Salesforce, Inc. (CRM) Earnings Call Transcript & Summary

Hi, everyone. Good morning. I hope you're all doing well. I welcome you all to this session where we are conducting this webinar on the future of self-service, how AI agents and the evolution of customer portals can actually lead to the next generation of customer experience and customer service. My name is Rajiv, and I'm joined by Sri Nihita, who will be running this session today. Before we get started, I just want to quickly give you forward-looking statements that Salesforce is a publicly listed company, and we urge you to make your purchasing decisions based on the currently available products and services. And in true Salesforce culture, I would like to thank everyone on this session who has taken time out and joining this session, all our customers, our partners, and everyone who has made Salesforce what it is today. Before I start, I just quickly want to -- a quick introduction. My name is Rajiv. I'm a Principal Solution Engineer at Salesforce. I've been with Salesforce for over 3 years. I focus on Service Cloud and AI capabilities within the Salesforce portfolio that we have. And I'm also joined by Sri Nihita, who is part of -- we are part of the same team and she is Associate Solution Engineer, and she will be also helping in driving this webinar today. Before we start, I quickly want to set the agenda and the expectation from today's session, what is that you can expect out of today's session. So we will start with the state of self-service today, where that market trend is going, how companies are innovating, what are the changes that are happening in self-service. What is the future of self-service, right? Because we all heard about AI agents or the agentic platform, as you might know. So we will also take a look at the capabilities of AI agents. Then we will take a quick demo. We will take a look at the self-serve capabilities as well as the AI agent demo, how they can come together and provide autonomous capabilities to the customers, how you can do self-service without getting in touch with the contact center any time of the day, whenever you want from whatever channel that you want. And then we will talk about the growth journeys with AI agents and self-service. If you want to start with similar capabilities with similar kind of experience for your organization, how you can start what are the stepping stones to get yourself or your organization enabled with AI agents and self-service. And finally, we will pause for Q&A as well. So while we will take Q&A towards the end of this session, my humble request that any time during the session, if you have any questions, feel free to put it in the Q&A box that should be there somewhere on your screen. And during the Q&A session towards the end of the presentation, we will take as many Q&A -- answer as many questions as possible. With that, what I'll do is I'll hand it over to Sri Nihita, who will walk us through with the self-serve capabilities, and then we will proceed from there. Sri Nihita, over to you.

Thank you. Thank you, Rajiv. Good morning, everyone. So I'm Sri Nihita, Solution Engineer here at Salesforce maybe for the last 1.5 years. And I have been here to drive digital transformation for businesses, and I've seen n number of customers in conversation with them. And in today's digital era, I think most or primarily the customers' requirement is to reduce their customers coming in for self-service and eliminate the cases that can come in, which can be dealt by the customers themselves. So that is where we're entering the era of self-service. So as a solution engineer here, I've worked with numerous customers to implement the self-service solutions and not only this has met the expectations, but has gone over and above. So basically here, what it helps businesses do is it reduces support queries and costs, increases the customer satisfaction and basically loyalty, enhances your brand reputation and competitiveness and gains valuable insights basically to have a better self-service customer experience at the end of the day, right? So quickly to start off. So as I already spoke about, today, the customers' expectations are reaching limits of -- expecting faster query resolutions. So basically, to continue to rise for faster and more reliable support, these service organizations are challenged today to meet these demands, right, while also maintaining the cost. So if you have to meet these growing demands today, it all comes at a cost. So as you can see these numbers, 81% of the customers are expecting faster service as part of the technology that advances today to meet today's demands, right? So most of the customers are turning towards self-service today. In order to get quick responses to both, it can be like to the simplest of their queries or even like most complex questions that can come up without needing to wait in a queue or like wait for a service representative from your end to take up a query or a question and to resolve it end-to-end. So without the help of a service representative is where we're looking at service -- self-service, which comes into the play. So with this increasing desire to self-service, businesses are today looking to improve their self-service capabilities and experiences to meet these customer