Global European Bank (unnamed)

Trader surveillance required continuous monitoring of trading data, e-communications, voice recordings, and P&L information to detect market abuse including manipulation, insider dealing, and MNPI misuse. Risk analysts manually reconstructed investigation scenarios and documented findings from surveillance alerts — a highly labor-intensive process prone to human error that consumed significant analyst bandwidth.

HCLTech implemented a GenAI-enabled trade surveillance platform featuring context-specific content extraction from structured and unstructured datasets queryable via natural language prompts. The solution was developed through five co-innovation workshops and prototyping with LLMs for regulatory functions, then scaled via HCLTech's Enterprise AI Foundry onto a Google Cloud platform using Google's Gemini model for enhanced performance.

Data preparation time for data scientists dropped 80%, from several hours to approximately 15 minutes per dataset. Overall analysis time fell from 8.75 hours to just 25 minutes. New analysis turnaround improved by 90% to ~10 minutes through AI-powered news aggregation and summarization, and alert resolution time decreased from 11 to 9 minutes.

• Natural language querying of surveillance datasets dramatically reduces analyst preparation burden, freeing capacity for complex investigations.
• Co-innovation workshops between vendor and client teams are critical for validating LLM applicability in regulated financial functions before scaling.
• Migrating from open-source LLMs to a production-grade model (Google Gemini) via an Enterprise AI Foundry framework enables reliable scaling of pilots.