Skip to content

Latest commit

 

History

13 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 

Repository files navigation

Smart Complaint Management Email Automation System

ProjectBackground

Customer support teams often face high volumes of repetitive email requests, including information inquiries and invoice-related complaints. These processes are typically manual, time-consuming, and prone to human error, leading to inefficiencies and inconsistent service quality.

This project was developed to address these challenges by leveraging automation and AI-powered classification to streamline the handling of customer emails. The solution integrates tools such as AI Builder, Power Automate, Excel, and OneDrive to create a structured, scalable workflow.

As shown in the workflow diagram, the system automates the full lifecycle of a request—from email reception to classification, validation, task creation, and resolution, ensuring improved operational efficiency and service consistency.

Modern Pitch Deck Presentation Template

ExecutiveSummary

This project presents an end-to-end automated workflow designed to manage customer support emails, specifically focusing on information requests and invoice complaints.

The system is triggered whenever a new email is received and uses Natural Language Processing (NLP) to classify the request type, detect sentiment, identify urgency, extract key data (e.g., TIN), and generate a structured summary.

Key functionalities include:

  • Automated email classification (Request Type, Sentiment, Language, Urgency);
  • Intelligent request summarization for operator support;
  • Automatic storage and processing of invoice attachments;
  • Validation rules for complaints (e.g., invoice within 12 months);
  • Integration with Excel for structured data logging;
  • Task creation for operational tracking;
  • Urgency-based alerts via Microsoft Teams;
  • Approval workflow with customer feedback loop (including satisfaction survey).

The solution significantly reduces operational workload, minimizes human error, and accelerates response times, ultimately improving customer satisfaction

imagem imagem

FIRST CONDITION - HAS ATTACHMENTS

imagem

SECOND CONDITION - REQUEST TYPE

imagem

ADD ROW TO EXCEL

imagem

ADD TASK

imagem

THIRD CONDITON - URGENCY

imagem

FORTH CONDITION - APPROVAL

imagem

SURVEY

imagem

Mail Classification

Prompt used: You are a natural language processing assistant. Your task is to analyze the text of a received email and identify the following information: Request Type Classification: Determine whether the email is an "Information Request" or a "Complaint".

Use the following criteria:

  • If the text contains questions, requests for details, or explanations, classify it as an "Information Request".
  • If the text contains dissatisfaction, criticism, reported issues, or complaints, classify it as a "Complaint".

Sentiment Analysis: Classify the overall sentiment of the email as "Positive", "Negative", or "Neutral".

  • Positive: The text shows satisfaction, gratitude, or praise.
  • Negative: The text shows dissatisfaction, frustration, or complaints.
  • Neutral: The text has no clear emotions or uses objective language.

Language Identification: Identify the language in which the email was written. Return the language in ISO 639-1 format (e.g., "pt" for Portuguese, "en" for English, "es" for Spanish).

Tax Identification Number (TIN) Identification: Search for the Tax Identification Number in the email text.

  • If found, return the identified number.
  • If not found, set it as "N/A".

Urgency: Yes or No: Determine whether the email indicates urgency. Use the following criteria: If the text contains words such as "urgent", "immediate", "priority", or mentions tight deadlines, set it as "Yes". Otherwise, set it as "No".

Input:

Expected Output: the following fields in the format:

  • RequestType: "Information Request" or "Complaint".
  • Sentiment: "Positive", "Negative", or "Neutral".
  • Language: ISO 639-1 code of the identified language and the language name in European Portuguese.
  • TIN: Identified number or "N/A" if not found.
  • Urgent: "Yes" or "No".

REQUEST SUMMARY

Prompt used: “Structure the task body based on the following email in European Portuguese.

Identify the customer, include a clear and objective summary of the reported situation, indicating whether it is an information request or a complaint.

Specify any relevant data, such as the customer’s Tax Identification Number (TIN), and detail any additional information or inconsistencies present, such as missing or incorrect attachments.

  • Highlight elements that facilitate the operator’s work, such as urgency, message details, or documentation. Ensure a structured and clear presentation to facilitate understanding.”

Recommendations

Based on the implementation and identified limitations, the following improvements are recommended:

  • Fully integrate AI Builder outputs into the workflow to enable automated validation of invoice data (e.g., Issue Date filtering);
  • Enhance data consistency by including all extracted invoice fields (especially Invoice Date) in the Excel database;
  • Improve system robustness by adding error-handling mechanisms for missing or invalid attachments.

Additional Resources

You can consult the full project presentation here:

🔗 Project Presentation PDF

About

AI-powered system to automate customer support emails, improving efficiency, reducing errors, and enhancing response time and service quality.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors