Artificial intelligence for supplier scouting: an information processing theory approach

The COVID-19 pandemic, the global chip shortage of 2020 and 2021, and the scarcity of raw materials are just the most recent disruptions plaguing supply chains (Manupati et al., 2022; Chen et al., 2022) and making procurement increasingly complex. Buyer firms have aggressively searched for alternative solutions for obtaining necessary resources, often through scouting for new suppliers (Scoutbee, 2021). This phenomenon is very likely to remain a valid concern in the future as firms seek to streamline the scouting and identification of new short-term suppliers.

Even before the latest emergencies, supplier scouting was relevant, as buyer firms have always needed to gather information on potential suppliers, assess their offerings and select the best one (Monczka et al., 2016). Thanks to structured, responsible supplier scouting, procurement can contribute to value creation. Indeed, supplier selection impacts the sustainability of a firm (Bag, 2020) because suppliers need to demonstrate their reliability to buyers.

In today's volatile and uncertain business environments, firms are reshaping the way they manage their supply chain by developing the current supply base (e.g. Calvi et al., 2020) and scouting for new suppliers (e.g. Zhan et al., 2021; Saghiri and Mirzabeiki, 2021). Supplier scouting, defined as a buyer firm's market exploration to identify potential new suppliers (Luzzini et al., 2014; Lee et al., 2011), has become crucial and requires the development of new models and digital tools for data handling and analysis.

The strategic management of procurement in general and supplier scouting in particular (Bienhaus and Haddud, 2018) requires the adoption of digital technologies (Batran et al., 2017; Lorentz, 2021). In this vein, artificial intelligence (AI) allows buyers to upgrade their scouting for new suppliers through the automation of activities and through AI's predictive power (Handfield et al., 2019). Although scholars have not yet delved into the topic, the role of AI in supplier scouting is a clear avenue for expansion among industry professionals. The most striking example is the success of Scoutbee, a start-up that uses AI to transform the way organizations discover and connect with suppliers. In 2019, Scoutbee was recognized in the technology category at the World Procurement Awards; in 2022, it ranked 31st in Procurement Magazine's top 100 companies ^[1].

However, for most buyer firms, the efforts to digitize supplier scouting are still limited, thus preventing digital integration (Richey et al., 2007; Singh et al., 2018). Thus, the object of this study is to investigate the role of AI solutions in the supplier scouting process in terms of existing applications and the algorithms behind them. We focus on the information required for supplier scouting and the capabilities necessary for the exploitation of such information as enabled by AI technology. For this reason, information processing theory (IPT) is the appropriate theoretical lens for this study. The validity of this theory's application to AI in supplier scouting is corroborated by previous research (i.e. Bensaou and Venkatraman, 1995; Cegielski et al., 2012; Lorentz et al., 2020) although the role of AI in supplier scouting has not been investigated so far.

The paper addresses one exploratory research question:

The research was conducted through 12 case studies of procurement technologies being used by information providers, with a view towards leveraging their expertise on AI-based scouting solutions and their use by buyer firms.

To address this research question, this paper relies on the two constructs of IPT: we first identify the buyer firms' information processing needs (IPNs) in the supplier scouting process; then we consider the information processing capabilities (IPCs) enabled by the adoption of AI-based supplier scouting solutions.

Supplier scouting consists of market exploration used to identify potential new suppliers (Luzzini et al., 2014; Lee et al., 2011) and to increase knowledge about the supply market (Spina, 2008).

There are several factors that trigger the scouting process among buyer firms. The first and most frequent trigger is the lack of a qualified supplier that can supply what is required. However, scholarly opinion regarding this phenomenon varies. According to Spina (2008), if the development of a new product requires a component that can be supplied by qualified suppliers, it is preferable to rely on them again. In this way, fewer burdens are incurred in scouting and qualifying new suppliers, and the risk of establishing the relationship is lower (Spina, 2008; Bartezzaghi and Ronchi, 2005).

Sundquist and Melander (2021) discuss the new interfaces between buyer and supplier that are triggered by new product development. Indeed, the unavailability of a needed resource triggers scouting for a new supplier that can provide it. This affects the characteristics of the new product and the configuration of the network of the focal firm that is looking for the new resource, which can be material, financial or intellectual (Park and Lee, 2018).

These relationships and interfaces affect buyers and suppliers and lead to the involvement of other actors in the scouting activities (Waluszewski et al., 2019). Indeed, in the exchanges between buyer and supplier, scouting activities are often the triggering event, with the IT provider playing a bridging role between the two firms (Bartezzaghi and Ronchi, 2005; Sundquist and Melander, 2021).

