From Fast Start to Real ROI | Agentforce Keynote, Dreamforce 2026

Please welcome GM Agentforce and EVP Customer Deployed Engineering, Salesforce, Mark Wakelin.

All right, all right, all right. Welcome, everybody. Welcome to Dreamforce, and welcome to the Agentforce keynote. How are we doing today? Everybody all right? All right, not too many hangovers? We were just chatting. Better to present on Wednesday morning than Thursday morning, maybe, right? All right, well, welcome to all of you, and welcome to the people joining us online on Salesforce+ and on LinkedIn. Now, we are going to spend the next 50 minutes talking about ROI, return on investment. How do you get real enterprise value from from AI and agents. Now, the stories that we're going to tell, though, are not going to be told by me. They're going to be told by some incredible customers that are joining us here in the room. And I want to thank them for taking time out from their business to come here and join us today. Now, my name is Mark Wakelin. I'm the EVP and GM for Agentforce. But actually, I've spent the last 2 years working in the field leading our forward-deployed engineering team. That gave me the opportunity to spend time with our customers implementing Agentforce. And I have to tell you, it was an incredible learning and development journey. Now, I think we all know the opportunity in front of us is absolutely enormous. But I think we also know that— doesn't it feel hard sometimes to really know how we can generate true business value? It's difficult to know how to get started. Which use case are you going to pick? And once you get started, how do you customize these agents without creating increasing complexity across your organization. Well, then the agents move into production. That's great, but how do we monitor them? How do we manage them? How do we know what's working and what's not? Well, fortunately, many of our customers have cracked the code, and they're here today, and they're going to tell you what they've done. They're going to tell you what they've learned. And they're going to tell you what they're going to do next. Now, let's break down this ROI equation. Got any accountants in the room? Obviously not. We've got a lot of shy accountants. Any shy accountants in the room? All right. There are really only 2 ways to generate ROI. The first is to increase revenue. And the second is to reduce cost. Everything else is a derivative of those. And you know, the only thing that will actually get in your way is your ability to execute, and that is a byproduct of focus and speed. Now, this is why we have put together a proven Agentic ROI playbook. 3 simple steps. Step number 1, pick a use case and deploy it fast. Step number 2, customize and extend that across your enterprise. And step number 3, observe and optimize these agents in real time in real life. And we have come so far. I want to say thank you again to the 30,000 customers that are now live on the Agentforce platform, seeing incredible results across their business, and for the customers that voted us the number one AI agentic platform on G2 this year. And you know, you will see this success across every industry, across healthcare, across retail, consumer goods, travel, and hospitality. And the best thing is these customers are right here at Dreamforce. You don't even need to go very far. Just wander down to the amazing Agentic Enterprise City. We have 30 customers down there, and they are ready to show you their agents. They're ready to let you get hands-on, and they will tell you their stories. But you know, it was these same customers that inspired us to build a suite of use case-focused, outcome-based agents. So whether it's Hunter doing outbound sales prospecting, or Casey helping customer success, or even Marshall now in the back office, these agents are ready to go. You can deploy them now. They're grounded in your enterprise data, and they are underwritten by scalable Salesforce trust. But you know, it doesn't end there. So many of our customers have used the tools within the Agentforce platform like Studio and Script to customize and extend these agents, creating truly unique agentic experiences. Then leveraging the tools in our optimization— observation and optimization suites to drive an ever-compounding virtuous agentic loop. So if you're ready to come on the journey with me, I'm going to get off the stage. We're going to bring up the customers and my colleagues. 3 chapters. Chapter number 1, you're going to hear about Fulton Bank, how they drove massive productivity for their employees using Coworker in just a matter of weeks. Our friends from Southwest Airlines are here. You'll hear how they customized and extended their agents from simple FAQ agents moving to agents taking actions across their enterprise. And our friends from Vivint on the journey observing and optimizing and now expanding the coverage of their agents and driving ever-increasing ROI. So, are we ready to open it up? Yeah, come on. All right, let me bring up my friend, Avanthika Ramesh. Take it away, Avanthika, you've got this.

