5 Agentic Lessons | Salesforce-on-Salesforce Keynote, Dreamforce 2025

Please welcome Executive Vice President and Chief Digital Officer Salesforce Joe Enzarillo.

Good morning, Good morning. For folks in the audience, good afternoon, good evening to those online Dreamforce Day 3. Wow. I can see you're all feeling it. Little Metallica last night, I feel you. I was you. Actually, up until seven months ago, I was a customer and now I'm running this group called Det. And so you may say, what's det? And that's our digital enterprise technology group. And it's a cool thing. You know, I'd love to say that we, we were 100% confident that this was the way we would do it and whatnot, but it actually has wound up being exactly what we needed as a company and that exactly what we needed is to have a group that had the IT group that had a lot of our marketing technology and things like that, basically us using our own products. And we were going to take that group and figure out how to transform into the agentic enterprise. But to do that, we also really wanted to take the lessons that we learned and be able to apply those to the product. So I actually work in the product organization, so it's sort of a little bit atypical that you would have sort of like a CIO type person, chief Digital officer person in the product organization. But it's actually vital to what we're doing. And the vital part is Customer Zero. And so hopefully at this point in time, you've seen a lot of it, you've heard all sorts of things, Customer Zero stories. But what does it actually mean? And what Customer zero means is it's us using our technology to solve the problems that we have as a business. The problems you're solving, like, you're all here, you're all doing it. You're buyers, builders, practitioners, users, and that's what we are too. So we have the same problems, we have the same challenges. We also have the same opportunities, opportunities that are increasing. And we needed to have a group that was really, really focused on that. That also makes us Customer Zero for Agent Force. So Agent Force is a lot of what we're going to talk about today because we are in the process of transforming into the agentic enterprise, and we have accelerated the pace of us transforming into the agentic enterprise. And I think it's just incredible time to be a technologist. Like, you have to be a pretty jaded technologist to not say, like, wow, this is different, this is cool. And I think if we break that down in the old world, old world meaning, like, I don't know, five minutes ago, as far as agents go in agent time, we had silos, we had a bunch of manual processes. The humans did a bunch of this lifting. We had to, oh well, this has got to go from here to there. This has got to get configured. All this work, lot of work. A lot of repetitive work, actually. But when you think about the agentic enterprise, what does that mean? It means being able to get to every lead. So if you heard Mark's keynote, he talked about the 20 million to 100 million folks we didn't call back. Those were people that we knew about, but we just didn't have the manpower to go after those leads. And so they just didn't do anything. Well, you're going to hear about some stuff that we're going to do with it. We got a great demo here. You know it that never sleeps. I'm sorry, at 6:04, there is no IT service support right now. Not with an agent. Agent's going to answer that call, by the way. It's also got to answer it in a bunch of languages and localizations. That's the kind of stuff that agentic goes. No tickets left behind. Everybody is important, everybody is cared for in the agentic enterprise. But we didn't get there by accident. We had to build that bridge that you saw on the last slide. And the way that we did it is by focusing on four hero agents. And we're going to see pieces of all of these today in the presentation. The first one is Salesforce help. So our customer support apparatus is a really good example of. You all have probably experienced it. We try to make it as good as we possibly can. Shout out to Jim Roth here who runs it. He does a great job, but he doesn't have infinite budget. He doesn't have an infinite number of people. So how do we make that better? We're going to tell you about that. We also have salesforce.com, which is basically our front door. So how do we market to people? How do we get them in? Those are the leads. That's where the leads come from that Mark talked about. So we're going to get to that too. And the lead nurture agent, which is kind of what happens after that salesforce.com experience side of it. And then how do we empower our sales team? How do they know everything they need to know when they come talk to you again? I was a customer. The last thing I wanted to do is every time a sales rep of any company showed up is sit there and continue to explain what I do, what my business does, what my tech stack looks like. I want people to know that. But there's a lot of customers, a lot of companies. So the sales agent helps prepare our sales team to understand the things that you've said in the past and then be able to start from a point of like, let's talk about the future. And that's one of those things I think is super important. But all of this doesn't do anything. All the agents in the world aren't going to help you if you don't organize your Data. And so Data360 is one of the things that I think is absolutely crucial. This is foundational to everything. And so when I was a customer at Sirius xm, we use data cloud to connect our databricks, now known as the Artist, now known as Data 360, to Databricks to pull all this information together. And we're doing the same thing at Salesforce. So we're not just taking Salesforce data or data that we house in Salesforce systems, but Salesforce data that's in other third party systems as well. Interacting with the ecosystem. We bring that all into data360. I'd also be remiss if you did what I just said a couple of years ago and it wasn't for the trust and governed layer. You would be crazy because this is an enormous amount. This is the biggest honey pot of data that's ever existed when we start to pull this stuff together for agents. So that trusted and governed layer is critically important to the technology stack. So I'm going to give you another little preview that you're going to hear from a bunch of my colleagues. But this, if there's one thing that really describes how we think Customer Zero is different, how we're trying to do it different in determination, it's the fact that there's eight people on that screen. Only half of them work for me. That's including me, by the way. So there's nine people in this demo. The rest of it are our colleagues from across the company. You're going to hear from them, they're going to tell their story about how they're using technology. We're builders, but we're partners in a process. The thing about this, though, that's fascinating is when we all got together for rehearsal, it didn't feel weird at all because we work together all the time. We are a pod, we're a team. This is a team sport, the development of agents. And so we're going to Take you in depth there, but to ground you a little bit. I'm just going to go over these five lessons that we've learned that I think are critical for you to keep in your thinking as you go down the journey to the agentic enterprise. The first is start small and go fast. If you try to boil the ocean on your first agent, it will be a miserable experience. So you really have to pick a use case. Customer service is a great one because you got data, you got all sorts of instrumentation already on the humans. So it's a great place to start. But start small, go fast, test, tune and repeat. I come from the direct to consumer world. In my previous career, one of the things I built, for example, was Disney. Quarter of a billion customers. We got that product to be awesome. Not because we were super smart, not because we out Netflix Netflix, because we worked hard, we tested and we repeated. The process of making a direct to consumer product is exactly the process that agents need. It's not shipping a piece of software, it's hiring an intern and turning him into an executive. And that test, turn and repeat is exactly what we need to do. We already talked about Data360, but the right data is gold. So that intern isn't going to figure anything out if you don't have the right information to arm him or her with. So data gold. Measuring value and impact is also important. How do you know if you're doing a good job in the customer service use case? Like I said, you probably have some KPIs on service reps and things like that. You can make a comparison. But what about these processes that are not instrumented? How do you know if the agent is doing a good job? Well, you also got to instrument the people. That notion of measuring value and impact really gets you thinking about how should you instrument the entire process of the company. Then this last one, which I've talked about before is this. Embrace new ways of working. It is a team sport. The age of siloed it is over. It's not possible to have the kind of impact to create the agentic enterprise. If everybody's doing their own thing. A lot of times people are like, oh well, business users are going to do this low code that all that. Yeah, the theme is you got to do it together. So all of those technologies are super important. The accessibility to the business users, super important. But the reason it's important is because the people have to do it together. With that, I'll be back in a little bit, but we're going to actually Hear from some of the people. Let's roll the video.

