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

Hello, all. Welcome to the webinar today. Just a quick reminder that we're a publicly traded company, and please base all purchasing decisions on products and services that are currently available. With that, I will say that nearly everything that you see today is generally available. We'll call out if we're going to talk about or show any road map capabilities. And just before we get started, I would like to begin by acknowledging the traditional owners of the land in which we meet today, the Gadigal people of the Eora Nation, and recognize their continuing connection to land, waters and culture. And we pay our respects to Elders, past, present and those that are in the room today. Okay. Let's get started. So for those who haven't met me, my name is Matt Watson. I'm a Customer Service Transformation Director at Salesforce. I've been with Salesforce for about 12 years. I have pretty much worked in the contact center and customer service space for the entirety of that time and have worked in customer service and contact centers for about 20 years. And I'm delighted to be joined today by my colleagues, Maddi and Rob.

Hi, Matt, and hi, everyone. I'll take a quick second to introduce myself as well. Nice to virtually meet you all. My name is Maddi Deans. I'm a Solution Engineer here at Salesforce. I've been with Salesforce for about 4 years, but prior to that, I was a customer working in financial services as well. So excited to show you some really cool capability today, some interesting demos as well and hopefully get some questions going as well. Rob, over to you.

Thanks, Maddi. So I'm Rob Whitaker. So I lead our advisory practice within financial services here at Salesforce. I've been here for 4.5 years now, but I came from the industry. I came from the customer side. I sat in many of your shoes, working through the challenges of the industry and how to apply our solutions in your businesses. So I'm going to cover today, after Matt, just how do we apply some of the learnings and insights that Matt brings on the contact center and our latest product announcements and how do we bring all that together for you from an industry perspective there. Back over to you, Matt.

Fantastic. So yes, we've got a lot to cover today. As per the title of the webinar, we're really focusing on the impact that AI is having on the landscape within the contact center and customer service teams. And in the first part of our webinar today, we'll be focusing on effectively the front office, how we bring work in and how we can supercharge your assisted channel productivity. We'll show a demonstration of what that looks like and then sort of transition to that middle and back-office and how we can help with automating and supporting the journey end-to-end. So really excited, lots to cover. I'm going to jump straight in and get started with a little bit of research around the impact that we're seeing AI have on the service landscape. A lot of what I cover in this first session is from our 2024 State of Service Report. That's a research report that we do with thousands of service organizations around the world, around what their priorities, challenges and focus areas are. We'll be sharing a link to all of you to be able to download this report in full as part of the resources at the end of this webinar. If you're really keen to get stuck in, you can actually use this QR code as well. But before we jump in, we're going to just ask a couple of questions just to sort of level-set with everyone in the audience and get a little bit of insight as to kind of where you're from and your background. So the first that we're keen to know is how many of you work within the contact center or customer service team within your organization or are part of the teams that support that IT architecture, project management, et cetera. So you should see this pop up on your screen now for you to be able to select an option. We'll just give a few seconds for everyone to hopefully select one of those 3. Then we'll get a bit of a breakdown on the next slide. [Voting]

Obviously, we can't see you, so we'll assume that everyone's had a chance to select an option now. Okay, so the vast majority here are supporting the contact center, working in projects or technologies that support the customer service team. So that's great to know. The next question we have is very specific around AI and where your organizations are at in terms of their use of AI in the contact center and with your customer service teams. So again, we would love to get your input on this in terms of are you using it today, are you trialing it, or are you an active user of AI in your frontline teams. Okay, just a few more seconds. [Voting]

Okay, so not using AI today. It should be a lot of interesting insights for everyone then in our subsequent slides. So just to set a bit more context about some of the stats and the research that I'll be sharing before we get into our demo. We run this research every 2 years. And it's not necessarily Salesforce customers that we're surveying. We surveyed over 5,500 customer service professionals, so executives, decision-makers, frontline agents, people who support the contact center. And they could be using any suite of frontline tools, whether that be Salesforce or any other servicing platform. And over 400 of those were from financial services organizations. So we have some stats here that are filtered very specifically to what we've heard from your peers in financial services around their priorities and challenges in the customer