Awesome. Thank you, Mark, and good morning, everyone. Now, let's talk about what it looks like to get started fast with Agentforce. Now, in my conversations with you all, here's the one question I hear most often. How can I get my agents to do real work for my business fast without having to spend months building these agents? Now, it's a fair question, but here's the truth. Speed doesn't come from skipping the foundation. It comes from starting on the right foundation. Now, most other agents force you to manually stitch together domain knowledge. They forget prior context, they act without your business context, and they hallucinate on what they don't know. And that's exactly why Agentforce is different. Agentforce enables you to remember your business context. It understands your business and your customers from day one, and it acts within the context of your business and your guardrails. And that is the foundation that helps you get started fast. In fact, we even have a customer who's actually felt that difference. and put into practice. Fulton Bank turned to Agentforce to drive client growth, win more opportunities, and deepen their relationships with their customers, all while staying true to their core value of changing their client lives for the better. Now, from the moment they first heard about Agentforce to when they got it up and running, guess how long that took them? 1 hour. They rolled it out to their power users to test within a week. And finally, they had over 3,000 of their employees augmented with Agentforce in a matter of 6 weeks. I mean, don't just hear it from me. We actually have Brandon Bernard, SVP of Technology from Fulton Bank, right here in the room with us today. Brandon, where are you?

Hi, everyone.

There you are. Everyone wave to Brandon. Say hi, Brandon. And Brandon, thank you so much for joining us all the way from Pennsylvania. All right, so let's show you how Brandon and his team did it. So now Brandon and his team used Agentforce Coworker, your autonomous AI teammate. Coworker knows your business from day one. It takes action on your behalf, and it's multisurface. It shows up wherever you work. Now, we're going to show you how Fulton Bank's employees are using Coworker today. And as we go through this, I want you to pay very close attention to just how well Coworker understands their business. Now, my product teammate Liz and I are so thrilled to show you what our team has been building over the last couple of months. So Liz, take it away.

Thanks, Avantika. Fulton Bank set out to give relationship managers deeper insight without losing the personal touch of banking. But as we all know, finding real opportunities is difficult. It means hours spent searching through endless amounts of information instead of spending them with your clients. And that's why Fulton Bank activated Agentforce Coworker. In just a few clicks, they embedded an AI teammate straight into their core operations. Let me show you. This is Coworker. So let's pretend that I'm not in product, and today I'm a commercial sales leader. It's nearing the end of the quarter, and I'm starting to get a little bit nervous that I'm not going to hit my numbers. What that means is I need a quick pulse check on my business fast. Now, normally this means a manual process, looking through financial statements, going through all of my customers, pulling up loan files. But instead, Coworker can do the heavy lifting. Now, I want you to pay attention here. Coworker is not just pulling basic reports and static numbers. It's scanning through my pipeline data, going through my entire team activities, and scoring relationship health across the entire portfolio. And it's doing this all in a matter of seconds. I can scroll through and understand what are my biggest stalled relationships, what's my action plan to get my quarter back on track. And our friend Brandon from Fulton Bank calls this directional decision-making, where leaders get to spot those high-leverage opportunities, cut through the noise, and understand where they need to focus Wow, Liz, this is pretty neat.

I mean, it's amazing how Coworker literally just knew Fulton Bank's business right out of the gate. Now, I'm curious, this end-of-quarter review that you just showed us, isn't this a repeatable process that, let's say, Fulton Bank may want to empower their leaders with? So, like, how do we take this back and forth and help more users do it more consistently?