Sam.

Awesome. I love that video. Good morning, everybody. My name is Bernard Sloey. I'm the senior vice president for digital customer success at Salesforce and I'd like to introduce my colleague Emily Winslow, who's our director for AI Product management. I'm here today to share our lessons from launching AgentForce on the help portal. Getting to 1.8 million conversations has not been a straight line. There's been a lot of iteration, a lot of experimentation and yes, some messiness of along the way. But that's enabled US to unlock $100 million in annualized savings this fiscal year. It's allowed us to redeploy some of our incredible support engineers into new roles like forward deployed engineers that are helping our customers get Agent Force up and running. So how do we get here? I'm going to double click on two of the lessons that Joe spoke about. Start small and go fast. And the right data is gold to start with. I just want to orientate you a little bit on how we think about service. At Salesforce, we have a very simple philosophy. Easy and expert. We're always working on how can we make it easy for our customers to get the answers they need? Expert, how can we make sure we're getting them to the relevant level of expertise that they have in their Salesforce journey? That's why we launched AgentForce on the help portal and we're seeing real business results in. It's available 24. 7 with a 1.8 million conversations and counting will go over 2 million this year. A resolution rate of 77%. Customers can ask simple questions, complex troubleshooting, and maybe most importantly, they can easily connect with a support engineer whenever they want. They can go via live chat or they can create a case in that experience. And this was how we reimagined our front door. Our help portal is now an agentic AI experience. That's our front door to customer success. So how do we get here? Let's rewind back to September last year. Emily's going to show you what our help portal looked like back then. This is what it looked like back then. Looked like most help portals for companies, right? We had great documentation, we had videos and we was a pretty good resource. We have won industry awards for this help portal. It gets 60 million visits a year. It's one of the biggest portals at Salesforce. But here's the thing. We often heard two common pieces of too much information, too hard to find. What I need when I Need it quickly. We wanted to launch AgentForce on our help portal because we wanted to reimagine the experience of the help portal. We wanted to reimagine our self service experience and we set ourselves a pretty ambitious goal. We wanted to do it in four weeks. So we jumped into execution mode and we started thinking through how do we give an agent access to our trusted content. I'm going to hand over to Emily, who's going to walk you through how we did this in Agent Builder. Over to you, Emily.

