AI PB: A Grounded Generative Agent for Personalized Investment Insights
Daewoo Park, Suho Park, Inseok Hong, Hanwool Lee, Junkyu Park, Sangjun Lee, Jeongman An, Hyunbin Loh
TL;DR
AI PB tackles regulatory and privacy challenges in deploying generative AI for retail finance by grounding all outputs in verified enterprise data and enforcing zero PII egress through component-based routing. It combines a dual-subsystem architecture with a deterministic orchestrator, a hybrid retrieval pipeline that anchors generations to evidence, and a layered recommender for proactive personalization. In production at Shinhan with on-premises deployment across 24 Nvidia H100 GPUs, it achieves scalable latency (p95 latency under $5.9$ seconds) and strong safety metrics (factuality $>90\%$, safety $>98\%$). This work demonstrates that grounded generation with explicit routing and safety layering can deliver trustworthy, auditable AI insights in high-stakes finance and offers a blueprint for regulated industries.
Abstract
We present AI PB, a production-scale generative agent deployed in real retail finance. Unlike reactive chatbots that answer queries passively, AI PB proactively generates grounded, compliant, and user-specific investment insights. It integrates (i) a component-based orchestration layer that deterministically routes between internal and external LLMs based on data sensitivity, (ii) a hybrid retrieval pipeline using OpenSearch and the finance-domain embedding model, and (iii) a multi-stage recommendation mechanism combining rule heuristics, sequential behavioral modeling, and contextual bandits. Operating fully on-premises under Korean financial regulations, the system employs Docker Swarm and vLLM across 24 X NVIDIA H100 GPUs. Through human QA and system metrics, we demonstrate that grounded generation with explicit routing and layered safety can deliver trustworthy AI insights in high-stakes finance.