Banks are investing heavily in AI-assisted contact centre transformation, yet a critical misalignment threatens to undermine these efforts. Deloitte's research reveals that whilst 74% of US banking executives are either currently deploying or planning to deploy generative AI in contact centres, a significant blind spot exists: customers consistently rate their support experiences lower than executives believe they're delivering. This gap manifests across resolution ease, follow-up quality, transparency, and response speed—precisely the areas where AI implementation could either amplify frustration or unlock genuine value. The Bitdefender analysis compounds this concern by exposing the security and governance risks that emerge when organisations rush AI adoption without proper controls. When chatbots access unnecessary customer data, handle sensitive requests without verification, or operate without human oversight, the damage extends beyond operational inefficiency into reputation and trust erosion. For CX teams already managing Zendesk, Salesforce Service Cloud, or similar platforms, this creates an immediate tension: the pressure to deploy AI quickly conflicts with the need to establish governance frameworks that prevent data exposure, ensure accurate responses, and maintain human judgment where it matters most.
The path forward requires deliberate architectural choices rather than wholesale automation. Deloitte's framework identifies four critical considerations that directly shape how CX leaders should approach AI integration. First, banks must prioritise fixing the "repeat yourself" problem—the channel handoffs where customers lose context—before layering AI on top of fragmented systems. This means starting with high-friction, high-volume areas like fraud alerts or payment disputes, then systematically improving data flow across existing platforms rather than replacing core systems wholesale. Second, success metrics must shift from cost-per-contact and average handle time to customer retention and first-contact resolution, which requires breaking down silos between fraud, disputes, servicing, and contact centre teams. Third, AI should be deployed in tiers: simple, low-risk requests move to self-service; moderately complex issues to agentic AI with human oversight; high-stakes interactions like disputes or hardship requests remain human-led with AI providing context and recommendations. This tiered approach directly addresses Bitdefender's warning that AI cannot safely handle sensitive requests involving account access, payments, or password resets without human verification. The fourth consideration—embedding governance early rather than bolting it on at approval stage—becomes non-negotiable when customer data is at stake. Should your team be running Agentforce, Freshdesk's AI agents, or similar agentic solutions, the question shifts from "can we automate this?" to "should we automate this, and what guardrails prevent escalation loops or data leakage?" The research indicates that 70% of banking executives expect AI to shift agents toward higher-value roles, yet only 25% of customers report that self-service resolves 50% or more of issues without human intervention. This gap suggests that many CX teams will spend the next 18–24 months not replacing agents but redefining their role as resolution specialists who supervise AI decisions, provide judgment in ambiguous situations, and maintain accountability for outcomes. The real business advantage lies not in cost reduction but in the 91% of customers interested in proactive support—a capability that requires clean data, integrated systems, and human-AI collaboration to identify and prevent issues before customers contact you.
7 AI customer service mistakes that can put your business at risk Bitdefender
How banks can turn AI-assisted customer service into a business advantage Deloitte