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

Good morning, everyone and thanks for joining this webinar. Some of you might have already seen the slide before. It is a gentle reminder that Salesforce is a publicly traded organization. And we request you to take your purchasing decisions based on the information that is currently available. I would like to start with an introduction of myself. I'm Kevin Domadia. I'm the lead solutions -- Specialist Solution Engineer working with Salesforce, and I specialize in service cloud, AI and field service operations. Srinihita, do you want to introduce yourself?

Yes. Thanks, Kevin. Good morning, everyone. I'm Srinihita and I'm working as an Associate Specialist Solution Engineer as part of the Service Cloud team here at Salesforce. I've been working here for close to 2 years now. So that's about me.

So Srinihita and I are going to talk about the AI solutions, which are available for field service. Before we start this webinar, I just wanted to -- I wanted to start with a small story of Raj and Ria, who recently had a baby. Now we all know how babies are, right? So whenever they are hungry or whenever they get bored, they ensure they get full attention of their parents, whether it's day or night. So Raj and Ria had already lost hours of sleep and they were almost acting like zombies. Summer was approaching and we know that summers have really started becoming very bad, and it's not possible for us to stay without air conditioners. So Raj and Ria, they used to keep their AC on 24x7. And behind this AC, there was this AI model who was constantly getting this telematics data and ensuring that Raj and Ria gets the AC up and running 24x7. In a few days, what happened is that this particular AI model just realized that there is a potential gas leak because of which the cooling operations might get impacted. And because it had the entire context of Raj and Ria, it understood the behavior and it was also aware that they prefer talking on WhatsApp and they mostly respond in the evenings and over the weekends. This AI model automatically went ahead and send that particular notification to Raj on WhatsApp. Immediately, Raj, of course, acknowledged it and he accepted this work order. Now the entire issue was then transferred to a field service engineer who then visited Raj, who had -- and the AI model also ensured that all the important necessary parts required for servicing that particular AC were already available with that technician. So the technician was able to solve that issue in the first go. And even before the issue occurred, even before the AC stopped working, Raj and Ria were able to get it serviced and get that problem solved. Isn't this the kind of experience you would want for yourself and for your customers? Well, now it's possible. So we've entered the third wave of AI, which is popularly known as Agentic AI. It all started with predictive AI and basically AI learning from past data and giving meaningful predictions. When this was given to humans, we started realizing that while it's very good at the work it does, it doesn't understand human language and context. So the application of this AI became limited to folks who are tech savvy and who knew their way around technology. And that's where a few years back, generative AI became very popular because it was able to converse with humans in natural language. And it was also able to generate content, which was in context to what the customers were asking for. It was built with billions of parameters, which ensured that most of the queries, especially in the consumer industry, the customers asked for, it was able to answer. And with the -- with this AI becoming popular and this AI becoming -- generative AI becoming dependable, that is where now these AI models also have the ability to take actions. AI autonomous models are available everywhere. So let's take an example of Waymo, driverless vehicles, which are dropping people to office or if we take an example of Wiley and OpenTable who have deployed these AI models on their websites and on digital channels, which are able to respond to their customer queries and solve most of them even without human intervention. Now whatever stage our customers are today in their AI journey, we are helping them adopt the new wave of AI. Most of the customers who have implemented AI today, they believe that AI will bring in a lot of efficiency, thereby reducing costs. It will also start assisting the agents and humans with simple and complex tasks thereby reducing the time taken for resolving that particular issue. And also, it will empower the customers to self-serve and solve a lot of issues on their own. So the customers now can reach out to businesses at their convenient time via a convenient channel, which they prefer and get their issues solved without any human interaction. This will thereby improve their overall customer experience as well. While there are instances of robots performing tasks autonomously, if you look at field service operations, even today, they are heavily dependent on skilled labor. Now for some businesses, this is the only face time that they have with their customers. I want you to close your eyes and think about the last interaction that you had with any of these businesses. Were you able to easily schedule an appointment? Did the agent reach on time? Was the agent able to resolve the issue in the first go? Well, these