demands. So with the help of various automation tools that we have here at Salesforce, it can be like bots or which we will talk about later, the Agentforce Service Agent. Basically, service leaders can start to deploy these more efficient self-service capabilities in order to deflect cases from coming in for simpler use cases and scaling support at a lower cost, right? As you can also see the numbers state that 61% of the customers would rather use a self-service than to actually approach another service representative or reach out to someone else if they can like resolve the queries on their own instead of having to reach out to someone else, right? So that's a quicker way of resolving queries. So what we want to look at here is basically pairing AI with our self-service capabilities as well. So not enough having self-service as a portal on the portal, you want to pair it with AI because AI is the new term today and everything that we do, everything that we think of requires AI in it, right? So -- but to keep up with these growing demands, the businesses need to continuously enhance their ways to always support these customers and meet them and provide faster resolutions from anywhere, right? It doesn't have to be like someone has to reach out. You want to ease the -- or bridge the gap rather between customers and your brand in order to have a seamless interaction with them and overall customer experience. So with the introduction of AI here in this era, service leaders are finally able to achieve this faster resolution that we're talking about, a resolution that is a lot more reliable and this support is supported from self-service, right? So self-service is more than just a place for customers to find like answers. It's not just like a direct FAQ sort of a situation. So you can actually find answers from the simplest of questions to complex questions as well. So it's a place where your customers can actually reach out to get help, to take actions on whatever help they require and to also manage their accounts, all of it in their own time and place. They do not have a specific situation or time that they can -- they can handle it at their own place, at their own time, at their convenience. And the advantage here is about pairing self-service with the autonomous AI, which brings in that element of personalization, speed, of course, that we're looking at, proactive service and overall higher customer satisfaction, which is key to all of us at the end of the day, right? So as you can see, 84% of our service leaders say that AI enables better -- a much better customer experience. So there is a bigger need than ever before today to invest in this sort of a technology and stay up to speed with the growing expectations of today's world. But while I say all of this, we all know that actually putting that into action or implementing this AI-powered self-service technology is easier said than done, right? So to achieve that sort of a superior or faster resolution and actionable support is a growth journey in itself and every business needs to be ready to make that step, right? So while we are ready to make that step, I'm sure there are a lot of hurdles that come on our way. This could be a few barriers that we'll touch upon. So while these could stand as barriers to achieve those goals and how we help -- we can look into resolving that. So since AI right now is still very new and it's changing rapidly, there's a lot of gap. There's a trust gap that happens around adding more AI into the support journey, right? You can have hallucinations or how much of those resolutions are reliable and not -- there's no amount of toxicity added into these answers that you're giving out to your customers without having any interference, right? So moving towards AI and strategy is an investment and not all companies today are ready to take on that additional cost because it is very expensive. So to have this effectively used, data and knowledge are key because all of your information resides on your system. So data and knowledge act as a key foundation. And many companies today have a lot of unstructured data here and there and data is not in a single place or a unified system. That is something that we hear from most of our customers that come in today that their data resides in siloed systems, right? So this kind of makes it very difficult to set up a smooth seamless experience between various applications, be it sales or service. And now when you bring in AI, to put everything into one single place is the task, right? So that's where that you want all of your data to reside in one single source of truth so that you can extract information and maybe drive insights based on all of the information that resides in one single place. So this is where Salesforce is here to help you. So what we'll talk about now is the Salesforce Service Cloud, which is a customer service management solution that we have that brings all of that what I spoke earlier together in one single place, right? So every service representative can act and use this to his advantage. And all of this is in an AI-powered workspace, right? So the solution here is such that, as you can see, this is a complete AI-powered workspace that gives the service representative who's taking up queries or talking in