The relevance of procurement is well established in supply chain management, serving as the connective tissue that enables inter-organizational coordination. At the heart of this connection lies the buyer–supplier negotiation process—the arena where firms interact to secure profitable agreements and establish enduring relationships. Traditionally, supply chains have been built around human interactions, with buyers managing direct relationships with suppliers. However, the advent of Artificial Intelligence (AI) fundamentally challenges this assumption. From traditional machine learning to generative AI and autonomous agents, technological advancements are reshaping not only intra-organizational processes but also inter-organizational ones such as buyer–supplier negotiations. The growing capabilities of AI introduce a new actor at the negotiation table, compelling a reconsideration of how interaction, decision-making, and value creation occur across firm boundaries. Yet, existing research has primarily focused on algorithmic optimization or negotiation support systems, while a proper integration the negotiation process must be examined across three interrelated levels—technical, operational, and social. This thesis investigates how AI integration reconfigures the buyer–supplier negotiation process when conceptualized as a sociotechnical system, also considering how operational performance are affected. The thesis first decomposes the integration of these components by examining the intersections of the three dimensions in pairs, technical and social, operational and technical, and social and operational, and ultimately synthesizing them in a final study that crosses all three. To address the four research objectives, which collectively examine the reconfiguration of buyer–supplier negotiations in the age of AI, the thesis adopts a multi-method research design combining conceptual, qualitative, and quantitative approaches. This design reflects both the complexity of the phenomenon under investigation and its multidimensional nature, spanning technical, social, and operational perspectives. The technical dimension examines how AI capabilities, learning, reasoning, prediction, perception, interaction, adaptation, and creativity, support the negotiation process across pre-negotiation, meeting, and post-negotiation stages. The operational dimension focuses on how these capabilities influence supply chain performance by altering the cost structure of negotiations. In particular, the analysis emphasizes mundane transaction costs, the often-overlooked frictions inherent in everyday inter-organizational exchanges. Findings show that different bundles of AI capabilities produce distinct effects on specific clusters of frictions occurring when inter-firm interaction are activated named mundane transaction costs: attribute defining costs (costs of defining what is negotiated), process defining costs (costs of defining how it is negotiated) and outcome defining costs (costs of defining what has been negotiated). When combining the technical with the social dimension, the analysis reveals how AI capabilities interact with human oversight, giving rise to multiple configurations of human–AI collaboration: level 0 – human-only, level 1 – human in the loop, level 2 – human on the loop, and level 3 – full autonomy. These configurations are contingent on the underlying conditions of relative power and interdependence in supplier relationships. Contexts characterized by buyer dominance favor human-on-the-loop configurations with strong automation, whereas settings with medium or high interdependence favor augmentative configurations that enhance, rather than replace, human judgment. Strategic and relationship involving sensitive negotiations remain predominantly under buyer control. The thesis advances the theoretical understanding of buyer–supplier negotiations by conceptualizing them as hybrid human–AI systems. First, leveraging current literature and empirical insights, it formalises and extends a hybrid-intelligence theory for negotiation, framing the process as a socio-technical system in which human and artificial intelligences jointly create value across all negotiation phases under varying levels of cognitive delegation and power–dependence conditions. Second, it redefines negotiation efficiency through AI, introducing a process-based view of how bundles of AI capabilities shape negotiation structures and costs, and by theorizing a taxonomy of defining costs that explain the role of AI in reducing mundane transaction costs. Third, it integrates technology into power–dependence theory, explaining how structural interdependencies determine the optimal human–AI configuration, automation in low-dependence contexts and augmentation in high-dependence ones. Finally, it contributes a behavioral perspective on AI-supported negotiation, uncovering human-AI interaction (cognitive load, trust, and emotional engagement) mediate performance in hybrid systems, and advancing experimental methodology by incorporating physiological measurement techniques to capture implicit cognitive and emotional responses during AI-assisted negotiation tasks. From a managerial perspective, the thesis provides a structured pathway for integrating AI into buyer–supplier negotiations through four principles. Efficiency alignment urges managers to match AI capabilities to specific negotiation frictions, targeting attribute, process, or outcome defining costs rather than pursuing generic automation. Contextual contingency highlights the need to tailor AI adoption to category and power–dependence structures, balancing automation and augmentation according to relational sensitivity. Human–AI configuration stresses the design of explicit collaboration boundaries, clarifying task allocation, decision rights, and escalation thresholds between humans and AI systems to sustain control and adaptability as technology and interdependence evolve. Finally, behavioral readiness underscores the human side of digital transformation—organizations should invest in training that develops cognitive and emotional preparedness, enabling negotiators to critically interpret AI outputs, calibrate trust, and maintain engagement without overreliance.
La rilevanza dell'approvvigionamento è ben consolidata nella gestione della catena di approvvigionamento, fungendo da tessuto connettivo che abilita il coordinamento inter-organizzativo. Al cuore di questa connessione giace il processo di negoziazione buyer-supplier—l'arena dove le aziende interagiscono per assicurarsi accordi redditizi e stabilire relazioni durature. Tradizionalmente, le catene di approvvigionamento sono state costruite attorno alle interazioni umane, con i buyer che gestiscono relazioni dirette con i supplier. Tuttavia, l'avvento dell'Intelligenza Artificiale (IA) sfida fondamentalmente questa assunzione. Dal machine learning tradizionale all'IA generativa e agli agenti autonomi, i progressi tecnologici stanno riconfigurando non solo i processi intra-organizzativi ma anche quelli inter-organizzativi come le negoziazioni buyer-supplier. Le crescenti capacità dell'IA introducono un nuovo attore al tavolo dei negoziati, costringendo a una riconsiderazione di come l'interazione, il processo decisionale e la creazione di valore si verifichino attraverso i confini organizzativi. Tuttavia, la ricerca esistente si è principalmente concentrata sull'ottimizzazione algoritmica o su sistemi di supporto alla negoziazione, mentre una corretta integrazione del processo di negoziazione deve essere esaminata attraverso tre livelli interconnessi—tecnico, operativo e sociale. Questa tesi indaga come l'integrazione dell'IA riconfiguri il processo di negoziazione buyer-supplier quando concettualizzato come un sistema socio-tecnico, considerando inoltre come le prestazioni operative siano influenzate. La tesi innanzitutto decompone l'integrazione di questi componenti esaminando le intersezioni delle tre dimensioni in coppie, tecnico e sociale, operativo e tecnico, e sociale e operativo, e infine sintetizzandole in uno studio finale che attraversa tutti e tre. Per affrontare i quattro obiettivi di ricerca, che nel complesso esaminano la riconfigrazione delle negoziazioni buyer-supplier nell'era dell'IA, la tesi adotta un disegno di ricerca multi-metodo che combina approcci concettuali, qualitativi e quantitativi. Questo disegno riflette sia la complessità del fenomeno sotto investigazione che la sua natura multidimensionale, coprendo prospettive tecniche, sociali e operative. La dimensione tecnica esamina come le capacità dell'IA, apprendimento, ragionamento, predizione, percezione, interazione, adattamento e creatività, supportino il processo di negoziazione attraverso le fasi pre-negoziazione, incontro e post-negoziazione. La dimensione operativa si concentra su come queste capacità influenzino le prestazioni della catena di approvvigionamento alterando la struttura dei costi dei negoziati. In particolare, l'analisi enfatizza i costi di transazione banali, gli attriti spesso trascurati inerenti agli scambi inter-organizzativi quotidiani. I risultati mostrano che diversi pacchetti di capacità dell'IA producono effetti distinti su specifici cluster di frizioni che si verificano quando le interazioni inter-organizzative sono attivate denominati costi di transazione banali: costi di definizione dell'attributo (costi della definizione di ciò che è negoziato), costi di definizione del processo (costi della definizione di come è negoziato) e costi di definizione dell'esito (costi della definizione di ciò che è stato negoziato). Quando si combinano la dimensione tecnica con quella sociale, l'analisi rivela come le capacità dell'IA interagiscono con la supervisione umana, dando origine a molteplici configurazioni della collaborazione umano-IA: livello 0 – solo umano, livello 1 – umano nel loop, livello 2 – umano sul loop, e livello 3 – autonomia totale. Queste configurazioni sono contingenti sulle condizioni sottostanti di potere relativo e interdipendenza nelle relazioni con i supplier. I contesti caratterizzati dal dominio del buyer favoriscono configurazioni umano-sul-loop con forte automazione, mentre i contesti con interdipendenza media o alta favoriscono configurazioni aumentative che migliorano, piuttosto che sostituiscono, il giudizio umano. Le negoziazioni strategiche e di relazione che coinvolgono negoziati sensibili rimangono prevalentemente sotto il controllo del buyer. La tesi avanza la comprensione teorica delle negoziazioni buyer-supplier concettualizzandole come sistemi ibridi umano-IA. Innanzitutto, sfruttando la letteratura attuale e gli approfondimenti empirici, formalizza ed estende una teoria dell'intelligenza ibrida per la negoziazione, inquadrando il processo come un sistema socio-tecnico in cui le intelligenze umane e artificiali creano congiuntamente valore attraverso tutte le fasi della negoziazione secondo livelli variabili di delega cognitiva e condizioni di dipendenza-potenza. In secondo luogo, ridefinisce l'efficienza della negoziazione attraverso l'IA, introducendo una visione basata sul processo di come i pacchetti di capacità dell'IA modellino le strutture e i costi della negoziazione, e teorizzando una tassonomia di costi di definizione che spiegano il ruolo dell'IA nel ridurre i costi di transazione banali. In terzo luogo, integra la tecnologia nella teoria della dipendenza-potenza, spiegando come le interdipendenze strutturali determinino la configurazione umano-IA ottimale, automazione nei contesti a bassa dipendenza e potenziamento in quelli ad alta dipendenza. Infine, contribuisce una prospettiva comportamentale sulla negoziazione supportata dall'IA, scoprendo come l'interazione umano-IA (carico cognitivo, fiducia e coinvolgimento emotivo) media le prestazioni nei sistemi ibridi, e avanzando la metodologia sperimentale incorporando tecniche di misurazione fisiologica per catturare le risposte cognitive e emotive implicite durante i compiti di negoziazione assistita da IA. Da una prospettiva manageriale, la tesi fornisce un percorso strutturato per integrare l'IA nelle negoziazioni buyer-supplier attraverso quattro principi. L'allineamento dell'efficienza incoraggia i manager a far corrispondere le capacità dell'IA alle specifiche frizioni di negoziazione, mirando ai costi di definizione dell'attributo, processo o esito piuttosto che inseguendo un'automazione generica. La contingenza contestuale evidenzia la necessità di adattare l'adozione dell'IA alle strutture di categoria e dipendenza-potenza, bilanciando automazione e potenziamento secondo la sensibilità relazionale. La configurazione umano-IA sottolinea la progettazione di confini di collaborazione espliciti, chiarendo l'allocazione dei compiti, i diritti decisionali e le soglie di escalation tra gli umani e i sistemi di IA per sostenere il controllo e l'adattabilità mentre la tecnologia e l'interdipendenza evolvono. Infine, la preparazione comportamentale sottolinea il lato umano della trasformazione digitale—le organizzazioni dovrebbero investire nella formazione che sviluppa la prontezza cognitiva ed emotiva, abilitando i negoziatori a interpretare criticamente gli output dell'IA, calibrare la fiducia e mantenere il coinvolgimento senza un eccessivo affidamento.