are very complex processes which are riddled with inefficiencies. Having said that, we have a lot of data, which then can be used to improve the customer experience. And that's why we've introduced Agentforce. Agentforce is a platform which will get humans and AI together, thereby solving customer issues. On this particular platform, you would be able to build agents for marketing, for sales, for service, for field service and any other department. And these AI agents will become the front line for your teams. They will not only be able to preempt issues but they have the ability to plan and reason as well. They will be able to trigger end-to-end workflows, automate a lot of the processes, also make decisions like how humans do and solve a lot of work on their own. While, of course, AI agents will be able to do that autonomously, they would still keep humans in loop. So wherever required, either it will keep the human informed or it will easily transition work to these -- to the human agents. When we talk about this Agentforce platform, it sits on the Salesforce metadata -- it's built on the top of the Salesforce metadata layer. These Agentforce agents, because they sit on your CRM applications, they already have the entire context of the customer. So they exactly know your customers' preferences. They know what orders they have placed in the past. They know the products that they have bought or the opportunities that you're running with those customers. Also, they are aware of all the service issues that your customers have faced in the past. So all of this is available to this agent on day 1 itself. Along with that, you have also built a lot of automation using Flow, using Apex or using Prompt. Now all of these automations also will be available to this agent on day 1. So you don't have to rebuild anything. A lot of this context and these automations can be reused by these agents from day 1 itself. And when it comes to our unstructured data, so when you want these agents to respond back with knowledge around your business by uploading unstructured data in the form of PDF, in the form of voice and so forth and so on, that is where you would be able to easily ingest this information in data cloud. And with the help of RAG, these Agentforce agents will be easily able to answer those questions. And all of this is happening on top of our trust here, which will ensure that there is no data leakage. There is no attack which happens on your AI models. And also whatever answers are given back either to the agents or to your customers directly, they are nontoxic. There is no hallucination. And it is all in a good presentable format. This is the power of Agentforce. While of course, with Agentforce, we can build multiple agents for different use cases. What we are going to focus today is on field service. Now think about -- when we talk about field service, it always starts with booking an appointment. Now this appointment, of course, in an ideal world, we would want the customer to be able to book on their own. And what if we are able to even before booking -- even before the customer coming and booking the appointment, what if we are able to preempt the issue and automatically inform the customer like it happened for Raj and Ria that there's a possibility of a failure and you need to schedule a visit as soon as possible. So this is where AI can start helping the customer, right? So in terms of preempting the issue, then booking the appointment, and once the appointment is booked, the next part is to find the right agent with the right skill set, with the right tools to then go ahead and solve these issues for them. On the day of that particular appointment, the field service technician would then want a brief because these technicians typically, they go to multiple locations. So they would -- normally would want a brief in terms of how they're going to -- day is going to look like. And even before they reach the customer, they need to be aware of the issue that the customers are facing or the maintenance work that needs to be done on top of it, which will then ensure that your customers don't have to repeat themselves and your agent is able to build a good relationship with them. And finally, in a scenario where they're not able to resolve the issue because of some technical challenges on the field, that is where, again, they would need AI to help them in identifying those challenges. And then in a scenario where it's not even possible to solve the problem, start the exception handling process. Now when we look at this entire journey, at every stage, we believe that AI will go ahead and help the customers in solving their problems. And that's where we've introduced Agentforce for field service. The idea is to enable the field service agents, your customers, your dispatchers and their supervisors so that they can focus on getting the work done as compared to just working on trying to do the paperwork. Now one of the research that we did, right, when we spoke to multiple customers who have been leveraging, using field service since a long time, we realized that 65% of their time that the field service engineer devotes is on paperwork, is on basically filling all the information, trying to order the parts and so forth and so on. And it's just 35% of the time is actually working on the field. So with the help of Agentforce for field service, we -- the agents would now be able to focus more