The shift from reactive scripted automation to agentic AI represents a fundamental recalibration of how contact centres operate, moving beyond deflection metrics toward genuine resolution. Jonathan Barouch's framing at Zendesk articulates this distinction clearly: agentic systems reason through context, handle multi-step workflows end-to-end, and adapt to customer intent rather than following predetermined decision trees. This isn't incremental improvement—it's a structural change in what automation can accomplish. The practical barriers, however, remain consistent across enterprise deployments: data silos fragment customer context, authentication layers create friction in agent handoffs, and brittle infrastructure collapses under the weight of edge cases. Teams attempting this transition without addressing these foundational issues will find their agentic pilots stalling at the pilot stage, unable to scale beyond controlled test environments.
The metric shift from containment to resolution carries significant implications for how CX leaders should evaluate success and allocate resources. Containment—the traditional measure of deflection rates—incentivises pushing customers toward self-service regardless of outcome quality, whereas resolution-focused measurement rewards systems that actually solve problems, whether through AI, agent escalation, or hybrid workflows. This reframing suggests that teams currently optimising for containment rates may be gaming metrics that no longer reflect operational health. The related finding that only 20% of companies using AI in customer service cut agent headcount indicates that agentic automation is augmenting agent capacity rather than replacing it—a critical distinction for teams planning headcount strategies around AI deployment.
The rollout methodology Barouch advocates—piloting agentic workflows, running A/B tests, and building learning loops from real interactions—positions knowledge management as the competitive lever. Rather than betting the operation on a single platform migration, teams should treat agentic deployment as iterative capability building, where each interaction feeds back into improved self-service, better agent tooling, and refined automation logic. For Zendesk administrators and CX consultants, this raises a practical question: does your current knowledge management infrastructure support this feedback loop, or would agentic automation expose gaps in how you capture, structure, and distribute customer intelligence? The answer determines whether agentic AI becomes a force multiplier or an expensive experiment that highlights existing operational fragmentation.
From Reactive to Agentic: What an AI Native Contact Centre Actually Looks Like CX Today
In this CX Today interview, Rob Wilkinson speaks with Jonathan Barouch, GM of Contact Center at Zendesk, about what “AI-native” and “agentic” actually mean in practice, where enterprise teams get stuck (data silos, authentication, brittle infrastructure), and how to roll out automation without betti