# Agentforce Service Agent: What Is It and How Does It Work Auto-transcribed by https://aliceapp.ai on Wednesday, 18 Sep 2024. Synced media and text playback available on this page: https://aliceapp.ai/recordings/m7bOC0_f9LjZiVtNARld4rpq1QkWPDXe * Words : 3,172 * Duration : 00:19:56 * Recorded on : Unknown date * Uploaded on : 2024-09-18 19:11:57 UTC * At : Unknown location * Using : Uploaded to aliceapp.ai ## Speakers: * Speaker A - 69.7% * Speaker B - 30.3% ---------------------------- Speaker A [00:00:00] Hi, everyone. Welcome to our session today. Thank you for being here. We're going to be talking about agent four, service agent, what is it and how does it work? We want to first start by saying thank you, uh, for attending this session. It is the first day of Dreamforce. There's so much happening right now. So we appreciate you taking the time to be here today and for everyone tuning in online on Salesforce plus, thank you for joining as well. We hope this session is extremely valuable for all of you in the room and online. Now we have to start with our forward looking statement. Just as a reminder, please base those purchasing decisions on your products that are currently available. And we want to kick off by introducing ourselves as your speakers today. My name is Christina Cohen. I'm on the product marketing team here at Salesforce. I've been at Salesforce for a little over six years and this is my 6th Dreamforce and this is a really exciting one. I think we just have extra innovations this year. A lot going on when it comes to AI and autonomous agents. So really excited to be here with all of you. And I'm joined by Jen. I'll let her introduce herself. Speaker B [00:01:05] Hello, Jennifer Solarious. I am a principal solutions and service cloud in the automotive pillar here at Salesforce. And I want to send a special shout out to those of you across the globe watching at watch parties today or asynchronously watching this on Salesforce plus. Speaker A [00:01:22] Awesome. Um, thanks, Jen. All right, so we have a jam packed 20 minutes. There's a lot that we're going to be covering, a lot of great content just to ground you all. We're going to be talking about autonomous agents and customer service. And then we're going to deep dive into our new innovation agent four, service agent. And Jen is going to take us under the hood to show how this actually works within Salesforce. Finally, we will leave you with some takeaways. There's a lot more to do in the next couple days at Dreamforce, so more that you can see, get hands on and check out. Okay, autonomous agents. You have heard this term a lot at Dreamforce, outside of Dreamforce. When people hear autonomous agents, I think the first thing we usually ask ourselves is, isn't that just a chatbot? And the answer is, actually, no, it's not. Because when a chatbot is still very good at handling those simple questions, those easy questions coming in, we've all experienced the conversation with the chatbot, where you're typing and doesn't understand the script, and then it sends you off to a service rep. That's where autonomous agents are really paving the way for the future. They are dynamic agents, so you can have a much more conversational, complex experience with them. This is a great example that you can see here. If I have an error with my air fryer, instead of having to type in my error, I can actually send an image to my autonomous agent. And instead of getting tripped up by not understanding the dialogue or not knowing the image, the autonomous agent can understand again that complexity and it can continue on in the support journey. In addition to being dynamic, it is also proactive, so it can recommend new products that can help me. Maybe I want to exchange my air fryer for something new that's not broken, but where it really takes the cake is it can make an action. So instead of just providing that exchange recommendation, it can complete the exchange for me in that conversation. Now, as a customer, this is effortless service and this makes it so much easier for me to get the support I need. And on the business side, autonomous agents are helping you by allowing you to reduce your costs with less resources and less time spent to launch them because they're easy to build and maintain. Where the bot experience, you have to create, you know, those numerous dialogue trees and you really have to train that model multiple times before you can launch it. An autonomous agent is easy to set up and you can use clicks rather than code, right? So you have that natural language instruction that can handle that complexity, and it's a much more mature AI reasoning model. This means there's less intent building and there's less maintenance that it takes and training that it takes to build and launch. Finally, your autonomous agents will still ground on your knowledge base, right? Knowledge is so key to providing the right information to your customers. But in addition to being able to connect to your knowledge base, it can connect to generative AI, your LLMs and your company data. So there's so much more information that it can take into account to provide the richest, most relevant response to your customer. So on the business side and the customer side with these value props, this is really why Salesforce has launched our trusted autonomous agent for every service experience, which is called Agent Force service agent. Now with agent force service agent, it's able to engage with your customer on any channel of their choosing. That's huge. And it's 24/7 in that natural language, whether that be SMS, WhatsApp, uh, or messenger, it's meeting your customer where they are. It can also ground all of those responses in your trusted data. And because agent four, service agent is built on top of the Einstein trust layer. You can rest assured that those responses that are being generated for your customers are relevant and they're accurate. Like I mentioned in the last slide, you can set this up in minutes. That's a huge time saver for your business, right? And less resources. You can use these pre built templates and your existing salesforce objects like flows or prompts that