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