The financial services industry has deployed AI at scale across customer service, lending, and security, yet the dominant measurement framework—cost reduction and automation rates—obscures a more fundamental problem. Wells Fargo's Fargo assistant and Bank of America's Erica represent divergent philosophies: one optimized for deflection and cost containment, the other designed around customer need with efficiency as a secondary outcome. The distinction matters because institutions measuring success by call avoidance rather than trust indicators are optimizing for the wrong variable. When AI is architected to restrict human escalation as a cost control, it signals to customers that the institution prioritizes its own economics over their outcomes. For CX teams already managing these systems, this creates an immediate tension: your deflection metrics may be masking erosion in the trust signals that actually predict retention and lifetime value. The question becomes whether your current measurement dashboard captures what your business actually depends on.
The second dimension—product development through AI-driven credit inclusion—reveals that efficiency and fairness are not opposing forces but often aligned. Upstart's model demonstrates that incorporating 1,600+ variables produces both lower default rates and broader access to underserved populations, yet this advantage evaporates without explainability. Regulatory frameworks in the EU and Korea now mandate that automated decisions be auditable and intelligible to customers, transforming explainability from a nice-to-have into structural obligation. An institution that cannot articulate why an algorithm denied a loan has forfeited the legitimacy to make that decision. For support teams, this means AI-driven decisions will increasingly flow through your channels as customer disputes, and your team's ability to explain the system's logic becomes a competitive differentiator.
The security dimension compounds these pressures. Adversarial AI—deepfakes, synthetic phishing, model poisoning—is evolving faster than legacy defenses, yet institutions like JPMorgan Chase and Mastercard have proven that AI-native security and frictionless customer experience are not trade-offs but complementary. The deeper implication is that security failures are not operational risks to be managed separately; they are trust failures that cascade through every customer interaction. For CX professionals, this means your platforms are now critical infrastructure in the trust equation. A breach or security incident does not merely create a support spike—it undermines the institutional credibility that your entire function depends on. The institutions that will lead are those treating customer service, product transparency, and security not as separate functions but as expressions of a single commitment: earning trust through explainability, inclusion, and genuine customer outcomes rather than through cost optimization alone.
[ECONOMIC ESSAY CONTEST] AI in financial services: Efficiency is not enough The Korea Times