Nanovest has shifted operational risk management from fragmented service resolution to an integrated governance model in which customer interactions function as real-time risk signals. According to Nur Vitriani, Head of Customer Experience and Human Resource and Culture at Nanovest, customer engagements are treated as indicators of operational, fraud, regulatory and AI-related risks rather than isolated service events. This approach reflects Indonesia's regulated environment, where issues may involve multiple authorities, including OJK, Bappebti and Bank Indonesia. The framework consolidates customer service data, complaints, know-your-customer (KYC) processes, chatbot interactions and fraud alerts into a single risk management process. It replaces siloed escalation with a four-layer governance model comprising AI interaction capture, structured risk classification, risk aggregation and service level agreement (SLA)-based remediation. Customer service and AI are integrated into enterprise risk governance, using customer interactions to identify, escalate and monitor operational risks while supporting regulatory compliance. Customer interactions become operational risk signals The CX-Risk Signal Governance framework consolidates customer tickets, complaints, KYC issues, chatbot interactions, fraud alerts and VIP escalations into a single risk management process. It replaces fragmented workflows across customer service, compliance, fraud and product teams with a four-layer architecture comprising AI interaction capture, quality assurance classification, risk signal aggregation and SLA-based remediation. Customer service functions as the intake layer for risk detection. Jasmadi Ramadhan Beruh, Project Head of AI and CX-Risk Governance at Nanovest, said tickets, complaints and customer enquiries are assessed for operational and regulatory risks before being routed to functions such as finance or product teams for resolution under defined SLAs. This shifts customer service from resolving individual incidents to identifying emerging risks. Jasmadi Ramadhan Beruh Project Head of AI and CX-Risk Governance Nanovest Risks are recorded in a central register, assigned to accountable owners and monitored until closure. Internal audit and regulatory oversight are integrated into the same workflow, while chatbot interactions are subject to the same governance controls as human-assisted customer service. Naura classifies chatbot errors to improve risk governance Nanovest's AI chatbot, Naura, launched in August 2025, operates within the same governance framework as customer service. Chatbot failures are classified into two categories: BOT-ERROR, where the chatbot provides incorrect or misleading responses despite having the required knowledge, and BOT-GAP, where it lacks sufficient knowledge or fails to interpret customer intent. Each category follows a different remediation path, with BOT-ERROR addressed through correction and model retraining, and BOT-GAP through knowledge expansion and improved intent recognition. The framework also redefines how chatbot performance is measured. According to Vitriani, "Compliance alone is not a sufficient measure of success. Customers must feel both safe and satisfied, which is why the target combines stronger compliance with an equally high level of customer satisfaction." Rather than relying on Bot Containment Rate, which measures whether interactions remain within the chatbot, Nanovest uses Resolution Accuracy Rate (RAR), which assesses whether customers receive a correct resolution without returning with the same issue within a defined period. The shift reflects the view that a contained interaction can still create operational or regulatory risk if the response is inaccurate. The framework therefore treats AI as an operational risk function rather than a standalone automation tool. Weekly governance reviews identify error patterns, retrain models and address knowledge gaps, creating a continuous feedback loop between customer interactions and AI performance. CSAT recovers as governance improves The CX-Risk Signal Governance framework produced measurable improvements in customer experience, operational efficiency and risk management. Customer satisfaction (CSAT) fell from 80% in January 2026 to 63% in February 2026 following chatbot accuracy and product-related issues, before recovering to 86% by June 2026. Operational performance also improved. Average resolution time fell from 10 hours and 16 minutes to six hours and 12 minutes, while first contact resolution increased from 86% to 90% year-to-date. KYC approval time decreased from four hours and 35 minutes to 35 minutes, and SLA compliance rose from 83% to 99.55%. AI governance metrics also strengthened. BOT-ERROR repeat rates declined from 86.60% to 66.40%, while RAR reached 87.90% in May 2026 following model retraining and routing improvements. Fraud rates fell from about 2.0% to 1.46% year-to-date, while false positive rates remained between 0.01% and 0.06%, indicating more precise fraud detection. Customer interactions become risk intelligence Nanovest extends customer interaction data beyond service operations into enterprise risk governance by linking customer signals to financial impact tracking and governance reporting. According to Vitriani, issues identified through customer interactions are monitored not only for resolution but also for their operational, compliance and financial implications, creating a common reporting framework across customer service, risk and management. This shifts customer interaction data from a service metric to a governance input for operational risk oversight and organisational accountability. The broader implication is that customer service may increasingly become part of financial institutions' operational risk infrastructure as AI adoption expands. Institutions that integrate customer interactions, AI oversight and risk management within a single governance framework are likely to strengthen early risk detection, governance and operational resilience while improving their ability to monitor and manage AI-related risks.