conversation with your customers. He can leverage these core tools that can be made more productive for his use case day-to-day. Everything starting from like case management or incident management or even to like resolve cases based on prior cases that were resolved to fetch information from your knowledge database and collaborate all of this information, right? So basically, what you can do here is you can -- as you can see, you can boost efficiency here by managing all of these cases, knowledge articles and incidents all in one single place. This, in fact, resolves complex cases faster with AI assistants that we have, which is the out-of-the-box automation that we have, right? So with AI that has come in, it can perform a few repetitive tasks for you or if you train it in a certain way, it can actually resolve cases without having a service representative involved directly, right? Along with this, as prior to like bots with AI coming in and Agentforce coming in, you can personalize every interaction with not just like personalization, you can also have intelligent recommendations that can come in for your customers. Since like I said earlier, all of the information is in a unified customer data, which gives a 360-degree view of the customer to your service representative and to the AI to act on this data, right? So the benefits strongly would be that it increases your service representative's productivity because, of course, a lot of his time has gone down instead of having to actually look into simpler cases, which are deflected. So a lot of his time can go into much complex use cases. So it increases his productivity and naturally increases case resolution speed as well because it is deflected on its own. AI takes it from there. You don't have to wait for a representative to attend your cases and take it from there and resolve it. It increases your overall CSAT and NPS scores. So basically, our customer service management solution would help all of your representatives work more efficiently and ensure that each of your customers gets the right support immediately, quickly and in a much more personalized way to them. So while we deploy this AI approach, right, in order to deploy this sort of a powerful and efficient AI-driven self-service, first, we need to create a foundation, a strong foundation. Like I mentioned earlier, most of these companies today have data in a lot of unstructured or structured format. So the foundation would be a strong foundation of knowledge and data that resides at the base. So data would be key to powering this most accurate AI responses. So all of your AI responses are driven based on the data that you put in. So that has to be a strong foundation. And your AI responses can be more accurate and personalized with the inquiries that can come in through these various self-service channels that we have. It can be through your messaging or chat or voice, right? Across various channels, you embed these AI solutions. So you have a much quicker response and more accurate response. So while you can see there are multiple systems where companies are storing their data today, with this unified knowledge in data cloud, we kind of make it easier for our customers to connect their third-party systems across various channels seamlessly and bring all of that data into the Salesforce's system, right? So this data powers the Agentforce Service Agent, which we will talk about a little later, and your autonomous agents for every service experience that we have, which is grounded by the answers residing in the knowledge database of your system and the company's trusted data. So Agentforce Service Agent meets these customers wherever they are, in whichever preferred channel of the customers, like you can see the various channels that they have on their preferred favorite channel to provide this autonomous support needed to resolve these simple or even like complex use case questions rapidly in a self-service capability that we have. So as I said, data is key to launching this sort of an AI self-service capability and providing this proactive customer service expectations that we have. So as I mentioned across various channels, so these are the few that you can see what is so important about AI and why I'm talking about it coming in via various channels, right? So what is so key about it is that it covers every part of the self-service experience, not just like from the start or you hand it over to someone else. So end-to-end AI takes care of it, which is why it is so valuable to the customer journey and plays a pivotal role from the start till the end of a conversation with your customer. So from features like AI-powered search answers that you can see or to recommendations and even next best actions like they have come in and you have an opportunity to upsell or cross-sell and you have next best action based on the interaction that you had with the customer and of course, Agentforce Service Agent, customers can get that fast and reliable support that they are looking for no matter where they are interacting, on whichever channel they are interacting right there on their self-service portal. So we'll dive a little deeper on these components of self-service. And I think what makes up this solution so strong. I'll hand it over to Rajiv.