service space. We also ask everyone that responds to self-evaluate whether they see themselves as a high performer, a moderate performer or an underperformer within customer service. Interestingly, when we first ran this survey 10 years ago, we run it every 2 years, we first did it back in 2014, we had almost 1/3 of respondents back then classify themselves as an underperformer. And as we've seen a shift away from siloed channels to focusing on the customer, there's been this consistent improvement in the way that people see themselves as they've been able to focus on a more connected customer experience. And in the full State of Service Research Report, there are 3 key priorities that service organizations are focusing on. We're not going to talk about all 3 today. For those of you that came to Agentforce World Tour event in Sydney a couple of weeks ago, I led a session where we focused a lot on priority 2. But today, we're going to lean into the third priority around the investment that organizations are making in automation and AI. And to kick that right off, a stat that is really surprising to me -- I mean, I guess not surprising from doing lots of ride-alongs but surprising that it's so high, is that agents on your frontline contact centers are spending nearly 2/3 of their time not focused on the customer. So that's doing things like manually taking case notes or after-call work, internal meetings and training, copying and pasting between lots of different platforms. And all of that is not spent in the service of the customer, right? You put the customer on hold or you're in your after-call work before you've picked up your next call or next e-mail or next chat. It's fascinating to look at this and go, that 61% of time is a heap of agent capacity that could be unlocked to allow you to focus on the customer more. And part of what's driving such a high stat, and particularly within financial services, is the average frontline agent in a FINS contact center has 10 technologies that they're navigating between to solve the customer's request. More than half of all of the financial services agents that we surveyed have to toggle between multiple screens, copying and pasting data to be able to get information to do their job. The session that we ran at World Tour that I mentioned earlier was focused on the revenue generation priority. And what was really interesting there is that most organizations have trained their frontline agents on how to upsell and cross-sell when they're servicing their customers, but less than 1 in 4 agents are able to actually effectively do so. And again, some of the challenges shown on this slide are a reason for it, but another big driving factor there is the lack of cross-company data. So again, really common, and you see the difference here between high-performers and underperformers in the scale. But those organizations that aren't doing it as well haven't connected their service teams with their organizations' sales and e-commerce interactions, the marketing campaigns that customers are on, or even interactions from other channels within the contact center, right? So can a phone agent see all of the messages or chats that your customers have had with you? Do they have that full visibility of customer interaction across channel? Because what we are seeing is that the type of work that's coming into the contact center is getting more and more complex. Customer expectations are rising. There's often more products and services that are being supported. And that classic contact center business case of do more with less is as prevalent as ever, right? Deal with increasing case volumes, more complex case volumes, increased customer expectations, but we're not going to grow the agent workforce or we're going to shrink the agent workforce, so a huge focus on efficiency at a time where there are already some challenges in terms of how we're able to effectively focus on the customer and their needs. And AI investments are increasing. So going back to that poll question earlier, we do see that amongst those organizations that rate themselves as a high performer in customer service, nearly all of them are investing in AI within their customer service teams. Only a little over 1 in 2 in the underperforming segment are. Across the board, the vast majority are planning to increase their investments in AI in this coming year. And part of that is, as they're seeking those efficiencies, AI does, of course, promise a lot of scale. We see -- and I'm going to show a bigger version of this soon, but lots of organizations have either fully implemented or are experimenting in AI within their customer service organization. And the top use cases, interestingly, aren't about trying to tackle really complex problems, right? They're left for those human-to-human interactions where you need that level of connection. The top use cases that we're looking at, and I'm going to expand this actually to look at the top 6 across financial services, the first 3 of them are all about supporting the agent: how can we provide assistance that support the agent; how can we quickly automate conversation summaries, call summaries, case wrap-up so that the agent is not having to manually type that; how can we quickly generate service responses for the agent. Then we get to the customer-facing use case around an intelligent assistance and then things like intelligent offers, next-best action, next-best recommendation, et cetera. So these are the top 6 use cases that we see service organizations in financial services investing in AI. I'm curious as to whether or not these use cases are ringing a bell with anyone on the call. Now the previous question around not having implemented AI may mean that the answer to these are no. But from an interest perspective, are you considering any of these use cases in your organization in the way that you want to make your frontline more efficient? I'll just pop this one up on the screen now. Hopefully, some of these use cases are relevant to what you're looking at doing. [Voting]

And let's jump in and take a look. Okay, so automated summaries and reports is the highest response there, nearly 44% of respondents are looking at that. And that's certainly an area that we see a lot of organizations starting with, particularly things like call wrap-up and the like. Agent-facing, intelligent assistance, knowledge article creation sort of round out the top 3. What's interesting with those use cases is that it very much aligns to where we see the future of service. And I use this slide a lot with customers that I engage with because I really want to stress that we don't see AI replacing the human frontline agent. We see that AI is about unlocking that human potential, freeing up that agent capacity, that 61% that we showed earlier, freeing that 61% up. So automation, self-service and AI, taking those simple one-and-done tasks and being able to take them off the human so that they can focus on those interactions that are higher in complexity and add more value both to the customer and to your own organization to try and meet those increased expectations that customers have. So in order to free up that agent capacity and unlock that human potential, it sort of comes in two areas. How can you automate that simple work that doesn't add a lot of value and allow agents to work faster and therefore, be more efficient in your frontline operations? And for those things that are perhaps more cognitively difficult, like reading 5 pages of messaging transcripts to figure out where the customer is at in their conversation, how can we use AI and automation to augment the human experience and allow them to work smarter and then driving and improving that agent productivity? And so when we think of AI, particularly when we think of AI within a customer service environment, everyone loves to focus on one use case or generative AI as being the buzz, but we very much see it as a continuum, right? And we're going to showcase a lot of these capabilities today. But there is value in the whole spectrum and life cycle of a customer conversation, right? Even traditional predictive AI, where you can save 4 or 5 or 6 clicks for an agent in classifying their interaction or classifying their case, the way that you can use CRM data and AI to make a more informed and smart routing decision or place an autonomous agent at the front of that conversation to either solve the customer issue or at least triage it, becomes incredibly important. And then when things do come through to an assisted agent, following that continuum through to the generative AI to be able to summarize conversations and generate service replies, things that agents can select from a library of content that agents can select from to speed up the way that they're engaging with the customer and, again, free up that capacity for the agent to spend more time focused on what the customer actually needs. So we don't have one single AI to rule them all, right? We look at these capabilities as a continuum, and we bring a lot of them to bear in the agent and the customer experience to really drive and enhance that frontline capability. And to bring this to life, I'm going to hand over now to Maddi, who's going to showcase a demo of this frontline experience through a customer scenario. Over to you, Maddi.

Thanks, Matt. All right. We're going to take a quick look at how we can drive productivity in a contact center through connected data and AI. We're going to focus on reducing that 61% of time that our customer service representatives aren't able to focus on the customer. And also, we're going to look at how we can use AI to detect signals from our customers and uncover some of those revenue-generating opportunities in your service conversations. To do that, we're going to follow along the journey of Julie Morris, who is a long-time customer of Cumulus Bank. She's been waiting for some money to be refunded to her account following a fraudulent transaction. She starts on the Cumulus website, but she can't get the answer that she needs, so she turns to her preferred engagement channel, chat. Now when she starts the conversation, she's greeted by the dreaded chatbot, the chatbot, a gate put up by Cumulus to prevent all but the most desperate of customers from getting help. But fortunately, this is not a chatbot. It's an Agentforce service agent, our customer-facing intelligent assistant whose job isn't to deflect but to help. Julie is going to go ahead and dutifully let our agent know what she's looking to do, and our agent correctly understands and classifies this request. Because of the type of request, it's asking Julie for some additional verification information. But once the agent is satisfied, the security questions have been answered, it's able to offer up an answer to her question based on Julie's profile and her open transaction dispute case. Importantly, this information is grounded in the company's policies around when she can expect to receive that payment. Now Julie is pretty pleased with the answer that she's gotten here from the agent, but she decides she's going to push the boundaries a little bit and asks whether anything