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

Hello, everyone. A very good morning. I'm Srinivas. I am a specialist solution engineer from the Service Cloud team, and I'm thrilled to share our latest innovations with you. I would want to start with a thank you. Thank you for being our customers. We hope to make valuable use of your time today. Your success is a priority to us. Before we start: a quick reminder that Salesforce is a publicly traded company and the customers should base their purchasing decisions on products and services that are already currently available. The last 2 years have been remarkable. Especially, there has been a lot of excitement ever since ChatGPT was launched some 2 years back, around November. And in just 5 days, it had reached 1 million users, faster than any other technology before, including social media. While the gen AI revolution started with text generation, today, we have capabilities to create images. We have speech and songs being generated by AI in different voices. And today, we are also seeing a lot of new technologies which are also creating videos for AI. And what is happening is that the technology waves, if you see, are becoming shorter and shorter; and nowhere is this change more profound than in customer service. We are in midst of a pivotal movement -- moment for customer service. And as service leaders, it is your moment to lead your organizations and your entire company. And what's creating this moment is the third wave of AI, the AI agents. The reasoning ability of these AI agents married with the ability to orchestrate action drives a step change in automation. That is a game changer for customer service. [indiscernible] new to customer service, AI from earlier waves has driven scale for service leaders, reducing costs and improving customer satisfaction. In fact, our customers have reported nearly 10% to 20% improvement in customer [ seg ] deflection and -- which is translating nearly up to $40 million in average cost savings, but what if you could get more than that? With AI agents, the pace has significantly accelerated. We are seeing customers get 50% to 70% increase in deflection or more in just 3 months. This is truly a game changer. This will result in $120 million to $125 million in savings per year for any -- for an enterprise company. With AI agents, you can achieve this, but it is only a part of the strategy. Let us first understand how AI agents will -- agents help with key service tasks. AI agents grounded on the right data can be armed with instructions to not only just answer questions but also handle all sorts of frontline support tasks. For example, it can be troubleshooting. It can be scheduling appointments or it can be handling returns, but certain tasks need some human nuance and judgment. For example, how -- there may be instances where you need to handle a complex troubleshooting that will require involvement of [indiscernible]. Or you may be required to de-escalate an angry customer. What if you had a platform that would seamlessly blend the nuance of humans with the scale of AI agents? Well, now you can. And I'm excited to present to you the complete AI service platform, Service Cloud. Service Cloud enables seamless collaboration between autonomous agents and humans for every type of service on any channel and across any industry, to deliver effortless experiences from first contact to final resolution. With autonomous agents, you can transform low-touch, high-volume interactions into no-touch experiences, seamlessly handing them off to your human agents only when needed. For high-touch and complex interactions, AI integrated into the flow of work helps your service teams achieve new levels of productivity, drive better customer satisfaction and more importantly, drive growth. And all of these capabilities are built on the Salesforce platform with Agentforce and Data Cloud. So what is Agentforce? Agentforce securely connects your real-time unified business data with intelligent action, with the right trust, security and data governance requirements directly supporting [indiscernible] which all the companies and all of the people absolutely need to run their businesses. Today, we will walk you through 4 steps to really seize the opportunity of the AI movement. Now the first step is, of course, to deploy the Agentforce agents across every channel, but before we dive deep into this in detail, let's talk a little about the technology that came before the agents, the technology that so far we have all used, bots. And to get us started on that, let's play a quick video. [Presentation]

So that was a really frustrating experience. And all of us would have had similar experiences in the past with -- when interacting with a bot. And this -- really because bots are really great at answering simple requests, but they often and more than often miss the mark when things go off script. They work on the principle of defining utterances and extensive training that is not only time consuming but also sometimes also not very accurate. That is because bots are just fundamentally different by their programming, by their static values. And that's why I am so happy to introduce you to something that takes a totally different approach, Agentforce service agent. Now let's get one thing straight. This isn't a bot. It's an autonomous AI agent that resolves complex requests with empathy and care. Unlike a bot, it can easily navigate pivots in conversation. And it doesn't just understand text. It will soon have the capability to understand images and audio too, and it does it across all the channels. And it can take action, actions that in many cases you have already defined in Salesforce through things like Flow and Apex. That ability to understand, reason and take action is what makes it so different from what came before. And to add to that, in just a few months, we are introducing 2 more new agents: scheduling agent for the field service teams, releasing anytime now; and an employee service agent for human resource teams. So Agentforce service agent is simple, as it's only compromised (sic) [ comprised ] of 3 components: topics, instructions and actions. Topics group actions into jobs to be done, helping Agentforce service agent focus on the relevant actions for the topic at hand, for example, order management, repair queries and general FAQ topics. Topics help constrain what Agentforce service agent will attempt to reply to, ensuring that an external-facing agent doesn't address off-topic subjects. Behind every topic, there are instructions that you write in natural language to guide Agentforce service agent on how it chooses which actions to run or what clarified questions to ask. Each topic has a specific subset of actions that Agentforce service agent can choose from. So let us see how all of these come together through the agent builder. [Presentation]