on getting the work done. A scheduling agent will first help the customer schedule their appointment. And then when it comes to dispatching, the dispatchers will get a complete view of all the appointments, which are going to happen in that particular day. Now agent, AI agent will then help them identify the bottlenecks, will help them reschedule the appointments in a scenario where it's required, and also create filters on that particular Gantt chart, which they are able to see, so that they are able to do their task efficiently. And finally, the field service technicians will get complete visibility and a view of all the appointments that they have to cater to. And when it comes to issues, if they face any issues, it will -- AI agent will then help them with knowledge articles. So this is the Einstein for field service offering that we have recently launched, which will then cater to your dispatchers, your field service technicians and also the operations manager as well. Let's look at what the dispatcher does, right? So typically, when the dispatcher starts their day, what they are looking at is first in terms of how my day is going to look like. So they would want to see all the agents and their appointments. But along with that, in most of the cases, someone or the other is not available or there has been an emergency issue, which has cropped up and another dispatcher has to now schedule that meeting accordingly. So it becomes very important for this dispatcher to get a complete summary of what is going to happen. Now if there are only 2 or 3 agents on the field, it's easier for them to just look at the Gantt chart and get the details. But imagine a scenario where there are 10, 15, 20 agents, who are working on the field, it is practically impossible for the dispatchers to everyday go in the morning and see -- look at each and every work order and then get a view. So this is where the schedule summary will help these dispatchers get a summary of all the appointments, which have been booked. Now once they get this, they would then be able to create a filter with the help of Agentforce again. So instead of actually going ahead and then trying to create this filter, use, put their logic, they would be able to converse with this AI agent in natural language, which will help them automatically built these filters and then finally, wherever the gaps have occurred, right? So the service gaps, this is because of unavailability of an agent or it could be because there is an emergency work, which has come up, and now it needs to be filled up someone else, right? So this particular AI assistant will then help the field service technicians to solve those schedule gaps as well. And finally, once all of this is done and the field service dispatcher is happy that they are now -- my day is going to look perfectly fine. It goes ahead, the system would then notify the technicians. Now from a technician standpoint, when they start their day, they would want to look at all the appointments. Where they have to visit, what kind of work they have to perform. And what are the tools and systems that they would need to ensure that they are able to cater to each and every appointment. So that is where a pre-brief, a pre-work brief will be presented to those agents and they would then be able to get all of this information. Now we understand that when we talk about these technicians, they are on the field and a lot of times, there is no Internet connectivity. And that's where we've started piloting the prework brief in offline mode as well. So the agents don't have to be connected to Internet all the time to get that information. Now imagine a scenario where the agent -- the field service technician gets stuck with some issue and they are not able to get help from their supervisors because they're busy. So in this case, now the AI assistant is now becoming multimodal. So the field service technicians will be able to take a -- click a photo of that particular gadget, the system will identify the issue. It will -- with the help of image processing, it will try to understand the root cause of the issue and then assist the field service technician in resolving the problem. It will go through all the knowledge articles, it will go through all the standard SOPs and then eventually go ahead and help the customers -- help the field service technicians solve the problem. And finally, the most boring part, right, once the work is done, is to create a summary of the job that has been done by these technicians. This is where post-work summary will eventually go ahead and automatically create a summary for the work order, which has been completed and it will ensure that all the fields are populated correctly. Now this is one of the challenges, which a lot of organizations face that because this information is not correctly entered in the system, they are not able to drive analytics and get the right insights on the types of issues that the customers are facing. Now going forward, because this is generated by AI, your -- the businesses will be able to get this information easily and they would be able to automate more and more processes. So now what we'll do is, right, we've spoken a lot about these features that are -- that we're going to come up with. I'll start a small video or a demo of the features that I spoke about. So Srinihita is going to take us through the entire demo of the solution that we spoke about.