Thank you, Sri Nihita. So thank you, everyone, for staying for the session. So what I will now talk about is the AI agents and the whole agentic layer and the autonomous agent capabilities that everyone is talking about. So what we are in right now is the third wave of AI, right? So if I just go back a little, the wave 1 was predictive AI, right? And Salesforce has pioneered the predictive AI era and Salesforce actually started their AI practice back in 2014, where we set up our in-house R&D team for AI, and we also launched our AI capability called Einstein, right? And as part of our predictive AI capabilities, we actually launched multiple capabilities across our products, whether it is Sales Cloud, Service Cloud, Marketing and so on. So I'm sure some of you who have joined -- some of you who are actually using Salesforce must have experienced the predictive AI capabilities like case classification, case routing, case wrap up. On the Sales Cloud side, also, there are a lot of capabilities like lead scoring, activity insight and so on and so forth. So we have been doing a lot of predictive AI capabilities already, and it is not something that has happened overnight that everybody started talking about generative AI capabilities within the last 18 to 24 months. This wave 2 of generative AI is where we started using large language models, which are trained on billions of pages of public data and so on. And it started to generate responses or generate the answers for the customers. Now the shift between generative AI to AI agents was very, very quick, right? In less than, I think, 12 months, we actually moved from generative AI capabilities to AI agents. Now how AI agents are different from generative AI? One, it not only can generate answers or responses or generate output for the customers, but can autonomously reason and take actions as well. Now what I mean by that is when I ask a particular question to an autonomous agent, it can actually reason and react or take action just like a human would do, right? So imagine you are talking to a contact center agent and you describe your concern or problem to that customer -- to that contact center agent. Now that contact center agent will understand your request, it will try to reason it, create a plan on what needs to be executed or what needs -- what actions needs to be taken to solve for that problem and then go ahead and execute those actions. An autonomous agent can exactly do the same thing just like a human would do. It can reason, it can plan and it can take actions on the query that a customer is asking. And this is going to change the whole perspective of how we actually look at contact center capabilities because it's a pivotal moment for customer service. When we talk about the customer service capabilities, you can actually bucket them into 5 or 6 broad categories. It could vary from technical support to appointment scheduling, to general questions. Some of the contact centers also focus on cross-sell and upsell, billing inquiries and return and exchange. right? So there are a few categories or a few areas which can autonomously be resolved by an autonomous agent. You don't need a human intervention there. They can handle most of the support tasks from tech support to answer general questions to scheduling appointment. What you need humans the most is to focus on high-value tasks. You want areas where they can, let's say, solve more complex tasks, which requires human intervention. And that's where you need some mix of human cognition and autonomous agents, right? And that's where you can actually use human capabilities or human intervention to address unknown complexity, right? If there is a system that has gone down or let's say, if there is a bug with a product that requires a product manager or an engineering team to get involved to solve a particular issue, that's where you would need a human to solve. Or the other area could be when you want to deescalate tough conversations, a customer is very unhappy with the product or service that they have purchased, right? They want to return or they want to exchange or they want to a treat, for example, right? So how you can deescalate those conversations. That is when you need a human cognition, human touch, human intervention for those customer queries. But for other autonomous, low-value redundant, repetitive tasks like appointment scheduling, general questions, cross-sell, upsell, billing inquiries, you can have somebody like an autonomous agent, which can give answers to the customer. It can execute actions on behalf of a human like making -- fetching the data from the back-end system, fetching the data from your order management system or CRM system and giving it to the customers. Those tasks can be automated completely with the help of autonomous agent. And that's where we see that humans along with autonomous agent can deliver seamless customer experience on Service Cloud. The name of our autonomous agent capability is Agentforce, where it can take care of all the low-value tasks on channels like your portals, your websites, your digital channels like WhatsApp, mobile app, and all of those digital channels, right? And then you can also have humans to focus more on higher-value tasks and they can work on Service Console, they can work on analytics, field service operations, so on and so forth. All of these capabilities are built on the common Salesforce platform, which has a common metadata across the capabilities. So you have data cloud at the bottom of the layer where you can bring in all the back-end data, create that unified golden record of the customer, harmonize the data and then activate for different use cases. And then you have Customer 360, where you have all the Salesforce apps connected, whether it's a Sales Cloud, Service Cloud, Marketing, Commerce, Slack, all of those are connected with one unified platform. And on top of that, you have Agentforce sitting, which can actually access data across these different applications. So one logical question that you might have, and that's a very reasonable ask from most of our customers as well that how an autonomous agent or AI agent different from your traditional chatbot because a traditional chatbot can also do, let's say, a lot of these redundant tasks. The biggest difference is that a traditional chatbot is very repetitive and reactive. You take any chatbot that is there -- out there. You will have to build your flows, you will have to build your dialogues, and it can only perform actions or give answers that you have added as part of your flows and dialogues. And it has to go through a scripted flow of record that you have created. It cannot dynamically pick the answer or cannot dynamically generate the answer, switch the context and give information, right? And at best it can only provide the information that is there available to