You're right, and that's exactly why we built Skills Together. Now, everyone in this room has processes that they run day in and day out. And now with a skill, you can actually package those repeatable set of instructions and turn that into something that Coworker can run and rerun. And we're going to build one together right now. So what's happening behind the scenes is Coworker understands that I need a quick pulse check before the end of the quarter. It's searching through the context of my conversation, and it's packaging this all into a reusable set of instructions. It's looking through the context, the logic, and the format, and actually turning this into a markdown file that Coworker and other agents can reference in the future. And I'm able to do this conversationally Inside of the chat. So as you can just see, it created this skill. With one click, I now have this end-of-quarter team review that I can run and rerun every week until the end of the quarter. This is going to allow me to get my quarter back on track. And what's great is that this skill is shareable. So I can share this with you, Ivanska, or other leaders across the bank so that they can turn their processes into the same level of excellency. So we're going to take it one step further. Like I mentioned, I'm in a little bit of trouble, so I don't want just insight. I need real-time intervention to make sure that my deals close. And that's why Fulton Bank created an opportunity coaching skill. So we're going to run it right now for one of my bankers, Elena. And what this skill is doing is it's telling Coworker to orchestrate to an already built and specialized opportunity agent. And this agent is going to take the records that Coworker has found. It's going to scan through the information, gather more activity, and turn this into an amazing brief That I can share these coaching tips with Elena. All I have to do is click Post, and this information is attached to the record. Now, here's the kicker. Elena does not need to be inside of Salesforce to view this information. Agentforce Coworker is completely open And headless. So in this case, Elena is inside of Claude, and she can call Agentforce coworker to understand where she needs to work and which deals she needs to close by the end of the quarter. Now, Fulton Bank has completely transformed their business. They're not just managing data, they're running an agentic enterprise, at scale, and they're able to do this all in record time. That's the power of the Agentforce platform and Agentforce Coworker. Back to you, Avantika.

Wow, what an incredible demo. Let's give it up for Liz and Colton Banks. I mean, you all just saw how Coworker, fueled by the power of the Agentforce platform, just knew Fulton Bank's business so well and helped them scale their team's knowledge and workflows with skills. Now, we just saw how Coworker helps every employee get work done in the moment, and these tasks take a matter of seconds. But some of the highest value work in your organization might be a little more complex, which means these agents need to run over an extended period of time. To do this type of work, agents need a fundamentally different runtime because these tasks may span, span days, weeks, or even months. But most agents weren't built for this type of work. Most agents lose context over time. They're difficult to steer as conditions change, and they're simply just not capable of adapting as situations evolve. And that's why agents need a fundamentally different type of runtime. Now, a long horizon runtime enables agents to remember where they left off, reach for the right tools at the right moments, and run over days or even months. First, they start with a goal. Let's say we want to reengage some of our at-risk deals. Then they formulate a plan to get there using the tasks, tools, context, and guardrails available. And finally, they continue to adapt as they work towards that goal using what it observes and the feedback it's given, because you are always in the loop. And we on Agentforce have now built that. This is Hunter, your outbound sales agent. You simply turn on Hunter, and it gets to work right away. It starts learning about your business, literally hunting for new opportunities, and it starts working towards your business goals. And the critical piece to remember here is that a long horizon runtime doesn't mean a long implementation. In fact, we made Hunter so easy to set up and out of the box. So now we're going to show you what that would look like for Fulton Bank. Let's roll the— let's roll the demo, Liz. All right, so here I am in Hunter, and I'm going to simply start by asking Hunter to help me set some goals for myself. As you can see, it's starting to curate an intelligent set of goals based on my profile. Now, as I look through these goals, I'm going to go ahead and select the first goal right here, to reengage some of my at-risk deals. Now, Hunter starts looking through some of the opportunities I have that I haven't engaged with and curates this list for me. I like what I'm seeing here so far, and it makes it a lot more concrete. It helps me reengage $340K of viable at-risk pipeline by the end of the next quarter. Now, as I go through this, I can actually see the tasks that Hunter has created to achieve that outcome. I can see when those tasks are scheduled. I can even see which ones require my approval, and I can conversationally continue to improve this task list if I wanted to. So far, I like what I'm seeing. So let's go ahead and have Hunter lock in this plan. Next, I want to connect Hunter to various surfaces where I work so it knows my style of communication. So I can connect Hunter to my emails, like on Gmail, so it knows my voice. So now Hunter knows my voice, my style of communication. It's established all the different background tasks. It's going to show me a draft of the goal it's generated. Now, this is where the human-in-the-loop element really comes into the picture. As you can see, I can actually see a preview of one of the deals and the plan it's created, and if I wanted to, I can go into the plan details and update some of the information manually. I like what I'm seeing, so let's go ahead and activate that goal. Hunter gets to work right away. And let's see, in a couple of seconds, it's actually going to create a draft that's already ready for my review. There we go. And because it knows my voice, I only have to make a couple of quick tweaks before this email looks good, and I'm going to go ahead and send it off. Now think about the power of this. You have agents running in the background. They know your voice and they act on your behalf. And anytime new context comes in, Hunter will immediately update its plan because this is a continuous self-improving feedback loop. And that's the experience you get with Long Horizon agents. And that's Hunter. And the beauty of all of this— Yes, that deserves a huge round of applause. And the beauty of all of this is that it's built on the world's most powerful agentic platform, Agentforce. And the results for Fulton Bank speak for themselves. I mean, 80,000 hours saved for bankers and clients, over $389 million generated in new loans and deposits, And complaints resolved 4 days faster. Again, huge shout out to Brandon and his team for truly blazing the trail here. So we just saw what it looks like to get started fast with Agentforce. Now, once you do get started, you may want to start customizing your agents. extending some of their capabilities, and really achieve better outcomes for your business using the rich capabilities on the platform. And there's no one better to speak about that than our very own Chief Product Officer. Please join me in welcoming John Kusera.