According to Melnyk et al. (2010), the scouting process – and supply base management more generally – must become a source of leverage to strengthen the competitiveness of the buyer firm. The scouting phase must be managed strategically, as the supplier base, like the business environment in which it operates, is dynamic and changes over time. According to Melnyk et al. (2010), scouting activities should also aim to enhance the buyer firm's attractiveness as a potential partner for suppliers. Supplier scouting also fulfils a competitive intelligence function within the supply chains of major competitors (Spina, 2008).

Scouting for new suppliers can be conducted in many ways. Often, the suppliers themselves approach the buyer firm with a commercial offer (Melnyk et al., 2010). Otherwise, the buyer will look for new partners by attending industry expositions, surfing the internet, consulting industry-specific journals and/or meeting informally with professionals from other firms (Spina, 2008; Lee et al., 2012).

Sometimes, the buyer firm leverages business agencies and websites (such as Alibaba or Amazon Business). These solutions are gaining traction in the industry, as they foster connections among players of any size and location (Sundquist and Melander, 2021). However, they are mostly exploited for less relevant purchasing categories, as they often provide very general and succinct information about potential suppliers (Lee et al., 2012). Buyer firms can also resort to services operators for industrial companies (Bartezzaghi and Ronchi, 2005). These operators equip the buyer firm with the electronic platform and software applications necessary to support the scouting activity. In this way, the buyer can collect additional information about the suppliers regarding product descriptions, certifications, etc. (Bartezzaghi and Ronchi, 2005). Thus, in most buyer companies, a structured supplier scouting process does not exist, despite its strategic importance.

In the supplier scouting process, high-quality information is crucial (Hazen et al., 2014). However, buyers often lack access to up-to-date databases, which are expensive and may be incomplete. Alternatively, buyers must gather information manually, thus wasting their efforts given the amount of data required to perform supply market scouting. In this way, the scouting process often results in a high degree of uncertainty. Therefore, buyers need to be aware of suppliers' information processing needs (IPNs), including the type of data required and their sources (Handfield et al., 2019). Nowadays, the amount of information available is enormous, but data processing requires appropriate techniques and capabilities (Zhu et al., 2019). For this reason, AI can play a fundamental role in guiding supplier scouting.

AI is still missing a unique definition: computer science focuses on creating intelligent systems capable of replicating human behaviour, while engineering focuses on AI for problem-solving (Guo and Wong, 2013). The same holds true for the application of AI in business: a single classification is difficult to delineate. In his definition, Min (2010) emphasizes more precisely the cognitive aspect of AI and the support it provides in solving practical problems: “Artificial intelligence is referred to as the use of computers for reasoning, recognizing patterns, learning or understanding certain behaviours from experience, acquiring and retaining knowledge, and developing various forms of inference to solve problems in decision-making situations where optimal or exact solutions are either too expensive or difficult to produce” (pp. 13–14).

Table 1 presents the key definitions of AI techniques and the algorithms relevant to the present research.

In supply chain management, AI facilitates supply chain analysis by processing a wide variety of data sources to identify market trends and predict customer preferences (Srinivasan and Swink, 2018). Many applications of AI are designed to support the selection of the best supplier, often functioning as multi-criteria decision models (Ho et al., 2010). Looking at previous research, Pitchipoo et al. (2013) introduced a hybrid decision-making model to evaluate and select the supplier based on a multi-criteria approach. In the same vein, Zair et al. (2019) designed an agent-based model in which negotiation and supplier selection are conducted by a purchasing dyad, i.e. the buyer and the buyer's customers, with the aim of including customer preferences in the negotiation algorithm and thus leading the buyer to choose the best supplier. Scott et al. (2015) proposed an integrated method to deal with multicriteria and multistakeholder supplier selection using a combined analytic hierarchy process/quality function deployment. However, these applications perform the selection of the best supplier from a list of potential partners already available to the buyer firm rather than scouting new suppliers. AI-related research currently neglects supplier scouting, which is certainly relevant to practice. Moreover, evidence suggests that the digital maturity of firms is at an early stage (Wang et al., 2016), and the potential of AI is untapped in many procurement activities, including supplier scouting.

This research takes information processing theory (IPT) as the overarching lens to study the adoption of AI in supplier scouting. As mentioned by Galbraith (1974) and Roßmann et al. (2018), this theory is especially advantageous in the context of technologically triggered changes in business organizations. Moreover, scoping out the information requirements and the means of enhancing information is critical in knowledge-intensive processes such as supplier scouting (Lorentz et al., 2020).