it. Autonomous agent can quickly switch context between answering general FAQs to providing information to taking actions. And you don't need any kind of flows or topics to be created or any dialogues to be created, right? It can use the power of LLMs to generate dynamic responses for the customers. So if a customer is using a casual language, your AI agent can give an answer in a very casual language as well. If a customer is using a professional language, then agent can understand that and give an answer in a very professional tone as well, right? If you have created, let's say, 5 or 10 use cases within your autonomous agent and I switch my context or my conversation from a general FAQ to, let's say, order-related query or any technical-related query, the autonomous agent can quickly switch the context and have a human-like conversation with me. And finally, it can not only provide you information, it can also execute actions on behalf of the human. So every time you don't have to transfer the chat to a human to do those actions. Now those actions would be multiple, right? Let's say, in financial services, I want to change my nominee or change my address or authenticate some of the transactions, so on and so forth, right? You don't need a human intervention to do that, right? So let's say, if I want to change my address or nominee, I can upload the document. It can see the document that is uploaded, see the information of that document and then go ahead and change the address or change the nominee or block the credit card transaction right in the CRM and in my core system as well, right? So that is the basic and the biggest difference between a traditional chatbot and an autonomous agent. And that's where we have announced Agentforce for Service. Agentforce is basically the term that we use for our autonomous agent. And we have actually introduced this Agentforce layer across our capabilities. But today, we are focusing for service. So this is an autonomous agent which can engage with the customers across different channels, 24/7 in its own natural language. It can resolve your cases. It can accurately give responses, which is grounded in your data, right, in your enterprise data, which is there in the CRM. It could be there in your back-end systems as well, but we can bring it all together in data cloud, create that golden record of the customer and grounded in your enterprise data. Building this autonomous agent is also very, very easy, which you can do it within minutes with prebuilt templates that comes out of the box with autonomous agent. So it is essentially a very low-code, no-code kind of a platform. You can give the instructions, topics, scope in your natural language. And you can use the existing capabilities of flows, Apex, API calls, prompts and provide them as part of the actions for autonomous agent to go ahead and execute. One very important layer of this autonomous agent capability is the security and privacy because we take security and privacy very, very seriously. So we have essentially built Einstein Trust Layer. So we ensure that none of your enterprise data is actually going out or compromise with any of the LLMs, right? So we have masking and demasking capabilities built as part of our Einstein Trust Layer. And we also have agreements with LLMs like OpenAI, where none of the data that goes from Salesforce will be stored by these LLMs for training purposes, for monitoring purposes, for learning purposes, right? So none of the data will be compromised with LLMs. Plus we have also created the security guardrails where you can define the scope of an autonomous agent. It is not your B2C or any kind of publicly available LLM, which -- where you can ask any question and it can give you answers anything that is under the sun. So you can define the security guardrail. You can define the scope of the agent under which it will perform its actions and give answers. So it will not give you any random answer that is available. So now what I want to do is take a pause. I know Sri Nihita and I have spoken a lot for the last 30, 35 minutes. So what we want to do is show you everything in action, how you can actually do self-service on the self-service portals and you can also engage with the autonomous agents to have a conversational experience with an AI agent to get your answers -- to get your queries resolved and get the answers. [Presentation]

All right. So what you saw is how a self-serve portal can answer the questions and it can also work hand-in-hand with an autonomous agent where you can actually get a response from autonomous agent in a very conversational way. It can understand the context of the conversation or the journey that a customer has taken and then give a response based on that. And it can also execute actions like, let's say, printing a return label or issuing a refund and so on, using autonomous capabilities that are assigned to it. So how do you get started on this autonomous agent capability, right? What is the foundation or what is the starting point for that? The first is building the data, right? Your AI will be as good as your data. So how good is your knowledge base, right? It's okay if it is in different shape and form. It is okay if it is in different systems, right? With the capabilities of Salesforce, you can actually bring in data from all your back-end systems. You can actually bring in data both in structured and unstructured form. Using the data cloud capabilities, you can create the vector database. You can actually create those search indexes, which can be used by the autonomous agent to give answers to the customer. So as long as you have the data structured, unstructured and different database, we can bring it all together and create that data foundation that will be the core of your autonomous agent. And then how you will want to enable it on your portals, on your help portals, on your channels, right? So how good is your self-service portals, what are the different channels on which it is available on the website, on the mobile app and so on. And then you can go ahead and deploy your autonomous agent on different channels. Now those channels can be your traditional channels like website and mobile app, but then you can also extend it to other digital channels like WhatsApp or Facebook Messenger and so on, which are more popular among the new generation of the customers that we have. So once you have built your data foundation, you have sorted your self-service portal, from there, you can actually start deploying autonomous agents on different channels and create an effortless experience for your customers. So that is what we wanted to cover in today's session. What we'll do is now take a pause. We'll take a look at some of the questions that are there in my Q&A section, and I will try to answer those questions. So let me take a look at the questions.