All right. Thank you, Avanthika. So we just saw how Fulton was able to get started fast and see real results. And now we're going to dive into what can you achieve when you customize and extend with Agentforce to deliver not just seamless customer experiences, but also real business results. That is the Southwest story. So, with Agentforce, so many different customers can achieve fantastic results that you can't get on all the other platforms. Where other platforms have overzealous escalation, oh yeah, I'll just send you to a person right now. With Agentforce agents, you get qualified intake. With other platforms, they can drift off mid-task and not handle the process. With Agentforce agents, they can execute flawlessly every single time. With other platforms, they might say, oh yeah, you can contact us to schedule that appointment. With Agentforce, you schedule in line. And other places might say, you know, we're going to get back to you in 3 days for your return. But with Agentforce, you get those instant results. So how do customers do this? Well, Agentforce customers can customize and extend with AgentScript to deliver the flexibility through the LLMs that you need, while also having the repeatability through all the rules and operating procedures that you create. And this is customers like DataSight. DataSight does 16,000 M&A deals across the globe every year. And they needed to make sure that their agents were qualifying the intake requests before they came in. And the results? They improved their customer satisfaction to 4.8 out of 5. And it's customers like Adecco. Adecco sees 300 million applications every single year. And at that scale, reliability is not a nice to have. That's the entire job. Applicants that are screened through Agentforce are twice as likely to get hired, and they're seeing everybody go through the process 28% faster. It's customers like 1-800-Accountant. Imagine it's tax season for those of you in the US, and you have a flood of all these questions coming in at a stressful time. They're hitting sensitive tax information. You have no room to make mistakes. Well, with 1-800-Accountant, they saw 98% or 90% resolution rate when they needed it most. And you also have Kogan. They're an online retailer that does thousands of returns for e-commerce across a patchwork of systems. They were able to stitch those together, make a seamless experience to have instant results, and they tripled their resolution rate. That is real productivity powered by Agentforce and AgentScript. But it's also Southwest. So Southwest set out to ask and answer a simple question. How do you deliver instant answers when the travelers need it most, and free up the care reps to deliver legendary hospitality that Southwest is known and loved for? Southwest turned to Agentforce, and they've seen incredible results. Huge amounts of productivity, millions of dollars, huge ROI.

That's right.

And I could tell you about it all day, but I'd rather bring somebody to the stage that is an expert in this, who leads VP— who is the VP of AI for customer care, operations, and back office. Welcome Justin Bundick from Southwest.

Thanks, John. Appreciate that.

