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

Hello, everyone, and welcome to today's session, Agentforce: Humans with agents empowering public service. Thank you all so much for joining us today. I am Richa Rajput, and I am a Lead Solution Engineer at Salesforce. Before we begin, I'd like to share a few quick notes about our webinar platform. Please don't worry about taking notes from this webinar. We will e-mail the recording to you in a few days. Alternatively, feel free to use the same link that you used to register for this webinar to gain access to that on-demand video. You can enlarge the slide or media player too by clicking the square button that's located in the top right-hand corner of the widget. If you require any technical assistance, please use the chat functionality and we'll do our best to respond as soon as we can. Once again, thank you for everyone who's joined us today. I would like to begin by acknowledging the traditional owners of the land on which we meet today, the Gadigal people of the Eora Nation, and recognize their continuing connection to land, waters and culture. We pay our respects to elders past and present. During today's presentation and the demo, you may see some capabilities that are on the Salesforce road map. If you are making any purchasing decisions, please do that based on the products that are available today. And I'd like to start by firstly saying thank you to our customers, partners and public sector workers. Thank you for tuning into the webinar today and to those who will be watching the recording. In the session, you are going to see how humans and autonomous agents work side by side and help government do more with less, using our deeply unified Salesforce platform. So today, we are going to explore this unique opportunity we have today to empower public service with data and AI and AI agents that can provide personalized service and support, expanding equity and access to information and benefits that enhance government efficiency and public trust. With that, let's first see some of the challenges that are faced in public sector today. So every single organization leader is thinking about how they can integrate agents into their business. We know that 41% of time is lost to low-value and repetitive tasks. Workforces are overwhelmed with low-value tasks that stall productivity and lead to burnout. Meanwhile, citizens' expectations continue to rise. Zero hold time, they want personalized and empathetic service and would like to work with an expert. So organizations are stuck in fixed capacity. So we wanted to dive deeper to understand more our customers and the citizens' perspectives on the use of AI to modernize and transform service delivery to really identify where the opportunity for impact lies and what's possible. We just released our second edition of our global connected government report, and that reveals those insights. So I'm going to share some of these insights and what's really interesting there. So what we found in the report is that government experiences are held to private sector standards. The public expects government digital services to be as user-friendly as those offered by the private sector. For many, positive experiences matter as much as the services themselves. Poor navigation and self-service options on public sector digital channels, if digital channel options even exist, fall short of citizen expectations. The public sees AI as an opportunity for government services. The public sees AI as a chance to improve government service delivery. Overall, the comfort with government use of generative AI is positive with use cases that enhance access and equity being the most popular. At the same time, the public looks to government to provide guidance, policies and safeguards against misuse. The public-facing government employees are tasked to do more with less. So government services -- service workers' #1 priority is improving the public's experience. At the same time, they are overwhelmed with rising workloads and increasing expectations with limited resources. So the public wants a personal touch, but service workers are often don't have the bandwidth or the tools to meet the modern demands. And finally, the technical teams modernize for a digital-first agentic AI era. Government IT teams are racing to modernize in the face of an increasingly digital-first public and rapid innovations in AI like agentic AI. 83% of government IT workers say they must do more to address the changing customer expectation. So let's take a second to understand doing more with less a little deeper. With rising expectations from the public, the dynamics of customer service are shifting. Workers report that people are more pressed for time while demanding more personalized attention. So this dual mandate creates a challenging environment for government service teams who are already strained by rising case volumes and limited resources and envision their case volumes to only increase with directly conflicts with public appeal for a more personal touch. So what if public sector workers had no limits and you could augment and scale your team with the digital labor force. Introducing Agentforce for public sector. Autonomous AI agents are here to strengthen and reinforce your greatest asset, your people, so that they can focus on high-touch, high-priority mission-critical work. At Salesforce, we pioneered enterprise AI and are comfortable shifting into this new wave of agentic AI, which is the third wave. And our deeply unified platform allows us to scale the use of agents while remaining focused on the Customer 360 and solving real-world problems for our agencies and achieving the right outcome for our citizens. Our research has found that by automating administrative