is, I know where I've stored it and I can delete it. If, but that's not how humans work, that's not how brains work. So if I just learn about you, where did I learn French in high school? Like which part of my brain is it stored in? Which part of my brain is John stored in? We don't know, right? Because it's not stored as John anywhere in my brain. It's a bunch of neurons that are connected that each have a little piece of this, right? And when we send PII into an AI model or train an AI model with PII, we lose that control over the data that we need to protect. And so that's where I get back to that trust layer. It's really important for a company to be thinking about, okay, I'm gonna use the AI to solve these problems. I'm gonna use it to accelerate me. I'm gonna use it to go faster, but I don't need to send John's name to the AI, right? I don't need to train the AI with John's name. I need to protect and think about where does my data live? How do I own and control that so I can have the right to be forgotten work? I can manage it, yeah. I love the you own your own data. I think that's what Salesforce has always to have. I've known Mark Benioff for years. He's always been users on their data. Here's what's interesting. If you look at the cool things around neural networks, you mentioned the brain and the vector database trend, people are using these technologies on their own data. So if I can connect the dots here and I want to get your reactions to this, if I'm running the Salesforce platform, I'm saying, okay, I got a lot of data and sometimes it's a small set of data compared to the big models, not huge petabytes. It could be pockets of small data, but it's important, clean data or discrete domain specific. And then I got embeddings going on or I got a development team. I might want to build with the data. So if you have a platform, you guys have a robust ecosystem, you mentioned trailblazers. So I can imagine that Salesforce must be in thinking about enabling your customers and the ecosystem to code the data. So what's your vision on that direction? I can see the ecosystem looking at this and saying, okay, I could have a power law, a bunch of small language models, SLMs we call them. Okay, maybe interface via an API through other LLMs and to get formatting, maybe inject prompt engineering into the API call, not just direct, all kinds of new user code or experience and user experience is coming. How do you see this? What's your reaction? So John, I think you made a really good point there. This is unlike anything we've ever seen in technology. I've been working in technology for over 20 years. I have never seen it move this fast. And when I say move this fast, just the pace at which the technology is changing what's possible. Technology's moved fast before, but not moved fast, not moved this quickly in terms of what is possible and what it's enabling. And what I see when I talk to IT leaders and CTOs about what they're doing is they're also seeing it move quickly and they wanna leave a little bit of space to be able to give themselves some time to understand what's gonna happen. If I go back to February, it felt like the only LLMs that we're gonna be were those huge giant ones, right? The open AI. And then we had Palm and Palm II and all these different anthropic and all these different models come out, different ways of doing models, small models, LLMA, big models. And I think we're still figuring it out. I think we are at a place where the technology's moving quickly and people are trying to understand, am I gonna be able to do this cheaply on a small model that is really tailored to my situation? Is this gonna require a gigantic model? Do I wanna use a public model? Do I wanna use something internally? And so what we're doing at Salesforce is we're building an open platform where they bring your own model as part of the core foundational piece so that you can play with these things, right? And as a developer, you can say, hey, I'm gonna try training this with our data, with our specific scenario and see what that does. But the other thing we've seen, John, is that in a lot of cases, you don't need to. It is amazing what grounding with a generic large language model can accomplish. I didn't expect it when I first started playing with these things, right? You look at it and you know, hey, how could it do it with just, if I send the schema, right? If I send the data, if I send the information about the JSON format that I want something to return to, the large language model without any specific training can return content using that JSON format. That's crazy. It's amazing, it's amazing what it can do. The co-pilot and the augmentation to humans is really amazing. I think it's going to create a creative class in tech we've never seen before. My final question is, is that as these platforms come out for you, Salesforce is well stacked up in terms of the product. They got plenty of aspects of the product. They've done a lot of acquisition of the agency, Slackum is one of the big ones, a bunch of the core product. It's already kind of easy to use. And so the question is, how are your customers preparing for the workforce? Their workforce are users to take advantage of the AI keepers. So you have customers that have users of Salesforce. I can imagine that the generative AI is going to help them be better and open up more users, lower the bar for usage or automate away the access where it's just voice activated. Give me the forecast, integrate, what's going on the Slack channel? Can you connect Slack, text that message out? So I see more user democratization. What's your take on how customers should prepare their workforce to take advantage of the generative AI keepers? Lean in, wait a little bit, wait for the tech to emerge. How should people prepare? If I were an IT leader right now, which I am, but if I were one of our customers, I would lean in, right? You want to be at the forefront of