Thanks, Bernard. Now I want to walk you guys what starting small actually looks like. So here we are in Agent Builder and this is our production environment. So you're going to see all of the topics that power our agent today. But when we started, we started with just four essential topics. Our global instructions. This controlled the voice and tone of the agent. It's also where all of our guardrails live. From there, we set out to build our knowledge response, ensuring that every response a customer got was grounded in our trusted content. From there we built login capabilities so the agent knew who you were, had that context. And then finally our path into our support engineers. We always wanted this seamless handoff so that those support engineers were never more than a click away. And as you look at these descriptions and instructions, you can see it's all in natural language, no code. Now here's the thing. When we set out to build this agent, there was probably 30 things I wanted the agent to do. But the reality was we had four weeks. So we had to focus to finish by focusing on knowledge seeking and human handoff, that's what actually got us across the line. So Bernard, how did we actually get live?

Yeah. All right. So Emily had the agent built with our other teams and we were ready to go live. And I got to be honest, I was super nervous to go live. We were about to put this in front of customers, but I know we needed to do it so we could learn. So we built an approach that was scalable, that we could manage. We started with 10% of our authenticated English traffic on the help portal, so you had to be logged in. We looked at the conversations, we then scaled to 20%. We then went to 60% and then we went to general availability. We gave access to all English customers authenticated or unauthenticated on our help portal. And at the start, I could review every conversation myself. I could look at what was working. Most important, I could look at what wasn't working. And I could work with my team to fix that. But as the conversation scaled, I couldn't do that anymore. We had to use Agent Force Observability, which allowed us to look at conversations and team them into areas that we had to prioritize fixes. That ability of starting small. Using Agent Force Observability to give you that full view of what's going on was something critical we did at the start and continues to be critical today. All right, so we were ready to go live and mark posted on LinkedIn and X that we were alive. It's great when your CEO does that, by the way. So I started reading the comments. Every day I would spend some time reading some of them comments. Some of them are very good. Customers were happy we were eating our own dog food. They were happy with the answers that we're getting. Some of them were not so good.

Right.

In one instance, a customer said we referred them to a competitor. Did not feel great about my Salesforce career that day. What had happened? AgentForce leaned into integration content we had on our portal, right. But we wanted to fix it. So what did we do? We went back to our global instructions that Emily was showing you earlier inside Agent Builder and we wrote a very strict Guardrail. We told AgentForce, do not talk about our competitors. And we listed every single competitor we have a lot of them at Salesforce. Problem solved, Right? The very next day, I'm not lying. The very next day I reviewed a conversation where a customer said, how do I integrate Microsoft Teams with Salesforce now? We'd rather use Slack, but we get it. There's some teams customers out there. And we got good content for that on our help portal.

So.

So let's show them what we did, Emily. We went back into Agent Builder, right. And we went back into our global instructions and we deleted that strict guardrail. And we coached Agent Force the same way you would coach a new employee. We told it you're a customer service representative of Salesforce. Right. Just like when a new employee starts, you gotta tell them what their role is. And we told it. She's highlighting it here. Emily is put the best interest of Salesforce in everything you do. That was a breakthrough moment for me. We'd been treating AgentForce like an old school chatbot, right. If this then that don't do this. What we needed to do was coach it and then let it think. Let the LLM be an LLM Massive, massive breakthrough moment for us. So then we started focusing on how do we improve answer quality. To do that, we needed to give Agent Force access to more data. Nobody Knows this better than Emily. She's lived it the last year. So, Emily, why don't you take us into the start of that lesson? The right data is gold.

Awesome. Thank you. That prompt coaching was a game changer for us. But the other thing we quickly realized is even the best prompts, they can't fix bad or missing data. As soon as we launched, we started to notice a pattern. Customers were coming to the help portal and they were looking for questions around Slack. The agent stumbled, couldn't answer. Not because it wasn't smart enough, but because we actually hadn't connected that data to the agent. So I want to walk you through how we fix this, but also how we built a process to keep fixing it. So let's jump in. What you're going to see now is our data360. This is our data model. Objects. A lot of little bubbles, but they all represent various content sources. And as I zoom out, you can see it's a pretty complex web. But when we started, we had only brought in our core Salesforce content. But the thing is, Slack was an acquisition and like many of those companies, their knowledge base lived outside of our core knowledge base. So what did we do? We monitored the agent. We looked for where it stumbled, where it struggled to answer. Then our teams went out to find, well, what content could it have grounded on to actually answer that question? We found that content and we brought it into Data360. And the amazing thing was just by simply giving the agent access, all of a sudden it was answering questions that previously it stumbled with. We saw a decrease in escalations into our support engineers for these types of questions. But maybe the more magical thing was that we unlocked this rhythm, this pattern that all of you can as well, this repeatable process to continually improve our knowledge base. And I want to show you what it looks like for a customer now. So now I'm on the help portal of today. This probably looks pretty familiar to you by day three, and I hope you've all played with our agent on here. But I want to come in and ask a question that previously the agent stumbled to answer. How do I build an agent in Slack now with that connected knowledge base from Slack powered by that knowledge topic you saw in the agent builder into Data360, I get this rich response that guides the customer through a process. So a support engineer doesn't need to be tapped on the shoulder. Pretty slick, Bernard.