So good to see you here, Justin.

Great to see you.

So you all have been on an incredible journey. You've seen some great results. Why did you turn to Agentforce? And how's it going so far?

Well, John, as you and maybe many of you in this audience know, Southwest has been on an incredible transformation journey over the last 24 to 36 months. But one thing has been true from the beginning, and that is that we are focused on taking care of our customers and providing them exceptional service and great hospitality that sets our brand apart from our competitors. Through all of that change, we know that our customers were having to really look into a lot of policies and information to understand what was changing and how it was changing and impacting their day of travel.

Yeah.

We saw that as an opportunity, John, an opportunity for us to provide really high-quality information at the moment that our customers needed it most. So we focused in on an AI assistant, the Southwest AI Assistant, that allowed customers to engage on our digital surfaces to be able to find information and knowledge in the moment that they needed so that we could answer their questions, allowing our representatives, our care representatives, to maintain focus on those high empathy, high hospitality moments when our customers needed that as well. And I have to tell you, the results to date have been really, really exceptional. We have over 20,000 customers engaging with our AI assistant today.

Wow.

Our CSAT scores on those engagements are at 60% or higher and growing every week. And of all of those that rate our AI assistant, over half of them rate it with 5 stars and let us know in real time that it's answered their question.

Those are real results. And I think that's results that a lot of you would love to live up to. So I'd love to ask you, what are some of the lessons that you've learned and what are some things that you'd recommend to people in the audience here?

Oh, John, you're going to make us have to show our laundry. I have to tell you, there are 3 things that I'd love for this audience to walk away with, and that is, that when you think about implementing an AI solution, especially that's one that's customer-facing, in this world of AI, you might want to go really bold. And I would highly suggest that you start with something that is very focused and learn and learn rapidly, because you'll be able to build off of that and actually reach your vision, but do it in a very incremental way that meets your customers' expectations. The second thing I would bring up and focus in on is that your customers will provide you their needs a lot faster in real time than all the planning sessions you could have had internally leading up to the launch of your product. You do that by analyzing the conversations that are happening between the customer and that AI product, and then making tweaks and changes on a very rapid basis. It's a brand new operating model that most companies are going to have to adopt, and it's one that we've been able to morph into as we've implemented some of these products. And the last thing is actually tied to the first, and that is that your business partners, in this case, our customer care organization, and our organization had to create new teams and new structures around this operating model that allowed us to analyze conversations that were happening in real time, plan tweaks and changes that we needed to make to either the AI assistant or to the context that we were providing that assistant, and see and value the changes that it's making to our overall results.

Well, I think you all have delivered incredible results. It's been really inspiring, and I just want to say thank you for the partnership on this.

You bet.

Thank you.

Thank you, Justin.

Thank you, everybody. All right.

