Introducing Agentforce 2.0 with Marc Benioff | Salesforce

There's a lot of talk about Agent Force. Agent Force, Agent Force. Of course, Agent Force will be the key. Salesforce unveiling Agent Force, shaking up the AI game. What if workforces had no limits? What if Agent Force redefined business? The time is now for a bold new horizon.

300 million people reach out to us every year. We need to scale the way we interact. We want to help a billion people improve their health, so a billion people have a better life. If you can work with people on an individual basis, a personalized basis, with Agent Force, the sky's the limit on what you can do with them. A digital labor force that's instantly scalable. Switch on Agent Force and it's always available. Agent Force is going to free up our people to work in a different way with their clients. Agent Force can answer a thousand questions all at once, and so that's something that no amount of humans can do.

They've done a good job of making it so easy to use. You don't even need to know how to input code or anything. I was amazed because I don't have a technical background, but I was able to build an agent so easy and so fast, and more importantly, it actually worked tirelessly. Driven, the insights are clear, from data to action. The future is here. We have a plethora of data that sits within our walls, one petabyte of data every single day. We take a multitude of data sources to create the unified data, data profile, and 360° view of our business. Agent Force is going to be knowledgeable of everything that's going on in their world, allowing for us to have deeper, more impactful interaction with our customers. Agent Force will enable us to service customers in ways we couldn't imagine before. Productivity rises, margins expand. With Agent Force, growth is always at hand.

Salesforce is constantly innovating and constantly creating new value. We set up Agent Force in a week. We saw an over 40% increase in case resolution. The ability to double our revenue. Agent Force will change the way that we live, the way that we learn, and the way that we work. Agent Force will revolutionize our business. We want to be the first to welcome you into the future, an era of abundance as far as the eye can see. This is Agent Force. This is what AI was meant to be. And now, please welcome to the stage, Salesforce Chair and CEO, Marc Benioff.

All right, all right. Good morning, everybody. Aloha. Aloha. All right, good morning. I am super excited to be here today. Uh, it only feels like what, a couple months ago it was Dreamforce, and here we are. And and we're talking about Agent Force 20. And a lot has happened, uh, since then, and it's been an amazing journey for us. Uh, time for us is definitely accelerated, and the progress that we're seeing made is incredible. So now is really the time for us to kind of take a step and to kind of walk into the future together and to show you what we have been building, what has been happening, and what we're really excited about.

There's so many things that have happened, but perhaps the biggest thing that has happened, uh, happened about a week ago, or maybe it's two weeks ago now. I've lost track of time completely. And that's when we hit the switch on our own site, help.salesforce.com. How many of you already seen help.salesforce.com? So what we're going to show that to you today, but help.salesforce.com has always been our primary customer support. It's where our customers go to get help on Salesforce answers. It's run by our Service Cloud, does about 32,000 interactions with customers every single week. And behind help.salesforce.com is, of course, our thousands of support agents that provide then the kind of quality support that's needed that help.salesforce.com was not able to provide.

So imagine this, that our team went to Dreamforce and heard this idea that there is a vision that humans with agents can drive customer success together. Humans with agents drive customer success together. It's kind of a fantastical idea. And yet, when we hit that switch two weeks ago, and what is happening right now in our current moment, right now, there are thousands of customers working with thousands of agents working with thousands of humans all in tandem right now. You can go to it at help.salesforce.com and see it actually play out.

I guess for 25 years as the CEO of Salesforce, what have I really been doing? We've been building software that helps our customers manage and share information, customer information, sales and service and marketing and commerce, analytics, Slack, Tableau, MuleSoft. It's been amazing. But this is different. And what is happening is really different because all of a sudden, as a CEO, I'm not just managing human beings, but I'm also managing agents. That there is an agentic layer around the support today for Salesforce. It's not some vision, fantasy, in the future idea. It's what is happening right now.

Why is that matter? And why is that significant? I think there's a couple of things that are important. We've seen the movies for decades on AI and where the future is going. Of course, you know, Minority Report and WarGames and Her and on and on and on. We really enjoyed them. But we have haven't really been working with agents ourself, using these kind of next-generation AI models grounded in the data that allows us to do our jobs and actually making our businesses much, much better, much easier, much lower cost. So suddenly, what I realized is Salesforce is providing, you know, we have Slack. It's an amazing, incredible collaboration and search layer, helping you to run this incredible capability in your business. I use it every day myself. It's probably a hundred billion dollar TAM. And then there's our CRM products. And there's plenty of analysts in the room who are sizing the CRM market all the time. And I don't know what the size of the CRM market is, but it's probably a couple hundred billion dollars. Maybe it's a little bit bigger. Very exciting market. It's been very extensive for us. It's why we're going to do 38 billion in revenue this year in our startup. So now we'll get to that. But it's kind of, kind of still on startup mode.

And then there's this new opportunity that we all are still getting our head around to the point that even at Dreamforce, I didn't use these words, digital labor. Because not only were we managing all this information, but we're also now managing this digital labor. And we can see that a new world is opening. When we look at digital labor, when we're talking about agents, we're talking about digital labor. When I took the limo here this morning and I got in the car and I hit the button, start ride, it's digital labor. It's the robot is bringing me here. It's all happening digitally. There was nobody in the front seat. And that idea that then robots are physical manifestations of agents, that these agentic platforms are the fundamental enabling technology for the robotic layer.

When we think about what Dreamforce looks like in 2025, know, of course, we're going to have help.salesforce.com. We're going to be demonstrating our entire agentic layer. But you can see how already how voice is becoming an incredible new user interface. You can see how avatars are becoming incredible new new user interface. And you can see how robots will be an incredible new user interface. There's no way that Dreamforce 2025 will look like Dreamforce 2024. The, we've already crossed the bridge. We've already crossed this bridge. And what the bridge is, is this bridge to this new world of digital labor.

I'm not sure that when we started this journey, even we fully understood where we were going. Because when you look at the digital labor TAM, it's not in the billions or tens of billions or hundreds of billions. It's in the trillions. And in fact, it's so new and so avant-garde that many of the analysts and firms that we have the highest respect for and work for every day haven't even fully sized it yet. So this is an incredible new opportunity for all of us.

Now, why I'm excited is when we finished up Dreamforce, we all got together as a tech team and then we said, okay, we're going to get this working for ourselves, not just for our customers. Customers were excited and got going, and we've seen a lot of customer success. But our team said, no, we're going to deliver this at scale. And the reason why that was important is because, and what was really on my mind, is by the time we got to Agent Force 20, and here we are, that we wanted to say, hey, go to our website now. For those of you who have your phones, like I do, I have my phone here somewhere. And you go to Salesforce, you might know we have a little experiment running on the front of our website that we have not been promoting where we put an agent on the front of our website. So you can talk to that agent about Salesforce. All of our, uh, product information and competitive information, it's all in that agent. It's not currently a product that we're offering. It's fronting our WordPress site. You know, probably know Salesforce owns 10% of Automatic, so we love WordPress. So here it is, that I can talk to that agent about everything going on with Salesforce.