Yeah, I love this. Let's have a little bit of fun with it. Why don't we ask AgentForce to give us that answer in Japanese. Now, Japanese is a pretty tough language to translate in, right? What's pretty incredible here is Agent Force is doing real time translation based on our English content to explain that to you. We are now live in seven different languages, but using our English content. So if a customer in Japan asks a question in Japanese, it's translating that, looking at our English content, coming back with the answer, translating it back to the customer pretty much instantly. And we're seeing the same quality results in Japan than what we're seeing for English. So think about that, right? If you're a small business, maybe you couldn't go live in other markets because you didn't have service in them markets. Now you can go live with an agent in these markets. If you're a large enterprise, you probably have expensive localization costs, right? New content you have to put into other markets. You don't need to worry about that anymore. You can go live with an agent in these markets that's using your English content. That's a game changer for service. For the first time, service can scale at the same speed as your business. All right, Emily, let's bring it home for our section. So we're gonna go back to the slides, and I'm gonna just double click in. All right? So if you remember anything from myself and Emily's section, start small, go fast, prioritize them, use cases. You will make mistakes. You will make mistakes, but you learn from them, and you scale with confidence. The right data is gold, right? In the world of AI agents, you need clean, trusted data in order for it to work well. And these aren't just principles we have in service. They're principles we're putting across the whole of Salesforce on how we transform our business. Now, I got to be honest. I wouldn't have a mission to serve customers if I didn't have customers in the first place. So I'm going to hand over to my incredible colleague Vanessa, who's going to talk about how we're driving agentic growth. Over to you, Vanessa.

Thank you, Bernard. Hello, everyone. I am Vanessa Tabert. I lead our Amer SDR organization, which means that I get to say that 80 of the most driven, curious humans at this company I get to call mine. But over the past few months, I have been hiring and ramping a completely different type of employee agent force for lead nurture. And it's been an incredible experience. So I'm so excited to be here today and and share with you how our customer zero story and our journey has shaped our key learnings of measuring value and impact and embracing new ways of working. The reality at salesforce is that last year at this time, every single chat that came to our website went directly to one of my SDRs. Whether it was tech support, non sales inquiries, anything, simple questions that created a lot of noise. It created lost focus. It just slowed us down completely. And so we knew that we needed to do something about it. But at the same time, even though every single chat was coming directly to my team, the reality also was that every single day, more people were coming to our website, filling out forms and becoming leads than what my people could ever feasibly follow up with. In fact, for every one marketing qualified lead that my team worked, three more were left behind. Three leads were left behind. We never put them in front of somebody. That means that 75% of the people that wanted to speak to us, we were doing nothing about. Now, as the person who leads the team, whose job it is to connect, that number lit a fire under me. I knew that it was going to be incredibly important for this to be how we used agentforce to solve that problem. We started small, we went really fast. You heard it from Bernard, you're going to hear it again. We started in June, we sat down, we did our vision, we decided what we wanted this to look like. Less than 30 days later, we were live. And almost instantly we were seeing real value, we were seeing real impact. Absolutely incredible. Every customer journey begins with curiosity. Somebody visits our website, they want to know what is salesforce all about? How can salesforce help me? But at that point, they don't want a pitch, they want clarity. And so the first thing that we had to do is we had to move our chat experience over to agentforce. Absolutely incredibly important. Because now when somebody is ready, when we see that somebody has an intent to evaluate, AgentForce is able to have a human step in with complete conversational context. When they're not, when it's not a sales opportunity, AgentForce immediately searches thousands of pages of content. It's able to deliver instantly the most relevant personalized resources possible. And when it's somewhere in between, that's where nurture agent is really able to shine. So for two decades, over two decades, at this point, our pipeline engine has run off of two different things. Marketing creates leads, sales closes deals. But what we're seeing now this year is something phenomenal. AgentForce is blurring those lines completely. What we're learning is that agents don't care about if something is sales generated. They don't care if it's marketing source, it just cares about growth. So for the first time, my team is not just waiting for marketing qualified leads. We're engaging prospects at the earliest place in the lead funnel. We're building trust and we're creating pipeline long before traditional lead routing would have allowed us to do so. For the first time, we're not waiting. We've made the shift to creating. But how do you measure the impact of an AI teammate? For that, I am so thrilled to be able to introduce Daniel Zaleski from our incredible data science team.