So now we're going to go a little bit deeper into how Southwest has been looking to customize and extend this agent. And I want to introduce Ray, or as his friends call him, Regent Force, who's Who's our demo driver for this. All right, let's see it, Ray. So this is one of the many processes that so many of us have, where you were built in a little bit of a different era. It asks you all that information over and over again. And for Southwest, they get requests for name change 25,000 times every month. And so what they knew is they could transform this to be a better experience. So what we're going to show you is that new transformed experience, powered by Agentforce and AgentScript. We have our newlyweds here who come to the site and need to make a name request. Instead of having to retype their name and put in their information, the agent knows who they are. They're able to upload a document fast and get that validated in real time so that you avoid that dreaded, I'm sorry, your document wasn't valid. Can you please resubmit this? These are some of the seamless experiences that are possible through Agentforce. So, let's go take a look at how this was built and how Southwest achieved this. So, here is Southwest's account assistant agent, and you have the name change sub-agent that they built. We look in here, and we see the different rules and the business policies that Southwest has. One of the key ones is, do you have any travel coming up in the next 7 days? We wanted to make sure that there's time for all the approvals and make sure that everything goes off seamlessly. And then after you upload the document, then it goes deterministically to a MuleSoft action. This checks to see, is the document valid? It extracts all those fields so that you don't have to do any retyping. And we also have the agent create a case so that we have seamless next steps for that human touch to make sure that everything's in order. This is the power of AgentScript to create truly seamless experiences. And it delivers real business value. This is 25,000 times every single month and takes up to 5 to 10 minutes. Think about the productivity that you can save with tasks like this. But Southwest didn't stop at name change. There's also a whole host of other use cases that Southwest is automating. We have things like receipt requests. We have things like travel vouchers. You have status matches. Tip for any of you travelers out there, if you've got status on another airline, you can go to Southwest and have your status matched. That is now powered by an Agentforce agent. But to the eagle-eyed of you out there, there's something a little bit different about this agent. This is a subagent that's part of the super agent formation for a unified experience for Southwest. So Southwest, like many of you, has different teams responsible for different aspects of the experience. So, they have one team that's focused on all the flight information, all the FAQ, and all those use cases, and they need to continue to make those great. And there's a second team focused on all the account management. So, they set to unlock how can you have these 2 teams work independently, make the agent use cases great, while presenting a unified brand experience. We call those super agents, and we are thrilled that Southwest is using our multi-agent orchestration to deliver this consumer experience that's fantastic while also delivering real business results. So that is a Southwest experience that's been retransformed. Let's give it up for Southwest.

Woo!

All right.

So now I want to recognize somebody from the community that's been instrumental to building this experience for Southwest and also contributed a lot to the Agent Blazor community. Let's give a warm welcome for Ishrat Bhatti. Thank you so much. Ishrat, you lead Southwest Center of Excellence for AI, and you've been super inspiring to me to see everything that you've created. But I have a little surprise for you. Congratulations, you are the Golden Hoodie Award winner. Let's give it up for Ishrat. We're going to pose for the camera.

We'll get a little picture.

So Ishrat, thank you for everything you've done for Southwest. Thank you for what you've done to the community. It's really inspiring what you've done. Thank you so much. All right, so we just saw how Southwest created a unified experience and delivered real results, millions of dollars of productivity, 7x return on investment, and did that with one platform that could grow with them. That's super agents, and that's a great consumer experience. And so, after you customize and extend your agents, next, you need to observe and optimize. And I want to introduce somebody who's been living this with the community every day. Let's welcome Suzanna to tell you about it.

Thanks, John. And congratulations again, Ishrat. So well deserved. When I'm speaking with customers and with our community, fabulous folks like our Agent Blazers over there, There are 2 questions that I hear most often. How can I make sure that my agents are behaving like I expect? And how can I ensure that they're going to continue to drive value over time? Now, with traditional software, we've solved this before with a predictable lifecycle. IT teams build a new feature, they write some tests, and when those tests pass, they deploy over the course of weeks and months. But building agents, is different because it's the outcome that's what— the outcome that what drives the build process. So IT teams need to work with the business together to observe and optimize over the course of hours and days, not weeks and months. Now, the Vivint team has been one of these customers who are leading this new way of working, and they're here with us in the audience today. Hello, Team Vivint. I see you right there. Hi. Welcome. Thank you. So now I want to introduce David Porter, Managing Director at Vivint, to tell us about their story. Mark and Dave, over to you.

All right, fantastic. Dave, it is great to see you. We're so grateful that you've taken time out. I think it was hearing what Susanna said, I just want to clarify. I think she said you expanded customer service to 2 million customers.

Almost 2.5, actually. So we're growing.

And you didn't add any humans there. Tell us about that.

Well, you know, the thing to understand about all of this is we started with the customer experience. What matters most is creating a great experience for our customers. You know, at Vivint, our customers on average, you know, we're providing them 15 individual smart home devices in their home that have to all work together.

Yeah.

And they're in their app every day to use them 5 or 6 times a day. And so the natural place for us to go to provide that help is in the app. And it really has created a tremendous customer experience, not just doing simple Simple tasks like I need help paying my bill, or maybe I want to change the name on an account, but really into troubleshooting and maybe adding more products if they need them. It's really been fantastic.