Okay. So I want to start with an example of how easy it is to get started with Agentforce for field service customers. This is something I hear all the time. It sounds great but where do we start. The good thing is that Salesforce makes it super easy to jump in, so what we're actually looking at here is a demo environment. Now I'm going to go ahead and dive into a templated flow that Salesforce provides out of the box to get started. Specifically, we're going to be configuring a pre-work brief. The idea is we want to get information to our technicians quickly and easily so that they have everything they need to start their job once they're on site or in route. So what's the flow actually doing? It's pretty cool. So this is using a standard flow action that most of our admins are already familiar with. Specifically, we're looking at the work plan, the service appointment, the record, the associated contact, et cetera. It can be extended with any other data that lives in Salesforce or can even pull an external data from your ERP, or other data warehouses through data cloud. Again, this template works immediately out of the box or you can take this as a starting point and modify it to your own unique needs. Lastly, it's actually taking all these inputs and formatting custom prompt instructions that are going to be sent over to the LLM and use to generate the actual pre-work brief. So what this actually looks like? Let's go ahead and take a peek at the prompt builder. And what this does is it makes it easy for any Salesforce admin to define, configure and deploy custom prompts right inside Salesforce. So for a pre-work brief could not be easier, I'm simply pointing to that flow that I created a minute ago. And that is going to be simple because that's going to compile the instructions from me. I could just as easily provide additional instructions of rounding data right here. But let's keep this one simple. Next, I have the ability to deploy this to the standard models. Some of the LLMs that Salesforce has officially partnered with or even bring my own custom LLM via Einstein studio. Within the standard model, I have the flexibility to deploy to different models or vendors. No other solution offers this type of flexibility. So what this actually looks like if we preview it right inside the builder, we see the prompt and we see the response that are generated based on the sample work order I have provided, right? I click the sample work order and it shows up. Pretty cool. Let's see what it actually looks like for a technician. Again, it could not be easier. It fits right into the work order they are already working on. When they open up the Field Service mobile app and upon selecting a service appointment, that pre-work brief is displayed right away. The tech understands the job at hand and can get started, delivering the service that your customers expect. More on this in just a moment. So where does this Agentforce fit in? How can we take this a step further. Okay, so let's go ahead and jump back into the setup and talk about it for just a few more minutes about how Agentforce works. So Agentforce has 3 name building blocks, your topics, your instructions and actions. Topics define your jobs to be done. Instructions are used to help the reasoning engine to understand what clarifying information it needs to then execute the appropriate actions. I'll spend just a couple of minutes highlighting a couple of examples out-of-the-box actions that are part of Einstein for field service. Specifically, let's see how we can arm those field technicians with a post-work summary to save even more time wrapping up a job. And let's use Agentforce to arm our dispatchers with conversational AI to help them manage intra-day scheduling. Again, these are already built for you. So what we see here is adding them to Agentforce deployment using point and click. And as they are added, we can see these topics, instructions and their corresponding actions. When we go ahead and test this out in the builder, we can see in real time is it identifies as a proper topic and then execute the instruction set to identify and run the necessary actions. First, we see a successful summary generated. And then it is able to pull out information around the days schedule. This is what makes Agentforce so powerful, right? Not only native access to all of the needed data but also the reasoning engine to identify the proper topics and execute actions on our behalf. So what is actually like for the same job that we saw earlier. Instead this time, let me show you. Let's open up our Field Service mobile action app once again. And now we can interact with Agentforce to save some time and some money. So first things first. I do want to call out Agentforce also natively has the ability to pull out data from your knowledge base or technical documentation that you already have existing in your system. So as issues are cropping up, we're not just digging through documents but Agentforce is sourcing answer on the fly, and it gets me the information I need to complete this job, saving both time and money. When it's time to wrap up, Agentforce quickly generates a post to work summary, saving time once again, an add stack to the completed work order. As you can see, this is done in a few seconds. What about that dispatcher action that we spoke about a few moments ago. Let's have a look at that as well. Same idea, it fits right into the work order dispatch team that is already doing. In the field service console, we can interact with Agentforce and understand what the schedule looks like today. Any jobs with rule violations or jobs that have gone into jeopardy are immediately surfaced and scheduling lists are created so that we can give these jobs the attention that they need and make sure that we are meeting the obligations that we actually set out with our customers. So of course, we do not want to disappoint them at the end of the day, right?

Thank you, Srinihita. So just to summarize, what we saw in the demo was the dispatcher agent, right? So we were able to see how AI will help the dispatchers in understanding what the appointments are for that entire day, reschedule them and also identify the gaps in the appointments and fill them. We also saw the...

So I want to start with an example of how easy it is to get started.