Okay. So there's a question from Anurag Srivastava. He's saying any use cases for AI-driven self-serve solutions for real estate? The answer is absolutely yes. We have actually implemented and we are actually working with multiple real estate companies for multiple autonomous AI use cases. So those use cases expand from presales to post-sales experience. On the presales side, those use cases could be booking an appointment for site visit, getting the information about the new project that you have -- that a company has launched. And in fact, you can have an autonomous agent which can actually work as a human. So I give you an example of one of the use cases that you are implementing, right? So just like I would interact a sales agent, we are building an autonomous agent for a real estate company where I can go to that autonomous agent and ask questions such as assuming -- I'll take an example of Bangalore because that is where I'm based out of. So I can say that my office is in Bellandur, my wife's office is in Whitefield and my kids is studying somewhere near Mahadevapura, for example. So recommend me the property, which is closer to all of these 3 places, right? And then autonomous agent can actually look at the database, look at all the prelaunch or ongoing projects within the vicinity of Bellandur and Whitefield area and suggest that these are the few properties that are available in this area. But do you have any specific requirement? Like are you looking for a 3 BHK or a 4 BHK, right? So if I say, okay, I'm looking for a 4 BHK, then it will filter out the properties, which doesn't have 4 BHK and it will only give me options. So just like a human would reason, plan and give answers, autonomous agent can also do that. And then on the post-sale side, there are multiple use cases, let's say, I want my statement of account or registration, post registration, handover, so on and so forth, right? So there are multiple use cases for the real estate industry. How to collect customer data to have a 360-degree view of the customer? Okay. So it's a valid point, and I don't think we have enough time in today's session to answer this question. But in short, what I'll tell you is with the capability of Data Cloud, not only you will have access to the CRM data, the customer sales data and service data and so on. But with Data Cloud, you can also bring in the data from other sources, right? It could be your S3 or Databricks, Snowflake, Google. So customer website data, right? What is the journey that they're taking on the website, their mobile app data, their clicks data, their past purchase from OMS system, everything can be brought in together to Data Cloud and link it with that customer to see what is their affinity to a specific product or a specific brand. If you want to create any kind of view from those data, you can actually bring in all of your back-end data to Salesforce and create that 360-degree view of the customer. There's a question, a lot of interactions -- a lot of interaction happens with customer for cases by using WhatsApp, how this will get integrated? So in case you don't know this, but Salesforce has a partnership with Meta. So we provide WhatsApp as a channel natively from Salesforce. So customers who want to interact with the brand via WhatsApp as a channel, you can actually deploy WhatsApp for your customers. And on the WhatsApp as a channel, you can deploy agent as well. So you can have Agentforce as a first point of contact for the customers where they can get the queries resolved. If the query is not resolved, it can then transfer it to a human or create a case and so on. So we have the capabilities of WhatsApp within Salesforce. Mahesh Kothari is asking, so data cloud setup implementation is mandatory or prerequisite for setting up service agents, right? See, data cloud is an important part of the autonomous agent capabilities because, as I mentioned earlier, your AI will be as good as your data, right? You can still configure the data cloud and not bring in the data, it's completely fine. But then the responses from autonomous agent will not be as elaborate or as accurate or as grounded as you would want it to be, right? So setup and implementation is -- while it is not mandatory, but it is highly, highly recommended. Okay. What other questions that -- what is the difference in architecture technology of agentic AI when compared to LLMs, basically what enables it to have? Yes. So there's a very valid question from [ Sridhar ]. He's asking what is the difference in architecture technology of agentic AI when compared to LLMs. Basically, what enables it to have decision-making capabilities. So see, I mentioned about the agentic AI, what we have built at Salesforce and I can talk about what we have built. So we have built something called as Atlas Reasoning Engine. See, the job of the LLM is to send the prompt and generate the response back, right? That is what an LLM would do. The response that they will generate would depend on the kind of prompt that you are building and the kind of grounding that you are doing, right? So the first part is grounding where you are grounding the data in CRM, which is already