this. It is a disruptive technology. It has a major impact. It's going to have a major value at, you don't want to be watching it go by and looking and saying, hey, you know, all my competitors are accelerating. They're able to do more. We already see the capability here. We already see that you can get more done. You can do it more quickly. You can be more effective. And I think a lot of what it does is it takes out the drudgery. It takes out the tasks that are, you know, we all say AI is amazing. It's amazing, but it's not brilliant, right? What it does is it automates away a lot of the repetitive tasks. A lot of the, hey, I'm writing a synopsis of a meeting. I'm writing a sales email. These are things that we spend a lot of time doing. And by having the AI help with it, you unlock the creativity so that your workforce can spend more time working on the things that are interesting, working on the challenges, overcoming the problems. And the workforce likes it. It's great. I just write bullets like main message sentence, bullets. Write a memo, done. Thank you very much. Exactly, exactly. And across the board, they like it, right? They like it, you know, you like it. But, you know, I was talking to a bunch of developers who are playing around with the new tools to be able to generate code for Salesforce using text to speech. And, you know, it's not like every thing we generate is ready to ship out of the box, right? You say, you know, you look at it and you're like, hey, I want to generate some code that tests for this condition and does a trigger. And you get some code and you get a starting point and then you edit it, but they're like, this is great. Right? It gets started, it's faster, it's easier. I'm not spending a bunch of time with the basic like, hey, I'm fleshing out my test cases. It just gets you started. It makes it easier to do that. You know, Alice, you brought up data earlier on the conversation. I think to me, the data paradigm is, the script is flipping. You know, days when you had a siloed data set or a walled garden, data warehouse is kind of like, oh, you bring it to the cloud. And now with open formats like Iceberg and Parquet and SQL, you can have SQL languages talking to each other across data. You can create a horizontal data plane, but yet vertically domain specific to be highly active and real time, which has been a hard problem. If that continues to happen, then the data becomes the fabric of the coding. It's like data ops. It's like, that's going to be kind of a data engineering data app role in IT that we kind of never saw before. We had data administrators, but we didn't have like data engineers like SREs. Like we didn't have large scale data architects that think. What's your reaction to that? Oh, 100%. Yes, I think that what we see. So at Salesforce, we've built a data cloud product that does exactly what you said, right? It allows you to bring in that structure and unstructured data. It allows you to bring in the engagement data, all of these data from other sources and leverage that as part of your AI, as part of your CRM and the interesting questions that we have around that it opens up. What are those data sources that I want to use? How do I want to leverage those? How does knowing what somebody's been browsing on a website change the kind of email or offer that I want to give them? How does knowing what somebody is interested in change the way I want to personalize marketing content for them? And being able to connect those dots together from the data into the CRM is a space where the engineer, the IT thinker, but also the line of business thinker, right? A lot of these tools are low code or no code can really think about strategically how do you connect these things together? And I think it's a fascinating space. I think we're moving in general where to see change and what development looks like from a set of procedural steps that are top to bottom and very defined experiences to being able to use the engineering thinking to say, hey, what is the job to be done here? What data do I need to do it? What do I want? What question do I need to ask the AI to help me do that? What question do I want to ask the user and how do I connect this into the flow of work for a user? And those building blocks, being able to put those building blocks together, I think is a really interesting space for engineers as we go forward. And not only do engineers put them together, but the AI does too. So when our co-pilot, once you've created these building blocks, the co-pilot can leverage these building blocks. So when I ask it, hey, can you help me figure out what John would like, it can go grab that building block that you built that connects into the data so we know about John and then surface that to the user as part of the co-pilot. It's going to be an exciting time, Alice. We can unpack a lot more of this. We'll definitely do a follow-up. There's so much that this has to be a two-hour conversation. The impact on IT, what that enables, the security aspect, the data operations, the new roles that are going to emerge, the opportunities and new applications, the startups are going to emerge. And certainly for Salesforce with the robust ecosystem, people can now create integrations at an API with Next Gen Cloud. You got data APIs, subscriptions to services, like data services, like language, unbelievable opportunity for you guys. So congratulations on that and it sounds like you got a fun job and definitely want to talk more about it. So thank you for coming on the SuperCloud 4, Genevieve AI sequence. Thank you. Thank you, John. It was fun to be here. Thanks for having me. Right. SuperCloud 4, Genevieve AI, Salesforce got the perfect platform for this next wave. I can do all that heavy lifting, put things together, merge data, talk to data, talk to each other. This is the Next Gen Cloud and actually with the data, this is SuperCloud 4. We'll be right back after this short break.