IPT relies on the concept of uncertainty, which is generated by missing information about decision-making situations and related outcomes (Duncan, 1972). Many different types of uncertainty such as environment, task and partnership uncertainty determine information processing needs (IPNs) (Galbraith, 1974). IPNs can be managed through information processing capabilities (IPCs), which are classified by Bensaou and Venkatraman (1995) into structural, process and information technology mechanisms. Thus, firms manage uncertainty by reaching the fit between IPNs and IPCs (Tushman and Nadler, 1978; Bensaou and Venkatraman, 1995).

In the context of our research, uncertainty is generated when a buyer firm scouts for new suppliers. Supplier scouting activities, which involve several decision-making variables and stakeholders, generate uncertainty in the buyer firm, leading to IPNs. To counter these IPNs in the process of scouting new suppliers, buyer firms can resort to AI-based solutions accessed through procurement platforms to leverage the IPCs of the IT provider delivering the scouting solution. In this way, the IT provider's IPCs match the buyer company's IPNs through the AI-based supplier scouting solution (see Table 2).

Confirming our choice of IPT as the foundation for this study, many previous studies in the supply chain domain have been designed based on the IPT constructs, including Cegielski et al.’s (2012) study of cloud computing in supply chains, Busse et al.’s (2017) study of sustainable supply chain management and Lorentz et al. (2020) study of supply market intelligence.

Since this study applies IPT to a new research domain, we adopt a case study methodology due to the exploratory nature of our research (Eisenhardt, 1989; Yin, 2018) and due to this methodology's capacity to assist in building theories (Voss et al., 2002).

Empirical data were collected from twelve case studies involving IT and information providers, relying on previous studies in the same context (e.g. Handfield et al., 2019; Yarramalli et al., 2020). The sample size is in keeping with the suggestions of Handfield and Melnyk (1998) and the methodological standards of Eisenhardt (1989). Twelve is considered a good number of respondents, resulting in good comparability of results while allowing for an in-depth analysis of each case – both of which are fundamental for theory-building research. IT and information providers are best suited for investigating data processing in supplier scouting activities for many reasons:

1. They are key informants about the information processing needs (IPNs) in supplier scouting, having served the IPNs of the buyer firms directly, including accessing their data and analysing and improving their scouting process.

2. They offer supplier scouting solutions, often customized based on the reference industry and supporting buyer firms with different characteristics (e.g. size, industry and purchasing categories), thus giving them significant experience in different applications of AI-based scouting solutions.

3. They develop the supplier scouting solutions implemented by the buyer firm, which involves knowing the required capabilities better than the buyer firms themselves, as said buyers simply adopt the digital solution they need without delving into the technical knowledge behind the solution's capabilities.

4. By offering their scouting solutions to several buyer firms, IT providers pool good practices from all of them and standardize the scouting process, which is poorly formalized and highly influenced by specificities. Therefore, their solutions are scalable, as well-structured AI-based solutions become viable for large and small firms in different industries.

Aiming at heterogeneity among respondents, we selected providers with strong experience in AI-based solutions for supplier scouting. Leading IT providers' solutions are used by major buyer firms around the world that are oriented towards technological innovation. Given the novelty of AI, start-ups are also relevant as they are agile players that pioneer digital innovation. AI-based solutions also require information to fuel the algorithms; therefore, the sample also includes information providers (see Table 3).

To ensure construct validity, we collected data while triangulating different sources of information. We conducted a preliminary review of the solutions offered through provider websites, including informative sections, whitepapers and case studies. Where available, we also ran a demo to test the solutions. The information was cross-checked with reports from industry analysts (e.g. Gartner). These insights supported the use of the semi-structured interview approach (see Table 4) for primary data collection and helped ensure the reliability of the construct. The interview protocol was intended as a checklist rather than a strict guideline for the interview, thus leaving room for interactions between the respondent and the interviewer. To ensure validity, at least two researchers were present during each interview to take notes about the answers. The interviews were recorded and transcribed and then sent back to the primary informant for an additional check that the information was accurately recorded. After the validation of the transcript, the results were traced in a structured database for the within- and cross-case analysis.

We assume the AI-based supplier scouting solution as the unit of analysis in order to investigate the dynamics between the IT provider and the buyer firm adopting the solution.