could be done to speed up this process. Now the Agentforce service agent, unfortunately, has to say no to Julie, that no, it can't do that, because it's been configured with guardrails which strictly prohibit it from deviating from policy. Now Julie is going to drop her guard a little bit here and mention that this whole fraud issue has been quite stressful for her, and she wants to know if there's anything that she can do to protect her account in the future. And our agent, sensing that Julie is feeling a bit anxious, offers to help but also offers to put her in touch with a human as it's sensing that, that may be the next best move. In this case, Julie lets the agent know that she would, in fact, rather speak to a person. So our agent is going to go ahead and transfer the chat to a customer service representative after collecting a little bit of feedback on how it's performed today. Now let's jump into the Financial Services Cloud service console to see what this looks like for our customer service representative, Tim, when he picks up the chat. As a customer service representative, Tim uses Salesforce as a single pane of glass for all things relating to servicing his customers. And he starts by making himself available to receive work through the various engagement channels that he's trained on. This omnichannel capability is incredibly powerful. All demand from all channels can be presented to your customer service representatives in exactly the same way. It doesn't matter if it's a phone call, a chat or messaging conversation, a social media post or any other piece of work. And here, we can see Julie. Her conversation has been assigned through to Tim. Now you'll notice that when Tim accepts this chat, he's presented with the tools to engage in this conversation, but he can also see Julie's full 360-degree profile, which has been launched for him, giving him everything that he needs to be able to handle her query, whatever it's about. We have access to Julie's previous and open cases being managed through Salesforce and we can see her bank account and transaction data, if we need it. This is data that doesn't live in Salesforce but is being surfaced up via a zero-copy integration through Data Cloud. We can see recent engagements and marketing data. And because it's one unified CRM across Cumulus, we've even got access to sales data. But let's go back to Julie's chat and see what she's after. Tim is presented with the whole history of the conversation Julie has had with our Agentforce agent, but we also get an AI-generated summary to help him quickly get his head around the conversation without needing to read through the entire conversation to get up to speed. This is an example of the automated summaries and reports use case that we discussed earlier, and it's a great example of accelerating customer service representative productivity and allowing them to focus more quickly on the customer. Now let's introduce ourselves to Julie and just find out a little bit more, make sure that we've understood. When Julie replies, you'll notice that these AI-generated conversation suggestions are updating in real time, offering Tim some options on the most relevant and meaningful thing to say next. These recommendations are grounded and based not only in the conversation that we're having with Julie, but they're also grounded in Cumulus' knowledge articles, helping us to ensure that responses align to the policies of the company. Importantly, these requires a refresh to each step of the conversation, and this is to ensure that they remain relevant to where the conversation is currently at. We've also got access to in-built messaging components and can help Julie out by sending her a link to one of our handy guides about protecting your account online. Julie is very thankful, and she mentions that she's been worried about her money and her card because she has an overseas trip coming up. And this is where our AI, based on the conversation with Julie, is able to pivot. But it's also surfaced a next-best action for Tim or an intelligent offer. And this isn't a generic sales action. It's been surfaced up because the predictive AI has detected that this might be a conversation about traveling overseas, and Julie doesn't have insurance included in a Cumulus credit card. And now it can guide Tim through exactly what he should say to do and qualify and nurture this potential revenue-generating opportunity. This next-best action isn't just a piece of text. It enables this customer service representative to take action right in the flow of work here in Salesforce. And if Julie replies that this is actually something that she's interested in, we can, of course, continue the conversation from there. But in the interest of time, we won't go through that whole process. The point is that we can use AI and guided workflows to detect these revenue-generation signals, subtly surface those to our service representatives and then guide them through the sales process as well. Let's wrap up the chat there and let Julie go. The last thing that you'll notice is that as soon as this conversation ends, we automatically take that chat transcript and use AI to generate a summary of that conversation, including the reasons for the conversation and the resolution proposed by the agent. This again shaves off more of that noncustomer-facing time and enables your service agents to spend even more time engaging with your customers and focusing on their needs. That's all from me for the minute. Hopefully, that sparked a few ideas for everyone. Matt, do we have any questions?

