This work investigates a novel application of Large Language Models (LLMs) to enhance Business Process Model and Notation (BPMN) diagram generation within the context of Digital Process Automation based on the Hyperautomation platform Appian. The research addresses the complexities present technical BPMN modeling, which demand deep technical knowledge and manual effort. The diffusion of advanced Generative Models, such as Gemini, has brought new opportunities for aiding modelers and developers tackle these challenges. This research is situated at the confluence of digital transformation, advanced Business Process Management (BPM), and generative AI. Its objective is to explore the feasibility of leveraging state-of-the-art LLM capabilities, particularly those of the Gemini family, to assist Appian developers in efficiently building Appian Process Models. This way, the HyperMod (Hyperautomation Process Modeling) tool is born. This work explores a direction that has received less attention up to now, the use of LLMs to model intricate technical processes that integrate data flow, database interactions, user interfaces, and Appian-specific nodes, as well as standard BPMN 2.0. The work pushes the boundaries of LLM capabilities by requiring them to construct entirely custom BPMN 2.0 diagrams that incorporate "made-up" new elements derived from Appian logic and implemented inside BPMN.io as an extension. A significant challenge overcome is maintaining the structural integrity, logical cohesion, and overall soundness of these diagrams, following both pre-existing BPMN rules and newly established conventions. The research uses state-of-the-art language models like Gemini 2.5 Pro to enable a more unconstrained approach to building BPMN process model generation tools. It does not require user queries in constrained language or a simplified version BPMN 2.0 XML output. Finally, this research explores the human-AI collaboration in process modeling. It allows the user to manually modify the process models while transforming the LLM into an interactive chatbot that assists in the building, brainstorming, and analysis of technical process models; it shifts AI’s role from a simple query engine to an active, intelligent partner in the design process. This could help Digital Transformation, where AI can aid developers in more efficiently conceptualizing and implementing.
Questo lavoro esplora un'applicazione dei Large Language Models (LLM) per migliorare la generazione di diagrammi BPMN (Business Process Model and Notation) nel contesto dell'Automazione Digitale dei Processi, basata sulla piattaforma di Hyperautomation Appian. La ricerca affronta le complessità della modellazione tecnica BPMN, che richiede una conoscenza tecnica approfondita e un considerevole sforzo manuale. La diffusione dei modelli generativi avanzati, come Gemini, ha aperto nuove opportunità per supportare modellatori e sviluppatori nell'affrontare queste sfide. Questa ricerca si colloca all'incrocio tra trasformazione digitale, gestione avanzata dei processi aziendali (BPM) e intelligenza artificiale generativa. L’obiettivo è quello di esplorare la fattibilità di sfruttare le capacità all’avanguardia degli LLM, in particolare quelle della famiglia Gemini, per assistere gli sviluppatori Appian nella costruzione efficiente di modelli di processo Appian. Nasce così lo strumento HyperMod (Hyperautomation Process Modeling). Questo lavoro esplora una direzione che finora ha ricevuto poca attenzione, l'applicazione degli LLM alla modellazione di processi tecnici complessi che integrano flussi di dati, interazioni con database, interfacce utente e nodi specifici di Appian, oltre agli elementi standard del BPMN 2.0. Il lavoro spinge i limiti delle capacità degli LLM, richiedendo loro di costruire diagrammi BPMN 2.0 completamente personalizzati che incorporano elementi “inventati” derivati dalla logica di Appian e implementati in BPMN.io come estensione. Una delle sfide principali affrontate è stata mantenere l’integrità strutturale, la coerenza logica e la validità complessiva di questi diagrammi, rispettando sia le regole BPMN esistenti che le nuove convenzioni introdotte. Questo lavoro usa modelli di linguaggio di ultima generazione come Gemini 2.5 Pro per un approccio meno vincolato alla generazione di modelli BPMN, che non necessita di query utente in linguaggio semplificato o output XML BPMN 2.0 ridotto. Infine, questa ricerca esplora la collaborazione persona-AI nella progettazione di modelli BPMN. Consente all’utente di modificare manualmente i modelli di processo, trasformando l’LLM in un chatbot interattivo che assiste nella costruzione, nel brainstorming e nell’analisi di modelli tecnici di processo. Si passa così da un semplice motore di interrogazione a un partner intelligente e attivo nel processo progettuale. Ciò potrebbe aiutare la Trasformazione Digitale, dove l’IA può aiutare gli sviluppatori a concepire, implementare e distribuire flussi di lavoro automatizzati in modo più efficiente.
HyperMod: BPMN model generation as an aid to Hyperautomation
Pardo Gutierrez, Sergio
2024/2025
Abstract
This work investigates a novel application of Large Language Models (LLMs) to enhance Business Process Model and Notation (BPMN) diagram generation within the context of Digital Process Automation based on the Hyperautomation platform Appian. The research addresses the complexities present technical BPMN modeling, which demand deep technical knowledge and manual effort. The diffusion of advanced Generative Models, such as Gemini, has brought new opportunities for aiding modelers and developers tackle these challenges. This research is situated at the confluence of digital transformation, advanced Business Process Management (BPM), and generative AI. Its objective is to explore the feasibility of leveraging state-of-the-art LLM capabilities, particularly those of the Gemini family, to assist Appian developers in efficiently building Appian Process Models. This way, the HyperMod (Hyperautomation Process Modeling) tool is born. This work explores a direction that has received less attention up to now, the use of LLMs to model intricate technical processes that integrate data flow, database interactions, user interfaces, and Appian-specific nodes, as well as standard BPMN 2.0. The work pushes the boundaries of LLM capabilities by requiring them to construct entirely custom BPMN 2.0 diagrams that incorporate "made-up" new elements derived from Appian logic and implemented inside BPMN.io as an extension. A significant challenge overcome is maintaining the structural integrity, logical cohesion, and overall soundness of these diagrams, following both pre-existing BPMN rules and newly established conventions. The research uses state-of-the-art language models like Gemini 2.5 Pro to enable a more unconstrained approach to building BPMN process model generation tools. It does not require user queries in constrained language or a simplified version BPMN 2.0 XML output. Finally, this research explores the human-AI collaboration in process modeling. It allows the user to manually modify the process models while transforming the LLM into an interactive chatbot that assists in the building, brainstorming, and analysis of technical process models; it shifts AI’s role from a simple query engine to an active, intelligent partner in the design process. This could help Digital Transformation, where AI can aid developers in more efficiently conceptualizing and implementing.File | Dimensione | Formato | |
---|---|---|---|
2025_07_Pardo.pdf
accessibile in internet per tutti
Dimensione
2.3 MB
Formato
Adobe PDF
|
2.3 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/240856