Now, I become a Salesforce customer after working with this agent. And I go to help.salesforce.com and I log in. And now I'm grounded into my information, which means it's looking at my data. It's looking at my metadata. The Data Cloud is federated to my other data sources. And all of a sudden, what I can do is start to work with that agent to resolve my customer service and support needs. Suddenly, overnight, it was stunning. Escalations to humans dropped by 50%. That's a big thought. So we have about 32,000 conversations a week on that site. And then just about 5,000 were now going to humans. And before it was about 10,000. 83% of all the questions now on help.salesforce.com are being resolved by the agentic layer. Incredible. Totally unique and completely different. So we have to reconceptualize our entire business. We're thinking about how is AI transforming Salesforce, and what does this mean for our company, and also for the industry.

Now, while we've been rolling out Agent Force, and it's been a lot of fun, and talking to customers, and running our Agent Force World Tours all over the world, doing our launch pads, and building agents with everybody, it's very interesting because this is true for me. I'll be talking with so many reporters and press and doing podcasts and doing all these things. And people will say, well, what about this competitive situation? Or I heard about that. Or did you know that Microsoft has already been in this market for two years? They co-pilot. And I had some funny things to say about co-pilot. You know, I don't know how I came up with a couple little things. Clippy 2.0, everybody liked the Clippy 2.0 comment. But the most interesting thing is, is that when you go to Microsoft's website, and you look for co-pilot, or you look how they're automating their support, or you look at how they've taken the technology over the last two years and used it for themselves, you can't find it. It's exactly the same as it was two years ago. That Salesforce is a company today who doesn't just put on the outside of Moscone, right there, humans with agents driving customer success together. That's actually what's happening inside Salesforce right now. Humans with agents driving customer success together. And since Dreamforce, so many companies have pared our words and said they also believe that and are also trying to do that. But when you go to their sites, because I think the website actually is a pretty good place to see like how real it is, seems to me that there aren't as many companies doing this as saying that they are. You find a lot of forms, you find a lot of FAQs, but you don't find a lot of agents. Because this does require a level of computer science that I think a lot of people still don't fully understand. It's why we hear so many kind of very pithy and provocative things from so many people about this, and then they themselves cannot get it running. And you'll hear all kinds of amazing and fantastical things about the future of software, and what things are going to look like, and how it's going to work, and what is going to happen, and who's going to do what to whom, and how, and then where is it? Because at the end of the day, we have to deliver customer success. It's one of our core values: Trust, Customer Success, Innovation, Equality, Sustainability. It all has to play out right now.

So, Silvio, are you here? Did you make it here this morning? We just come up here. Please welcome Silvio. Before we get going, I, I just want to thank Silvio and his team because without them, I don't think it was really possible. So thank you very much. And Silvio, just tell us a little bit about what is going on. We watch your research, we're trying to implement your vision. It's hard. You're, you have so many other things going on, even at the robotic layer, and at Stanford, the Mahalo models and robotic models. Your wife is doing incredible work with her new company, with her new world models, and the demonstration she's already showing just grounds us right now based on what you see Salesforce deployed. What we're about to show everybody, just show, just give us that open, that door to the future for us.

Yeah, thank you, Mark. Well, I think it has been really an honor and excitement to to work, um, at the, at Salesforce and leading the research organizations. We've been really looking at how the future is going to shape up in in the space of AI, uh, for the next few years. And and what we have seen clearly, that we were, uh, looking at the transition from GenAI, which is a tool for helping productivity and efficiency, into something that can lead to automations, can lead to, uh, helping, um, users and stakeholders to really automate their jobs and makes, uh, things much more scalable. Uh, and this is exactly what we've been pushing collaboration with with product team, engineering team, to see what are the opportunities for us to really, uh, uh, help and transform in this amazing landscape. So the concept of agent is the concept of, um, enable an AI to plan, um, a set of actions to be able to perform these actions on behalf of humans, and, and the ability to really, um, make, thank you, Mark. I want to remember the moment. Make this happen, uh, with with a level of accuracy and, uh, and the trust that we need for these applications. So trust is, is definitely a very important component as we build these agents. And this is exactly one of the focus we put as well, uh, looking ahead. So we are really excited to see how this is paved the way for this new robotics layer that Mark mentioned. Um, so in a way, you can think of an AI agent, a digital agent, as a robot operating in a digital space. Uh, those actions that these agents perform are in the digital space, but the exact same tools, the same capabilities that we are looking at now, they also be powering the future of physical agents, which are the robots. So we still going to have a planner, we still going to have a reasoning, we're still going to have a memory. Uh, the difference is that now these, uh, new generation of agents will operate in the real world, and interact with the physical, uh, humans and physical, uh, components. So this is going to be very exciting. It's going to pave the way for this new, uh, era of of digital labor and physical labor.

Okay, as we've been getting ready to deploy help.salesforce.com, and it's been a couple months since Dreamforce, and we've now have this running and live. Even have your own microphone now. So what has been the hardest part? You know, we have our apps, so the apps are working with the agents. We have the Data Cloud, it's ingested all the data, it federates to all of our data sources. We have the agentic layer. What has been the hardest part?

Well, I think the hardest part has been really to, to, uh, connect the dots between what agents can do and the data that it's actually sitting there in our Data Cloud. I think being able to retrieve all the information, um, and be able to make it useful, you know, for these agents to operate has been very, very important, uh, because an agent, you know, without having, without being grounded, without, without being contextualized, it's kind of useless, right? You don't want to have an assistant that doesn't remember, you know, what's your, what are your preferences, doesn't remember, you know, what are the data connect to the customers. So having this ability to actually feed into the, into Data Cloud, into the ability to contextualize and associate all the business logic, all the business metadata that we have from there, is going to be, has been very, very important, very critical.

Now, there's a lot of companies that have been saying they can do this too, you know, since we announced Agent Force, there's lots of agent announcements. We really unleashed a whole set of Misha's. You've seen all, it's been incredible, right?

Yes. Okay. Now, why do you think they're having such a hard time building and deploying their scaled agent systems?

Well, I think that first of all, uh, again, the fact that we do provide the cloud has been a big enabler. So many other companies don't have that. Don't have also don't have the, uh, the know-how, the business logic associated with all these processes. We've been working with customers for 20 years, so we know exactly how, what customers want, what are the flows, what are the, the business logic. And this is what we're inheriting and injecting, embedding into our agents. Uh, and it's actually a huge differentiator, besides of course, you know, being a technology that it's also superior in terms of accuracy, thanks to our engineering team, our research team, uh, that we are be able to really make, uh, those reasoning more accurate, and more factually correct.

And what is your dream for Dreamforce '25? What do you want to see have happened with the whole agentic layer and what we will be able to deliver to customers?

Well, I think I would really love to see an actual robot movie on stage be able to come talk to you, uh, and we are hope we can make this happen.

I agree. Nice job. Thank you. Thank you very much. Please thank Silvio.

