The acceleration toward always-on customer service is creating a structural problem that most CX leaders have misdiagnosed. Rather than automation itself causing agent burnout, the real culprit is the operational friction that AI was supposed to eliminate—disconnected systems, poor visibility, and the constant cognitive load of assembling customer context across multiple tools. Zendesk's Nuri Gocay points to a stark benchmark: 87% of call centre agents report high workplace stress, with 74% at ongoing risk of burnout. The paradox is that AI removes the lightweight interactions—simple password resets, routine queries—that once provided natural breathing room between complex calls. When AI is bolted onto fragmented systems rather than deeply integrated, agents inherit not just the complexity but also the responsibility for cleaning up AI failures. This creates a structural shift that demands rethinking staffing, coaching, and escalation models entirely. The question for teams already running Agentforce or similar agentic platforms is whether they've addressed the underlying data and integration problems first, or whether they've simply added another tab to an already fractured workflow.
Knowledge retrieval remains a hidden tax that most leaders underestimate, consuming an average of 2.7 minutes per call as agents search for answers while customers wait. This search time compounds stress because it signals to the customer that the system has failed to travel context forward—a failure that often manifests in the agent's opening question, "How can I help you?" which Gocay describes as "one of the biggest insults to our data and systems." The solution is not more AI tools but better integration, knowledge hygiene, and a fundamental shift in how workload is designed. Routing by cognitive load rather than skill alone prevents agents from stacking emotionally heavy interactions back-to-back. Creating simple feedback loops that allow agents to flag low-quality AI outputs builds ownership and reduces the sense that AI is something happening to them rather than something they shape. For CX leaders seeking immediate progress, the practical move is to identify a single friction point that repeatedly forces context switching, fix the integration path, and measure the effect on escalation and agent sentiment before scaling further.
The deeper implication is that AI-driven CX will ultimately be judged not by automation rates but by whether customers make progress and whether human agents retain the capacity for empathy at moments that matter. If empathy is the last meaningful human differentiator in service, leaders cannot afford to drain it through poor system design. This reframes agent wellbeing from a culture initiative into an operational metric that directly affects service quality and customer outcomes. Teams that treat integration, knowledge management, and workload design as prerequisites to AI deployment will see the promised benefits of reduced manual effort and faster resolution. Those that treat AI as an overlay on broken processes will simply shift the burden from customers to agents, creating a false economy where automation metrics improve while human capability deteriorates.
Always‑On Without Always‑Burned‑Out: The Human Cost of AI‑Led CX CX Today