Research Updates

Procurement and AI: applications, outcomes, and future prospects

In most Italian companies, the AI ecosystem is capable of managing and supporting procurement decisions. But there are roadblocks on the path to disseminating AI linked to the company culture, people’s lack of competencies, and the inability to re-engineer internal processes.  

The questions

Procurement processes are currently experiencing tremendous change due to huge investments in digitalization along with the transformation dynamics driven by new technologies. Artificial Intelligence (AI) applications in particular can go a long way in helping companies process historical procurement data, establish partnerships for sustainable growth, and identify negative events and the consequent risks. 

The impact of the pandemic has spurred companies to speed up the redesign of their supply chains and prompted CPOs to focus on adopting intelligent technologies. The aim here is to boost end-to-end productivity and visibility in supply chains, maximize the value of procurement data, and create responsible partnerships for long-term sustainable growth. 

To give a more complete picture of the changes that are underway, and to support procurement management in strategic decisions, SDA Bocconi and the School’s Business Procurement LAB carried out a research project in 2021 focusing on rolling out AI technologies in various stages of the procurement process. 


In our study, we utilized a broad classification of AI, encompassing more complex methodologies (machine and deep learning), as well as more deterministic applications (Robotic Process Automation and Optical Character Recognition). Through a survey of Chief Procurement Officers (CPOs) we collected data from over 130 Italian manufacturing and service companies that have adopted a number of AI projects. 

Of our sample, 37% reported they were currently running AI projects in procurement. These companies present a strong proactive orientation aimed at integrating and absorbing competencies and expertise from third parties, in particular consulting firms, software vendors and other players in the supply chain (customers and suppliers). These same companies say they’ve invested more than six people in AI technologies (49%), with 41% employing over 10 people in AI. 

The AI projects in question involved a range of procurement sub-processes, primarily Vendor Management (39%), eSourcing & Tender Management (33%), Contract Management (30%) and Spending Analysis (31%). What’s more, the projects often entailed several sub-processes: 33% of cases more than 4, for 25% from 3 to 4, and in 42% of cases 1-2 sub-processes. This finding highlights the tendency to use technologies as a tool for integrating activities end-to-end. In 51% of projects, AI is relevant and decisive in entire stages start to finish, underscoring the capacity of these technologies to fully support tasks and run analyses that were previously delegated to buyers.  

Our study also delved into the level of support that AI provides for management, distinguishing three macro-activities (or stages) to describe the decision-making process. The impact of AI is relevant in every one, with a peak in data analysis to get a clearer vision and understanding of the context in question. 

  • Data Generation & Collection, supported in 49% of cases 
  • Data Analysis & Processing, 87% 
  • Output Validation and/or Managerial Decision, 51% 

We also noted that in 23% of the cases, AI supports all phases of the decision-making process, from data collection to management decision-making. 

In the end, we analyzed the autonomy of algorithms, from complete AI to a mixed human/machine contribution. We found total autonomy in Data Generation & Collection in 24% of the cases, with an additional 36% with a marginal contribution from the people involved in the process. Machine autonomy drops markedly in the next two stages (Data Analysis & Processing, and Output Validation and/or Managerial Decision), with only 4% in the latter. What emerges in these two stages is that the logic model used in most projects is based on a human/machine balance with reciprocal support. More generally, we can see that as far as the optimal mix in processes, as of now the key to their optimization and overall success lies in close human/machine interaction. 

AI, especially in some methodologies like machine learning and deep learning, can enable companies to build systems capable of self-learning through autonomous, automatic feedback systems (26% of the cases). In most cases, “human-in-the-loop” systems integrate manual feedback from people (41%). Lastly, in 33% of the cases, the system has no self-learning and is reprogrammed occasionally for updates and upgrades.  

Several different AI technologies have been implemented, with particular focus on the ones that help recover efficiency (basic data analysis and visualization, robotic process automation (RPA), natural language processing (NLP)/text analytics, and optical character recognition (OCR)), and those that serve to understand and support more complex decision-making contexts (advanced & predictive/prescriptive analytics, machine learning, and deep learning). 

The outcomes of these projects are overwhelmingly positive, meeting or exceeding expectations (in 51% and 10% of the cases respectively). These findings confirm the maturity of the AI ecosystem for managing and supporting procurement decisions.  Among the criticalities that companies face in AI implementation, various major organizational obstacles come into play: company culture (39%), the competences of personnel (36%) and the ability to re-engineer processes (25%). 

Our data and analyses show that AI is a mature technology that is available to companies, and that it can fully support their procurement processes and stages. The use of AI is decisive both for automatizing low-value-added activities, and in more evolved and autonomous applications, to govern complex phenomena. That said, the various applications of AI paint a picture of machines supporting people in the different phases of the decision-making process, rather than replacing them. 

Looking ahead

  • From an analysis of these projects, what clearly emerges is that the companies in our study have already made progress on the learning curve in terms of using AI in procurement. What’s more, they are currently expanding their innovation activity to other critical processes in procurement, and paying attention to possible evolutions of initial applications. Investments in the coming years will be channeled into developing more competences inside the company and bringing new qualified professionals onboard from the outside. 
  • Companies that are still lacking a clear vision on AI in the context of procurement risk losing a competitive edge with respect to their more advanced counterparts. These laggards should concentrate more on new positions at the level of corporate strategy and company culture, to become more open to new AI technologies and move toward a more proactive leadership, willing to collaborate with external players. 
  • Questions of company culture and internal competences are critical ones which must be addressed before embarking on an AI project. What also appears to be salient is to set up in-house teams whose members include experts in both business matters that must be dealt with on one hand and in big data ingestion and advanced analytics on the other.  
  • Today AI is a rapidly growing technology in terms of procurement applications. As with every other technology, the innovation adoption curve shows several companies that are early adopters. Thanks to projects with positive results, and the development of excellent benchmark practices, AI applications will eventually proliferate to the majority of companies, even those have less exposure to the phenomenon.