Thanks, Vanessa. Hi everyone. My name is Daniel Zaleski and I run a team of data people who are focused on improving the quality of our internal agents. And I'd like to introduce you to the lead nurture agent, Email quality Dashboard. This is just one of many tools that we've made as part of a customer zero measurement program. This dashboard grabs data from sales cloud, from marketing cloud, mashes it together with Data360 and presents it in tableau. Take a look at the upper right hand corner. You'll see a time series graphic that shows our average email quality score over the last couple of months and you'll notice that the pattern is up and to the right. We've improved the quality of the lead nurture agent's emails by 12.2% just in the last month. But what's a high quality email? Take a look below. We grab the last two years of data from Vanessa's team. The real emails that our SDRs or human SDRs send to customers, we mined through them with statistical machine learning text mining algorithms. We found a pattern and we turned that pattern into five heuristics. For every email, we measure the quality of the opener, the call to action, the level of personalization, the tone, the clarity. You'll notice just below the tone and clarity, we're knocking it out of the park. But the opener, the call to action, the level of personalization, those are watch items. So we're working with Vanessa's team. We rip the agent open, we change the configuration, we add additional data points, we change the prompts, all to increase the overall email quality score. And my team has proven with experiments that that score is statistically linked to an increased likelihood that the emails will be opened, that the leads will turn into opportunities, that the opportunities will turn into dollars. So let's scroll down a little bit. This is the details section of our dashboard and this is really where we nailed trust. In the details section, our executives, folks on Vanessa's team, everybody at salesforce can read the individual transcripts of all of the emails that go out through the agent and they get scoring on the right hand side so that they can see the heuristics brought to life. That's made it so our executive team can build trust in the agent. They can compare the way that the agent writes to the way that they write and they've seen that the agent is as good or sometimes better than them. And get this, we're taking this dashboard and this reporting and we're shifting it into the out of the box agent force observability product so that your teams are going to be able to test, tune and iterate in exactly the same ways that we do. Back to you, Vanessa.

Thank you, Daniel. This is absolutely hypercritical because this type of measurement is what allowed the agent to build trust with me, the business owner. It's what allowed the agent to build trust between sales and marketing, the executives and even the end users. Now, I've spent over a decade building sales teams the very traditional way. Hire, ramp, train, get roi, promote and repeat. But now I am leading a completely hybrid workforce. One that, yes, blend sales and marketing, but also one where agents are extending the impact of every single one of the people of my team. Because when you allow agents to focus on the noise, my team can focus on the things that are uniquely human. Judgment, persuasion, storytelling. Right. So it's absolutely incredible what we've been able to do here. Now here's the real magic of all of this. If salesforce can do this in 30 days, if I can do this in 30 days, we're four months in at this point. We got it going live in 30 days. Any company can do this. If you're an SMB and you don't have an SDR team, this is your team. If you're a global enterprise, this is how you get your message everywhere and you scale it. Because the goal is the same. It's every lead, every time. With agentforce, you don't have to pick between customer focus and scale. You get both on day one. So now that we've talked about how agentforce is helping us recoup left behind leads and revenue and how it's helping us nurture our prospects, let's learn how AgentForce is helping us grow and close our deals. For that, please welcome Marci Lederman.