First, we conducted a within-case analysis to understand the relationships existing among the variables within each individual case. Then, the cross-case analysis allowed us to focus on the convergence or divergence among the twelve cases. To obtain solid results through a detailed coding process, in keeping with Gioia and Pitre (1990), we built a coding tree based on the IPT constructs ( Annexure). In this way, the uncertainties were assessed as high or low and the mechanisms as strong or weak, as we diligently followed classical IPT formulations (e.g. Bensaou and Venkatraman, 1995). Furthermore, each uncertainty's underlying IPNs and each mechanism leading to IPCs was qualified through more descriptive codes developed from the in vivo analysis of the interviews.

To address the results of the study, the available AI-based solutions for supplier scouting are identified and summarized in Table 5.

Summarizing the findings from the case studies, the IPNs used in supplier scouting are described in Table 6, which also considers the nature and level of underlying uncertainties.

Environmental complexity is fully realized in the customization of the supplier scouting solution, which buyer firms require of IT providers. Case studies reveal how, during their scouting activities, buyer firms have many specificities linked to the characteristics of their business and the requirements demanded of the suppliers. By increasing the parameters to be considered in scouting for new suppliers, both uncertainty and the information to be processed increase. All this converges into a high level of customization required of the IT provider. In the required solutions, each specific piece of information must be collected, analysed and enhanced through a solution that is tailored to the buyer and its potential supplier base. In this way, the level of customization of the scouting solution increases the IPNs.

As described by Provider B, buyer firms that resort to scouting solutions typically require very specialized products available through a limited number of suppliers.

Describing collaboration systems that enable scouting for new suppliers, Lee et al. (2012) refer to the development of solutions embedding all the functionalities needed to meet user demand. However, following the aim to improve and accommodate user needs, the resulting solutions become complex, reflecting all the buyer firm's requirements for functionalities or services.

Environmental dynamism is affected by the low maturity of AI technology for supplier scouting, whose potential has not yet been fully exploited and whose actual implementation is still limited. The case studies highlight how the maturity of AI technology and the adoption of advanced supplier scouting solutions are in a virtuous circle that struggles to get going. Indeed, innovative supplier scouting solutions need a reliable and trusted underlying technology; on the other hand, AI cannot develop further if it does not provide the field with actual applications. Moreover, the spread of AI applications in scouting is still too low to justify massive investment in this kind of solution, despite the huge potential growth shown by adoption rates (Provider G). Thus, the maturity of the underlying technology depends on the potential achieved by AI and also on the maturity of procurement in embracing the change triggered by AI.

According to the traditional formalization of the IPT (Bensaou and Venkatraman, 1995), the maturity of the underlying technology within the environmental dynamism is an external variable. However, given the key role of the buyer firm in the deployment of AI solutions, technological maturity also depends on the digital readiness of the adopting actors (Kosmol et al., 2019).

The environmental uncertainty can also be exacerbated by sociopolitical issues, in terms of regulatory control over the industry. When scouting for a new supplier, buyer firms must pay attention to several requirements imposed by local or international regulations. Thus, a lot of detailed information is required about potential suppliers, especially regarding their production processes, raw materials and certifications. This is necessary for building a regulation-compliant supplier base. Of course, the frequent change in regulations results in a high level of uncertainty plaguing the buyer firm, leading to higher IPNs (Dubey et al., 2015).

From the empirical analysis, a further source of uncertainty emerges that was not considered in the IPT framework: the strategic relevance of the purchasing category. In many case studies (as reported by Providers C, D, E, F, H and I), the respondents stated that the strategic relevance of the purchasing category had a high impact on the scouting process. Managing a core purchasing category increases the level of environmental uncertainty perceived by the buyer firms, as more complex data gathering and analysis are required due to the higher strategic relevance, which amplifies the importance of finding the most appropriate supplier. Although this source of uncertainty was not made explicit in the formulation of IPT, it is significantly reflected in the procurement literature. Addressing sourcing and supplier scouting solutions, Bartezzaghi and Ronchi (2005) describe a fundamental role attached to IT providers when dealing with highly specialized purchasing categories: they are entrusted with an advisory role in supporting the buyer firm during the scouting activities.

Task uncertainty is impacted by the skills of the personnel in the purchasing department: according to the case studies, digital competencies that support the scouting activities and help conduct the purchasing process are lacking. Indeed, people involved in these activities still lack the technological skills to fully understand and exploit the support of AI in supplier scouting. This finding is in line with Bals et al. (2019), who did not include digital competencies among current buyer skills but recognized them as fundamental for future development.