tasks, providing 24/7 support and personalizing the service, government employees can take advantage of agents to quickly make sense of large amounts of data, thus freeing up time for more complex, high-impact tasks. Agents use conversational interfaces that are intuitive and easy to adopt and do not require extensive training. For the public, agents provide an easy way to interact with the government organizations, promise a 24/7 service and a seamless experience, whether it's filing taxes or requesting road repairs, tailoring that experience for the citizens. So some of you may have seen these terms before, such as chatbots, Copilots and might be wondering and comparing that with agents. So I just wanted to quickly touch on this as well. So simple chatbots have been around for a while. They were good at fixed rules and repetitive tasks. Then we have copilots. Now they are way smarter and provide more tailored assistance than chatbots. Now we have autonomous agents or Agentforce, and is different to these 2 because it can reason and it can take action. Agents can engage in natural language, and they can retrieve data on demand and they can build action plans and execute on those action plans without intervention. Most importantly, they can understand the guardrails on when to escalate to a human staff as well. So how can you onboard your agent in 5 easy steps. The first is the role, which is the agent's purpose. This defines the job to be done and the broader goals the agent should achieve on your team. The next is data. This refers to the knowledge an agent needs to be successful. This could include your company knowledge, articles, your knowledge base, CRM data, external data via data cloud, your public websites and so on. So think of structured as well as unstructured data. Actions. These are the goals an agent can fulfill. This is predefined task an agent can execute to do its job based on a trigger or an instruction. So for example, it could run a flow, a prompt template and take any actions. Guardrails, most importantly, are the guidelines and agents can operate under. These can be natural language instructions, telling the agent simply what it can do and what it cannot do and when to escalate to a human. And then finally, we have channels. So the applications where an agents get the work done. This can be your website, your CRM, your mobile applications, Slack and more. So most importantly, all these 5 components are supported by built-in security features in the Einstein Trust Layer. While the benefits of agentic AI are clear, the public sector must also navigate the delicate balance between security, innovation and ethical use. In short, the trust of its constituents and citizens is paramount. Good news, trust is also our #1 value here at Salesforce. Since we started our journey into the building of AI-enabled services over 10 years ago, we have had at the forefront of our efforts a focus on the ethical and trusted use of AI. We have built an entire trust layer into the underlying AI platform that automatically handles bias and toxicity, implements strict prompt defense and ensures your data is never used to train AI models. Of course, your data is the key to a trusted AI agent. Agentforce is grounded in that data in your CRM, data lakes, websites and static policy documents that are the cornerstones of the services that you provide to the public, always ensuring that any answers that are provided to the questions are from the knowledge you control. So the impact of AI and automation is not just theoretical. It is already being realized in various government operations. So I just wanted to give you some examples of the Agentforce use cases we've seen in public sector. We've seen in child welfare, AI can expedite access to critical information, provide case summaries and offer policy and process guidance, ensuring that decisions are made faster and more accurately. In public safety, AI can generate incident report summaries and provide procedural recommendations, reducing the time needed to complete these tasks. Social insurance processes can be expedited through auto adjudication and diagnosis code mapping, leading to quicker and more accurate service delivery. So hopefully, that gives you all the context around Agentforce. So now I'd love you to see Agentforce in action considering a public sector use case. So in today's demonstrations, we are going to take you through 2 personas. First is from the point of view of a citizen who is in distress after some changes to his working arrangement. He then visits a government website and is exploring and looking for the assistance in his scenario. So we'll see how Agentforce converses with the citizen in natural language. It's able to gather the information and suggest potential benefits based on the information that the agency already knows about that citizen and then can create an application for rental assistance. A government employee will then pick up that application to view that citizen's unique -- unified profile and then use Salesforce to determine the eligibility for rental assistance and how much the citizen is going to be paid. So with that context, I will first introduce our first persona, Adam. He's a citizen. Adam lives in regional New South Wales with his wife and 2 kids. And until recently, he worked at a local travel agency. However, the company is downsizing its physical store presence and Adam is a sole breadwinner and he's found himself out of a job. Next, I will take you through the demo video. So you'll hear the voice of my colleague Rasika and Swathi to take us through the demo video. So we'll start with Adam, citizen, who's exploring the benefits the government provides on a benefit finder portal. So I'll just roll that video now.