Okay, all right. Well, that kind of sets up how I've been thinking about this, which is Agent Force is a digital labor platform. It's not words that we use to Dreamforce, but as we've been deploying this to customers, I've been using it myself. I'm like, wow, I am a CEO of a company that is managing agents and humans, and I have a digital labor platform at my disposal to augment my support, my sales, my service, my marketing. I'm doing already 32,000 conversations per week, 83% of these are being resolved. And then 50% of non-resolved, basically 50% of this kind of, what I would say, um, escalations to humans, that's kind of the rate where we're like, okay, we have knocked this down where it's only like where we would have deployed, let's say maybe 10,000 of those 32,000 humans before Agent Force, 5,000 are now getting deployed. So let's look at some of that. We really start here. If we're not eating our own dog food, we can't really get at this point on another podcast or have another whether meeting with another analyst or a reporter, or even show our face with a customer and say, this is a great product, you should use this. So this has been a huge priority. That number one, we're using this, and it has been a huge amount of work. So to Silvio, to his team, to our incredible engineering organization, our product managers, the folks you have, we have here, to you can really talk to the experts. Will you guys just stand up so I can just say thank you to you? Stand up if you've been working on deploying this technology. It's been great. So we have our top technical leaders. Thank you. All right. We're going to get into all of them, and I'm so grateful to each and every one of them. And as you see that we've deployed this technology, you're also going to see it's been built with our core values: Trust, Customer Success, Innovation, Equality, Sustainability. And your data is not our product. And in the world of AI, I think we all understand how a lot of these other AI companies are operating, right? You're on their models, they're taking your data to train their model, to make their model better. That your data is their product, and your data is how they're making their product better. We're not doing that. We've also maintained during all this our core 111 model where we've been able to give away almost a billion dollars, done 10 million hours of volunteerism. We run over 59,000 nonprofits and NGOs for free on our service. Over 20,000 companies now have already copied our 111 model. So that's very cool. We've given away almost a billion dollars, and those companies have given away more than two billion dollars. It's an incredible part of our industry. And as we're going forward, we're really encouraging these next-generation companies to copy this. And we all know that AI is also has an energy issue. That it requires a lot of energy to train these models and to run these models. And so our commitment to having a net-zero data center and net-zero systems and being a net-zero company is as important to us as ever. We continue to focus on our core business of doing well and doing good. And we just delivered our goal to our investors that this little San Francisco startup that's in the AI space here, just down the street from a couple other AI startup companies, will do 38 billion in revenue this year. We're going to deliver 33, is that 33.1% operating margin this year? Is that right, Michael? Pretty good. Okay. I think 12.9 billion in cash flow. It's not bad. It's good. Good for a startup.

And then I kind of get back to this. It has been interesting even hiring. And we're going through a big hiring surge right now at Salesforce ourself, and we're adding another couple thousand salespeople to help sell these products. But we already have 9,000 referrals for the, for the 200,000 positions that we've opened up. It's amazing. But it is pretty clear to us, you look around on a global basis, there is a global labor shortage going on. That labor has been mostly stagnant, is in the United States and also other markets. It has to do with a lot of different things, declining population growth, an overall labor shortage, you know, reduced, basically reduced production. It's all together, really has kind of showed, hey, we have slower, lower GDP growth. But then what's happened is this really other interesting chasm that has to be crossed, which is that, you know, we talk to customers and they're like, yeah, we're very overwhelmed. We're trying to grow, we need that labor. We have a certain level of fixed capacity, but productivity is stalled. And we're also going with some level of burnout. Even yesterday, I was at UCSF, I was getting worked on. I've been doing all kinds of things to reboot my health after I had a little problem and Fakarava, uh, a couple months ago. And when you talk to the doctors there, really pretty burned out, actually. And even when I was like coming in for my medical procedure yesterday, there's a lot of people like calling me ahead of time and then calling me afterwards. And every time thinking to myself, wow, I can kind of see where that burnout is coming from. And I wonder how we could automate some of these pieces. You know, everybody wants a zero hold time. They want a more personal and empathetic approach to their experience. They want to work with experts. You start to put all of those things together, you can see the beginning of why this kind of agent idea is so attractive to so many companies.

I've never been more excited about anything in my career. This has been amazing. But talking to customers, I've never seen a fever like they have had in trying to understand the technology and also to acquire it. Brian sent me a text last night that he has now closed more than a thousand, uh, paid Agent Force deals, close to 200 in the first, yeah, amazing. It's great. More than 200 in the third quarter, and the fourth quarter off to a great start because so many customers get this, that this can really help solve their needs. This is the power of it. We can see that to unlock this GDP growth, we need a breakthrough technology. We have to become a digital labor provider. So while we for 25 years have been helping companies to manage and share their information, store their data, now here we are delivering digital labor. And digital labor is this new horizon for business. This idea that a door has opened for business, and that business will never be the same. How we architect our businesses and run our businesses and staff our businesses and think about our businesses will never be the same.

And when I started to look for the textbooks and the surveys and the reports and the analyses on digital labor, there's not a lot out there. And our top software analysts and top software experts, you know, they have written about things maybe called hyperautomation, but not really about digital labor. Like this is a big idea. That right now, there are thousands of humans working with thousands of agents responding to thousands of Salesforce customers right now. So where is this going? We can't exactly size the market yet. We think it's the biggest market, the most exciting market we've ever been in. And we also see that in this market, that a lot of our core values are still in place. It's about trust. It's about getting it out and making it easy to deploy. Just like look at Salesforce. Now, here we are, it's deployed. Everything, the good, the bad, you're on it. It's amazing. Here it is. And humans are in the loop. So if you're on help.salesforce.com and it's not going well, you can bring in a human agent. And the human agent is using our Service Cloud, and it's deeply integrated with the agentic layer. It's not two separate things. It's seamlessly integrated. And you have that zero hold time experience. It's integrated. It's open. You can add other services. Something has changed. Something has fundamentally changed and is different. It's not that Salesforce wasn't already a major enterprise AI supplier. We'll do two trillion Einstein transactions this week. All of our customers are running on Einstein. They use it every day in their sales and their service and their commerce and all of their these areas. And they use Einstein and they use this incredible, you know, artificial intelligence across the board. But just like in the limo that I came to this hotel today, you know, and when I go back, there's a lot of people talking about autonomous driving right now, and how great it is, and all the opportunities, and what it's going to mean to have a digital driver. But there is a lot of talk. But there's only one button you can push in San Francisco, limo. There's no other buttons to push. There's no other service to drive us home. And I would contend to you, there's only one company today that can deliver this at scale, as evidence by what we're going to show you today, across every industry. And I've had so many conversations with so many CEOs across all these industries, and the opportunities are just incredible. It's amazing. But the real opportunity for each one of these companies is to do something that they've never done before, or to try to get through an obstacle that has been blocking them. In one case, I was talking to the CEO of a large bank. They've already deployed part of our solution, part of Agent Force, been very successful for them. And when I was talking to the CEO, you know, he had this kind of latent desire, he always wanted to go into another geography, another huge market. And the reason he couldn't go is he didn't have the people. He had the systems, he had the brand, he had the trust, he had the regulatory capabilities, but he had not made the leap. And then when he realized that the agents could deliver the labor in that market, his mind opened up and realized this is a moment of abundance for his bank. He had been constrained by people. He couldn't go to this geography or that geography. He couldn't move into this product or that product because he didn't have the people. But when you have digital labor, you have a level of flexibility that you did not have before.

In the case of some of these early customers, like Addeco, where they're working with 300 million candidates a year, working on their resumes and consulting them and helping them to get into the right jobs, and working with employers, it, it was the only solution because who else is going to be able to operate at this scale and capability to be able to deliver for them what they need? They were already a huge Salesforce customer, and and they were just able to just turn this on. And in the case of RBC, they're already using Salesforce across the US and Canada, and now with wealth management, they're able to turn on these agents to work with their clients to deliver that next generation of capability. We've all known this was possible, but now we can see that with Agent Force, this is what AI was meant to be. And as Silvio said, it takes many components to get it right. The Data Cloud, the integrated apps, the full agentic layer, the zero copy on the RAG, deeply inside the Data Cloud, the, the fundamental Hyperforce platform. We're going to go through this in detail to show you why ours is working so well right now, and why you're not seeing it from other vendors. I'm sure that other vendors are going to try to fast follow us, of course. We don't want to be the only ones being able to deliver the solution at scale. But it is a noticeable fact that even some of them who claim that they have a two-year head start in AI, that they've delivered this incredible technology, but they don't use it themselves. It's not even on their website for their own sales or service or support or marketing. That is amazing to me. But it should be a sign to all, all of us, the different vendors are able to offer different levels of capability right now.