Thank you, Maddi. And yes, we do have a couple of questions in the chat, and I will get to them in a second. Just wrapping up what we just saw, I showed this slide before the demo and what we were able to have, working both in the background and very visibly in the demo, is a combination of a lot of these features all working together. The way that we front-ended the conversation with our Agentforce agent routed to the best available agent for that type of customer, and you saw both the conversation summaries not just used at the end of the conversation but also to allow the agent to catch up on the conversation that's already happened, so multiple uses of that; and obviously, the service replies dynamically adjusting as the conversation progressed. Very relevant to one of the questions that's come through about the impact of frontline AI tools like this on agent productivity, I've been working with customers where their standard after-call work or after-conversation work time is in the minutes, 5, 6 minutes sometimes, that agents are expected to manually capture what the conversation was about. I've worked with organizations where I've seen that decrease from 5 or 6 minutes down to 30 seconds on average because not only that summary that you saw at the end of Maddi's demo is being used, but the agent just needs to quickly skim through it, make sure that it's an accurate summary, endorse it and then move on. They're not having to manually go back and read the transcript themselves and make sure they've captured everything there. We've also got a question here around can we provide summarization on a real-time basis as the conversation goes between the agent and the customer. Great question. And yes, absolutely. We do that very deliberately at points of handoff. So when the chat or the message came through from the agent, from the autonomous agent through to the human agent, that's a great point where you need to summarize the conversation as it currently is for the agent that's receiving it. So we provide what we call a conversation catch-up then. And we also provide it to supervisors and team leaders any time they want to go in and see a conversation that's in-flight that one of their agents is working on. So summarizing the conversation halfway through or partway through is absolutely a key use case, but it makes sense to do it at the right time, right, at that point of handoff where a new human is being brought into the conversation to understand where it's up to. Okay, we have a couple more questions that have come through the Q&A. We will absolutely get to those a bit later in the webinar. We're going to hand over in a sec to Rob. But the capabilities that are shown on this slide that we just demonstrated also touch upon 4 of those 6 use cases that are the top priorities for financial services organizations in their contact centers. So we showed the automated summaries and reports. We showed the dynamic service responses that change at each step of the conversation. We obviously showed Agentforce on the front of that conversation with the customer. And then subtly, that next-best action popping up at the end picked up on a buying signal that perhaps wouldn't have been obvious to the agent. The customer didn't come out and say, "I want a new home loan," right, or, "I want to buy insurance from my credit card." They just said that they were traveling overseas and the system was able to leverage that connected data to surface that there's an upsell opportunity here. So 4 great use cases that can all be applied within a single conversation with a customer. But of course, once you've actually got the request from the customer, particularly if it's a complex request like a fraud or transaction dispute, there's also AI you can apply to how you actually solve that customer issue as well. So for the second half of this webinar, I wanted to hand over to Rob to talk us through how the Agentforce journey applies in the broader financial services context. Over to you, Rob.

Thanks, Matt. Great insights on what's happening in the industry and the impact on service. And great demo, Maddi, just bringing to life what Agentforce can be for our customers. So I'm going to spend the next little bit just really talking through how we're responding to the change in the industry, and how we're trying to leverage AI, in particular, to really deliver this agent experience, which we announced at Dreamforce about a month ago now under the banner of Agentforce narrative. So before I sort of start, like many of you that have been on these webinars before or many of our events know that we've been in financial services and supporting financial services institutions for 25-plus years now. 10 or a few more years ago, we actually started building industry-like solutions. So we started in the wealth space. We went through to banking and insurance and really making sure our solutions are fit for your purpose and your industry. The feedback we got about 5 years ago was, yes, that's great. We've got a great data model. We've got some great processes from a salesperson. But how do we solve the service experience along the way there? And we've really been investing heavily over the last 5 years, building out the service capabilities of Financial Services Cloud. One of those capabilities has been the workflow and automation capabilities through our acquisition of Velocity, which is now fully implemented in there, and building out prebuilt processes to have a prebuilt service library along the way, all built on the same platform we all know and love, all with the same security, trust and compliance that comes with the platform, and then making it available to a partner ecosystem so they continue to expand the capabilities of our platform into origination