Thank you, Vanessa. It's amazing to be here with all of you today. So my journey is a little bit different from Joe's. This is my 17th Dreamforce. I have been at Salesforce for 19 years. And this is our first ever Customer Zero keynote. So round of applause for that. And now the last year has been one of the most exciting of my 19 years because in my job I have the privilege of helping to define and drive the agentic experience for our sales organization. And it has been an amazing journey. Now you heard from Joe and Vanessa, Emily and Bernard, the journey that we've been on today. I want to focus for you how testing and tuning has helped to accelerate our agentic transformation and and drive more and faster business value to our sales teams. So what you will see here is that the agentic opportunity for our sellers is massive. We have 15,000 sellers at Salesforce. In fact 30% of our organization is either in sales or in a role supporting sales. But our jobs to be done analysis has shown us that 76% of their time is spent on non customer facing non selling activities like admin or prep. We also know that our top performers have a 4x deal velocity than the rest of their peers. This means that our sellers that understand how to navigate the complexity of a deal cycle are delivering forex deals and revenue for the rest of them than the rest of them. These two data points represent a massive opportunity for us. By us identifying the sales cycle we can reclaim selling time and make our sellers more productive but drive true revenue growth by scaling deals velocity. And that is exactly what we have done. We have started to completely identify our sales cycle. We are taking what our best sellers do and scaling it across the globe. And you heard from Vanessa and Daniel the great work that we are doing to help nurture and generate more leads. And as that lead has a seamless handoff from an agent to a person, our sellers now have an agent working with them hand in hand, step by step through the sales cycle to guide them, to enable them and to better equip them to have a meaningful rich conversation with the customer. We've seen great results. But let's not see it on a slide. Although I love this slide, let's see it in a real demo. So I am going to go ahead and welcome Andy White here. Andy is my technology partner. He's in Joe's organization. This is how much we love Andy. Andy runs the salesforce on Salesforce technology and that's how serious we are about this is we actually have an organization directly focused on this. Okay, so we are in Slack. I think it's important for all you to recognize that agents must be deployed in the flow of work and meet your employees where they are and that could be in CRM, it could be in Slack, be in the mobile. Our sellers are spending more and more of their time these days in Slack. So we're going to show you what their identity experience like is in this. Now, I just want to caveat. This is live, all right? So bear with us in case we hit any road bumps along the way. I am an account executive at Salesforce and I have a new account, Meridian Solutions. So I'm going to go ahead and ask the agent to help create a sales point of view for me. And what the POV is, is it's a content rich document that brings together content from 10 different places. It's looking at external information such as web searches or news articles about this account, but it's also pulling from all of our internal documentation across CRM Slack marketing materials and bringing it all into a document so that I am equipped, as Joe said, to have a meaningful conversation. And I don't go into a customer conversation saying, tell me about you. That is all done for them through the pov. So as we're going to go ahead and pull this POV up, what you see, it's searching through the different accounts here and folding that generate. Okay, we got it. So what you'll see here in Slack is a summary, right? But we can click directly into the canvas. Now, when we were starting to think about how we would identify the sales process for our sellers, we went and asked them what they wanted. The POV was the number one ask for theirs. They said that it took them anywhere from 1 to 15 hours to create this document. Now, what that meant is a lot of them just weren't doing that work. They did not have the time to go and do this. So we took what used to take hours down to minutes in the POV and in the canvas here, what you will see is it gives you a summary of the account. It shows us what products and services they have of Salesforce. As we continue to scroll down, you can see latest news information about them. But then it starts to actually create a point of view. It starts to actually call out for you potential challenges that this company is having. And it starts to guide the seller on how to sell to Meridian Solutions, challenges they may have and also how our products can potentially fit for them. What I'm seeing in this POV here though is that it's recommending Agent Force in Data360. We know that Agent Force is a pretty competitive space for us these days and so I want to make sure before I engage with them That I understand how we are doing against our competitors and so did a little demo trick here, but I'm going to ask how we are competing against sneaky software so that I am fully prepared. Now, this is a topic within the agent. We have deployed a series of different topics to help drive and help align our sales teams and get ready for these conversations within the competitive intelligence topic. We first launched it and we worked with the users and through the testing and tuning process, we actually heard from the end users. Originally, it wasn't very good. What we learned is that we were not connected to the right relevant data to bring in the context risk information that they needed. We worked with the business experts, our competitive intelligence team, and, and we tested and tuned the agent and we made it better. One of the things that we track through the testing and tuning process is utterances. And an utterance is essentially when you ask the agent a question and it comes back with an answer. A negative utterance is no answer or a partial answer. Through the testing and tuning of the competitive intelligence topic, we were able to reduce the negative utterances by 50%. It's a pretty amazing progress that we made in just a short period of time. Okay, so I have the pov. I feel pretty good. Talk to the customer. Now I want to try to secure a meeting. So I'm going to go ahead and ask the agent to generate a personalized email from me to the CIO of Meridian Solutions, and it does it here. And because it's in the same session as the pov, it's pulling information about this account to create that personalized email that speaks directly to the cio. So fast forward, let's just say a week. Play a little magic here. Sarah loved my email and we've secured a call, but I want to make sure I'm fully prepared for that call. And so I'm going to ask the agent to help me prepare for that call. And so it is going to generate for me a call prep document. And you can see it has. Let's load it. Let me go back up. Okay, there we go. It has very important information for me. A summary of the account, key challenges that we should expect to hear and talk to them about, discovery questions that I can ask, and then some other key next steps that I want to try to accomplish here. Okay, great. So the final thing that I want to do before we have this conversation is make sure that I have a bit of a secret weapon. And that's a customer story. Being able to tell your customer how other customers like Them used our products is that secret weapon. And so I'm going to ask the agent to go ahead and pull a customer story for me. Now, one thing that you should know is that our sellers used to actually have to go to our customer story library. They'd have to manually search through filter for a customer, pull up the story, read through it and then make a decision, is this a good one or not? So something that used to take a lot of time and really manual for our sellers is now down to a number of minutes. And all of this is possible because we are connected through the power of data 360 and that it's searching across both our structured and unstructured content to pull in the information that is relevant for me. So now I'm fully prepared for my conversation. I'm ready to knock it out of the park. So thank you, Andy, for that demo. There we go. So as, as you've seen, we've made a significant amount of progress in the past six to nine months in delivering the sales agent. But what we learned very early on is just launching the agent is a very first step. It needs constant love and attention and testing and tuning to be better. Do not expect when you launch an agent on day one for it to be good.

