Deloitte's research challenges a prevailing assumption in banking AI deployment: that customer value tier should determine automation strategy. Instead, the consulting firm argues that task complexity and customer intent should drive decisions about where AI augments agents versus where it operates independently. Lauren Littlefield, a managing director at Deloitte, highlighted a critical gap between executive perception and customer reality in contact center experiences, noting that banks often default to cost-reduction narratives when implementing AI rather than experience enhancement. This framing matters because AI acts as a magnifier—it can amplify existing service failures if deployed thoughtfully, particularly when knowledge management hygiene is poor or when different AI models across channels reference conflicting information sources.
The implications for CX teams are substantial and somewhat counterintuitive. The research dismantles the high-net-worth-only-human-touch model that many institutions have built into their tiering strategies. A lost debit card inquiry, regardless of customer lifetime value, benefits from fast automation; a mortgage discussion requires human judgment and empathy. This task-driven approach directly challenges how many teams have structured their agent hierarchies and routing logic within platforms like Zendesk or Salesforce—where premium agents are often reserved for top-tier customers regardless of actual interaction complexity. For teams already managing multichannel deployments, the warning about conflicting knowledge bases across AI implementations is particularly acute: if your chat AI, IVR system, and agent-assist tools are trained on different data sources, you're not just creating friction, you're actively eroding customer trust.
The research also signals a fundamental shift in agent skill requirements that CX leaders must prepare for now. As AI handles procedural tasks across all customer segments, the agents who remain will need to excel at empathy, complex problem-solving, and high-stress de-escalation—a different hiring and training profile entirely. This raises a practical question for support leaders: are your current performance metrics, coaching frameworks, and quality assurance processes designed to evaluate these capabilities, or are they still optimised for speed and first-contact resolution? The transition won't be instantaneous—Littlefield suggests human agents will remain central for at least five years—but the window to reskill teams and restructure workflows is narrowing.
Tasks, not customer tiers should drive bank AI use: Deloitte Banking Dive
New BBB study: Assessing AI and customer service thegazette.com