Now relying on their savings to make ends meet, Adam knows he needs to explore his options. With the job market tight, he decides to investigate government assistance and logs into the familiar government benefits portal. This time, though, he is greeted by a new feature, a benefit finder agent. Adam begins the conversation explaining that he's here looking for assistance. To verify Adam's identity, the agent requests a few details like his driver's license, e-mail and date of birth. Once Adam provides these, behind the scenes, our autonomous agent seamlessly validates this information. Once confirmed, the agent asks Adam what type of benefits he's most interested in. Given his current financial strain, Adam replies that he's looking for financial assistance. The agent then presents him with a list of potential financial benefits. Seeing the options and considering his rent stress, Adam selects Rent Choice subsidy. The autonomous agent then begins a tailored series of follow-up questions to determine Adam's eligibility for the subsidy and the potential amount. Notice how these questions are personalized based on information the department already possesses. For example, the agent asks Adam to confirm his weekly income. The data is drawn from Adam's unified profile powered by Data Cloud, which has ingested and harmonized data from various agencies. Adam answers the questions and the agent responds indicating that he appears to be eligible for the subsidy. However, to finalize this process, a case is going to be created for a benefits officer to review Adam's details and formally confirm his eligibility.

Now let's meet Claire, a dedicated Benefits Management Officer with the department. For the past 3 years, she has been passionate about connecting citizens with the support that they need, whether that's a social housing benefit, a rental subsidy or any other type of benefit and service that the government provides to its citizens. Today, we'll see how Salesforce empowers Claire to efficiently process Adam's rental assistance application. Claire's personalized homepage provides a clear view of the work that the department has been doing. Dashboards display key statistics on the benefits that the department has been able to provide to the citizens in strife today. Previously, what used to take weeks to process and respond to now has an average processing and payment time of 1.2 weeks, thanks to the prework and the assistance that our agents are providing the employees with, as you saw with Adam. Our Tableau Agent enhances the employee dashboards with automated and easy to understand natural language descriptions of the data so they can save time and understand what the data visualizations mean to them using rules-based templated natural language generation that updates as they click through the data. She can also see insights from Einstein relating to the different types of benefit cases that she's worked on, the breakdown by type and status. So she has all the information she needs at her fingertips and can ensure she's not only meeting her own employee KPIs, but also her customer service KPIs for the department. Upcoming appointments, critical tasks and recent applications are all readily visible, ensuring that nothing gets overlooked. Reviewing her list of open cases, she can see Adam's case for the Rent Choice Assist and opens this up. She can see that he's used the department's new autonomous agent to query details relating to the subsidy and that the agent has not only summarized this information on the case, it has also created the application record linked to the case with the details of his eligibility criteria based on the information provided. This gives Claire valuable context right from the start. She can see details related to the application as well as the status, which provides her with a visual display, but also guidance notes to help her complete the different tasks at each stage of the application, but also provides the citizens with consistent services regardless of which agent they're working with across the department. Claire can also see more information about Adam on his account record, which provides that 360-degree view of the citizen. We can not only see his address, data birth and history with the department, but we've also liaised with most government agencies across Australia to create that unified citizen profile so that we now have information gathered from these government agencies to create this unified profile, which consolidates information like e-mail addresses that Adam may have used across his different services like ATO and Medicare or social housing. His addresses if he's lived across different places, different states, his latest income statements, his citizenship status, his child support payments as well as any employment history. This eliminates the need for Claire to manually search across multiple systems. This cross-agency sharing ensures a comprehensive and accurate picture of Adam's situation, reducing the need for manual paperwork or evidence submission by Adam, so we can now work on supporting the citizens better and faster. Now how did we do this? Underpinning this platform experience that Claire has is Data Cloud, which serves as the central foundation for bringing all this agency-related data into Salesforce so you can activate it across the platform and then use it to power smart AI solutions. This means that we can now combine citizen data and engagement data from multiple sources, external as well as external agencies alongside that customer data that you capture in Salesforce like Service Cloud to fully deliver against our vision of unlocking a Citizen 360 for our customers. Now once Claire has an understanding of Adam's profile, and she has reviewed his application, she can go back and update this to In Review. Since Adam has applied for the Rent Choice subsidy, this triggers a calculation of the subsidy amount for Adam based on the information he's provided as well as the information that the agency has unified via Data Cloud. Once she makes this update, you can see that the amount calculated has been generated by the Business Rules Engine and populated on the application. Claire remembers that the policy relating to subsidy amount has recently changed, and she's not sure if the system has been updated. However, she has an agent that can help her with this on Slack. So she quickly opens up her Slack and asks the benefit and services policy agent what the appropriate subsidy amount should be for Adam. Based on a RAG search of PDFs and other policies that the agent has access to, it provides a response which Claire can now cross-check with the Business Rules Engine to ensure that Adam is getting the appropriate amount of subsidy for his circumstances. Now this is a powerful tool for employees like Claire who don't have the time to sift through pages and pages of policies, but also just want to ask a policy agent, an onboarding agent or even an IT agent a quick question and then get back to the citizen or customer. Behind the scenes, the Business Rules Engine has leveraged a combination of verified data like Adam's income and information gathered from his conversation with the autonomous agent like whether he is on the housing register or not to calculate this subsidy. This automated process ensures consistent and accurate assessments, freeing up Claire to focus on more complex cases and personalized support. The BRE's calculations provide a transparent and auditable record of decision-making processes. She can now even update the application to close after ensuring that Adam will receive his rental subsidy and move on to helping the next.