At the level of individual members' commitment in the transition to the new systems, Providers A, C, H and I mention the problem of change management, as buyer firms suffer from cultural barriers when it comes to implementing innovative technologies. In fully embracing the changes, the endorsement and prioritization of top management are crucial (Kosmol et al., 2019). Top management – in this study, the Chief Procurement Manager and the CEO – are called on to understand and appreciate the role of AI in the supplier scouting process in order to act as a catalyst for adopting valuable new supply relationships.

Task uncertainty is also impacted by the organizational functional and staff units component. The presence of conflict among units is common where the purchasing process is not centralized and is delegated among several units, each with a partial view of the situation (Kosmol et al., 2019). Providers D and H emphasize the inter-unit conflict arising within the organization when different units approach procurement separately. This affects the scouting process as well, including duplicated efforts in the search for new suppliers and information asymmetries and silos hampering the potential of AI to assist in supplier scouting.

Furthermore, many buyer firms still do not have a structured approach to data analysis for supplier scouting, making task analysability a source of uncertainty. Task variety is an issue, as buyer firms may lack complete visibility on these activities and scout for new suppliers without a structured process. Indeed, each firm has its own needs when it comes to scouting, and even within the same company, different requirements are managed through different approaches. Thus, variety increases uncertainty (Bartezzaghi and Ronchi, 2005; Monczka et al., 2016).

Considering partnerships, Providers B, C, F, L and K describe a low degree of comfort about sharing sensitive information between buyers and suppliers by means of the procurement platform. More precisely, these providers describe their scouting solutions as pooling information from different companies in the same data lake and making them accessible to any player using the solution. Thus, a buyer's supply base information is available to all the players who can access the same solution – with proper management of sensitive data. In this setting, suppliers are not interested in sharing their data publicly in the digital environment managed by the IT provider. On the other hand, buyer firms are not willing to share information about their supply base with competitors or other potential suppliers who might access the same services through the procurement platform. Therefore, mutual trust is missing from multiple players: buyers and suppliers trust neither each other nor the IT provider. The missing trust among the stakeholders involved is not new to the procurement domain (i.e. Shore and Venkatachalam, 2003; Cox, 2001), but it remains an open issue.

In dealing with supplier scouting, suppliers' asset specificity is relevant as well. Strategic suppliers are not easy to substitute, both due to high asset specificity and for practical reasons that require a significant amount of information to scout for alternatives (Bartezzaghi and Ronchi, 2005; Cox, 2015).

Summarizing the findings from the case studies, Table 7 describes the IPCs enabled by the IT providers and transferred to the buyer firms through the AI-based supplier scouting solution.

Structural mechanisms consist of the formalization of the scouting process. According to almost all the respondents in the sample, the collaboration between the buyer firm and the IT provider allows the former to increase the level of formalization of procurement processes. Indeed, IT providers typically support their clients in redesigning processes that are fundamental to effective supplier scouting, following a more structured approach. This approach increases the buy firm's IPCs. In most cases, the IT provider takes care of the scouting process on behalf of the buyer firm and redesigns the buyer's internal process to standardize as many tasks as possible, as addressed by Provider G.

Thus, only considering the buyer firm, the level of formalization is mostly low. However, the level of formalization increases thanks to the IT provider, which brings structure to the buyer firm's processes and thus increases its IPCs. This is in line with the advisory role of the IT providers established in previous research (Bartezzaghi and Ronchi, 2005): besides solutions, they also provide professional consulting services in structuring and performing the main activities of a procurement department.

The commitment mechanism is mainly exercised through the role of the IT provider in intermediating between the suppliers and the buyer. Indeed, the procurement platform guarantees the security of any information provided by the actors, as providers do not have any interest in disclosing data. Almost all the providers in the sample confirm that this mechanism solves the problem of confidentiality while ensuring an equal sharing of benefits among the parties. Providers F and L explain that the platform is built with a layered structure in which every supplier has a public layer – where the information that is publicly available in the network is stored – and a private layer – where the information shared with selected players in the network is stored. In this way, buyers and suppliers can store and share information while still protecting the data. Furthermore, strict non-disclosure agreements are key to ensuring the non-disclosure of sensitive data and fostering collaborative behaviour. In this way, only the entitled buying firms can acquire specific information about potential suppliers (see Provider G's statement). The IT provider thus has the role of catalyst and guarantor for the parties involved in data sharing, which is essential for pooling information and benefiting from a large dataset of supplier scouting information. However, these statements may bear biases due to the IT providers' comforting claims about data security. Data sharing is still a significant hurdle for many firms.

Providers also state that their databases are enriched with data coming directly from suppliers who are invited by the buyer firms to join the network and add more information, moving towards a community where different actors take joint actions in sharing relevant information. Indeed, the buyer firm is also required to communicate data about the supply base to provide the platform with sufficient information for conducting the scouting process. On the other hand, Providers E, H and K emphasize that the data gathering process is limited within their platforms and could be improved by a higher level of joint action, especially from suppliers (e.g. Provider E). However, Provider H does not trust the network effect coming from the joint action mechanism underlying a procurement platform: “The network of collaboration is limited within the platform. Every client is managed as a stand-alone instance without any communication with other use cases. This allows for a more tailored service to the customer, but it doesn't allow to exploit network externalities”.

The process mechanisms identified are related to data sharing and exploitation, mainly those impacting the relationship between the IT provider and the buyer firm, but also including suppliers. This is key in the proper adoption of AI in the scouting process. The issues related to data management and information sharing were already being debated in the procurement literature, as in Lorentz et al. (2020). In this study, the information processing interventions for data storage and management are presented as tools for process improvement and strategic alignment, in keeping with the concept of process mechanism identified in IPT.

Dealing with AI and its applications in business, the technological mechanism is the most important one, and the compatibility mechanism constitutes its initial step. Compatibility can be assessed through several features of the procurement platform, such as the ease of access to the platform (e.g. Provider B). Moreover, simple user interfaces allow buyers to easily navigate the platform.

Indeed, the transparency of the interfaces is described as relevant as well. Many IT providers state that they offer services allowing for a high level of process transparency, including sharing information among different business units and granting visibility throughout the procurement process. This transparency allows users to find consistent information quickly.

The compatibility mechanism is further empowered by the variety of data feeding AI-based solutions for supplier scouting.

In line with Cegielski et al. (2012), the accessibility of the solution, transference between interfaces and strong data integration create a virtuous circle that is crucial for the success of AI-based supplier scouting solutions. These dimensions enhance the compatibility of the solution as a means of controlling and coordinating communication between all the actors involved in the process: the buyer firm, the complex network of potential suppliers, the IT provider and the information provider.

In all cases, IT providers' analytics capabilities are fundamental and perceived as an intangible asset transferred to the buyer firm through the procurement platform. However, different degrees of capability were found among the different case studies, highlighting the varying maturity of supplier scouting solutions available to buyer firms today. For some of them, such as Providers F and I, analytics capabilities are high because the platform allows them to combine the data processing skills and supporting capabilities of the buyer.

However, Providers C, D and E report low analytics capabilities within their platforms. Provider D states that their platform is only able to provide stand-alone analyses of a selected supplier. Provider E says that their platform can handle multiple sets of data and information, but specific analytics-related skills are lacking, suggesting that there is still room for improvement. Therefore, analytics capabilities are crucial: according to the case studies, a lack of analytics capabilities constitutes a barrier preventing the adoption of AI in supplier scouting. The advanced analytics capabilities at the base of AI should be used to decompose, combine and integrate information (Srinivasan and Swink, 2018) while discovering useful insights for supplier scouting. However, buyer firms lack these skills (Bals et al., 2019) and rely on specialized IT suppliers. When IT providers have strong analytics capabilities, the collaboration with the buyer takes off, and AI becomes the arm of advanced supplier scouting. By contrast, when IT providers lack these skills, the potential of AI is untapped.

A provider that assures high analytics capabilities needs to guarantee high data quality as well, in terms of accuracy, timeliness, consistency and completeness. According to all the providers in the sample, the quality of the input data in AI-based solutions is crucial for reliable results.

As far as accuracy is concerned, the relevance is high. However, two critical factors arise from the case studies. For Provider C, the control for data accuracy is done manually. In the case of Provider E, no control is done since the platform involves only first-tier suppliers considered to be trustworthy.

Regarding the timeliness of the data, many providers in the sample state that the information in the platform is constantly being updated. Moreover, Providers B, L and K emphasize the capability of the platform to check for any expired data or new information and to notify the user about items of note.

Providers F, K and H emphasize the importance of dealing with consistent data to increase information processing capabilities. Provider H believes that data consistency is high, as the main input data come from suppliers, who are required to upload the information within a certain format. These consistent data allow for the application of AI in the scouting process.

The case studies emphasize the benefits of an effective integration between the procurement platform and the ERP systems of the buyers and suppliers. From a broad perspective, one of the most promising applications of AI in business is the support of ERP systems' machine intelligence. According to Hinova (2021), AI complements and optimizes the human factor in the interaction with ERP systems, guiding people in making the right decisions. This is corroborated by the case studies, as Providers A, F, G and K confirm the enabling role of AI in ERP integration, especially when it comes to the alignment between internal and external data.