Here we are in the agent builder, where we start by equipping our AI agents with topics. Topics are nothing but jobs to be done; for example, order management. Next, we give our agent its scope and instructions of what we want it to do. Now this is what's really cool about Agentforce agent. The instructions are written fully in conversational language. Just think about this: Your admins can partner with your top human service reps to learn how they do things. They can explain those best-in-class processes to your AI agents with simple instructions, and your AI agents will be able to execute those with any customer. Now the last step is to assign actions to the topics. Our agents have some out-of-the-box actions. And we can even create custom actions using some existing objects like Apex, flows and even prompts. And what -- you would have seen that we have not used even a single bot dialogue, only some instructions in natural language. So with a few clicks, our agent is set. And now let's see it in action. For this part of the demo, I am the customer. And I'm reaching out to Alpine group, a cable and Internet provider. I want to add more networks to my cable package, so I message Alpine through my favorite channel, WhatsApp. Right away, I am greeted by the Agentforce service agent that we just built. The service agent reviews my criteria and offers to create a custom package, but how can it do that? With the right data and right actions, Agentforce service agent is grounded in data and knowledge that's stored in Data Cloud, so that it can be pulled from that information to execute a custom package action. I -- when I accept the offer, I am instantly given personalized recommendations based on my responses and my customer profile in Salesforce. Let's pause here. This is awesome. I didn't have to navigate through a long bot menu to find the right package. The agent only shows what's most relevant to me, but let's throw it a curveball, a context switch. I am worried that mobile streaming might eat into my data plan, so I take a look at this. Unlike a chatbot, Agentforce service agent remembers our conversation history. It doesn't start from scratch. Now I am ready to go with the new movie package, but look at the text message I have received. Agentforce service agent proactively suggests an upgrade and is ready to schedule my field service appointment. I definitely want the upgrade, but I can't be late for my customer meeting. Normally this would be the end of a chat, but not with the Agentforce service agent. Here's the game changer. The agent can switch to voice conversations using the same generative AI so we can chat on the go...

So now that we have talked about how with self-service and Agentforce agents routine tasks can be taken care of automatically. Now with those aspects addressed, let's talk about how we can elevate the role of teams into truly trusted, highly informed advisers to your customers, so that your service teams can focus on the more complex and critical issues that truly matter. But here is the challenge. When your teams are overwhelmed in complexity and lack the data and institutional knowledge to address these issues, how do you support them in handling these high stakes situations effectively? Well, guess what. Agentforce helps here, too. Agentforce embeds AI-powered capabilities directly into your service teams flow. And this is -- this AI is deeply grounded in your enterprise knowledge base, which means it can proactively craft strategies based on case details, intent and engagement history. This means your service teams are not only armed with the right information but also empowered to drive personalized impactful solutions. Let's dive into another demo and see how Agentforce and Data Cloud elevate your service teams from overwhelmed to outstanding. [Presentation]