you already have built out. You can utilize those to launch your agent for service agent. Lastly, you can define those clear parameters. Obviously trust is big and you want to make sure that the agent Force service agent is following the conversation that you deem appropriate so you can create those guardrails. And once the conversation gets out of scope it will generate that directly to a service rep. And all of this is built on the trusted Salesforce platform, right? Which is allowing you to have all of your data, your trust in one place and again provide the most effortless experience for your customers. So let's talk for a second about how you actually create these conversations that your agent force service agent is going to be completing with the customer. Jen is going to give us a great demo in just a couple minutes, but I want to ground everybody in what you're about to see. There's really three key components to setting up these conversations. It's topics, instructions and actions. Now when you think of topics, those are generally just those jobs to be done. That can be something like order management or repairs, those overarching themes, right? That your agent force service agent is going to pick from based on the customer's initial question. Then you have your instructions. All of your topics are defined by a list of instructions. And these instructions help navigate the uH, agent force service agent, making sure that again, it's following those guard rails and those parameters that you set up and it's understanding the topic at hand. Lastly, you have your actions. Actions are your subset of your topics. So every topic is going to have a list of the action steps for agent four, service agent to take, again, based on inquiry coming in in the conversation. Now before we get into the demo, I just want to give you a quick example of what that initial conversation looks like and how the outcome of the answer happens. That uh, first initial response. So you can see here from the conversation that our customer Lauren is asking to return her last purchase. What happens from here is agentborst service agent will call your large language model and it's going to make a plan of action. It's going to plan by understanding the topic, which is order management. In this scenario, and it's going to select the right action which because Lauren wants to return her last purchase, that is going to be generating that return label. Once it has the plan, it can then plan on the execution. What's it going to do to actually take that action? So again, it's going to be pulling from your current or existing objects, like again, those flows or those prompts, APIs, whatever you already have set up. It's going to know what it can launch in the conversation and it's going to ground all those responses in your trusted data that shows you the initial response that comes from agent for service agent. It recognizes the question, right? It recognizes the product that Lauren's talking about and saying that it can send that shipping label if she wants to return it. So again, pending Lauren's response, it can then put that plan into actual action. So I know I've shared a lot, you've seen these pretty slides, but, um, let's actually dive into the product. Let's show you how this works within Salesforce and for that I'm going to pass it over to Jen. Speaker B [00:08:37] All right, without further ado, and we are going to see a lot in a few moments. Go ahead and start the video. Ready? All right. Welcome to agent four, service agent. We're starting our journey today with an embedded agent on a website asking standard questions like where is my order? And continuing with loose but contextual questions that are then answered with clear and concise business data or information grounded in knowledge using a sophisticated reasoning engine and natural responses. Not only were these questions understood, but the agent replies were appropriate. Agent force can be deployed with confidence through clear guardrails and seamless handoff to a live agent and is easily configured through declarative and low code tools. This is where we want to focus today. So we're going to lift up the hood and get out our wrenches and see how this agent is built. We're going to start with the primary building blocks. These are, as Christina mentioned, topics allowing administrators to group actions into functional jobs to be done. Actions. Think of these as, uh, specific tasks based on your business process logic. These can be flow apex or prompt templates and knowledge allowing focused content for grounding. In addition, the ability to select language at release agents are going to be available in seven languages. This is where you can set the overall agent response tone quickly, see what it's looking like. All right, but let's dive a little bit deeper into topics and actions. Remember those questions on the front that, uh, the customer asked where is my order in the conversation preview we can see that agent force classified that as an order management topic which has associated instructions and actions. The action find order flow has gathered inputs based on the logged in customer, user and resulting outputs, all of her orders and key relational data which are now available as response by the agent. Agent Force also has reasoning grounding the output data to determine responses as the conversation turns. When a question could be easily answered with knowledge, an out of the box standard action answer questions with knowledge applies, surfacing the previous response from the corpus of your predefined knowledge content. New topics can be created from scratch or added from a library. Let's take a closer look at the order management topic. There are no rigid dialogue trees, just a clear scope of the job to be done and natural language instructions. These instructions help guide which actions to run or what clarifying question the agent needs to ask. The builder makes it very simple to add, tweak or test instructions as well as add or remove actions. But what if I wanted to create a new action? It's as easy as referencing functionality that you have already built on your