Citi · Wealth · 2026

Client Insights AI Agent

Conversational AI with entitlements, plus a CRM summariser

Conversational AI over production client data, where every answer respects who is asking.

  • Python
  • GenAI
  • LLMs
  • RBAC
  • React
outcome classes for call notes
20
answers by managed segment
Role-aware

The problem

Sales and admin teams needed fast answers about their clients, and their meeting notes were unstructured text nobody could analyse at scale. Any AI on top had to respect strict entitlements.

What I built

  • Chat with session history for GenAI conversations on production data
  • Entitlement-aware results scoped to each user's managed segments
  • Role-based access for Admin and Sales users
  • A GenAI summariser that turns unstructured client call notes into structured data
  • Classification of notes into 20 predefined outcomes for quick client-level analysis

Built at Citi. Descriptions stay at the level of problem, capability and outcome. No internal data, algorithms or client details are shared.

Want to talk about systems like this?

Email me