Let's bring back the Alpine group again. So Alpine group does this real-time monitoring across all of their devices just to make sure that they are all working well. They noticed that there's one particular customer, Emily, whose main control panel is not working well. Agentforce service agent can quickly send her a note through the app, lets her know of the issue and also gives her step-by-step guidance to troubleshoot that, but Emily does not want to do any of that. She just wants to return the item. However, there is a small hitch. Agentforce service agent gives her a [ compliant ] response and lets her know that her panel is unfortunately back-ordered for 3 to 6 months. Emily is really, really frustrated about this. And this is a very routine conversation that you are seeing between an AI agent and a customer, but this is where it gets interesting. Service agent realizes that we need to now escalate to a human agent and does so in a very seamless way, so let's check in with our human agent, John. So this is a brand-new, slick UI. And this is the home base or the service console for John. This is where he gets to see all of his customer interactions across all of the channels, right here. And once he accepts Emily's case, you will see that this is Emily's unified profile powered by Data Cloud, with all her details, her name, her e-mail and even her customer sentiment, thanks to the customer experience intelligence. And on the right side, you have Agentforce right there in action. And there's a brand-new feature called Service Planner. Service Planner is giving John step-by-step guidance on how to resolve Emily's case. Step one is to just verify her issue and make sure that we have the right details. John has access to refining, editing right here; and that's humans in the loop in action. And because Service Planner has unified knowledge, we quickly realized that there is another panel that's available right here for Emily. And you can finish up the order right here. It does not stop because the Service Planner is recommending that we give her complimentary security camera that is actually in her cart. Emily is thrilled at this. And we've been able to bring in commerce and Service Cloud in one platform to complete that order right here. And the last step is to get a quick case summary. And we can see what happened. What was the issue? What was the outcome? And with one click, you can create a knowledge article...

So let us just reflect on the range of activities that were showcased just now. So firstly, Emily was not knowing that she had an issue. We were able to proactively tell her that there is an issue. Second, we were able to proactively give her options that she could pick on, what she wanted to do with her panel, whether she wanted to troubleshoot it, whether she wanted to return it. All the options were available to her. Third, Agentforce was able to transfer the call to a human agent on sensing that the customer is getting jittery or based on the -- and based on her displeasure and based on her conversations. And finally, we even gave her a complimentary security camera that is in her cart. In business-as-usual terms, this could have been a disaster -- disastrous customer experience. However, with Agentforce, we were able to turn this into delightful customer experience. And to think about John. His agent productivity just went up multiple times because Service Planner was there, was right there giving him the exact personalized steps to take in this particular case. Now this really showcased the power of Service Cloud bringing humans and AI agents on one platform across every channel. So for those of you with field service teams as well, the complete service experience means we must equip our field terms -- field teams also to be more proactive. While there are some instances of autonomous robots performing tasks in the field, by and large, we are still -- we still depend on our skilled workforce out on the road. For some organizations, this is the only face-to-face they have with the customers, so it could not be more critical. I want you to think about your last on-site visit. Was it easy to schedule? Was it on time? What is resolved in one go? Unfortunately, processes are riddled with inefficiencies, but we have so much data to inform a better experience. So with Agentforce for Field Service, your teams are able to focus on people work, not paperwork. Everyone from your customer to your dispatcher, technician, your operations team management, they can take advantage of this AI that augments, automates and optimizes. It empowers your customers to schedule appointments at their convenience and on any channel. Agentforce service agent for scheduling is an autonomous agent that can manage complex scheduling scenarios using natural language. It can assist your dispatchers with scheduling exceptions, surface-up details about the schedule to help them work faster and do -- and improving the -- and maximizing the utilization. In the field, it can greatly minimize the clicks for your technician and also help them focus on the work they are there to complete rather than the administrative task around it. It does this through providing complete and comprehensive summaries of the purpose of the job and the work completed, conversational assistance and more. And for managers of field service operations, Agentforce will transform how they derive insights from unstructured data by analyzing all service notes captured by the technicians; summarizing this unstructured data into common themes like cancellation reasons, issues encountered, missing parts and more. So while we -- while the above showcase some of the innovations that we have launched with Field Service, let me take you to the top 3, the more interesting ones. First one, the dispatchers today have Agentforce -- can have the Agentforce to help them scale and take control of the schedule from one place in the most efficient way. Secondly, asset service prediction can use AI and Data Cloud to anticipate additional work that will be required soon so that you can maintain uptime for your customers and be efficient with your engineers. And thirdly, techs have multimodal troubleshooting, which means that grounded techs guidance now understands image and voice to get quick answers. So let's see them all in action in the next demo. [Presentation]

