In Italy, 2023's Legislative Decree No. 36, better known as the new Public Procurement Code, opens the door to a complete automation of public contracts' entire life cycles. In line with the techno-solutionism advocated by the Brothers of Italy (AlgorithmWatch reported in 2022), it encourages Italian contracting authorities and grantors to find “technological solutions.” It expressly advocates for using Artificial Intelligence and distributed ledgers (blockchain) to automatically define the parameters necessary for participation in a tender and to choose bids' winners. The new rules apply to all procedures initiated from 1 January 2024.

The possibility to automate parts of public procurement is not new. “E-procurement” was introduced in law in 2016. The revision of the Public Procurement Code was prepared in a 2022 decree, issued by the previous government, which made clear that sorting bids should be handed over to algorithms. With the new Code, however, there is a leap forward. 

Francesca Petullà, administrative lawyer and procurement expert, explains that “while in the previous Code we had digitization solely related to the presentation and evaluation of bids, this new Code envisages the possibility of automating the entire life cycle of public contracts. The government really took sides and chose automation.”

This commitment is driven by a significant investment from the National Recovery and Resilience Plan (PNRR), Italy's implementation of Next Generation EU, the European Union’s post-Covid recovery program. Italy has been allocated a whopping €191.5 billion. PNRR's Mission 1, which covers digitalization, innovation, competitiveness, culture, and tourism, was allocated €50 billion, with €11 billion dedicated to the complete digitalization of the Italian public administration. There is a considerable amount of money available to automate procurement but unknown how much has already been earmarked for it.

Certified platforms

With the new Code, contracting authorities and economic operators are now required to use digital procurement platforms that are vetted by the Digital Italy Agency throughout the entire life cycle of public contracts. Currently, 46 such platforms exist, but not all cover the entire procurement life cycle.

Entirely automated procurement has yet to be implemented, but it is likely that private companies operating these platforms will determine how the algorithms work. Such a privatization of public-service algorithms already took place in the education sector, with disastrous consequences, as AlgorithmWatch reported in 2023.

Problems are already emerging. Some hospitals reported that the national, centralized platform that delivers “contract identifying codes” necessary to pass any kind of tender sometimes is out of order. As a result, hospitals cannot buy the medication they need, and patients won't get treatment.

Efficiency

Francesca Petullà explains the reasons for such problems: “The platforms we operate on are all certified but use different languages and systems. Economic operators who need to submit a bid are struggling because the platforms do not share a common standard.” 

According to Luca Forteleoni, advisor to the National Anti-Corruption Authority (ANAC), the new Code would instead streamline the chaos of contracting authorities in Italy and improve efficiency: “The new Code outlines a system that aims to reduce the number of contracting authorities, and moves towards a concentration in favor of increasingly extensive, qualified, and equipped central purchasing bodies that also manage tenders for local authorities,” he said. There are currently 3,700 authorities that can issue tenders above €500,000 in Italy. An additional 8,000 are affiliates (small municipalities, for instance), which must rely on the procurement infrastructure of a qualified authority. In total, public procurement in Italy amounted to nearly €200 billion in 2021.

However, according to Francesca Petullà, the reduction of contracting authorities could weaken the power of smaller and local public administrations. He is critical of ANAC’s position. “We must ensure that these tools don't create two tiers of administrations, especially that smaller administrations aren't excluded, which would affect constitutional prerogatives and undermine the rights of the community.”

Corruption spotting

Full automation has yet to be implemented, but the National Public Contracts Database (BDNCP) could get things rolling. “Since the establishment of ANAC, we have recorded tens of millions of public procurement bids. Although our corruption measurement indicators already rely on statistical algorithms, the new Code could really unlock the potential of our database. Through predictive tools, there will be the possibility to predict not only corruption risks but even the outcomes of the bid,” Forteleoni says. However, as of now, this form of automated prediction hasn't become reality yet.

Confronted with the possibility that the new system could malfunction and produce discriminatory or erroneous decisions, Luca Forteleoni replies: “A technical commission will be established, as provided for in the AI Act, to assess whether the proposed technological solutions are fit for purpose.“ This makes him him confident that the machine won't break the law.

Little evidence

There is little empirical evidence available on the effect of public procurement's automation. LIUC-Università Cattaneo's Healthcare Datascience LAB is doing research on this field in the STEINBOCC project (Forecasting Tender Needs: Impact Variables and Predictive Data Analytics). Emanuela Foglia, one of the the project's researchers, explains that they “started by working on drugs, specifically classes of antibiotics, anesthetics, and oncological drugs, all off-patent, to study behaviors in the procurement phases in public tenders. A predictive framework emerged that highlighted significant deviations between actual consumption and procurement requests by contracting authorities at specific historical moments.” The software the project is building will intervene in cases where demand is overestimated or underestimated, helping suppliers and healthcare facilities.

However, the project is still in the experimental phase. “In a few months, we will publish new results with updated data for 2023. If the model is valid, we should find ways to access additional data from contracting authorities because our decision-support tool could have worked in other contexts, in line with the new Procurement Code.” If the national platform delivers their contract identifying codes, that is.