Gauri Gupta 开源 agentic_ai_customer_copilot：银行客户增长多智能体 AI copilot

Agentic AI Banking Customer Growth Platform

A multi-agent AI copilot that helps bank executives identify and act on customer growth opportunities in real time — during live calls.

What it does

During a customer call, a bank executive pastes the customer's profile data. Five AI agents fire in sequence:

1. Customer Intelligence Agent — extracts behavioural signals and spending patterns
2. Product Match Agent — searches a vector database of bank products using semantic similarity
3. Opportunity Discovery Agent — identifies the highest-value growth opportunity
4. Offer Construction Agent — builds a tailored value proposition with benefits and objection handling
5. Executive Coach Agent — generates a privacy-safe personalised opening line

The result is a live Opportunity Card shown on the dashboard — ready before the executive even finishes their greeting.

Tech stack

• Claude claude-haiku-4-5 (Anthropic) — powers all 5 agents
• ChromaDB — vector database for semantic product search
• FastAPI + uvicorn — REST API connecting the UI to the pipeline
• Vanilla HTML/CSS/JS — professional dashboard UI with dark mode

Project structure

agentic-banking-platform/ ├── notebooks/ │ ├── customer_intelligence_agent.ipynb │ ├── opportunity_discovery_agent.ipynb │ └── banking_pipeline.ipynb ├── app.py # FastAPI backend ├── dashboard.html # Live call dashboard UI ├── .gitignore └── README.md

Setup

1. Clone the repo
2. Install dependencies:

   pip install anthropic chromadb fastapi uvicorn python-dotenv sentence-transformers

1. Create a .env file: ANTHROPIC_API_KEY=your_key_here

2. Start the server:

   python -m uvicorn app:app --reload

1. Open dashboard.html in your browser

Built by

Gauri Gupta · MSc Business Analytics