Sachin Prasad¹, Priya Ranjan Sahoo², ^{1 2}USA

Sentiment analysis has transitioned from academic research to a critical enterprise capability for customer intelligence and decision support in regulated industries. Despite significant advances in deep learning accuracy, organizations face substantial challenges in deploying sentiment analysis models on a scale, particularly around trust, governance, explainability, and operational reliability. This paper examines architectural requirements for operationalizing sentiment analysis within large, multi-tenant enterprise environments. We present a platform-centric architecture that integrates sentiment models into governed data and AI ecosystems, enabling consistent lifecycle management, policy enforcement, and observability. The approach emphasizes the separation of concerns between data ingestion, model execution, governance controls, and monitoring, allowing model evolution without compromising compliance or stability. Drawing on analysis of over 2,000 enterprise deployments, we examine explainability mechanisms, runtime monitoring, and trust establishment through standardized pipelines. Practical considerations for scalability, resiliency, and cross-domain applicability are highlighted. This work demonstrates that well-designed platform architectures are fundamental in transforming sentiment analysis into sustainable enterprise capabilities.

Sentiment Analysis, Enterprise Architecture, AI governance, Explainability, Platform engineering