Okay, so I'm very excited about this, as you can probably tell. And I want to introduce you to my good friend, Adam Evans. And Adam, you may remember, we bought his company about a decade ago, RelateIQ. He was with us for six or seven years, left to build his new company, which we have the opportunity to invest in, called Airkit. Airkit. And then we bought it back about a year ago. Is it? I'm a boomerang. You are a boomerang. And now, um, you're going to show us this incredible next level. So thank you, Adam, and welcome. Please welcome Adam Evans.

Thank you, Mark. Thank you. So the title of Mark's last slide, Agent Force is what AI was meant to be. This is something that we think about a lot. We think about it when only three months ago was when we debuted Agent Force. Yet at Dreamforce, we launched over 10,000 prototype agents. Just heard Mark say 1,000 customers, paid customers have started building their agentic layer, and many more have gone live. And how is this possible? How is it possible to have such speed and precision to build Agent Force? And that's because it's built on the Salesforce platform. So all of the 25 years of product investment between the metadata layer, all of the things from search to Data Cloud to understanding flows and processes and much more are all there that we're building on top of. And our customers are loving it. How can they just turn it on? As you just heard, that's because all of their data exists. Their security models are in the system already with permissions for the right guardrails for agents they've deployed and defined their processes across all of the clouds. So Agent Force can be there. And the results have been phenomenal. Hundreds of customers have gone live. They've started their agentic layer, not just purchased, but delivered it. Customers from every industry and all sizes. Customers like FedEx, that's using their Agent Force to streamline operations in customer service. Customers like Saks Fifth Avenue, that are using Agent Force to answer customers' questions around retail during kind of these peak moments and the holiday season. Customers like 1-800-Accountant, that are taking 90% of inbound requests with Agent Force to do things like quite complicated, like looking at tax assistance, and also scale their sales team with products like Coach to make sure that everyone is enabled. Uh, in my personal favorite, uh, companies like Pristina, Pristina Health helps customers, or helps patients with chronic disease management, right? This is like diabetes and these kinds of things as well. So you have agents with empathy that are helping the customer along their way in their own journey, and bringing in providers at the right moment to escalate so they have quality of care and lots of context in doing that. And when we talk to these customers, we heard one thing loud and clear, and that is that co-pilots are not enough. And don't get me wrong, co-pilots are a generational leap between chatbots that we've all seen that are programmed robotic. Co-pilots are the next wave. The only problem is that they're reactive. That they're 100% human in the loop, where you have to ask to get a response. And if you think about a customer service rep at any kind of contact center, there's a thing called average handling time, okay? This is measured in seconds. And that extra task of asking every time is immediately not a productivity boon, but it actually takes more time and hurts average handling time. And what we need to really break through to the next level productivity is agents that can take action, not like co-pilots that are reactive, but agents that can be proactive, that can work the case before it even arrives, and maybe even solve the issue with the customer so that the case never had to arrive to begin with.

And so as all of these companies are investing out and building their agentic layer with Agent Force, they've been asking us, how do we take it to the next level? They're looking for a platform for digital labor. And it's very exciting to announce Agent Force 2.0. Agent Force 2.0 has everything we need to not just scale our clouds, but to help every team and every workflow beyond CRM to take actions outside of Salesforce. And it's built with a deep integration to Slack. So we have a connection between agents and humans where we want to work. And additionally, it's built with more trust. It's built with more trust because of the Atlas reasoning engine, ability to look at more complicated tasks, to do more, but also do it with consistency and auditability that enterprises demand. And there's much more in Agent Force 2.0, but we want to talk about these three things today. And I want to do a little double-click on this last one about trust and the Atlas reasoning engine.

So there are two things here that you should understand. Organizations have lots of data, and we talk about unstructured and structured data in Data Cloud. And one of the techniques that's very important is the ability to study that data. Agent Force and the Atlas reasoning engine indexes this information and creates a semantic understanding, a semantic understanding using things like knowledge graphs and recursive trees and more. But it's about studying the information ahead of time. And this is including not just your data, but your actions, your policies, everything that you've been investing in Salesforce for the years, uh, to be. And then during the reasoning loop, where the agents are thinking over what to do in the decision, this is the moment where we're doing reflection and inference scaling, where an agent can think about what it's doing when it has all that awareness and have a supervisory role that makes sure that it's taking the right choice, taking the right action, following the policies, and having the guardrails that you need.

Ultimately, agents take action. And we think about the kinds of actions that Agent Force 2.0 can take. 2.0 can take. It's not just the CRM, of course, we have that Sales Cloud, Service Cloud, Marketing Cloud, etc. But it's about what is beyond CRM. It's about every team in every workflow. So whether that's connecting to SAP to process a refund, or maybe you're trying to onboard an employee and you need to connect to Workday, Agent Force 2.0 can now cross the boundaries beyond CRM. And if you think about the packaging of these actions, how do we bring this kind of action into an agent? This is what we call skills. Skills is a fundamental concept of being able to create new capacity, new agentic capacity inside your Agent Force. And of course, there's skills for our clouds, our AppExchange, and our partners have skills. And you can think about what's inside of a skill. It's not only actions within the CRM and outside, but the instructions coupled with those actions about how to take them. And you can think about an agent also not having one skill, but instead many, so that an agent can help service your customers and also sell at the same time. And in a world where multi-agent comes, you can configure it so that skills could be specialists of agents working together as well.

