Lab for AI-Agents in Business and Economics
Lab Directors
Prof. Ye Luo
Associate Director, Institute of Digital Economy and Innovation
Associate Professor, Economics and Finance
Prof. Xin Tong
Professor, Economics and Innovation and Information Management
Mission
Our mission is to pioneer AI-driven solutions for business challenges by developing multi-agent systems and domain-specific AI architectures, while guiding organizations through ethical, scalable, and transformative AI adoption.
Core Objectives
Multi-Agent System
To develop platforms for AI agents and multi-agent architecture for business, management and AI-transformation of organizations
Domain-Specific AI Agents
To study and design AI agents for specific business applications with deep domain knowledge
AI Impact Research
To study and understand the economic impact of AI transformation on economic behavior of human, business and organizations
Major Projects
Econometric AI
This project is an international collaborative open source initiative aiming to build an AI-agent system that provides solid econometric knowledge, code-generation, and execution capability for education and research in applied econometrics. The system would consist of the following key features:
Econometrics AI Agent
We develop an “Econometrics AI Agent” which exhibits outstanding performance in planning econometric tasks strategically, generating and executing code, employing error-based reflection for improved robustness, and allowing iterative refinement through multi-round conversations. It establishes a testbed for exploring AI’s impact on social science research and enables cost-effective integration of domain expertise, making advanced econometric methods accessible to users with minimal coding skills.
Econometrics Helper
A chat based agent for explaining and suggesting econometric method and identification strategy.
Econometrics Recommender
An RAG based agent with recommendation ability build on theoretical and applied econometrics papers published at top journals along with datasets.
Econometrics Writer
A latex copilot writer agent that helps thesis writers with citations and references, mathematical formulas, table compiling, etc., in applied econometrics.
LangGraph Agent Rollback System
Explore a live, sandboxed demo of conversation rollback and branching. Create checkpoints, branch timelines, and see how tool calls auto-checkpoint. Use this to understand how AI agents can safely explore alternatives while preserving original state.
Interactive Demo
🌳 Session Branch Tree
✨ Key Features
Full Snapshot
Complete conversation state saved at each checkpoint
Branching
Create multiple timelines from any checkpoint
Tool Rollback
Automatically reverse tool operations
Auto-Checkpoint
Smart saves after tool execution
💻 Code Example
from src.agents.rollback_agent import RollbackAgent
from langchain_openai import ChatOpenAI
# Create agent with rollback capabilities
agent = RollbackAgent(
external_session_id=session_id,
model=ChatOpenAI(model="gpt-4"),
tools=[calculate_sum, fetch_data, save_to_db],
auto_checkpoint=True # Auto-save after tools!
)
# Have a conversation
response = agent.run("Calculate 42 + 58")
# Auto-checkpoint created after tool use!
# Create manual checkpoint
checkpoint = agent.create_checkpoint_tool("Important State")
# Continue conversation
response = agent.run("Now multiply by 2")
# Rollback when needed - creates new branch!
new_branch = RollbackAgent.from_checkpoint(checkpoint_id)
# Original timeline preserved, new branch active
response = new_branch.run("Let's try division instead")
🎯 Real-World Use Cases
Customer Support
Undo mistakes, explore solutions safely
Research & Testing
A/B test conversation strategies
Interactive Games
Save/load NPC conversation states
Education
Safe exploration for students