Generative AI has made financial services chatbots amazing. Yet, India’s diverse demographics can complicate AI security and regulatory compliance

Supratik Nag, Vice-President, Business Solutioning and Engineering, Maveric Systems

This accessibility helps in serving diverse customer demographics more effectively. Conversation AI analyzes various aspects of human language, such as syntax, semantics, pragmatics and morphology, and thereby transforms linguistic interactions as rule-based, machine learning algorithms. By understanding the nuances of language, conversation AI systems can interpret customer inquiries more accurately and provide relevant responses. Some of the prominent use cases in banking:

    • Automated document processing, which helps banks to retrieve information from large volumes of customer-submitted documents — such as identification cards, income statements, or utility bills — with minimal manual intervention, facilitating faster and more efficient account opening or loan application processes.
    • Reviewing of regulatory compliance documentation to identify discrepancies or non-compliance issues. By analyzing documents such as Know Your Customer forms, and Anti-Money Laundering documentation AI-powered systems can help banks ensure adherence to regulatory requirements and mitigate compliance risks more effectively.

SN: Achieving AI integration in the BFSI sector poses challenges due to the diverse requirements of customers and regulatory considerations.

    • Human oversight still required: Successful banking operations involve striking the appropriate balance in integrating AI technologies with human expertise to achieve optimal results. While AI can streamline processes, automate labor-intensive tasks, and reduce operational costs, human interface is still the cornerstone of banking operations. Given the stringent regulations surrounding data privacy, security, and customer protection, human oversight is crucial to ensure compliance with laws and regulations. In India, navigating regulatory frameworks and obtaining permissions for AI-driven solutions can often be cumbersome and challenging.
    • Generational and socio-cultural gaps:While millennials and Gen Zs lean towards digital banking, segments such as senior citizens and technology-averse groups, a visit to the bank holds social significance beyond transactions: transitioning them to a fully automated banking experience will not happen overnight. Therefore, effectively attracting diverse demographics and gaining their trust remains a critical focus for the industry.
    • Fraud/cyber risks:While AI can aid in fraud detection, the ultimate responsibility for ensuring compliance and ethical conduct lies with humans. It is imperative to establish clear accountability frameworks for AI systems and define the roles and responsibilities of individuals handling these accounts. This proactive approach helps mitigate risks and uphold trust in the banking system.

SN: This can be achieved through various mechanisms:

    • Real-time monitoring, pattern recognition, and anomaly detection: Identifying financial fraud is a multifaceted task that requires banks to analyze vast amounts of data from diverse sources, both proactively and reactively. This includes structured data such as transaction records and customer profiles, as well as unstructured data like emails, call transcripts, and social media interactions. AI-driven authentication methods, such as voice recognition and biometric verification, enable banks to accurately identify the right customer during interactions. By verifying the identity of customers, AI helps ensure that sensitive information is only provided to authorized individuals, reducing the risk of account takeover fraud and unauthorized access.
    • Recognizing patterns of fraudulent activities in conversations or transactions: AI can be used to analyze documents and conversations to identify suspicious activities and provide immediate alerts to all parties, on a 24/7 basis.
    • Strict regulatory compliance: In spite of the utility of AI automation where collecting more data is more viable, the industry has to adopt a principle of data minimization, collecting only limited data necessary for the intended purpose. Additionally, it is important to classify AI models based on the level of risk associated with the data they handle. Stricter security measures and observation are required for high-risk models dealing with sensitive financial information. Furthermore, BFSI firms should anonymize or de-identify personal data used to train AI models to minimize the risk of re-identification

CybersecAsia: Looking ahead, what are some potential future developments in conversation AI that could further transform India’s BFSI services in the coming years?

SN: Some of the following advancements will revolutionize the way banks interact with customers, optimize operational processes, and deliver value-added services:

    • Multimodal/Multilingual AI: By enabling multilingual and multimodal capabilities can enable conversation AI functions to provide personalized services to customers in their preferred language and across various communication channels. This enhances accessibility for diverse customer demographics and improves communication effectiveness.
    • Predictive personalization: Offering personalized banking experiences tailored to individual customer preferences and financial goals can enhance the overall customer experience.
    • Ethical/Responsible AI: This involves implementing robust data governance and other frameworks to ensure fairness, diversity, equity, inclusion, safety, accountability; and vigilance of any proliferation of AI abuse, biases and unforeseen challenges.
CybersecAsia thanks Supratik Nag for sharing his insights on India’s conversation AI landscape.