So what are some of the skills for us for Salesforce and the CRM? We'll start off with Sales Cloud. So Agent Force has the skills within Sales Cloud to help alleviate, you know, tasks for sellers, so they can scale themselves to be able to follow up with leads, to qualify, to kick over with the sales development skill. Additionally, once all of those deals and opportunities that are warm and qualified come in, you can use Coach to practice your perfect pitch as a seller in the context of the opportunity and all of the information that you've gotten. It goes into Service Cloud with the ability to help customers 24/7. We heard a lot of these stories right now where you're looking at 90 plus percent resolution rate. The tier one tickets can be handled now by agents. And when things ultimately do escalate, or if they escalate, or if somebody needs a human in the loop, it's not a cold handoff. All the information that's happened gets written up in a nice way for the agents, saving time. So again, it's proactive, not reactive. That's a key difference. And it's not just about sales and service. Marketers get on in the action too. We have the ability to take all of those interactions with customers to create personalization for better segmentation, or marketers can use agents to help them create the perfect campaign. And in Commerce, the same concept. Whether you're doing merchandising, or maybe you're thinking about promotions, as somebody that owns the storefront, or with a personal shopper agent, bring the agent right to the customer on the website, helping them find the right product for their needs. And because it's tied into the order management system, after the sale, be able to help service with them. You're going to have a lot of orders. And if those orders are enterprise orders with revenue and Orders Cloud, you can now get agents to help you with more complex quoting and complex contracting that understand how to ask the questions to build these enterprise quotes with Revenue Cloud. And with all those orders, you're going to have more data. And what do you need with data? You're going to put the data in Tableau, and agents to help you understand it, that deeply understand the relationship between the Tableau, the metrics, your terminology, so that you can ask questions in natural language, not needing to know how to use SQL and everything like this, and be able to get visualizations and the answers that you need very quickly. And where are you going to do this? In the place that we all love to work in, Slack. An Agent Force has a deep integration, as I mentioned. We're going to dive into that a little bit later. But I also want to talk about beyond CRM. This is about the entire enterprise, okay? And so for CIOs, for admins, for developers, I want to just speak to you for just a second here. When you think about all of the information that you're connecting into Agent Force, there's a semantic layer, so structured and unstructured, so Tableau and understanding those kind of terminologies, unstructured data inside a Data Cloud, and bringing that back. There's also a data governance, so a permission layer here. So we talk about things like trust, but also what data do these agents have access to? That is all part of this, as part of the guardrails. Additionally, we think about moving and connecting to actions to be able to take things outside of the CRM. MuleSoft has connectors out of the box for every major ERP. But more importantly, you can connect to any API. So if you spent years building your own systems and your APIs in place, you can connect to it. In fact, you can connect to it with natural language by describing what you want, and the agents will build an integration for you, also supporting things like open API standards and other schemas as well. It's just drop in. And with all this power about being able to understand more data, more actions, and connect to things within the CRM and outside of the CRM, you are going to need more tools to make sure that your agents are consistent. So we have Testing Center, a brand new product specifically designed to test agents, everything from do they follow instructions, are they grounded, latency, and so much more. We've opened up the entire state of what's happening inside of Agent Force so that you can audit every part of it. And what's happening with agents? Agents are becoming the new apps. When you think about building apps, you think about all kinds of years of tooling for the software development lifecycle. So your test in your Testing Center to make sure that you have confidence in the agents that you're launching can also be used to make sure it's a guardrail or a gatekeeper. So that as agents have more skills and can do more in the future, that before you release the new version, you can rerun the same test and make sure that you don't have regression, keeping those controls as part of the guardrail.

By the way, I, I could talk to you about Agent Force, uh, literally for hours. Uh, and so what I'd love to do, though, is if there was one takeaway, it's that Agent Force 2.0 goes beyond CRM. It allows you to take, uh, any action for any team and any workflow. And I think a great customer story to emphasize this is what Mark mentioned earlier, Addeco. Addeco is one of the largest recruiting organizations on the planet. The scale is immense. They have over 100,000 companies, they represent 20,000 recruiters, they put to work 3 million people every day. But what's really daunting is that they have 300 million applications. 300 million applications they have to process. Now, last time I checked, we didn't have a recruiting cloud. Okay? That's not one of those clouds in the customer 360. So how does Addeco use Agent Force to go at scale to help their customers and recruits and connect them to employers? And the answer, uh, is that Agent Force 2.0 goes beyond CRM. So why don't we go ahead and show you a little bit here? Let's start off with something that everybody is probably at one point of their life experienced something like this. You're looking for a job, okay? And when you have 300 million applications a year, that's a lot of inbound, right? This is an experience that a lot of us, you know, hopefully you haven't had. You might send an email in, not get a response, maybe you have to follow up, maybe send a couple applications in, no response. It's not a great experience from a candidate. Nobody wants this. Addeco doesn't want this. But this is the reality because on the other side of this, what does it look like? Hiring managers and recruiters are inundated. Too much inbound, too many applications. How can they scale the personal one-on-one touch? It's impossible. 300 million is nearly a million a day. So how did Addeco do it? Well, they created an agent. So what we're going to do is we're going to create an agent right now, very similar to what Addeco did, to show you how this is an unlock for anything. So we're going to go ahead and hit new agent. And the first thing you're going to see is those skills. These are the bundles of functionality, the agentic capacity that you can start using to build your agent. Now, lead development from

Uh, Sales Cloud, this is very similar to recruiting, right? You've got leads that you might, uh, do outbound to. Then you've got to talk to about your products and services. You've got to qualify them and introduce them to a seller as a warm lead, right? Qualify them. It's very similar. It could be a candidate. It could be questions about a job. It could be qualifying that the candidate and introducing them more into a recruiter, right? Scheduling an interview. Very similar process, different words.

Okay, we could use that and remix it because all of these things are able to be remixed very quickly. But, um, that would be too fast of a demo. So what we're going to do instead is we're going to create one from scratch. Okay? We're going to create one from scratch because I want to show you something that's really cool. And that is that we can create an agent just by telling it what we want to do in natural language.

So in this case, we're a recruiter, right? We want to be able to talk to the candidates over email, interface to the applicant tracking system that's not inside of Salesforce, outside to look for the perfect job openings, figure out what those job openings are about, and ultimately connect them, schedule maybe with an inter with a recruiter, etc.

So let's go ahead and hit, uh, start or next. And what's happening in the background is that all of the data that we have, all of our enterprise data sources, all the processes that we have, the flows, the Apex, everything that we built in is being scanned for the use case that we typed in. And it recommends what you can include kind of out of the box for skills. But beyond that, in fact, let's actually talk about a couple of these.

So, analyze data from Tableau. That sounds pretty great. Maybe we want to understand what's happening in our pipeline for for candidates. There's that lead development, by the way, that's been now kind of modified. It's very smart now for candidates. So we could keep that in there. Appointment scheduling from Service Cloud understands my calendar, can schedule recruits as a hiring manager. That's fantastic.

Um, but I want to point out this one because this one's different. This one is AI generated. Okay? So not only is it really amazing, but I didn't have to code anything at this point. I described what I wanted. It scanned all of the, the library of skills that we provide, all my data, all my flows, all my existing integrations. But it understood for my use case, which is recruiting, that there was a gap. Then even I had a lot of stuff in Salesforce, there was a gap. It said, I'm going to generate a new skill for you right now for this one use case.

All right. And it is about what pre-qualification. So let's go ahead and open up that skill now. Inside of skills, you have actions and you have instructions, if you kind of recall. So we have a handful of actions that are getting suggested here. And we could, we can see that we've got some flows and a couple of Mof. There, so get job details. That is going to connect us through Mof to our applicant tracking system and recommended job postings. I think these look pretty great as a way to ultimately find the right job and, uh, qualify them.

Yep, let's go and hit next. The next step beyond actions is how to use them. Instructions again. Want to point out there's no code right now. Natural language. So this is when we get to describe how to do this job, how to do this skill with those actions. And the answer is is going to say, based on a description, find the right job. When you find a job, ultimately kind of look at the qualifications, get some of the questions. If they seem qualified, pass them on and schedule. That looks great.

So let's go and hit next. And then last with the setup here, what channels is this is going to operate on? Agents can operate over any channel. But in this case, let's connect it to Slack. Talk to recruiters. Let's connect it to email, uh, uh, to talk to the candidates. And hit next. And then data is the last thing. When we go ahead and move forward, and there's been two enterprise data sources here that the agent will have as well, that's been suggested, which is kind of generic information about the process.

So we are ready to create this agent just with natural language and combining skills that existed from Salesforce, from the App Exchange ecosystem, and even generating new skills on the fly. So we can hit next. And what this is going to do is drop us right into Agent Builder. But Agent Builder not with a blank slate. It's Agent Builder with the, uh, kind of the set of all of those skills together in the topics here that we could continue to do the last mile. So we can finish testing, developing, and ultimately publish this agent out. So building agents for every team and every workflow is faster than ever and doesn't require you to become a developer.

