Thanks to visit codestin.com
Credit goes to github.com

Skip to content

duuinh/InboxFlow

Repository files navigation

InboxFlow

Submitted for CWB Hackathon 2025

Automate and enhance the handling of incoming customer service emails in a banking environment using modular AI-powered agents that collaborate intelligently to streamline operations.

Overview

The solution uses Azure Logic Apps to automate the end-to-end workflow of handling service requests via email. The steps are:
  1. Trigger: When a new email arrives in the monitored inbox (When a new email arrives (V2)).
  2. Processing: The app invokes the EmailProcessor Azure Function (service-request-handling-agent-EmailProcessor), which analyzes the email content, determines intent, extracts data, and interacts with the SR system.
  3. Response: A reply email is automatically sent to the original sender (Send an email (V2)).

Multi-Agent Orchestration

The system uses the Handoff Orchestration pattern in Microsoft Semantic Kernel. Each agent has a focused role, and agents hand off control to the next stage in the workflow.

Agent Responsibilities

  • EmailTriageAgent Classifies the incoming email to determine if it’s a new service request or a follow-up.

  • SRCreatorAgent Handles service request creation based on email content, and passes control to ReplyGeneratorAgent.

  • FollowUpHandlerAgent Checks SR status for follow-up emails, and passes control to ReplyGeneratorAgent.

  • ReplyGeneratorAgent Generates the final customer reply using structured logic depending on the SR state (created, in progress, closed, or not found).

Orchestration Flow

[EmailTriageAgent]
   ├──> [SRCreatorAgent] ─────┐
   └──> [FollowUpHandlerAgent] ───┐
                                  └──> [ReplyGeneratorAgent]

This modular pipeline enables intelligent, dynamic agent collaboration to automate decision-making and response generation in customer service workflows.

Scope 

  • Read and classify customer emails using LLM.
  • Automatically create service requests via API.
  • Reply with personalized responses using LLM.
  • Handle follow-ups by checking SR status and sending updates.
  • Orchestrate the flow using Semantic Kernel deployed on Azure Functions.

Tech Stack

  • Microsoft Semantic Kernel
  • Azure Functions (.NET)
  • Azure Logic Apps
  • LLM (Azure OpenAI GPT-4.1-nano)
  • GitHub API
  • Azure Key Vault
  • Azure Application Insights

Try It Yourself!

Send a service request email to our monitored inbox at: duuinh [at] outlook [dot] com

Try sending:

  • A new service request (e.g., request a new credit card, report a lost card)
  • A follow-up email asking for status updates on your existing requests

Watch how you get a personalized response — powered by AI agents working behind the scenes!

About

Email Agent using Azure services — CWB hackathon 2025

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages