This dissertation investigates the evolving role of automation in supply chain management, focusing on how Robotic Process Automation (RPA) and Agentic AI can support medium-sized manufacturing enterprises in addressing operational inefficiencies, data fragmentation and rising process complexity. A dual methodological approach is adopted. First, a literature review reconstructs the technological evolution from RPA to Intelligent Automation, Hyperautomation and Agentic AI, highlighting capabilities, limitations and organisational implications. Second, an empirical investigation based on semi-structured interviews with eleven medium-sized firms examines three core and interdependent supply chain processes—demand forecasting, order processing and inventory and replenishment management. These processes were selected due to their repetitiveness, data intensity and strong cross-process couplings, which make them both highly impactful and well suited for automation. The empirical results reveal recurring inefficiencies driven by manual activities, fragmented information systems and heterogeneous levels of digital maturity. Through thematic coding, cross-case comparison and expert validation, the research identifies the organisational, technological and process-related conditions under which automation can be effectively adopted. Building on these insights, the dissertation develops an Applicability Framework that supports firms in assessing readiness, selecting appropriate automation paradigms and designing scalable implementation pathways. In addition, an agentic system prototype (Solution S12) is designed and validated, demonstrating how multi-agent architectures can operate within bounded autonomy to augment human decision-making and coordinate data- intensive tasks. The dissertation contributes a unified conceptual reconstruction of automation evolution, an empirically grounded framework for medium-sized enterprises and a system architecture illustrating the practical feasibility of agentic solutions. It concludes by outlining the study’s limitations and indicating future research directions, particularly the need for quantitative validation and for deeper integration of agentic systems with process mining, digital twins and physical automation technologies.
La presente tesi analizza il ruolo emergente dell’automazione nella gestione della supply chain, con particolare attenzione a come la Robotic Process Automation (RPA) e l’Agentic AI possano supportare le imprese manifatturiere di medie dimensioni nel fronteggiare inefficienze operative, frammentazione dei dati e crescente complessità dei processi. È stato adottato un approccio metodologico duale. In primo luogo, una revisione della letteratura ricostruisce l’evoluzione tecnologica da RPA all’Intelligent Automation, alla Hyperautomation e fino all’Agentic AI, evidenziandone capacità, limiti e implicazioni organizzative. In secondo luogo, un’indagine empirica basata su undici interviste semi-strutturate esamina tre processi chiave e interdipendenti della supply chain—demand forecasting, order processing e inventory and replenishment management—selezionati per la loro ripetitività, intensità informativa e forte interconnessione, che li rende particolarmente adatti e critici per l’automazione. I risultati empirici mostrano inefficienze ricorrenti legate ad attività manuali, sistemi informativi frammentati e livelli eterogenei di maturità digitale. Attraverso codifica tematica, analisi cross-case e validazione esperta, la ricerca identifica le condizioni organizzative, tecnologiche e di processo che determinano l’efficace adozione dell’automazione. Sulla base di tali evidenze, la tesi sviluppa un Framework di Applicabilità che supporta le imprese nella valutazione della readiness, nella selezione dei paradigmi tecnologici e nella progettazione di percorsi di implementazione scalabili. Inoltre, viene progettato e validato un prototipo agentico (Soluzione S12), che dimostra come architetture multi-agente possano operare con autonomia governata, supportando il decision-making e coordinando attività data-intensive. La tesi offre un contributo concettuale, metodologico e applicato, e si conclude delineando i limiti dello studio e possibili sviluppi futuri, tra cui validazioni quantitative e integrazione con process mining, digital twin e automazione fisica.
From RPA to agentic AI: enhancing Supply Chain process execution in medium-sized enterprises
ESPOSITO, IACOPO
2024/2025
Abstract
This dissertation investigates the evolving role of automation in supply chain management, focusing on how Robotic Process Automation (RPA) and Agentic AI can support medium-sized manufacturing enterprises in addressing operational inefficiencies, data fragmentation and rising process complexity. A dual methodological approach is adopted. First, a literature review reconstructs the technological evolution from RPA to Intelligent Automation, Hyperautomation and Agentic AI, highlighting capabilities, limitations and organisational implications. Second, an empirical investigation based on semi-structured interviews with eleven medium-sized firms examines three core and interdependent supply chain processes—demand forecasting, order processing and inventory and replenishment management. These processes were selected due to their repetitiveness, data intensity and strong cross-process couplings, which make them both highly impactful and well suited for automation. The empirical results reveal recurring inefficiencies driven by manual activities, fragmented information systems and heterogeneous levels of digital maturity. Through thematic coding, cross-case comparison and expert validation, the research identifies the organisational, technological and process-related conditions under which automation can be effectively adopted. Building on these insights, the dissertation develops an Applicability Framework that supports firms in assessing readiness, selecting appropriate automation paradigms and designing scalable implementation pathways. In addition, an agentic system prototype (Solution S12) is designed and validated, demonstrating how multi-agent architectures can operate within bounded autonomy to augment human decision-making and coordinate data- intensive tasks. The dissertation contributes a unified conceptual reconstruction of automation evolution, an empirically grounded framework for medium-sized enterprises and a system architecture illustrating the practical feasibility of agentic solutions. It concludes by outlining the study’s limitations and indicating future research directions, particularly the need for quantitative validation and for deeper integration of agentic systems with process mining, digital twins and physical automation technologies.| File | Dimensione | Formato | |
|---|---|---|---|
|
2025_12_Esposito_Tesi.pdf
accessibile in internet per tutti
Dimensione
3.4 MB
Formato
Adobe PDF
|
3.4 MB | Adobe PDF | Visualizza/Apri |
|
2025_12_Esposito_Executive Summary.pdf
accessibile in internet per tutti
Dimensione
958.56 kB
Formato
Adobe PDF
|
958.56 kB | 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/10589/246935