Provider G describes the support of processing images in the scouting activities: their AI-based solution compares the data in the buyer ERP (i.e. the bill of materials and other information about the procurement requirements) with the images in the catalogues of potential suppliers, speeding up the recognition of the right product from a wide range of available alternatives in the supply market.

Indeed, integration with ERP systems increases the IPCs of the IT providers, which are then transferred to the buyer firm. This approach allows IT providers to directly plug the supplier scouting platform into the buyer firm's available systems, guaranteeing a broader control over each process and procurement need. However, Provider K reported that this mechanism does not always result in the expected benefits. Nevertheless, this experience was limited to a few cases where integration had not been successful, and Provider K still recognizes the value of this mechanism for increasing the IPCs. This aligns with Huang and Handfield (2015), who found that the strategic sourcing activities of firms implementing ERPs as compared to non-ERP users lead to better performance for the procurement department.

In analysing the IPCs described in the case studies, an additional construct was found within the technological mechanism: integration with information providers.

The empirical analysis of the case studies reveals a rich overview of the uncertainties for the buyer firm regarding supplier scouting activities and the mechanisms that enable them to cope with this uncertainty, especially the AI-based solutions offered by IT providers (see Figure 1). The resulting framework fits well within the formulation of the IPT constructs developed by Bensaou and Venkatraman (1995).

We adopt the “fit as matching” perspective established by Bensaou and Venkatraman (1995) by comparing IPNs with IPCs (Premkumar et al., 2005). Following this method, common patterns and differences can be found in the AI-based scouting solutions we analysed.

All the case studies demonstrate a high level of environmental, task and partnership uncertainty, meaning that the buyer firms scouting for new suppliers face significant uncertainty, which leads to high IPNs. On the other hand, the level of IPCs, defined by the structural, process and technological mechanisms, varies across the cases.

Although it is not possible to identify a one-to-one relationship between IPNs and IPCs (Tushman and Nadler, 1978), AI supports and sustains the development of appropriate mechanisms to manage uncertainty in supplier scouting. Structural and process mechanisms are fundamental to the adoption of AI and to extracting maximum value from it, as they are enablers necessary for adopting the new technology rather than actual AI-enabled mechanisms.

Technological mechanisms are certainly more substantial and directed towards the adoption of AI. In many of the case studies, sophisticated data analysis and natural language processing algorithms enable web crawling to run the supply market intelligence and inform supplier scouting with the integration of data from different sources. These mechanisms also result in higher data quality, which is fundamental in supplier scouting: AI enables automatic data quality checks, triangulation with data from internal (e.g. ERP) and external (e.g. structured data from information providers or unstructured data from news and social media) sources, and the harmonization of all the information available. The most advanced AI-based supplier scouting systems have actual recommendations for the buyer firm.

Most of the providers in the sample have the capability to process the required information: they have high IPCs, which are transferred to the buyer firm through their AI-based solution. Thus, the solutions offered by high-capability providers match the high IPNs of the buyers. Among the providers studied, there are large technology providers (Providers A, B, F, G, H and I) capable of merging data from internal and external sources, which ensures a view of the whole process and results in intelligent recommendations during the scouting activities and the automation of ancillary tasks. Matching the IPNs and IPCs demonstrates that information providers (Providers J, L and K) also play a key role: they are mainly focused on gathering data from sources external to the buyer firm, enabling these providers to process data and guarantee a high level of data quality. Being involved as data gatherers and providers, they are sceptical about offering explicit recommendations to the buyer firm.

However, smaller providers and start-ups (Providers C, D and E) are not able to provide the IPCs needed to manage the uncertainty inherent in supplier scouting, leading to a mismatch with the buyer firm's IPNs. These providers do not support the development of technological mechanisms through adequate investments, resulting in a low capability to process information and grant the required data quality. Furthermore, providing their solutions mainly to small and mid-sized buyer companies, these providers struggle to gain a good level of commitment in deploying the structural mechanisms in the suppliers' onboarding and in the formalization of the scouting process.

This paper contributes to theory and practice by studying an under-investigated phenomenon that is relevant for companies, namely the role of AI in supplier scouting. This study identified buyer firms' IPNs and found a high level of uncertainty in scouting for new suppliers. This study also identified the IPCs enabled by AI, helping to understand how AI copes with high uncertainty. The case studies reveal that the most advanced IT providers achieve a match between IPNs and IPCs, providing the buyer firm with a sophisticated AI-based solution to support the scouting of new suppliers.