solutions like nCino in Q2, into onboarding KYC solutions like Finovo. So in that time also, we've also been investing in AI to support financial services. And we've seen it come through a number of waves. Back in 2016, when we introduced Einstein, the product, it was all about predictive, how do we give recommendations, next-best action, lead scoring, those recommendations to the agent or the banker experience. Late last year, we started to see the introduction of Copilots. So Copilot is augmenting the experience of the agent to really help them on tasks, like summarize some information, generate some content, to help them through that using the generative AI capabilities. And as our stuff has evolved in this rapid transformation in AI, we're now entering that third wave into agents. But before I sort of talk through agents, I thought we'll bring up another one of those polls around generative AI. And I know that the earlier question of people that are using it today, we might get some mixed results here, but really keen is how is generative AI perceived at work? Do you have any uses for it in your business? Are you using it? Are people encouraging you to use it? Or are they encourage you to discourage it? Or there's this sort of restrictions in your organization around AI tools. So let me just bring up the poll for you for people to provide their responses. I'll give it a few seconds just to collect those responses. [Voting]

Cool, maybe we don't have a challenge on the collection there. So maybe I'll move on from that one there. So as I said, we sort of had those first 2 waves. And now we're entering the third wave of AI, and that's about agents. So agents is all about providing that autonomous and action off the back of it. The first 2 waves are really giving you information or helping you out with something. Agents is about actually doing some piece of work on behalf of a human, taking the human out of the loop or really assisting a human in doing an action along the way there. And that's where we really see the focus of AI now shifting into this third wave. And when we talk about agents, they really unlock an extraordinary experience and drive ease of use. So if you think through today in a service or agent service center, it's all very manual tasks. You're copying and pasting stuff between solutions. You're completing lots of manual processes. Some of organizations may be using some AI to experiment, maybe more in the predictive than the generative side. And they're very fragmented experiences. Like Matt said at the front, the average agent in the contact center is using 10-plus screens. I know in my old organization, it was over 20-plus screens to do a work. And you're having to pull data from everywhere across the organization. And often it's out of date, not timely along the way. With the introduction of agents, you'll have them perform the human assistance. So you start to be able to drive hyper-personalization because you'll be able to have that data, giving recommendations of not only just the data but what's in the context of the work you're doing. You'll have digital agents be able to assist you in tasks. So think about rather than send it off to a back-office function to follow up and pull some stuff together for you to get some information that you can then complete the next step, having an AI agent to do all that for you 24/7 in near real time, it will have predictive responses, and it will give that instant issue resolution rather than relying on multiple teams and handoffs between teams along the way. So that's sort of where we see the value opportunity for agents. And how we see that applying in financial services, the use case is endless. I think the learning that we came all of the way from Dreamforce is the use cases. And the ideas that sparked in customers' minds, and people like yourselves, is what are all the different things in my business that I could start automating or having an agent assist, whether it be streamlining the onboarding processes in banking, helping a wealth adviser complete advice or execute advice faster through to insurance about simplifying the claims process, or in IT, just trying to modernize and simplify the solutions along the way. Those opportunities are endless. At Dreamforce, we had 10,000 agents built by our customers, sitting down at a desktop, articulating a business problem, and then having that agent built for them on the spot. And with that, all that agent element leads us to what we refer to as Agentforce for financial services. And Agentforce, for us, it all sits on the common Salesforce platform, that Marc and Parker established 25 years ago, which has all the security, the trust, all of those platform capabilities around flow and objects, all of that available there. The Data Cloud sits in the middle. That allows us to connect the platform to all the various data sources in your organization, leveraging capabilities like zero copy. So you no longer have to copy all your data into Salesforce, we can access that data in near real time from your Snowflakes, your Databricks, from your Google BigQueries. And then on top of all that sits all of the applications that many of you probably know and love, our Sales Cloud, our Service Cloud, our marketing experience, all the way through to our commerce experience. And what we've done and announced at Dreamforce is we're now amplifying that experience to now provide agents that can interact with that experience and perform tasks within those applications, using that data you connected through in an automated fashion, augmenting that experience that everyone knows today to keep that productivity improvements and unlock that productivity to drive revenue growth for your organization. So how is AI redefining