Chatting with the Middleware: A Graph-Based Agentic Al Assistant for Flow Discovery and Troubleshooting Support

A graph-based AI assistant helps SWISS navigate complex airline middleware. By combining knowledge graphs, agentic AI, and natural language interaction, it streamlines flow discovery and troubleshooting, proving effective in enterprise environments.

Franco D'Agostino, 2025

Type of Thesis Bachelor Thesis
Client Swiss International Air Lines AG
Supervisor Montecchiari, Devid
Views: 3
SWISS's middleware platform connects numerous software systems essential for airline operations, from booking to flight management to crew scheduling. This creates highly complex networks with intricate dependencies that are difficult to document, understand and maintain. Finding information, identifying problems, tracing system relationships, and training new staff requires significant time and specialized technical expertise. Traditional static documentation cannot keep pace with the constantly evolving system landscape and growing operational complexity.
Using Design Science Research methodology, we developed a comprehensive digital knowledge graph mapping SWISS's middleware components and their complex relationships. An intelligent AI assistant was implemented using agentic GraphRAG technology and Model Context Protocol (MCP) architecture to understand natural language questions and autonomously navigate this intricate system landscape. The solution underwent rigorous evaluation across varying complexity scenarios by middleware experts, systematically measuring effectiveness and efficiency for system exploration and troubleshooting tasks.
The proof-of-concept demonstrates significant operational value for SWISS. The AI assistant consistently exceeded expectations across all tested complexity levels, proving its effectiveness for enterprise middleware environments. Potential productive implementation benefits: Accelerated onboarding - new staff explore system relationships through natural language queries rather than weeks reading technical documentation; Knowledge democratization - business analysts and operations teams access technical architecture details independently without specialized expertise; Improved troubleshooting - simplified dependency analysis accelerates problem resolution and root cause identification in complex business flows; Seamless integration - modular MCP based architecture enables for sameless integration with existing tools e.g. Visual Studio Code. The prototype validates the approach while requiring additional development for production readiness. Beyond immediate SWISS applications, the solution establishes a foundation for automated middleware management and shows clear applicability to other enterprise environments facing similar integration challenges.
Studyprogram: Business Information Technology (Bachelor)
Keywords Knowledge Graphs, Agentic AI, Airline Middleware, Enterprise Integration, Model Context Protocol (MCP)
Confidentiality: vertraulich
Type of Thesis
Bachelor Thesis
Client
Swiss International Air Lines AG, Basel
Authors
Franco D'Agostino
Supervisor
Montecchiari, Devid
Publication Year
2025
Thesis Language
English
Confidentiality
Confidential
Studyprogram
Business Information Technology (Bachelor)
Location
Basel
Keywords
Knowledge Graphs, Agentic AI, Airline Middleware, Enterprise Integration, Model Context Protocol (MCP)