So thank you to my colleagues there, Rasika and Swathi for that great demonstration on how Agentforce guided Adam, the citizen, and how Claire, a government employee, was able to help see all the information related to the application, have the right context, the guidance at each stage of the rental application as well. So I hope this today's presentation has offered you more information on Agentforce and the building blocks of an AI agent, its application in public sector, and we presented to you a great example of a demo in public sector as well as to resonate and start get you thinking of different use cases you think might be relevant in your respective organizations. So just to recap the building blocks of the Agentforce agent, it's all built on our unified platform, and it offers a very powerful framework that allows you to build sophisticated grounded Agentforce agents, which will help automate tasks, improve efficiency and enhance your customer experiences while freeing up time to cater for and manage more complex and personalized experiences with our citizens. So with that, I'd like to leave you with some of the resources that you might be wanting to explore to find out more about Agentforce. So feel free to scan these codes. And I think these links are also available in the resource widget on the screens. You can click on those. There are many learning tools and trails that are available on our Trailhead, which is a free learning site. So please go ahead and explore that. If you would like to get your hands into building an agent yourself, I have provided a link where you can build your first agent with Agentforce and really experience how easy it is to get it started. And if you'd like to find out more information about public sector solutions, please click on that link in that resource section, and it will take you there to explore some of the other use cases that we see around that.

So with that, I'll now open up for Q&A. If you have any questions, thoughts, comments, feedback, please do leave that in the Q&A box. And I'll probably get started with some of the questions you've already received here. So I have a question here. Is Agentforce HIPAA compliant and SOC 2 compliant? So yes, our Agentforce for service is both HIPAA and SOC2 and SOC 3 compliant as well. Thank you for that question. How can we leverage our AWS investments with Agentforce? That's a really good question. So thank you for asking that. We are strategic -- global strategic partners with AWS and also other cloud partners with some top-tier providers as well. So Salesforce offers zero copy integration. And so our customers are able to access live external data without having to copy. So what this means is that data remains in its original location, and it eliminates the need for your extraction transforming load processes and data pipelines. So it also enables you to build your own LLM, so you can -- we offer the bring your own LLM capability as well. Perfect. And then I see one other question around testing Agentforce agents. So that's a really great question as well. So Agentforce, how do we go about testing an agent? I didn't cover that in today's session, but we do have Agentforce testing center, which enables your teams to test on these different topics and actions selection at scale. So using natural language instructions, you can auto generate hundreds of synthetic interactions such as requests from your customers while engaging with the Agentforce and then you can test them in parallel to see how frequently the result is the right outcome. So you can pressure test Agentforce through these Agentforce testing centers and automatically generated test cases or questions as well. There's more information that we can provide you after on this particular topic as well. So I have another great question here. So how can AI-driven agents assist in policymaking and crisis response? So again, a great question there. Agentforce can really help provide quick responses. So whether be it on an information that's available and given the power that it leverages large language models, you can very quickly provide the guidance as to how to support in this matter and what information to provide, which can be very much in line with how your government department wants to interact. The other way of a use case is to also assist your teams internally to quickly take actions as well. So Agentforce can also be -- have an internal use case helping your employees quickly come up to speed and provide support in the crisis respond as well. Sure. I will also send you the information about a testing center, and there's a video, I think, we have as well. So I'll include that in. Thank you. So I hope I answered that question there. Yes, so I think those were all the questions. I'll just give it a couple more minutes to see if there are more questions. Otherwise, I am happy to wrap up the session. And once again, thank you so much for your time for joining and looking into this solution in the webinar as well. We have a couple of more webinars planned out, which you can see on your console there. So please register for those as well. Perfect. Once again, some resources for you and -- that you might find helpful, and thank you again for your time today.