Automation of Incident and Inquiry Management with Artificial Intelligence in a Technology Company

 

Automation of Incident and Inquiry Management with Artificial Intelligence in a Technology Company

Author: Francisco Prats Quílez

Introduction

In the modern business environment, efficiency in managing incidents and inquiries is crucial to maintaining high levels of customer satisfaction and optimizing internal resources. This report analyzes the implementation of an automated system based on artificial intelligence to manage and classify inquiries and incidents received via email in a technology company.

Objective

The main objective of this project is to develop an automated system that efficiently manages received incidents and inquiries, classifying them and assigning them to the appropriate personnel using artificial intelligence. This system aims to improve response capability, reduce resolution time, and provide a detailed history of incidents for future reference.

Development

• Email Monitoring

Using the pywin32 library in Python, a module has been developed to continuously monitor the company's email inbox. This module detects new emails and extracts their content for further processing.

• Incident/Inquiry Classification

A large language model (LLM) is used to automatically classify emails into different user-editable categories in the Project Manager AI application. These categories include Technical Issues, Usage Inquiries, Customization Requirements, among others.

• Additional Data Management

The user interface allows the addition of relevant information for managing the inquiry or incident, such as product types, company description, and staff characteristics.

• Incident and History Tables

Open incidents/inquiries are displayed in an interactive table, while a detailed history allows access to past incidents, facilitating the analysis and resolution of future cases.

• Incident Assignment

The backend uses the collected information to break down and classify the incident, automatically assigning it to the appropriate personnel using the LLM.

• Analysis with History

Using Retrieval-Augmentation-Generation (RAG) techniques, the incoming incident or inquiry is compared with the stored incident and inquiry history. This allows identifying patterns and similarities with previous cases, providing faster and more accurate solutions. The system searches for similar past incidents and suggests resolutions based on previous experiences, continuously improving the effectiveness of responses.

• Integration with Project Managers

Using APIs from programs like ClickUP, Asana, and Jira, tasks are automatically created in the chosen project manager, facilitating the tracking and resolution of incidents.

• Database Storage

All incidents are stored in a database to maintain a historical record and allow the reuse of information in resolving similar future incidents.

• Report Generation

A .docx report is automatically generated with the incident data and a preliminary analysis conducted by the LLM. For example, the report may include data from the manufacturing process of that product, as well as the quality report.

Conclusions

The implementation of this system has proven effective in automating the management of incidents and inquiries, significantly reducing response times and improving the accuracy of task assignment. Integration with project managers and storage in a database facilitates thorough tracking and quicker resolution of recurring problems.

Future Development

Potential improvements for this system include:

• Prompt Improvement: Optimization of the prompts used by the LLM to enhance the accuracy and relevance of responses and classifications.
• API Expansion: Integration with more project management platforms for greater flexibility and adaptability.
• Predictive Analysis: Implementation of predictive models to anticipate possible incidents and propose proactive solutions.
• Advanced Customization: Allowing more detailed and specific configurations for each type of incident, better adapting to the company's particular needs.
• Enhanced User Interface: Development of a more intuitive and feature-rich user interface for managing and visualizing incidents.

Continuous improvement of this system promises even more efficient, adaptable, and proactive incident and inquiry management in the future.