We also saw the field service technician view. And we saw how pre-work brief and the AI assistant will work in terms of helping the field service technicians. Now what you're going to look at is the service agent for scheduling. A lot of customers today, they want to self-serve, right And they want to book the appointments on their own. They don't want to reach out to a customer service agent, call them or e-mail them, then there will be multiple back and forth to find the relevant slots and then go ahead and book your appointment. And that's where this particular service agent has the entire information of all the agents and their availability. It also understands that for a particular issue when the customer has approached, which agents will be able to serve them and is able to look at their entire schedule as well. So whenever the customer types in natural language that, I want to book an appointment for this weekend in the second half. This agent is smart enough to understand that when the customer is saying this, they mean that they would want to book an appointment on 24th or 25th of March. And then along with that, they would want to look at slots which are available after 12 p.m. or 1 p.m. IST. So it will then go ahead and only present those slots to the customers, which they can select and then go ahead and book an appointment. This way, your customers will be able to have this conversation in natural language, and they would get the experience like how they typically interact with a human and get their work done. Now let us quickly look at a demo of the scheduling agent and see how this will help our customers with scheduling their appointments.

Good morning, everyone. So what about Agentforce Service Agents, right? How we can exactly deploy this technology to interact with our customers over our digital channels. I'm going to highlight a couple more of these examples that I have built out to show exactly how Agentforce for field service can be in action. So let's talk about exposing Agentforce directly to our customers. And this one, I'm really excited to talk about today. Right. So if we hop back into the builder, right, we can see how exactly this works. Same idea here, we are exposing an appointment management topic directly to our customers. Behind the scenes, what we can see is it reaching out to our scheduling APIs and pulling my potential booking slots for our on-site work. What you do not see is a ton of other dialogue trees, intent, models, utterances, et cetera. So this stands as one of the big advantages of Agentforce. That is its ability to leverage an LLM to handle the reasoning, so that we are able to understand the context, understand relative dates and actually get an appointment booked without hitting any confusion throughout these dialogues. This is exactly where the traditional bots fell short and what makes Agentforce so unique. These are out-of-the-box scheduling actions that you can take and use immediately or continue to iterate on for something more unique later on, right? We can now see the same exchange from the customers' perspective. Let me show you how that would look like to you. All right. So this time, I have gone ahead and deployed it on the Experience Cloud site using Salesforce messaging. It will just show up to you in a minute. Let me switch. Again, this could just as easily be over SMS, WhatsApp or any other Salesforce channel. We see that our customer Lauren is now interacting directly with Agentforce. Again, no dialogue trees behind the scenes here whatsoever, right. Instead what you can is that we have a natural language conversation happening back and forth between Lauren and our Agentforce. Our agent here understands the request a time frame and simply ignores all other irrelevant information. Behind the scenes, it's still running that same plan of service. In fact, it's identifying the correct topic and ultimately running the action to return the proper time slots. Once that is confirmed, that service appointment is scheduled, and we have automated the process end to end and freed up a human agent to focus on more complex customers inquiries or use cases or complaints that can come in from your customers, right? So what about exposing Agentforce to contractor technicians? I'm sure that's a thought that has to all of you right now. So we saw a bit ago how Agentforce fits right exactly into the field service mobile app, right? So something that I most commonly hear from my customers is, what if you are utilizing a contractor network that's not using the app, then what do we do? Ideally, in that case, all of our field technicians would be working out of the field service mobile app. But if that's not possible, this is another example where Agentforce can unlock unique such use cases that drive massive productivity gains for you. Here, we can see that Agentforce has been deployed over SMS. The use case is such that I'm assigning these jobs to contractor resources. They have the ability to interact with this data over text. Agentforce have now identified who it's speaking with based on the incoming number and serves up information about the day-to-day schedule or jobs that the text has been assigned. As they communicate with Agentforce, such as letting them know they are enroute. The status is automatically updated by Agentforce, executing all of these related associated actions. But what about some of those use cases that we saw earlier. Do you recall any of them, right? So these can also be accessed through Salesforce services. I'm sure by now that is very clearly understood to you. Then you might ask me what makes this so powerful, right? So what makes this so powerful is it's truly build once and deploy anywhere concept. So basically, I have deployed it once and I can use it over and over again like you can see right now. As I already mentioned earlier that it can be reiterated upon. The exact same flows I built earlier and those same out-of-the-box actions are powering both the prework brief and the post-work summary exchanges for this contractor resource that you can see. Any other such updates are tracked as part of the service appointment itself, and all other such information will reside on our work order records. This basically saves us time of the service agent. And we can see that this time can be saved and this resource can get started immediately on the next job quickly and easily without any hassle, right? So this is how Salesforce, Agentforce can take care of a scenario end-to-end without having to involve a human intervention or to actually save the time of the service agent to focus on something that is primarily of a much more important cause. So that your service agent can focus on more complex use cases, thereby saving his time and actually focusing on more customer service and customer satisfaction, thereby having a great customer experience at the end of the day.

Thank you, Srinihita, for this wonderful demo. So now we've seen -- we've spoken a lot about the features which are either there or they are upcoming with field service Einstein. And we've seen the system in live, thanks to Srinihita. I'm just going to talk about one of the customers, ACG, who has found a lot of value with the help of Einstein for field service agents. So ACG is the second largest AAA club in North America, and it has 13 million-plus members. They provide roadside assistance, fuel services, tire services to their customers. And at an average, they solve around 6 million customer issues per year. And if we look at the nature of their work, it's mostly emergency services, right? Because typically, when the vehicle is broken down is when the customer would reach out to them. So when we started engaging with them, they said that to effectively solve the issue. The first thing that we need is dispatcher agent who would then be able to identify that which of these services are emergency, right? So like in a scenario where, let's say, there's an accident, which has occurred, then of course, that appointment has to take precedence over maybe a flat tire or maybe a fuel-related problem. So while, of course, everything is critical over here, but then how do you identify the ones, which are even more critical from this particular lot. So they wanted AI agent to first help these dispatchers in identifying those appointments and then prioritizing them over the other ones. And when the field service agent was on the way, right, to tow their car or maybe to help them jump-start their vehicle. That is the time they also wanted a brief of the entire issue and the context of that particular customer. So when they reach over there, because in most of the cases, the customers are frustrated. No, it's not because of the service, which ACG provides, it's just because their vehicle has broken down and they're stuck in the middle of nowhere. So they wanted that whenever -- even before the technician starts engaging with the customer, they should have the entire context of that particular issue, which they have been facing. And then along with that, they also wanted a brief overview of the Customer 360 profile, right, so that they can -- the technicians over there can then have a better conversation and can build that relationship and pacify the customer. So they started working with us and then they implemented our Agentforce solution. And they have seen that to begin with that and when they started the engagement, they saw at least a 5-minute reduction in terms of -- in solving the queries, solving the problems of the customer. And then they also saw that the customer satisfaction went up because now they didn't have to repeat themselves, the agents have better context. They were able to build that relationship with the customer and then solve the problem accordingly. So we had some customers in India, which are implementing this over here as well, and you could reach out to your account teams for more details on this. And finally, what I wanted to leave you with was these 7 use cases, right. So today, in the entire webinar, we spoke about these 7 use cases for field service. It started with the appointment management. So we spoke about the scheduling agent, which will then help your customers book an appointment or schedule an appointment. Then we spoke about the dispatcher agent, which will then help the dispatchers in rescheduling the appointment or finding those appointment gaps and then booking them. And we also spoke about the use cases for your field service technicians, which would then get a complete view of the work that they need to do. And when they are actually performing that particular activity, they would be able to talk to an AI agent and get help and assistance in terms of resolving that problem. Finally, the job wrap-up will ensure that all the information that needs to be fed in the system gets fed in correctly so that we get the right data and analytics, and the managers are then able to get better insight out of the work that their agents are doing. And the scheduling exceptions, right, that in a scenario where the technicians are not able to perform the work, then how would these exception handling work, right and how agents would then be able to trigger those workflows. So with that, I would like to open up the stage for any questions. While I see some of them over here, and we'll try to respond, please feel free to post your questions over here. So thank you all for attending this webinar and we hope you -- we were able to make good use of your time.