All right. So what we want to do is fast forward a little bit in time. All right? Because now what we want to show is what, what is it like to use this agent, not just to build it? And the D-code has millions of these applications coming in. And how do we want this to be proactive? Before, remember the spreadsheet, all right? That the recruiters are going through. Now they get a notification and Slack, "Hey, there are some candidates that I'd like you to review for a particular job." And I can see here that like, so say Simone is recommending it's understanding some history here that Simona's, um, applied for a similar role, looks like, and done well.

Um, and what can do is we can ask more. We can say, "Hey, for this Project Manager job, it looks like we'd like to to ask more, but maybe we just want to move forward with Simone." And say, "Can you help me? Can you help me schedule? Can you take the next step? Kind of be an assistant to me?" And the agent is going to happily do that. And when it reaches out, because it has that scheduling skill, it knows who I am, it knows my free/busy time and my calendar, it can now start having a conversation asynchronously with Simone about the right times that I'm available and ultimately book a meeting on my calendar on my behalf.

Uh, which is pretty great. That saves me a lot of time. But now when I'm actually doing this, and I know none of us go into meetings and prepare right before those meetings. I know that never happens. But in this case, uh, you know, we've got a lot of interviews. 300 million applications a year. Okay? So why don't we want to prep for that interview? So we can actually ask for some questions to prep now. The difference here is that it has a lot of context because this is about a Project Manager job. In the case of D-code, that job is from one of the 100,000 customers. Okay? So the job requirements and everything about it isn't one static thing. It depends on who that job is for. So it's contextual to that. It also understands who Simone is in terms of the resume and whatever happened in the email conversation to even get to this point. Maybe some of those questions were even kind of answered.

So I'm going to be able to pull all that context back to prep very quickly. And this is the kind of, uh, symbiotic aspect of humans and agents being able to work together, uh, that, uh, is just absolutely phenomenal. And of course, um, what we're showing you right now is is kind of like a single agent. But the reality is is this is a multi-agent future. You're going to have a team of agents. It's your agent force. And when you think about a team working together, there's no better place to work with us than where we already work, which is Slack. And there's no better person to tell you about how this is going to work than the CEO of Slack, Denise Dresser. Denise, great job. Great job. Great job.

All right. Well, of course, I could not be more excited that Agent Force is in Slack. Because Agent Force is fundamentally transforming Slack as the place that every company can bring digital labor to every employee and every team. And if you think about it, it's such a natural place for Agent Force to live because it's where we're already working. But it is also allowing you to not just tap into your structured data, your incredibly, incredibly valuable CRM data, but your unstructured data, your conversations about that data, giving it context so that Agent Force has more context, more relevancy, and more accuracy. And you heard from Adam now with with Agent Builder, you can build skills right into Agent Force, CRM skills, Tableau skills, and Slack skills. So that you can create a canvas in Slack through Agent Force, you can create a workflow, and it's seamless. If your employees can do it, Agent Force can do it. And it's happening right there where you're already working. And what I think about is in Slack, where you're already working, having Agent Force there is like having an expert on demand, 24 by 7. Think about that. I mean, that is a fundamental paradigm shift, as Mark said, to unlock so much potential.

And I want to tell you a story about Accenture. Because Accenture, I know there are folks from Accenture here in the room. Everybody knows them. They're a global leader in data and AI. But they see the potential of Agent Force, particularly in their Salesforce business unit. Because today, their client executives, their customer-facing executives spend thousands, maybe millions of hours on the work of work. It's the preparing for meetings, summarizing the meetings, aligning with their teams. And they are planning on leveraging Agent Force to be able to drastically reduce the amount of time that they take to do this work, freeing up so much potential for them. And I think about this not just as a productivity improvement, but really, if you think about it, this allows Accenture to tap into the intellectual capital they have, the talent they have, with no limits to deliver this incredible outcome and differentiated offering to their customers. So I think this is incredibly powerful. It's one thing to talk about it, but I think it's really compelling to be able to show it to you in action. And so we are GNA, I'm going to introduce Amy Bower, who's going to lead this demonstration for us. And so for purposes of this demo, I'm going to play the role of Claire, a client-facing executive at Accenture. Okay? And the scenario here is Claire has just come back from a vacation. You name it, Bora Bora, scuba diving, at an example, whatever it may be, right? But we all know vacation is great until the minute you land, or maybe right before when you're on the plane, and then you get the scaries. How many notes, how many messages, how many meetings did you miss? What is waiting for me on Monday morning? As a client executive, you probably have a forecast call. You have something going on, right? So it's that feeling, whether it's vacation or you've just been, you know, checking out a little bit for a Sunday. So I'm going to tell you how not only can Agent Force help all companies deliver differentiated service and truly have a workforce without limits, but also how it's going to remove the Sunday scary. So let's get started.

So first of all, we're going to start in Slack, which hopefully is where all of you are working. But what you're going to notice here is there's something new in Slack. And you can see that Agent Force is now here in Slack. And as Adam said, it's your team of experts. It's a library of experts with skills across all types of things that you may need help with, whether it's benefits, or it's IT, you name it, it's right here in the library. But right now, I'm Claire, and I need to figure out what is going on with my most important customer, Capricorn. So I'm going to ask for help. By the way, you know what this is like? You're reaching out to people, you're reading all these emails, or you try to call somebody on a Sunday night, nobody is available. Well, Agent Force is available. This agent is available 24/7. So I can ask, "What is going on with the Capricorn account?" right here. Notice I didn't have to say, "Since I've been out on vacation for a week or anything," because Agent Force, as an agentic layer, has the power to understand the context of my calendar and give me the relevant answer. And so in a matter of seconds, what usually takes hours and creates so much stress is done in seconds. And there's a couple of things I want you to see. First of all, there was an earnings call, which gave me really good insight to what is going on with Capricorn. Then there was a business review, a QBR, that I missed. The screen is so big that I missed. So now I know I missed that, and I need to get up to speed. But here's what I think is really powerful. Of course, saving me time, but also Agent Force is working for me proactively, anticipating things. And it's reminding me that there's a meeting that's coming up this week that I have to get ready for. Well, that's very, very helpful. So now, in a couple of seconds, I've gotten up to speed. Agent Force was available off-hours for me, and now I know what I need to focus on.

So now I'm going to ask Agent Force, "Tell me about the QBR, and I'd like to see this document." How many people have tried to find a document? Right? It's the bane of our existence. So let me show you how this gets better with Agent Force. So first of all, in the top part, what you can see, this is a summary of the QBR. This is Agent Force going out to your CRM system, to all of your conversations in Slack. We know that data plus that context is so important to give relevant answers. So this is a very quick summary. I could have called somebody. I could have called three people. But instead, Agent Force has done all of that for me to give me a summary of what happened in the QBR. So there's an inventory management issue, there's forecast accuracy issues that we need to address. But, but wait. Now what you can see here, Amy, you have to zoom in because it's like the screen is big, but not big enough for this. So what you can see here is this is the Business Review document. But notice it's a PowerPoint document coming from OneDrive. This is Agent Force surfacing federated search results, leveraging Slack's search. I mean, this is incredible. You know what it's like when you find it? You're trying to find the document. Is it in Gmail? Is it in my Drive? Did somebody send it to me? Do I have the right version? In a matter of seconds, I was able to get for myself, with the help of Agent Force, a summary of the QBR and the document. This is incredible.