there, right? So our responses are much more contextual and personalized to the customer who's asking the question. That is the first part. That is what LLMs would do. But what we have built internally within Salesforce is called Atlas Reasoning Engine. So when I ask a question to an autonomous agent that what is the status of the refund. The autonomous agent would first need to identify who is Rajiv, right? Is he authenticated or not? If he's not authenticated, what do I need to do? I need to ask the e-mail or verify the e-mail or send the OTP or whatever, right? And once I authenticate Rajiv, then I have to look at his last 5 orders and the order that he is asking for refund or return, right, then I have to check whether he's eligible for refund or return. Based on the certain business conditions, what is the category of Rajiv as a customer, whether he is a prime customer, whether he's not a prime customer or gold customer or whatever, right? All of those planning and reasoning is essentially done by the Atlas Reasoning Engine that we have built, right? Generating the response is the second part. But first, I need to understand the request, reason and then create a plan and then go ahead and execute. So this middle layer of reasoning engine that we have built is what is different from the basic LLM, right? I hope I was able to answer a question, so I'll move on. So Sanjay is also asking is conversational AI [indiscernible] agentic AI. I've answered the question what is agentic AI. So it can also -- beyond just conversation, it can also plan and reason and give answers. What else? Sorry, I missed a few questions. There's a question from Venkat. He's asking how do we build these agents? I mentioned briefly that you can actually build these agents in your natural language by giving the instructions and topics, defining the scope for the agent. You can get in touch with your account manager if you're already a Salesforce customer, and we will have a detailed discussion in terms of how you can build these agents. Again, there's a question from [ Dr. Venu Gopal. ] How do we contact you for more specific use cases for a contact center? Again, get in touch with your account manager who is aligned to your organization, and we can actually get in touch for a more detailed discussion. Same question from Anurag. Can you plan a specific case demo in upcoming days? The answer is yes. There's a very good question from Premjit, who is asking what would be the response if the customer query is out of the knowledge-based trained. Now this is where an enterprise autonomous agent or agentic AI comes into the picture as compared to using a ChatGPT or a Gemini or Perplexity and so on. That's where you define the scope and guardrails for an enterprise AI. You don't want an enterprise AI to give any random response that is out there in public domain or to just go ahead and execute any random action, right? Because that can have a lot of consequences, both regulatory and financial consequences. So that's where you define the scope that what are the use cases that will solve for. If a customer is asking a question which is beyond any of these use cases, you have multiple options. You have an option to connect them to a human. You have an option to create a case and then let a human get in touch with them or you can ask them to, let's say, to visit the nearest branch and get those queries resolved because there are certain sensitive industries like financial services and health care where you better off not giving an answer than giving a wrong answer, right? So that is where you need to define the scope, the boundary and the use cases what autonomous agent will solve for. I just want to see if we are good for now. There's a question from [ Shankaran. ] Can we connect with 2 live database for getting live price and shipping status? The answer is yes, you can. [ Shreeram Choudhary ] is asking what LLM models are being used in back end for these agents? So we have this concept of bring your own LLM. By default, Salesforce will give you capability to choose any of the LLM models that are available out of the box. It could be OpenAI, 4.0, Omni, Azure and all. But let's say, if you have built any LLM models in-house within your organizations, let's say, Amazon Bedrock or Google Vertex and so on, we have this capability of BYO LLM as well, which is Bring Your Own LLM. So if you want to bring an LLM, which you have trained on your enterprise data or modeled on your company data for company -- for your company use cases, you can do that as well. Is it live for any telco industries? We are actually working with a lot of telecom companies globally. I mean we don't have any referenceable names right now, but we'll let you know once it is available. All right. So I think we are good on time, and I have pretty much answered all the questions, unique questions. So that is all from my side. Thank you, everyone, for taking time out and joining this session. I hope you found it useful. Thank you, Sri Nihita, and thank you the organizing team for setting this webinar. I hope you all found it useful. Have a good day, rest of the week and a good weekend. Thank you, everyone.