core platform in Salesforce, your business process layer and logic, apex flows and prompt templates. At launch, agent force will be seeded with standard actions across sales and service use cases. And the beauty of the Salesforce platform is the reusability of these functional building blocks, allowing scalability and rapid time to deploy. Here is a look at the find orders, action configuring and clicks, the defined inputs collected and the outputs available for use in the conversation, boosting the data that's available for that sophisticated reasoning engine. So we've had a look at flow action. What about prompt templates? Again, Salesforce has seeded a multitude of standard templates, account summaries, contact summaries, messaging session summaries custom prompts are enriched by grounding your business data, whether that lives in Salesforce or we're pulling it from data cloud or mulesoft APIs, you can select test deploy against our out of the box LLMs that Salesforce has partnered with or custom LLMs through model builder. Power your prompts with the reference fields, flows and apex callouts and test these against actual records like contacts and orders. But most importantly, it's all secured within the Salesforce trust layer, which means end to end encryption, data masking and zero retention within our LLM provider. But we really need to focus on the foundations. At launch, Agent Force will be available on all of the enhanced channels, so we can use the power and flexibility of flow builder to define our routing rules and dynamically route to the best resource as well as grab the data outputs that will be needed for the context of those actions. But we can't forget about knowledge. It is a key grounding lever. With Einstein data libraries, you now have the ability to create and curate the sources that will help Einstein and your agent force agent focus on the right objects and fields as well as filter your knowledge by data categories, giving you all of the levers and declarative tools powering this agent force engine that surfaces those natural language responses for your customers. So you saw a lot. I want to take a step and show you that it was user interface agent layer building blocks of uh, the instructions, topics and actions of which they are built from flow and apex and prompt builders. And with the foundation of your business process data and the data layer and knowledge bringing together agent force. Now, my challenge to you is to think about the use cases. Pick those couple of use cases that you can now bring forth on your channels of choice. Speaker A [00:16:41] Thank you, Jen. That was awesome, guys. I mean, that's a really cool demo. So much innovation, so much that agent for service agent can do. And like Jen said, you can launch this across all of your industries. These are just some examples that we threw together for this slide for retail, auto travel and hospitality. You just saw manufacturing. But the beauty of agent for service agent is it can help across those use cases. So I know we have a few minutes left and we've covered a lot in this session. And if there's three things that we just want to give you as some takeaways in addition to what Jen just kind of recapped for the demo, it's first off, autonomous agents, they are revolutionizing customer support. And because of that, we are launching our agent for service agent. And that is really with an end goal to help you improve your productivity and your customer satisfaction. And finally, as you saw with that demo, um, uh, implementing your AI strategy in general as well as your agent force service agent strategy is a growth process. It takes time to think about your goals, how you want this to be launched, what you want it to be helping your customers with, and it will continue to change and ebb and flow, but you have the power to be changing that alongside your goals as well. So just think of it as that growth journey. There is so much more agent force service agent content and things that you can do around Dreamforce if you haven't already experienced it. One thing that we really wanted to call out is we are going to have a breakout session both in person and that's going to be filmed on Salesforce plus for agent Force service agent. That's going to be a 45 minutes session where we're going to deep dive even more into those use cases, hear from one of our customers and see another demo. So if you want to get more in the weeds there, feel free to join that session. We also do have some reports that we love to share with all of you. Um, excuse me. We launched our state of service report this year and it comes with a lot of statistics around a lot of the things that you saw today. So if you want to learn more about the customer support industry, where we're at those different metrics, definitely download that report. And then lastly, um, for some swag. Cause everyone loves swagger. If you want to get a service blazer hoodie, they are these very cute maroon hoodies that you might see people walking around with. You can scan this drawer card right here and leave us a g two review for service. Cloud and field service obviously helps us understand what we can do, uh, better and how we can improve and hear from our customers. Speaker B [00:19:05] Thank you, team. Speaker A [00:19:07] Yes, and last couple CTA's, because again, wouldn't be dreamforce without so many things you can do. Join the service laser community on if you haven't already done so, great opportunity to meet and network with other professionals in the service industry like yourselves. Uh, if you head to the campground in Moscone north, there is actually a booth where you can learn more about this. And then we want to say thank you. Thank you so much for being here in person, for being here online. We really appreciate you guys taking the time. I know it's after lunch, so hopefully we kept it exciting and energetic. And if you need more energy, you can scan this last QR code and get coffee on Salesforce. Um, if you want, get that caffeine high for the rest of the afternoon. All right. With that, I want to say thank you again and have a great rest of your dream force.