service? It's really changing it from a world which is very process rule-based, very prescriptive world, to a more conversational world, more intelligent world, using intelligence to complete the processes rather than a human having to interpret those processes and apply that contextual thinking. You're moving a world that is a pretty much 100% human-driven world to one where you still have got your humans and agents, but they're supplemented by digital agents to take those lower-value tasks, those tasks that can be automated across the world. And then moving it from seen internally as very much a cost center to through the productivity really being a profit center, freeing up time so agents now can have that true conversation around what next, how can I help you next, driving customer advocacy, which will obviously lead to NPS and other relationship benefits along the way. So we've talked a lot about agents and Agentforce. Just let me just level-set what is a digital agent along the way. So like when you hire an employee in your business, an agent is no different. When you hire an employee, you want to give them a role. You give them a responsibility, what are they actually going to do for your organization, so answer calls, respond to queries and process loan applications, assess claims. So you always have a role and the digital agent is no different. Also when you hire someone, you give them access to your organizational data or a subset of your organizational data. You will go, this agent has access to this customer set or this product set or these sets of policies, and that is the data with which that agent will work on. You also give them skills and responsibilities for what they can do. The agent can complete a complaint, a claim or it can process a credit card increase up to a certain level within the guardrails of processes and policies. So you would assign in a digital agent those skills, those flows, those processes they can attach to along the way. And you also say where they're going to work from a channels perspective. So you'll say, "Well, this agent is going to answer voice calls," or "This one is going to answer web calls," or "This agent is only going to focus internally and only support team A, B and C." So those are the elements that make up an agent. And importantly, they all need the trust and security to protect your organization. And that's how we ascribe an agent and think of an agent when it comes to Agentforce, which is what Maddi demonstrated before. And importantly, for financial services, we ground all our agents in the foundations of trust. Trust is our #1 value here at Salesforce. And when we started on this generative AI journey 18 months ago, it's the very first thing we put in place, put the trust framework in place. We made sure we've got the audit and reporting of any action that's done by AI within our platform. So you can go back to your regulator, your Audit Committee, your Risk Committee and be sure that you've got controls in place there. We put in what we refer to as the Einstein trust layer, making sure that we've got zero data retention, making sure we mask and protect your PII data, making sure we don't hallucinate and give results that are incorrect or invalid along the way. We've also then put guardrails around what an agent can do, what topics they can actually work on, what actions they can perform within your business. So like Maddi showed in her demo, they're only doing what you've let it do. If it goes outside of that, it knows it can't go outside of those areas. And then we've been very focused on the accuracy and the disclosure, so making sure that we have citations, making sure that it's grounded in your data, the prompts are providing that accurate information. And we did launch a LLM comparison accuracy benchmark and many of the models and capabilities that we show in Salesforce are at the highest levels of those accuracies. And we always make sure we've got the disclosure. We're always clear that here are the audit, here's the toxicity, here's the feedback along the way there. So that's sort of agents, the trust layer, and then finally, just sort of to touch on this is, like generative AI, not everything is agent, not everything is predictive or generative. We really think about, when it comes to financial services, AI for CRM across 3 key sort of use cases. The embedded AI, these are the prebuilt out-of-the-box AI capabilities that come with the platform, that perform common tasks, which you can just turn on and start injecting in your processes, such as please summarize this call, summarize the relationship or summarize this account, generate me a letter of credit. The assisted agent is providing that in-app experience of that conversational interface that you can start interacting and asking an agent to perform tasks on your behalf, such as, "Help me prepare for this meeting." It will go away and summarize and pull all the recommendations, information together for you for that meeting in a briefing note example. And those who are at Agentforce World Tour Sydney, you heard CBA talk about some of those examples of how they're exploring using those briefing note capabilities. And then you've got autonomous agents, which is what Maddi showed earlier, where agents are actually operating autonomously and taking responsibility for executing an action on behalf of a human, verify a person's balance, update a payment date, stop a check payment or stop a process along the way there. So that gives you a feel of how we are thinking about AI and our product, and all the new capabilities have now been launched as part of Dreamforce. I'm going to hand back to Maddi now just to bring just some of these other use cases to life in a bit of a demonstration. So over to you, Maddi.