Absolutely amazing. Now, Dave, I know the agent Ava.

Ava is what we call her.

She began life as a text agent, but you've moved to voice, and you've obviously seen so much here at Dreamforce. Tell me what's exciting you, where you're going here. Tell me about voice.

Well, you're exactly hitting the nail on the head because it does start with the app. Ava is what we call her, but we are also moving it into voice. And there are many customers who either prefer or need to interact with us over the phone. And so, you know, all of those simple use cases we talked about, you know, rolling a truck to help with a problem, to install a new product, all of that happens there. But what most excites me is when you get into the more complex troubleshooting tasks.

Yeah.

That's where you've got to nail it. You've got to get it right for the customer. And so we're able to do it. And that's where we're moving AIBA. And we do move to a specialist, you know, even an engineering-level technician if we need to. But we can do so much with AIBA. And one of my, you know, most fun experiences listening to one of those calls is, you know, AIBA introduces herself as a virtual agent. You hear the whole interaction and then you hear the customer say, well, thank you, AIBA. That was very helpful.

Well, Dave, we are loving being on this journey with you. We're looking forward to the future. Thank you so much for being here. Put your hands together. Suzanna, back to you.

Dave, so we've already heard so many new innovations today. We just heard some new ideas from Dave, but the funny thing about new ideas is that they usually don't start from scratch. Now, out of college, I worked at a call center, and every single call started like this. This call may be recorded for quality assurance.

Woo!

Yeah, I was pretty good at it. It was good, yeah. You can give that a hand, why not? Okay, so the thing is, when you're working with people, right, on the phone with your customers live, you don't just train them and hope. You listen for quality and for assurance. And we've been doing this at call centers for as long as they've existed. But the difference now is that with Agentforce Observability, we can do this for every single conversation across every channel all of the time. And the best part about Agentforce Observability is that it enables customers like Vivint to do their work at scale. But my favorite part of observability is that it's now unmetered. That means it's free, y'all. Yeah, yeah. So this means that you can get full visibility across all of your agents without a surprise bill. And now I wanna show this to you with a demo. So I'd love to introduce our amazing demo driver, Allison, And we're gonna hop right in. So you heard about, uh, Vivint's amazing voice agent Ava. They built it, but they also are improving it to continue driving business value over time. So to do that, they set up notifications with agent health monitoring. So we'll see here that this alert is about, uh, contact resolution. So if we drill into the session, we'll see a little bit more about what was going on. We'll see that this particular call was escalated. So let's give it a listen.

Hi, Allison. I'm Ava. How can I help you today?

Hi.

Yeah, I was just messaging you about an appointment I set up. Yes, I see you scheduled a technician appointment about an hour ago. Are you calling about that or something else?

Yeah, I'm actually at Dreamforce right now, and I didn't check with my family.

They're actually not going to be home, so I need to reschedule. That's all right. You're certainly not the first person to schedule something before checking in with the family. What time works better for you? Ulysses Friday at noon. Got it. Rescheduling your appointment now. Fantastic! Your appointment has been successfully rescheduled for Friday at noon. Before you go, I noticed that you're eligible to upgrade to our new. new Outdoor Camera Pro 3. It combines AI-powered detection with radar technology to help identify and deter potential threats before they become problems. Would you like to upgrade?

