Something revealing happened at Enterprise Connect last month. Virtually every major contact center platform vendor (Genesys, NiCE, Five9, Salesforce) was in one way or another talking about multi-agent orchestration, agentic automation, and artificial intelligence-native customer journey management. Not infrastructure. Not routing. Orchestration. There's a jostling of vendor positions for the right to sit above your contact center stack and decide what happens inside it.
That's a tell. The fight for customer experience isn't about who handles your calls or chats. It's about who controls the layer that decides what to do with them.
This layer is the control plane for CX. In a smart customer service setting, Opus Research refers to this as conversational experience orchestration (CXO). It's the system that listens to every interaction, understands what's happening, decides what should happen next, and executes across channels and business systems. It doesn't just process conversations. It learns from them and uses them to drive outcomes in real time.
This distinction matters because AI has changed the economics of orchestration. Large language models can hold context, reason across steps, and take action without constant human intervention. The value in a CX stack no longer lives in the routing layer. It lives in the orchestration layer, in the intelligence that decides which agent (human or AI) handles which task, under which constraints, and with which tools.
Two Layers. One Strategic Bet.
Think of the modern CX stack in two parts. The first is the interaction substrate: telephony, recording, routing, compliance, uptime. Operationally critical. Expensive to rebuild. Also, increasingly, a commodity.
The second is the control plane, where enterprises define what customer service actually means. Which tasks AI agents complete autonomously. Which workflows cross systems. Where human judgment is required. How performance gets measured. This is not necessarily a feature inside your contact center-as-a-service platform. It's an architectural capability you need to own.
If the control plane is the strategic layer, companies need to be deliberate about what they hold vs. what they rent.
Conversation data is the most obvious asset. Every interaction generates signal, from transcripts, intent patterns, resolution outcomes, escalation triggers. Treated as a first-class enterprise asset (exportable continuously, structured usably) it compounds over time. Locked inside vendor reporting environments, it creates dependency.
The action catalog matters equally. Customer service is a set of discrete actions: check order status, process a refund, reset a password, schedule a technician. Define those with stable interfaces so that whoever orchestrates the conversation calls your catalog, not the reverse. That's the difference between an operating model you own and one you're borrowing.
Policy, security, and escalation logic should never be fully outsourced. What AI agents can decide autonomously, what requires human approval, what must be logged. These are governance decisions. They belong in repositories you control, versioned, auditable, and portable.
Don't Cede the Layer That Learns
Agentic AI raises the stakes on every architectural decision. An AI agent with real-time access to CRM, order management, payments, and customer identity operates with entirely different consequences than a chatbot that stumbles. The control plane is what governs that scope of action. It's what accumulates the intelligence that makes it better over time.
AI features are already delivering quick wins: summarization, triage, real-time coaching, low-stakes automation. But what's next are advanced skills, workflows, and journey state that learns from your customers. This should stay in your hands. Treat the interaction substrate as a mission-critical utility and the control plane for CX as the strategic asset.
Derek Top is principal analyst and research director at Opus Research.