But it's not just about summarizing what is happening. As Adam said, Agent Force is smart and it is learning and thinking ahead. And so look, Agent Force, if you see here, is suggesting maybe we should engage the product specialist based on the business challenges that were identified in the QBR. Maybe we should bring a specialist in to help us understand what solutions might solve the supply chain issue and the forecast accuracy issues. So heck yeah, I mean, help a girl out. I want to get some help and figure out what I need to do. And what I would normally do in this case is maybe I'd call Patrick, or I'd call Adam, or I'd call Parker and try to get the answer, call a solution engineer, pull it all together. But no, I don't have to worry about who to call because Agent Force has that context and is bringing that to me. So you can see the product specialist has stepped in. This looks so easy, but what the product specialist is doing is tapping into this valuable data that you have in CRM, all your previous customer examples, all of the other ROI stories, and surfacing this for me right now. And it's saying, "For other customers, we would recommend is an analytic solution. And when we have provided this solution to other customers, here are the results: 15% improvement in forecast accuracy, 10% reduction in supply chain costs, referencing lower, lowering holding costs." So how does Agent Force know what holding costs mean and how do you calculate that? That is the power of what Adam was talking about about tapping into Agent Force, tapping into the Tableau semantic layer. So it's using common language to define metrics across all of your data sources. That is powerful.

So, and now I have here, let me just back up. Summarized the meeting, looked at the QBR document to see what the business problems were, leveraged Agent Force to help me figure out, find that document, figure out what the solutions might be because Agent Force is this expert for me. What product might solve the problem? So now where I started out, I didn't know what happened, and I just got back from vacation, and I have to get ready for this meeting. Now I actually have all the data I need to do the proposal. So I'm actually going to ask Agent Force, "I'm on a roll. Let's just keep going." So I'm going to ask Agent Force, "Create a value assessment canvas, that's a proposal, and add customer stories for me because that is really what helps customers understand the value." But that's also hard to do because you're searching, you're looking in different sources of information, you're getting a summary from one human, you're trying to articulate that. But right here, right in Slack, because Agent Force now has not only CRM actions, Tableau actions, but Slack actions. So, Agent Force has created that canvas, which is the beginning part of a proposal for me. So now I'm looking at it, it looks okay. The customer stories look great, that's very helpful. But it's a table, and I think a picture is worth a thousand words. So what I'm going to ask Agent Force to do is convert that into a Tableau viz. Because as we talked about, Agent Force now has Tableau skills to be able to do that. And in a matter of seconds, I now have the starting point of a great canvas for a proposal.

But I'm not going to work in isolation. I'm going to bring this to my team. Agent Force doesn't work in isolation. So I'm going to share this canvas, the beginning part of a proposal, right in Slack, where we're already all working, which is such a natural place to work. So humans, all of the team is there. You can see 119 people are there. Amy can zoom in. But look who else is here. Agent Force is here with the skills that we talked about. An Agent Force is just like another teammate collaborating right in the flow of work where we're already working. You can see that Agent Force is here, listening, answering questions. Sarah has a question. She's asking the product specialist about an architecture document. The account capabilities, all the account is being updated with insights. I don't have to do it. That's the thing. I don't have to update it. It's just happening naturally. And then Agent Force is also helping us bring our team together by scheduling a meeting. And all of this is happening right in the flow of work. And to me, this really articulates that power of digital labor and the power of Agent Force in Slack. And with that, Adam, I'm going to turn it back to you.

Awesome. All right, that was incredible to see skills from Tableau and Slack coming together. The interactions between all of your teammates with agents just in the mix feels like work is changing now. We've showed you how you can take actions for every team and every workflow within the CRM and beyond. We've showed you how you can work with agents just now with Slack as well. And we touched on a key tenet of this, which is the trust and the ability to tackle complicated tasks with more consistency than ever. And this is due to the Atlas reasoning engine and our incredible innovation there. And for that, to walk us deeper into it, I'd like to welcome to the stage, Claire Chang, the VP of Engineering.

[Applause]

Good job. All right, thank you, Adam. Agent Force is more trusted than ever with its upgraded reasoning engine. We call it Atlas. The story of Atlas begins with our AI research team, who have been innovating at the frontier of enterprise AI for over a decade. And we have Silv in the room. Thank you for your guidance and for leading the team at the forefront of AI innovation. Give a hand to Silv.

[Music]

[Applause]

Again, AI has been evolving so fast. Every day, you are hearing about new models. You're hearing about new technologies that build upon those models. However, not every single company has their dedicated AI research team that constantly innovates and gets AI ready for their business. With Salesforce, you gain that edge. When you're investing in Salesforce, you are effectively integrating our AI research team into your organization. And I think Silv agrees. Our AI researchers and engineers collaborate directly with our key customers to incubate on the new technologies and new techniques that help solve their real business problems. And then we bring all those learnings into our products to make them available for every single customer. Atlas reasoning engine, at the very core of Agent Force, is one of those examples. With Atlas, Agent Force is transforming tasks into trusted outcomes. With Agent Force 2.0, we upgraded our Atlas reasoning engine, which can handle more complex questions. With this deep reasoning process, we upgraded how we index data. You heard from Adam about the new technologies that we put into data handling, data graphs, knowledge graphs, and build all this comprehensive knowledge base to enable more and accurate data retrieval. We also improved from answering questions to producing more deeply researched responses. With advanced retrieval, we can handle more deep questions, do specific questions, complex and a multifaceted request with deep reasoning and a multi-step refinement. As a result, you get more accurate, reliable, and actionable responses with citations, providing the transparency and explainability. AI is only as good as your data. The real challenge comes when your data is scattered across your segments of your business. These data silos create difficulty. You are not able to get a full picture of your customer. If Agent Force cannot get a full picture, we are holding back the agents from taking more effective actions. That's why you need Data Cloud. This is your hyperscale data engine built right within Salesforce. It seamlessly integrates all your data and allows you to bring your own data lake with zero copy. It unifies all structured and unstructured data in one single landscape. And with Data Cloud, you also have the latest innovation on data privacy, security, and governance. Data Cloud is the foundation for Agent Force. It's not about managing data, it's about unlocking the full potential in the data to enable Agent Force to deeply understand your business and your customer and empower the agents to take more effective actions.

So you might be wondering, how does that work? I'm going to show you a recipe. And this is Agent Force. And at the heart of it is Atlas reasoning engine. We all know that this is the brain behind Agent Force. It combines all models, data, business logic, and workflows in one unified system. It simulates how humans think, plan, and act. It starts from evaluating the user's input and refining it with additional context. It taps into the extensive data from your Data Cloud and extracts the most relevant information. With our advanced RAG, it analyzes information, evaluates the quality of the generation, and reflects on the answers to make decisions on what's the next best action to take. With more complex questions, it enables Agent Force to think deeply and reason more. We tailor the parameters of handling simple questions differently from handling complex ones to ensure the agents can work efficiently, have more time to reason, and to deeply understand the business and the customer to really deliver a more personalized and trusted experience.

So with all that, let's see all of this in action. You heard Mark talking about RBC earlier, and I'd like to invite Gabe to share with you a demo to see how Agent Force helps RBC's financial advisors unlock every single client interaction more personalized. Gabe, you ready?

All right. This is customer record for Emma Reed. Emma is a customer of Royal Bank of Canada Wealth Management. This page helps RBC's financial advisors to work with Emma and achieve her financial goals. It also gives them access to Agent Force where they can ask a simple question or quick question like, "What's the current balance in Emma's portfolio?" Because Agent Force is connected to all the data, it will generate a very quick and helpful answer. However, not all questions are basic. Imagine asking, "How will falling interest rates impact Emma's portfolio?" This is a complex question, and Agent Force is thinking, reasoning, and trying to find the right action to execute on. As you can tell, this answer is not wrong, but it's not as good as it could be. As engineering building Agent Force, I'm actually very proud to see Agent Force was able to find the right knowledge article from RBC's knowledge base and generate this honest answer without any hallucination. It even recommends the financial advisor to take further actions and do more research. But after second thought, I feel a little bit awkward. Agent Force is assigning homework to its customer. And I want Agent Force to do this homework by itself. In this case, Agent Force should be able to have access to Emma's portfolio, and it also has RBC's guidance on how to identify which funds are sensitive to interest rate change. And it also has access to the effect on the trend of interest rate change on the long-term funds. We should be able to break down this complex question into multiple steps and orchestrate this multi-step data retrieval across multiple data sources and ultimately all of this deep reasoning and analysis on behalf of the customer. This is what motivated us to upgrade our reasoning engine in 2.0.

So let's switch to Agent Force 2.0 org and try this again. Let's ask the same question: "How will falling interest rates impact Emma's portfolio?" Let's see. Agent Force is going to take a little bit longer to reason. We switch from the basic mode to advanced reasoning and retrieval. And as you can tell, this answer being generated is much more rich and is well deeply analyzed with citations, providing the visibility and transparency on how reasoning worked. This is how RBC's financial advisors understand that Agent Force utilizes all the underlying data and then is deep analysis that they can trust. And this is an answer they can put in action by requesting a fund specialist to be invited and discuss further with Emma on those considerations. And Agent Force is able to leverage this analysis and find the perfect specialist with the right skills at the right location and help with scheduling. Expert answers require expert reasoning. With advanced retrieval, deep reasoning, enriched index, Atlas is enabling Agent Force to deeply understand your business and your data and your customer and empower your agents to deliver more trusted, reliable, and a personalized experience. That's what Agent Force 2.0 makes possible. Thank you, Gabe, for the demo. And back to you, Adam.

All right. So we just saw a lot of things. The ability to inject and understand data that far outperforms the traditional kind of vector database retrieval to generation that so many of our customers have tried to do on their own. And the ability to have more inference time scaling with more reasoning and reflection to get the right answer. So you can tackle more use cases. All right. So what you've seen here are three major, major parts of Agent Force 2.0 that brings agents beyond the CRM, okay, beyond the CRM to any workflow and any team. But this is only the tip of the iceberg. There's so much more innovation. And when we reflect that Agent Force was only launched three months ago, it's unbelievable to think about the pace of innovation that's continued past this point, built on the platform. And we couldn't be more excited to see what you build. Thank you very much. Back to you, Mark.

Great job. How about a hand to Adam and the whole team? Well done. Okay. Well, I'm sure you can see we've never been more excited about what we're doing at Salesforce. I'm so thrilled that everyone is here. And I have to call out one friend of mine I saw hiding in the back row who wrote the very first article on Salesforce way back in 1999 about how software was going online. But Don Clark, would you just stand up and be recognized? Come on, there he is. An institution of our whole industry. Very grateful for you being here. And you know, our industry has changed a lot in 25 years. There's no question. You know, we've seen the cloud, we've seen social, we've seen mobile, we've seen so many things. Uh, we, of course, saw the beginning of artificial intelligence. We saw the idea that, uh, we're going to have machine intelligence, machine learning, deep learning, and now, uh, generative AI. But now we're taking one more huge leap. We're crossing one more huge, uh, huge new expanse, which is agents. But this idea that we have to be able to integrate the data, we have to integrate the apps, we have to integrate the agentic layer, we have to deliver it in a hyperscale platform, we have to operate in a global capability, we have to deliver it with the reliability and the availability. And all of these things together, that is the big surprise. And it's, while I wouldn't say it's extremely difficult, it's also not extremely easy. But you can see after a couple of months of pretty hard work, um, we live. And when you look at help.salesforce.com, to kind of come back to my original thought, right now, right now there are thousands of humans working with thousands of agents working with thousands of customers on help.salesforce.com, resolving Salesforce customer service issues. And by doing that, we have dramatically changed how we are operating our company. I believe that that is a model and a vision and an insight and an idea for all future companies. And we are going to work passionately, feverishly, and honestly, as hard as we possibly can to deliver this to all of our customers all over the world. The team is doing extraordinary work. I especially want to thank out the incredible capabilities of our engineering team, our research team, our product team. I've never seen them work harder, do more, uh, and deliver more value. And now to see so many customers actually beginning this journey, we realize we have to be the first ones. We have to be customer zero. If we can't show that we're going to do it, it's not really going to happen. So, um, I don't have a clicker with me, so I think I gave it to Adam. So you want to click to the next slide? Okay. And then go. Anyway, he's, we're delivering this incredible new next generation of our platform to make all, all of it happen. That's all delivered to customers today. We have shipped that to all 135,000 Salesforce customers. Have it now. They just need to turn it on. They just need to make the decision that they're ready to go to agents, and then they're able to have deliver the same level of success. This isn't something new that it's not a bolt-on. They're not adding something on. It's already inside their implementations. It's inside our Salesforce platform. Remember one thing about Salesforce, of course, just to know, of course, we have this incredible hyper Salesforce platform. We call it Hyperforce. Our agents, you know, um, our apps, our models, our flow automation, our Omni Channel interface, security, privacy, analytics, Mof integration, or bringing your own model as well, integrating in other models, our trust layer to help protect your data from the models, then our Data Cloud with zero copy and RAG, our apps, our Agent Force. These aren't separate products. These isn't separate code lines. It's one piece of code. It's one platform. That's why it works because it's all able to operate as one unified system. That's the major advantage that we have. We're not shipping something, adding it on, plugging something in. It's all one integrated system. Okay. So we're going to encourage you to become an Agent Blazer today. We want every one of our Trailblazers to become Agent Blazers. This is an incredible career opportunity, I think, for so many of them. We have millions of Agent Trailblazers all over the world. We are working as hard as we can to inspire them to become Agent Blazers. We've been running multi-thousand person seminars all over the world since Dreamforce in New York, and in Japan, and in Australia, and in France, and in London, everywhere we possibly can go. It will all culminate again, uh, when we get to Trailhead DX back here in San Francisco on March 5th. So anyway, to that, we're pretty happy. We're excited. We're thrilled. We want to wish you a Merry Christmas, a Happy Hanukkah, a Happy Kwanzaa, a Happy Diwali, a Happy New Year, Aiki Mak, uh, Aiki Mak, and a Makahiki Ho, which is a Merry Christmas and a Happy New Year. And I think that we could not be more excited about this Agent Force vision of what Agent Force 2.0 is. You can see the beginning of what Agent Force 3.0, which will be ready to go in May. We're going fast. Um, this is an incredible opportunity, I think, for customers, for us. And we're excited to be leading the way here and to showing everyone what's possible. We have, um, all of this laid out in the hallway. We can get your hands on it, build agents. And so throw everybody's here. Okay. Thanks, everybody. We did it.

[Music]