Integrating artificial intelligence in buyer-supplier negotiations: a technical, operational, and social reconfiguration of a human-intensive process
Borsani, Camilla
2025/2026
Abstract
The relevance of procurement is well established in supply chain management, serving as the connective tissue that enables inter-organizational coordination. At the heart of this connection lies the buyer–supplier negotiation process—the arena where firms interact to secure profitable agreements and establish enduring relationships. Traditionally, supply chains have been built around human interactions, with buyers managing direct relationships with suppliers. However, the advent of Artificial Intelligence (AI) fundamentally challenges this assumption. From traditional machine learning to generative AI and autonomous agents, technological advancements are reshaping not only intra-organizational processes but also inter-organizational ones such as buyer–supplier negotiations. The growing capabilities of AI introduce a new actor at the negotiation table, compelling a reconsideration of how interaction, decision-making, and value creation occur across firm boundaries. Yet, existing research has primarily focused on algorithmic optimization or negotiation support systems, while a proper integration the negotiation process must be examined across three interrelated levels—technical, operational, and social. This thesis investigates how AI integration reconfigures the buyer–supplier negotiation process when conceptualized as a sociotechnical system, also considering how operational performance are affected. The thesis first decomposes the integration of these components by examining the intersections of the three dimensions in pairs, technical and social, operational and technical, and social and operational, and ultimately synthesizing them in a final study that crosses all three. To address the four research objectives, which collectively examine the reconfiguration of buyer–supplier negotiations in the age of AI, the thesis adopts a multi-method research design combining conceptual, qualitative, and quantitative approaches. This design reflects both the complexity of the phenomenon under investigation and its multidimensional nature, spanning technical, social, and operational perspectives. The technical dimension examines how AI capabilities, learning, reasoning, prediction, perception, interaction, adaptation, and creativity, support the negotiation process across pre-negotiation, meeting, and post-negotiation stages. The operational dimension focuses on how these capabilities influence supply chain performance by altering the cost structure of negotiations. In particular, the analysis emphasizes mundane transaction costs, the often-overlooked frictions inherent in everyday inter-organizational exchanges. Findings show that different bundles of AI capabilities produce distinct effects on specific clusters of frictions occurring when inter-firm interaction are activated named mundane transaction costs: attribute defining costs (costs of defining what is negotiated), process defining costs (costs of defining how it is negotiated) and outcome defining costs (costs of defining what has been negotiated). When combining the technical with the social dimension, the analysis reveals how AI capabilities interact with human oversight, giving rise to multiple configurations of human–AI collaboration: level 0 – human-only, level 1 – human in the loop, level 2 – human on the loop, and level 3 – full autonomy. These configurations are contingent on the underlying conditions of relative power and interdependence in supplier relationships. Contexts characterized by buyer dominance favor human-on-the-loop configurations with strong automation, whereas settings with medium or high interdependence favor augmentative configurations that enhance, rather than replace, human judgment. Strategic and relationship involving sensitive negotiations remain predominantly under buyer control. The thesis advances the theoretical understanding of buyer–supplier negotiations by conceptualizing them as hybrid human–AI systems. First, leveraging current literature and empirical insights, it formalises and extends a hybrid-intelligence theory for negotiation, framing the process as a socio-technical system in which human and artificial intelligences jointly create value across all negotiation phases under varying levels of cognitive delegation and power–dependence conditions. Second, it redefines negotiation efficiency through AI, introducing a process-based view of how bundles of AI capabilities shape negotiation structures and costs, and by theorizing a taxonomy of defining costs that explain the role of AI in reducing mundane transaction costs. Third, it integrates technology into power–dependence theory, explaining how structural interdependencies determine the optimal human–AI configuration, automation in low-dependence contexts and augmentation in high-dependence ones. Finally, it contributes a behavioral perspective on AI-supported negotiation, uncovering human-AI interaction (cognitive load, trust, and emotional engagement) mediate performance in hybrid systems, and advancing experimental methodology by incorporating physiological measurement techniques to capture implicit cognitive and emotional responses during AI-assisted negotiation tasks. From a managerial perspective, the thesis provides a structured pathway for integrating AI into buyer–supplier negotiations through four principles. Efficiency alignment urges managers to match AI capabilities to specific negotiation frictions, targeting attribute, process, or outcome defining costs rather than pursuing generic automation. Contextual contingency highlights the need to tailor AI adoption to category and power–dependence structures, balancing automation and augmentation according to relational sensitivity. Human–AI configuration stresses the design of explicit collaboration boundaries, clarifying task allocation, decision rights, and escalation thresholds between humans and AI systems to sustain control and adaptability as technology and interdependence evolve. Finally, behavioral readiness underscores the human side of digital transformation—organizations should invest in training that develops cognitive and emotional preparedness, enabling negotiators to critically interpret AI outputs, calibrate trust, and maintain engagement without overreliance.| File | Dimensione | Formato | |
|---|---|---|---|
|
Phd_Thesis_CB_final_reduced.pdf
non accessibile
Descrizione: Tesi
Dimensione
6.11 MB
Formato
Adobe PDF
|
6.11 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/10589/257998