Hillary Bursell outlines agentic CX, an AI operating model where systems perceive, reason, and act within guardrails to resolve customer issues autonomously. Most organizations waste conversational data on isolated features like chatbots and summaries. Real customers have cut repeat contacts and field visits by redesigning workflows for AI and human agents to work together.
Chapters
Hi, thank you so much for joining me today for this session called Agentic CX That Drives Real Outcomes: A Blueprint for Next-Generation Customer Teams. My name is Hillary Bursell and I'm Director of Product Marketing at Dialpad, and I'm really excited to talk to you today because I think we know every customer experience leader is really focused on how AI can transform their business. But in reality, we also know that most organizations are still trying to figure out just how to apply AI to really isolated tasks, making summaries, using Copilots, launching chatbots, or using AI to get better analytics. What's much rarer is designing the operating model itself around AI. And at Dialpad, that's really what we mean when we say agentic CX, not just systems that can analyze conversations, but systems that can perceive, decide, act, and improve in real time. Today, I wanna share a practical blueprint for what that looks like. The irony in CX today is that most organizations are capturing more customer data than ever, but they're operationally using very little of it. So we all have data living across fragmented systems, sometimes 5 to 10 or more per business. Most of that interaction data is never used to inform decisions, and QA programs within the contact center review only a tiny fraction of conversations with customers. The issue isn't a lack of information, it's that most systems still still don't have the ability to put that conversation data to work. And that's why most AI deployments are plateauing today. Organizations add AI summaries, QA pilots, analytics dashboards, and automation tools, but the under— their underlying operating model doesn't change. Insights are arriving too late, bots are disconnected from humans, analytics don't trigger action, and automation lacks government governance. So companies end up with AI features, not AI systems. And this is the shift that we're, we're in the middle of right now. For the past couple of decades, we've seen systems that can help you react. They record conversations so that you can make decisions later. Now we have AI-assisted systems that provide recommendations, but usually after the moment has already passed. Agentic systems are different. They operate in real time, they take action within defined business guardrails, they orchestrate across workflows and channels, and they continuously learn from outcomes. The important point is agentic doesn't mean uncontrolled, it means governed autonomy. And here's what that looks like operationally. In an agentic system, every interaction with your customers becomes a live signal. The system detects intent, sentiment, urgency, customer history, and likely outcomes, and that signal can then trigger actions, whether it's routing a customer to the right place, retrieving the right information to resolve the issue, executing a workflow, deciding to escalate to a human, or even autonomously resolving the issue in some other way. And critically, the outcome of that interaction then feeds back into the system so that future decisions improve over time. At Dialpad, we think about agentic CX as 4 connected layers. The first layer is perception, the ability to understand every interaction in real time. The second layer is reasoning, being able to determine the right next step to take or the right workflow to complete to meet that customer's needs. The third layer is action, the ability to execute tasks autonomously where appropriate or escalate to a human if that makes more sense. And the fourth layer is learning, using every interaction to improve routing, coaching, automations, and performance. Most organizations today are still primarily focused on perception and assistance. Very few have reached true operational agency that you see with the action layer and the learning layer. So to get to that point where AI agents are autonomously executing tasks and making decisions, enterprises first need to have confidence that it's going to work and make good decisions on their behalf. The question they have to ask is, how do we maintain control? The answer is that governance has to be built into the lifecycle of that AI agent. It has to help you build the right workflows to begin with. That could be finding the most common workflows that your customers are calling in for assistance with, Or it could mean saying, hey, this is a really complex workflow that is too hard or won't be satisfying if we automate it. Then once you've identified the right workflows to automate, the system will help you build that, that agent with safeguards built in. The next layer is testing. You want to test that layer, that agent before you deploy it. Making sure that it is actually going to perform as it's designed and also to withstand those edge cases where customers aren't doing what you expect them to do. And then finally, once it's deployed, you want to be able to continuously monitor that behavior in production. The goal there isn't unrestricted automation, it's governed autonomy. And that distinction really matters. Underneath all of this is the real prerequisite for an agentic system, a structured conversational data layer. With that type of layer in your system, every interaction is becoming structured intelligence. It's capturing intent, sentiment, outcomes, actions, and escalations. And that shared context is what allows humans and AI systems to coordinate effectively. Without unified conversational data, agentic systems will break. At Dialpad, our customers are already following this blueprint and seeing measurable impact. They're seeing reduced repeat scheduling contact— contacts. They're seeing fewer repeat field visits, and they're seeing large-scale autonomous resolution of customer requests. Importantly, these outcomes aren't coming from replacing humans. They're coming from redesigning workflows so AI and human teams operate together more intelligently. For most organizations, this transformation happens in stages. First comes establishing visibility. Just having the ability to capture and understand all of your conversations will bring a whole new level of, of intelligence to your operation. From there, you can use that visibility to assist your human teams in real time, giving them guidance, recommendations, and coaching. And then finally, agency— having the ability to create systems that autonomously and intelligently resolve and orchestrate work. Most enterprises today are somewhere between visibility and assistance, but the shift towards agency is already underway. The opportunity here is actually much bigger than just having more AI features. It's about turning your conversations into operational intelligence and turning that operational intelligence into action. The organizations that succeed aren't just deploying more AI, they're redesigning their customer experience around systems that continuously learn or coordinate and improve. And that's what agentic CX really means. Thank you so much.