Okay, Allie, let's pause right there because so far, I mean, that sounded like a pretty great customer experience. And the agent also used Vivint's enterprise context to offer an upgrade. That is Ava driving real measurable ROI. But remember that notification? If we look down here, we can see that the initial interaction— in the initial interaction, the customer asked for a device reboot, which the agent had not yet been configured to do. So instead, it gave this information to the customer about how to do it themselves. Now, this is a real example from the Vivint team. They were seeing across all of their sessions at scale a pattern emerge. Customers were asking for these device reboots. So they took this as an opportunity to add functionality to their agents. And like many of you, when they started to add this new feature, they did it headlessly in Claude. So you can see here, I've added my subagent that to the troubleshooting— I've added my action to the troubleshooting subagent. Looks pretty good, but we need to test it. So we are gonna use the agent force development lifecycle skills. This is a public repo that you can use to add structured advice for how Claude tests. And when these tests run and they pass, sometimes these test suites take a little while, but this one is nice and speedy. Okay, so this looks pretty good. So we're gonna deploy it so that someday Vivint might see an alert like this where customer satisfaction is actually increasing because of that enhancement. We, that enhancement we introduced. So Vivint is able to use Agentforce observability to do this at scale so that observing and optimizing their agents truly compounds their ROI. They're seeing real success. These stats, these stats are amazing. 20% chat containment. They're seeing less friction for their customers and a higher CSAT as well. But some of you might be wondering, this is great. This is an amazing example of business value, but it's just one use case. How do I handle my agent portfolio as it grows like we've all heard it's doing this week? Well, I'm very excited to speak with you today about a new feature because we actually have an agent for that now, and it's called Agent Optimizer. So we're introducing this to you for the very first time, and it's going to be a way for you to more easily optimize and observe your agents. So instead of having to find and fix every single thing by yourself, The optimization process becomes agent-led, and I'm so excited about this. This is truly the future of Agentforce. It's a sneak peek, and I'm going to show it to you myself. So what if instead of having to hunt and peck through all these reports and logs every day, agent managers started their days here in our reimagined Agentforce Studio experience? So we can see that in this experience, there are proactive alerts that describe the highest impact patterns across all of my agents. So I know exactly how to prioritize my time. And when I drill into one of these insights, I can see that we have an opportunity here to improve how this particular agent handles refund requests. But not only that, we're also able to see the potential business impact that would happen if I were to prioritize this change. So let's go ahead and open this up in Agent Optimizer. So here we are looking at Agent Optimizer, and I can see that we've already brought over the insight and the context from that issue we were just looking at. So we can ask Optimizer to get to work. Now, first, it's gonna start to understand the issue, and it'll work to identify the root cause. And from there, it'll start to draft a plan. Now, this plan isn't just the description of the configuration changes that need to be made. It gives me the context and describes why we're making these changes, and then also gives me a suite of tests that I can run to ensure that when I introduce this change, it's actually doing what I want it to. And when I feel good about this plan, I can kick it off here right in the same conversation. Now, you might be thinking that this looks like my favorite coding agent. How is this different? Well, this is all built on the Agentforce platform in one conversation, and it's an, it's an opinionated experience that's guiding us through. So instead of just a summary of the change that was made, I can drill in and actually see the inline agent script changes that were made. This is a really seamless human-in-the-loop experience. And once we've previewed these changes, we of course need to test. And if some of those tests don't pass on the first time, no worries, 'cause Agentforce Optimizer is gonna self-correct and continually optimize until it does. And when we feel good, when these tests pass, we can also then choose to activate our agent. Looks like someone got, got click happy and activated it first, but we're gonna activate it again 'cause I wanna show you we can actually gradually roll out these changes to our users so that we can have a better idea of how they're impacting folks in production. So what we just saw here is a big deal. Instead of 3 separate processes for observation, building, and test, we did it all in one conversation. And this is the power. Again, this is the future of Agentforce, and this is the power of Agent Optimizer built on the Agentforce platform. Mark, back to you.

Awesome.

All right.

Woo! Fantastic. Thank you, Susanna. I, for one, am glad that you're here with us and not in that call center. So team, we have got this. Let's get started fast. Use case focused, prebuilt agents. They are ready to go, grounded in your enterprise data. You can use them now. Customize them with the incredible tools in the Agentforce platform. John showed you those. Keep watching. Wasn't that amazing from Susanna? Optimizing the agents as you go forward. You've seen it. This is real results. This is happening right now with these customers. It's not over, though, folks. We have much more here for you to discover at Dreamforce. Go to the Agentic City, get hands-on, go to our learning labs, get in the campground, see the demos. We'll see you at Dreamfest tonight. Thank you all so much.

See ya.