Thanks, Rob. I'm excited to see this in action. Let's jump now into another quick demo. Now we're going to dive into the back-office at Cumulus Bank. We're going to see exactly what the experience is like for our customer service representatives who are handling those more complex cases. But first, a little bit of context setting before we get started. Rachel is a Cumulus Bank customer who has just moved to Sydney. She's been busy exploring, building new friendships and getting settled into a new job, which is all very exciting stuff. But moving to a new state and all of the life admin and change that comes with that can be pretty stressful. And to add to this already busy time, poor Rachel has just received a text from Cumulus about a questionable transaction. When she takes a look, unfortunately, it's a transaction that she doesn't recognize. So she goes ahead and submits a transaction dispute. Understandably, though, her anxiety level is starting to go through the roof. So let's get this resolved for Rachel, so she can go back to enjoying her new city life. Back at Cumulus, a case is created and routed to their fraud investigation team. Our customer service rep, Scott, has a clear line of sight not only into this case but into Rachel's full relationship with the bank. On the left-hand side there, we can see information like CSAT and NPS, and all of this information is being pulled in from Data Cloud, along with Rachel's full service history. All of the important information is packaged front and center here in the dispute record. We can see that Rachel has an unusual charge of $222 from Ampol in Raymond Terrace. And this has been flagged on her rewards card. Now this is really curious. With all of this context, let's put on our fraud investigation hats to see what we can uncover. I'm sure you can see the milestone that's ticking away in the top right-hand corner as well. That's the service level agreement, or SLA, for Rachel's case, and there isn't much time for what could be quite a complex job. Thankfully, Scott has an assistive agent ready to help. He can ask the assistive agent to summarize all of the details of Rachel's service needs with Cumulus. So let's go ahead and do that now. We can see that Rachel hasn't had any transaction disputes in the past, but she did replace a lost card. That's a pretty standard event, but it will help our agent to identify potential trends that are important to the investigation. Our agent is also providing Scott with guidance on the next steps and is suggesting that we might want to add Rachel to a marketing journey to keep her in the loop about how Cumulus is handling her case. Scott's first task is done, but the clock is still ticking. Let's go ahead and mark that milestone as complete. Now researching a transaction dispute is often a tedious task that could take hours or sometimes even days. But I'm sure you can see here that Scott doesn't have that long to update Rachel's case. Let's jump into her rewards card account and investigate the transaction trends further. To help with this next stage of our investigation, we're going to leverage embedded AI, specifically prompt builder and intelligent field generation. With a single click here on our AI-enabled field, we can activate an embedded prompt that pulls all of Rachel's relevant transactions and populate them here for us to review. Now this is much too much detail to be immediately useful. So let's use natural language to filter this list into something more succinct and actionable. We can ask the AI to summarize these trends using location, using amount and maybe date and time as well. Let's see what that gives us. All right. Now this is where the magic really starts to happen. We can see that Rachel appears to always fill up at the Ampol Foodary near her in Manly. And generally, the transaction is up to about $90. We have a few other transactions, but I think we can start to see a theme here. With all of this detail now nicely formatted into a trend report for us, we can save that straight against our investigation record. With the investigation wrapped up, our agent is also able to help Scott compare the data from Rachel's trends to the disputed transaction that we're looking at today. So let's ask our agent to do that for us now. We're going to ask it to compare the historical data to the transaction that we're looking at today. And maybe we'll ask it how we should proceed as well. Now Rachel's spend at Ampol usually tops out at about $90, which suggests that maybe Rachel drives a smaller or more efficient car. $222 in petrol seems a bit out of the ordinary. And as we saw before, she seems to have a habit of filling up just near home. So a transaction that took place about 150 kilometers away is also pretty unusual. From this, our agent determines that this transaction is likely fraudulent and guides our customer service representative, Scott, through the next steps. First of all, the agent is recommending that a provisional credit be issued to Rachel based on our findings but also considering Cumulus' policy, which the agent is referencing via Data Cloud. Scott can do that in a few simple clicks using this action here, which leverages all of the automation tools already available on the platform. And with this, we've helped Rachel get her cash back so she can get back to sipping lattes or maybe sipping cocktails. And because we've identified that this transaction is likely fraudulent, Scott is also prompted to escalate the case to Cumulus' customer protection team. That brings us to the end of our demo. Hopefully, you've enjoyed seeing how our customer service rep, Scott, together with AI and an assistive agent, have been able to help Rachel in her time of need but also how they've gotten this case to the customer protection team in record time to ensure that we're protecting all Cumulus customers.