Contact centers deploying AI agents are systematically neglecting the foundational data work required to make those deployments sustainable. The core problem is straightforward: organisations are racing to implement agentic AI without establishing the governance, measurement frameworks, and architectural rigor needed to manage it. This isn't a technology problem—it's a discipline problem. AI agents generate substantially more data than human agents, including failure signals and turn-taking patterns that traditional conversations never surface, yet most teams lack the infrastructure to capture or act on this data. The opportunity is real, but only for organisations willing to do the unglamorous work of building proper data foundations first.
The critical gap lies in what's being skipped: establishing "ground truth" as a reliable baseline for AI performance. This demands scientific rigour that most contact centers simply aren't applying, leaving them vulnerable to model drift and performance degradation they won't detect until damage is done. Agentic AI differs fundamentally from legacy IVR systems—the removal of linear flow constraints creates new design possibilities, but only if organisations layer proper orchestration and monitoring on top. For teams already running systems like Agentforce or evaluating similar platforms, this raises an uncomfortable question: do you have visibility into whether your AI is actually performing as intended, or are you operating on assumptions? The answer determines whether your AI investment becomes a competitive advantage or a liability masquerading as innovation.
The measurement challenge extends beyond traditional metrics. Generative AI now enables conversation-level appraisal at scale, transforming CSAT from a historically biased sampling tool into something genuinely representative. Yet this capability only matters if teams have built the data architecture to support it. For support leaders and CX consultants, the implication is clear: before your next AI implementation, audit whether your organisation can answer three questions with confidence—what data are you capturing, how are you establishing performance baselines, and what happens when your models drift? If you can't answer these, you're not ready to deploy.
Rhys Fisher, Associate Editor at CX Today, sits down with Dave Rennyson, CEO of SuccessKPI, to tackle one of the most pressing yet underexplored challenges in modern contact center operations: what it actually takes to build an AI-ready data foundation. As AI agents become a fixture of the contact
Contact Center AI Is Only as Good as the Data Behind It – So Why Are We Ignoring the Data? CX Today