You just don't.

And I like to compare launching an agent to hiring a new employee. You don't hire a new employee, give them a badge and say good luck. You have to onboard them, you have to train them, you have to coach them and performance manage them. An agent needs the same love and care and attention and that is the testing and tuning process that we go through. Another thing that we learned that I am incredibly excited about through the testing and tuning process is that we uncovered that if you saw in that previous demo, it has a bunch of topics and really rich information, but I had to ask for every single thing, right? Not every single one of our sellers knows what to do every step along the way. And what we did last week is we are now testing what we call proactive nudges that now that our sellers are in the agent, instead of them having to ask for a customer story, it's now going to say to them, hey, you have an upcoming call. Do you want a customer story that I can pull for you? It's pretty amazing progress that we are making. And before I hand it back over to Joe, I just want to highlight something that he talked about and Vanessa did as well. Agent force is essentially forcing our organizational barriers down. So whether you are part of the business User team that is defining the vision and the goals of what you're trying to accomplish, or whether you are the technology team deploying and building the agent, an end user using it, or the team's helping to test and tune and measure it. We are now working together in a seamless way to help drive better benefits for our customers. And as customer zero, everything that we do and everything that we make better for our product is going to be delivered for you, for your benefit as well. So with that, I'm going to hand it back over to Joe. Great.

Thank you, Marcy. Thanks, Marcy. So we just heard a lot about this journey, and I think it was great. The way that Marcy set it up is it's hiring an agent is like having an employee. But as it turns out, we still have employees. We have a bunch of humans. And to talk about how we're getting the humans into the agentic enterprise and scaled up for it, I'm going to invite Natalie to stage our chief people officer.

Hi, Joe.

Thanks for coming.

I know. Thank you.

So you have a really unique viewpoint because you see all of the employee base that's yours to sort of monitor and help and empower. You've seen our five principles. Which ones are interesting to you?

Well, it will be no surprise. Embracing new ways of work has to be at the forefront of that. And the second is, we know that agents need to be grounded in data. If you're going to try and drive this change management of the workforce, then the workforce needs to know that the data is trusted, that it's accessible, and that it has context. So data is gold. And embracing new ways of work, that's great.

Well, we've been on this journey now. I mean, we say, oh, we've been on this journey for a while. We've been on this journey for a journey.

Here is a gift that keeps on going.

No, totally. Yes. But we've been on this journey for a year. How have you thought about empowering our workforce?

Well, there are a lot of different ways, but we've talked here at Dreamforce about the customer expectations, the consumer expectations. They have changed forever. And the employee experience and expectation is no different. So we've been on a huge journey, and we kind of anchored it to some key learnings along the way. Number one, learning is a meta skill. There is so much coming at us that we have to create space for our employees. So the first thing that we did is every single employee at Salesforce is an agent. Blazer Champion, which is our online tool via trailhead as you know, we were at a management offsite last month and we would be disinvited if we didn't get to agent Blazer, innovator. So now 98% of our workforce has the AI fluency and the data fluency. So learning has to be number one. And our job, if you're a manager, if you're an employer, the job is to create that space, space and learning opportunity. And trailhead is a great way. The second is, and Vanessa said it well, and Marcy, you have to have agents in the flow of work. Gone are the days where again, going back to the employee expectation experience, I don't want to look for an agent. I want it to be built into my everyday. And for us, slack is that operating system trust. We know at Salesforce that those employees who work with AI daily are twice as likely to trust the AI. And trust drives adoption and adoption drives transformation. So the trust is critical and then the last is impact. Everybody is in the business of impact. And for employees, we look at retention, we look at engagement. Has it improved the accuracy of results and where people are spending their time. So it's this constant, constant iteration of impact.

Yeah, that makes sense. So we talked about how we're doing it, but part of this is to let people know how they can do it. So if you had some advice, what would you say? What are you going to say? Because they're here to our customers?

Well, I think that goes to our framework, which is the four Rs. We love a good acronym, as you know at Salesforce, but sometimes things can feel a little abstract, especially when you're going through major workforce transformation. So with the Workforce Intelligence team, we designed the for our framework and essentially it's every job is being redesigned. We've worked with your team extensively on looking at the task of every role, which parts need to be automated, which should be augmented. Bernard talked about it with the support Engineer team. So first off is every job is getting redesigned with AI. Second is reskilling. There are jobs that we're hiring for that didn't exist six months ago. Ethical AI architects, agent, orchestrator, managers, I mean, you name it. And every day there is a net new job. So we have to understand what those skills are and reskill our workforce. The third is the redeploying of talent. We have redeployed talent from different positions like support engineer into Florida forward deployed engineering teams. And this is now about unlocking workforce agility. People don't want to feel like things are happening to them. They Want to know how do I get prepared, what are the skills that I need and how do I get redeployed? And the last is rebalancing. Again, your team credit to all of the work around the command center. You need to know where the role is happening between the digital and, and the human. So it's this constant rebalancing of information.