This paper contributes to the advancement of scientific knowledge in several ways since the application of AI in the procurement domain is still a novel, little discussed phenomenon. Indeed, previous contributions about the role of AI in procurement are few, often missing the process perspective and the specific activities carried out by the buyer firm (Min, 2010; Nguyen et al., 2018). This is even more true when it comes to supplier scouting, which, although key within the procurement process, remains scarcely investigated in terms of activities, information requirements and technologies. Thus, the supplier scouting process is a fertile ground for a new avenue of research focused on AI technology.

From a theoretical perspective, this study demonstrates the applicability of IPT to the context of AI implementation in supporting buyer firms' supplier scouting activities, thus contributing to the purchasing domain and to IPT research. This paper illustrates how IPCs reduce the buyer firms' uncertainty in scouting for new business partners. In this way, the IPCs developed by IT providers and transferred to buyer firms in AI-based platforms represent a fundamental enabler for increased competitive advantage stemming from supplier scouting. In addition, starting from the original intra-firm (Galbraith, 1974) and inter-firm (Bensaou and Venkatraman, 1995) IPT formalizations, the case studies reveal how uncertainty arises within the buyer firm perimeter and how the boundaries are extended thanks to the capabilities of the IT providers, which match the high IPNs in supplier scouting.

IPT is fundamental in providing a solid and consistent structure to the findings: the theory supports the validity of the study from the early stages of research design by considering all the constructs relevant to the adoption of AI in supplier scouting. Additionally, IPT has been developed over several years through various formulations, additions and applications in different domains. Thus, applying IPT to a largely new area of investigation enables a renewal of its validity and further confirms the robustness of its constructs. By adopting the IPT lens, this paper contributes to theory by identifying the IPNs of the scouting process, the IPCs offered by AI-based solutions and the match between IPNs and IPCs in a structured way.

The empirical data gathered directly by relevant players yield highly relevant takeaways for practitioners. Generally, the value of case studies is the investigation of a contemporary and complex issue (Yin, 2018) – in this case, the adoption of AI in the supplier scouting process – in a way that is fully embedded in the reference context and that urgently calls for the involvement of the key actors – in this case, IT and information providers. The IT and information providers in the sample offer an insider perspective by considering actors directly involved in the design of AI-based solutions for supplier scouting. Working with several buyer firms, IT and information providers hold a stock of knowledge related to different applications in terms of industry, type of buyer firms and purchasing categories required. Thanks to this internal perspective, we were able to precisely identify the needs of buyer firms when they approach the supplier scouting process in terms of data required and IPNs. Moreover, we also focused on IPCs, identifying the capabilities that are fundamental for buyers and that providers offer to compensate for current deficiencies. The implementation of AI is a critical problem for companies, especially when it comes to internally demonstrating why such technology is relevant and which problems it is going to address. This paper offers insights into the elements to consider in this implementation process to reduce internal uncertainty by managing and simplifying the scouting process.

Therefore, when analysing the constructs of IPT in the study of AI adoption in supplier scouting, the fit between IPNs and IPCs represents a further contribution for managers. This fit may be considered a proxy in the matching of supply and demand in the digital procurement solutions market since IPNs elucidate the needs of the buyer firms and IPCs describe the mechanisms through which IT and information providers address business needs. This focus on fit may represent an important contribution for both users and IT providers: users are supported in identifying the contribution of IT providers, and providers are better able to present their value to the users.

This study also has limitations. First, although the perspective of information and IT providers contributes important insights into the application of AI in supplier scouting, this perspective is still biased in many respects. In fact, information and IT providers' point of view is often skewed by commercial intent, as was unintentionally expressed by respondents: according to someone who designs, develops and sells a digital procurement service, the solutions provided are often considered to be very powerful and highly innovative. IT and information providers tend to be optimistic about AI adoption in a buyer firm's procurement process, particularly in the case of supplier scouting. Moreover, although providers can boast of longitudinal experience across different types of buyers, actual application cases are still few. Certainly, engaging the buyer firms in the collection of empirical data would be extremely valuable and could be taken into consideration for further research. Furthermore, the buyer's perspective should be addressed in specific industries or business contexts, as the scouting activities are affected by industry contingencies and specific supplier selection issues, and the type of AI support may change accordingly.

Second, this paper is qualitative in nature, as it only uses a case study methodology because of the novelty of the research. The combination of qualitative and quantitative methods would better support future research and provide an opportunity to move from theory building to theory testing.

Figure A1 and Figure A2.

Table A1 and Table A2.