How to Manage AI Agents with Agentforce Observability

In this demo, I'm going to show you how AgentForce observability helps manage and improve your entire fleet of AI agents. This is Pronto, a company that recently deployed an agent for their entire user base. But this agent is engaging with numerous customers and is available every day at every single hour. Across all of these interactions, how do you know if the agent is helping you move the needle on your deployment goals? And if this agent runs into an issue, how does the Pronto team know where to improve? That's why we've built AgentForce Observability to help you track your agent's performances and use actual conversations to iterate on your company's agentic experiences. With agent analytics, leaders can understand how their fleet of agents are measuring up against deployment goals, validating success all of these metrics come out of the box to showcase session outcomes, but can be customized based on metrics important to you. Knowing directionally where to guide the team to start improvements is also crucial. With agent effectiveness, you can get high level session data to explore how customers are using agents, and with quality, you can see which specific agents need revision based on their engagement handling scores. Now, deployed agents don't look like one off closed sessions. Rather, their data enriches a comprehensive view that enables you to align their engagements with your goals and allow your teams to start moving quickly on improvements. For builders, seeing actual conversation data allows you to fully understand outcomes and to iterate on overall design. And instead of waiting for tickets to be logged or creating custom dashboards, you have access to Insights, which helps you identify performance trends on a session level based on the agent channel or timeline most relevant to you. With Insights, you can start your investigations on how topics and intents are being handled with your end users. We can spot here that there's an underperforming topic which we'll go ahead and dive more into. And what makes those investigations easier are sessions and Intents, which allows us to look deeper into actual conversations. These sessions and intents are not just your regular chat transcripts. These are fully contextual conversations with scoring given for each intent mentioned by the customer. They can also be viewed as fully processed with full scores and intents, or as unprocessed, which are just freshly logged sessions. Going into one of these sessions allows you to see an entire conversation and how the agent responded. But the best part? Clicking into one of these responses which reveals its underlying reasoning, which means we can find out the exact paths the agent took, helping us troubleshoot on the spot. Unveiling the black box that is agent customer interactions and on voice calls. We also get an additional playback feature which allows you to hear the full voice conversation for yourself. All of these tools allows you to make precise decisions on where to make improvements, opposed to guessing and spending extra time uncovering gaps. Which means now we can go into the builder and know exactly what part of the agent's configuration we need to rework to improve the end user experience. Now all these features in Observability helps to see how agents have already interacted with customers. But what about real time engagements? Well, that's where Agent health Monitoring comes in, our new feature that lets you detect issues in near real time across your deployed agents. Here you can create alerts based on important health metrics which lets you react live to ongoing sessions. Which means you now have a view with near real time metrics that you can monitor making sure that your fleet of AI agents are working as intended. And that's what AgentForce observability is. From being in the dark on how your agent fleet is performing to having a dedicated tool that constantly compiles insights, identifies optimization opportunities and monitors your agent team in near real time, helping you build a constantly evolving agentic enterprise. So now you've learned how AgentForce observability helps you effectively manage and improve your team of AI agents. To learn more, visit agentforce.com and as always, thanks for watching.