The fundamental tension in agentic contact centres lies not in whether AI can handle routine interactions—it demonstrably can—but in whether organisations deploying these systems understand their categorical limitations when customers present non-conforming problems. Martha Buyer's analysis, drawing on legal scholar Christine Pedigo Bartholomew's distinction between rule-following and situational reasoning, exposes a critical gap: AI systems operate as "software brains," flattening reality into structured databases and yes/no decision trees, whilst human expertise involves reading context, ambiguity, and circumstance to determine what actually works for the person in front of you. This distinction matters acutely for CX teams because it reframes the deployment question. Rather than asking whether agentic AI can replace human agents, teams should interrogate whether their current tooling—whether Zendesk, Freshdesk, or Salesforce Service Cloud—is equipped to recognise when a customer's issue falls outside the model's training patterns and escalate intelligently. The risk is not that AI will handle routine queries poorly; it's that organisations will optimise for efficiency metrics whilst remaining blind to customers whose problems don't fit the boxes, creating friction that drives churn.
The implications cut deeper than operational efficiency. Blair Pleasant's observation that AI cannot apply conscience, empathy, or common sense to exceptional circumstances suggests that contact centres relying heavily on agentic workflows without robust human handoff mechanisms are essentially filtering for customer homogeneity. This creates a secondary question for support leaders: if your agentic system is passing increasingly complex cases to humans, are you measuring the cost of those escalations against the revenue impact of customers who abandon the interaction entirely rather than wait for human intervention? The article's framing—that every business is fundamentally a customer service business—becomes especially pointed when AI interactions are perceived as annoying rather than helpful, particularly among younger demographics who actively resent the experience. For teams already embedded in these platforms, the strategic imperative is not to reduce human involvement but to architect escalation workflows that preserve human judgment for the ambiguous cases where conscience and contextual reasoning determine whether a customer returns or defects.
The operational takeaway is straightforward: agentic contact centres will succeed when they're designed as triage systems, not replacement systems. Teams should audit their current Zendesk or Salesforce configurations to ensure they're capturing the signals that indicate a customer's situation requires human reasoning—not just routing based on intent classification or sentiment analysis. The question becomes whether your organisation has the discipline to let AI handle what it handles well whilst genuinely empowering agents to reason with conscience when the situation demands it, rather than forcing non-conforming problems back into the system's existing boxes.
The question that any enterprise comms professional deploying an agentic AI workflow should be asking is simpler— are the existing AI tools doing the job of serving everyone effectively?