The fundamental shift in contact centre operations driven by AI automation demands an immediate recalibration of how CX leaders measure success. As agentic AI absorbs routine inquiries, human agents are increasingly reserved for complex, emotionally charged, and high-stakes interactions—yet most organisations continue optimising for average handling time and other speed-based KPIs designed for an era of high-volume, low-complexity work. This mismatch creates a perverse incentive structure: teams rushing through difficult conversations to hit targets inadvertently damage resolution quality, erode customer trust, and accelerate agent burnout. The panel consensus is unambiguous: speed metrics are no longer fit for purpose. Instead, leaders must pivot measurement frameworks toward resolution quality, first-contact resolution rates adjusted for complexity, and whether AI tooling genuinely augments agent decision-making in real time rather than simply reducing headcount.
The implications for teams already embedded in platforms like Zendesk, Freshdesk, and Salesforce are substantial. These systems were architected around throughput optimisation, meaning their native dashboards and reporting hierarchies still default to speed-first indicators. Administrators and support leads face a practical challenge: redesigning what gets measured without necessarily replacing underlying infrastructure. This requires moving beyond dashboard defaults to construct composite metrics that weight resolution quality and customer effort alongside efficiency, then embedding these into coaching, scheduling, and performance management workflows. The risk of inaction is acute—organisations that continue chasing speed whilst deploying AI risk repeating the volume-scaling mistakes of previous decades, where root causes went unaddressed and cost reduction became the primary driver rather than genuine service improvement.
The governance dimension adds another layer of urgency. Building organisational trust in AI requires demonstrable reliability and relevance—which cannot be established through speed metrics alone. Teams must establish clear protocols for when and how AI recommendations are surfaced to agents, measure whether those recommendations are actually used and whether they improve outcomes, and create feedback loops that allow agents to flag unreliable or irrelevant suggestions. This shifts the measurement burden from simple operational metrics to more nuanced assessments of AI utility and agent confidence. For CX professionals, this represents both a strategic opportunity and an operational necessity: those who redesign metrics proactively will position their organisations to extract genuine value from AI investment, whilst those who delay risk embedding automation into broken measurement systems that optimise for the wrong outcomes entirely.
AI is pushing contact centres into a new operating model, and the old metrics are struggling to keep up. In this CX Today roundtable, Rob Wilkinson sits down with Martin Teasdale, Steve Morrell, Sangeetha Rai, and Rebecca Wetteman to explore what “good” looks like when automation absorbs the simple
CX Metrics In The Age Of AI: Stop Optimising For Speed CX Today