Our dispatcher at Alpine is reviewing the schedule in the Field Service dispatch console. This is where you can see all your resources, people, assets and equipment. The schedules were optimized overnight, but now several technicians have unexpected openings, which means they are now underutilized. The dispatcher turns to Agentforce for insights. Based on the summary, there are higher-than-expected appointment cancellations in the last couple of hours. Our dispatcher needs a quick solution to fill these gaps. Agentforce is recommending to be proactive and do more work on the jobs we already have. With asset service prediction, it noticed there are upcoming maintenance jobs and related work items that they can complete on the site. The dispatcher approves those work items. This means we are preventing a future truck roll, extending asset life and ultimately enhancing the customer experience. Okay, let's shift to one of the technicians receiving this additional workload. When it's time for the job, the technician receives a real-time notification. She can tap to say that she's on the way. And now she can listen to the AI-generated pre-work brief that outlines the install and maintenance. On site, she goes through the work plan steps. And as she begins testing, she can go into the connected asset and review the asset health score. This is unlocking the trapped asset or IoT data, where she can see real-time signals. On the last step of work, she encounters an unfamiliar error on the control panel. In the past, this could have been a major setback, wasting time seeking help or leaving that job incomplete. But not to worry, she has the power of AI in her palms. The tech can speak to Agentforce and simply snap a picture of the error code. Agentforce interprets the code and brings back resolution. It just gave it a picture. It knew an asset model and history from the work order where we were doing an installation process. It reasoned with that and relevant product knowledge to return the precise response. It also knew we didn't have a part. We could order a quick delivery from somewhere nearby. This is what we mean when we say Agentforce assists and acts. And by using voice and images, our techs now can have next level of efficiency. Once the job is complete, Agentforce helps us summarize the on-site job into consistent, useful notes. Our customers are saying this is the easiest way to bring in knowledge from the field. And it can be used as a grounding for future troubleshooting. So you just saw schedules get optimized. Agentforce helped resolve unknown error with a snap of a picture, making this a first-time fix. That gives techs back time and improves work satisfaction. This is the power of humans with AI agents delivering customer success.

So now let's talk about our final step, which is turning insights into action. Today, we have service leaders who have to juggle way too many tools, too many screens, too many interfaces. And it is very hard to even get the insights you want, let alone be able to take or -- take action on them at the scale which is required, yet it's also never been as important. Nearly 85% of decision makers are saying that service is expected to contribute a larger share of revenue this year. We need to be able to spot the trends where we should take action quickly. We need a world where your service leader has real-time insights into their operations; can track performance, bottlenecks; and even dive deep into the root causes behind the customer sentiment. And that's why we have the customer experience intelligence. This new AI-powered capability unlocks the full potential of your customer interactions by providing "up to the minute" insights based on the content of your conversations themselves. With AI at your side, you can make fast, informed decisions; and personalize the experiences at scale. Let's dive into how service leaders at Alpine are using Service Cloud and customer experience intelligence to stay ahead of customer satisfaction right in the flow of work. And to do that, let's get into the customer experience intelligence dashboard. [Presentation]

This is where you can see all of your sentiment metrics organized by topic, by geos, by channels all right in one place. We can also see from the graph that there is one particular topic, warranty issue, that is also related, correlated with a lower CSAT, so we invoke Agentforce over here. So we type in Agentforce, "Can you tell me the top 3 reasons why the warranty and low CSAT are associated?" Because Agentforce has access to conversation mining data, very quickly, you can see the 3 reasons for why it is happening. You can also ask Agentforce to quickly draft a note, so that we can send to customers to -- letting them know of this change. You can edit and make changes to this draft. And in fact, you can even go further and, where required, offer customers a 10% discount for their future purchases.

Okay. So let's take a look at some of the customer success stories. Okay, the first one is OpenTable. OpenTable helps diners in restaurant reservations and discover new restaurants. They serve more than 60,000 restaurants and have nearly 1.7 billion seats to fill. They wanted to transform their user experience while improving efficiency. Agentforce allows human service reps to focus on more complex tasks and deliver superior service with Service Cloud's AI-powered recommendations and case summaries to reach [ switched ] resolutions. The other case that I wanted to take up was around Wiley. Now thousands of students use Wiley's services. And it is -- and the business is pretty seasonal in nature and there are sometimes a lot of spikes during the peak times. Agentforce empowers Wiley's agents with AI-driven data-powered recommendations and seamless integration into their workflow, enabling the team to resolve cases faster, optimize service operations. In fact, they have been able to report nearly 40% higher resolution with Agentforce than the previous chatbot, culminating into a total savings of nearly $230,000. And finally, before we ended, I would -- I wanted to emphasize that, while Agentforce is the latest offering from Salesforce for gen AI, this is not in isolation. Salesforce has been the front end of AI for nearly 10 years. And our customers have access to more than 60 features that also include use cases involving other AI domains like machine learning and predictive intelligence. So we help our customers with bots and recommendations, builders that can help customers find answers on their own or service team build workflows, case routing so that the right case gets to the right queue or recommending replies based on customer and case data. And also next best actions and surfacing personalized recommendations, so we have got a load of AI which is already powered on the platform. And the best part is that this is -- all continues to be available to our customers and exist with our latest Agentforce offering. So with this, I have come to the end of the presentation, so at this point of time, we can take any questions, if there are. [Operator Instructions]