Awesome. Well, you know, you brought up tools, so maybe we could take a little look, maybe a little demo.

Oh, we love a good demo. And we have Robbie Birbik here with us to talk us through it. So like many of you, we started with building a lot of agents. We went agent kind of crazy because it turns out it's quite simple to build the agent.

Exactly.

But we built onboarding agents, wellness agents, benefits agents, new hire agents. But it turns out that was a very fragmented experience because, well, how do I know which agent to use? So with Data360, we actually simplified the experience for employees and we created the manager and the employee agent. So this is the manager agent today. It knows who Robbie is. And Robbie's asking, what is the AI fluency of my team? How do I know that they're prepared? What are the tools internally that they're using? And the manager agent knows Robbie's profile, the data on the team and can tell them, well, you have a lot of agent blazer champions and 50% of your team are innovators. Then I'm sure Robby will want to know, well, what are the biggest barriers? And we know that time or lack of time to learn is a big barrier. And so we can ask the agent, well, how can I help my team remove those barriers? And right in the flow of work in Slack, he gets a very action driven plan for him to drive that adoption with his team because that is also a huge skill for managers right now. Then we move over to the employee agent. And this is one of my favorite. As you know, it's the most used agent at Salesforce now. Over 350,000 conversations. And I want to know how well am I using AI, what am I using it for? That's going to help propel my skills to the next level. I also want to know what are the courses that you could tell me that I need to develop that fluency even more. And right there in the flow of work, it will take you to data programs on trailhead to increase your fluency of AI. So right there it is personalized to you. And then careers reskilling, redeploying talent. 50% of our open Jobs have been filled internally, but it's hard to keep up with all of the net new demand. So Robbie is asked, well, what are those roles? And you can see customer agent, designer, AI, augmented account executive, revenue intelligence, architect. And then the natural question, well, how do I know more about those roles? And right in the flow of work, Robbie can get connected to a recruiter or somebody who is in that job today to learn more. And that's really the power of unlocking human potential in the flow of work with agents.

Awesome. Well, I'll tell you, one of the things that I'm most excited about that we built is our engineering agents. So back to the flow work. This is just a really cool example on the screen here that talks about the way in which we build code. I just want to flash this because you keep talking about in the flow of work. And here's something that's a highly technical thing that's radically been changed. Thanks so much for coming on.

Thank you so much.

Great to see you.

Thanks, Robbie.

We're going to bring it home here. We're going to talk about just a little bit. We only hit the tip of the iceberg. I showed you the engineering agent right now, but here's kind of like the big agents and there's a bunch of them that run in here. But that earning the right to go fast I think is really important. So by focusing on those hero agents, we built the muscle and the skill to understand how to then go faster, add more. You can see we have a whole bunch of stuff in addition to these hero agents. And I want to bring you back to the start small, go fast, test, tune and repeat. The right data is gold. Measure value and impact and embrace new ways of working. These are the lessons that we learned. I think about it this way, when people say, like, what is the agentic enterprise? If you had to sum it up, what is the agentic enterprise designed to do? And it is humans for impact, agents for scale. All the work that you saw here, all the people here represent hundreds of people that worked on this stuff sitting behind it. The agents didn't get good by themselves. They got good because the people provided the impact. The people were focused. They used these lessons and they created things that were awesome. And then they hit the button to deploy and it went wide. And that is the scale. And that is a scale that is unachievable by regular software and is only available in Gentix. But it doesn't work if you don't connect it all together. It doesn't work. If you don't connect all the people together. And so you've seen a ton of what our products are, and we have a lot of awesome products. But hopefully the thing I wanted you to take out of this is seeing how we as practitioners put it together to create an experience, to tie these things together, to drive impact. And to do that, we want to also give you a little bit of lead behind this QR code right here. We'll give you a guide to a lot of the stuff that we talked about so that you could take these lessons back. We also have a ton of stuff on the website in Customer Zero, but the goal here is enablement. We're no different than you. We happen to work at Salesforce. We carry a different badge, but I can say definitively as having a different badge last Dreamforce, everybody could do this. And that's exactly what we're trying to do. We're trying to democratize the technology. And we as customer 0 try to tell our story to make it easier for you to do your work. We're going to make 1,000 mistakes, hopefully, so you don't have to. So with that, I thank you very much for your time. Thanks for attending the session and we'll see you next. Three force.