So I'm just going through the questions. Okay, so I think there are questions. I will start from the first question. Okay, so there's a question from [ Nishant ], which is essentially whether we can understand on diverting agents [ suppose, based on ] the customer sentiments. Okay, so see. What we are doing is in -- when you create a topic and an action, okay, you can definitely give the Agentforce instructions as to when it should be able to transfer the call. So it can be based on the tonality of the customer. You can -- and the best part is, instead of getting into sentiments, what you can do is you can provide the instructions as to the points at which the case should be transferred. So there is a lot of flexibility out there. So there's one question on how AI can support in troubleshooting part. So in fact, this is something -- so the best part is that, as I covered -- so there is a multimodal support, for example, in Field Services that will be coming up. So the best part is that, once you are able to scan, say, an error code, it is able to identify the device, the device name. And also based on the error code, it will be able to correlate it to one of the knowledge articles which basically explains what that error code means, is. And then troubleshooting steps can also be provided as standard next options for that. Okay, one question which is coming, is Agentforce [ create for ] any programming language needed? I did not get much of it, but just what I am assuming is that the question is around what -- how is this Agentforce deployed or what is the technology behind it? So from the customer perspective, I can assure you that, if you have to deploy the Agentforce, primarily you have to create topics. And you have to -- which is classification of steps or -- of the work that is need -- to be done. And instructions can be given in conversational natural language. You do not require any programming language understanding for essentially deploying the agents, if that was the question. Okay, there's one question from [ Patricia ], which is asking, during the customer journey, can Agentforce help in creating flow of converting the new customers to [ EW or AMC ] customers or renew the existing contract? Yes, this is something which can be done. So based on -- again, we can have a topic which -- to deal with this. And then in -- we can create flows which are essentially going to convert or -- the typical onboarding flow which the customer -- or the warranty update flow which has to be taken. That can be invoked through the actions, so that is very much possible. So I think we will have time for maybe 1 or 2 questions more, so -- okay, there is a question, which is, when a user is conversing with Agentforce, can the agent or any internal person access the chat during the conversation or only after it is transferred? Okay, a very good question. So we also have something called Omni Supervisor. And the idea is that -- I think that is going to be GA sometime in February. So the only supervisor dashboard today, we are, anyways, able to monitor the conversations which are happening between the human agents and the customers. Now with Omni Supervisor dashboard, you will also be able to monitor the conversations which are happening between the customers and the Agentforce agents. So that is something which is -- that is a feature which is soon going to come up. So I will take the final question. It's a little broad question, which is what's the TAT for deployment. Okay, now this is again one other area where Agentforce brings enormous benefits. So in traditional bot deployment, typically what you needed to do is you needed to define a lot of dialogues. And also you needed to create the entire logic of which dialogue should come after the previous dialogue, so you had to define the whole hierarchy or the whole tree as to how the conversations -- you want them to be happening. With Agentforce, all this goes away. You give it a set of topics and you give it a set of actions. And basically the Agentforce reasoning engine basically decides how to match the customer queries with the actions that are going to be invoked. So the deployment times will go drastically. And second advantage is that you also do not require any explicit training when deploying the Agentforce. So all that is required is that your team, your SME and your Salesforce admin, they work together and then do the testing around the Agentforce conversations. And the entire training or creating utterances, that goes away. So that's again -- and it saves a lot of time. So I think this is probably last question that I could take. What I will say is you can post these questions. You are sending these questions. We have noted it. And we will try to -- we have noted the questions which are on the chat. What we will try to do is we will try to reach out to you or respond to these questions individually. Thanks for sharing your time today. Thank you.