With public procurement representing about 13% of GDP in OECD countries (OECD, 2024[61]), adoption of AI in public procurement is often driven by the need to enhance efficiency and operational decision-making and reduce costs (Friton et al., 2024[62]; Hickok, 2022[63]). It is also used to help address challenges such as workforce constraints. With AI, public procurement can become more dynamic and responsive, capable of meeting the demands of a rapidly changing environment throughout the procurement lifecycle (Hickok, 2022[63]; Gastaldi et al., 2024[64]) (Figure 5.3). This digital transformation also presents an opportunity to fundamentally rethink public procurement and administration, improve connections between public entities and suppliers, and enable more dynamic collaboration (Glas and Kleeman, 2016[65]). Yet realising AI’s full potential requires effective implementation, robust data governance and a user-centric approach.
Governing with Artificial Intelligence
Note: RFP = Request for Proposal
Source: OECD’s illustration based on desk research and (EC, 2020[66]).
Current state of play
Public entities are integrating AI and algorithmic decision-making into their processes, using these technologies to improve services, streamline operations and enhance decision-making, as well as strengthen risk management, oversight and accountability. A recent study mapped AI-based functionalities offered by public procurement platforms, revealing that 54% of solutions support pre-tendering and planning, 31% focus on tendering activities, and only 4% address the more operational activities of the supply phase (Guida et al., 2023[67]). Additionally, 11% of solutions, including digital assistants and automation of non-value-added activities, support the entire procurement lifecycle. Common functionalities include spend analysis, risk management, supply chain finance, supplier scouting and negotiation optimisation.
Streamlining operational tasks
AI can help to classify spending to standardise reporting. Classifying spend data into standard taxonomies in public procurement means organising spending into predefined categories, which helps in tracking and analysing expenditures more effectively. Its importance lies in the ability to better make decisions, enhance transparency and identify opportunities for cost savings. For example, Ukraine’s ProZorro e-Procurement system uses a ML solution to predict the correct Common Procurement Vocabulary (CPV) code for products and services (Box 5.23). These codes establish a single classification system for public procurement; aimed at standardising the references used to describe procurement contracts to increase transparency and make it easier for potential suppliers to identify opportunities (EC, 2020[66]).
Recognising the challenges raised by lack of fair competition and the risks of corruption due to procurement practices, Ukrainian civil society organisations, expert associations and government collaborated in 2014 to develop ProZorro. This open-source procurement system uses AI and advanced analytics to enhance efficiency, transparency and accountability. A decade after its inception, ProZorro has evolved into a data-driven ecosystem that integrates cutting-edge business intelligence (BI) tools to support evidence-based decision-making across government, oversight bodies, businesses and civil society.
The system’s BI ProZorro and ProBI analytics modules enable real-time monitoring and deep analysis of procurement data. The public BI Prozorro module offers 49 dashboards covering all procurement stages, allowing users to assess price trends, medical procurement, buyer performance and risk indicators. The ProBI module provides a report-building tool for advanced users to create customised analytics, facilitating regulatory oversight and strategic procurement planning.
The impact of ProZorro’s analytics is substantial. These tools have been widely adopted, with over 30 000 users analysing Ukraine’s procurement market annually, covering transactions worth the equivalent of EUR 21 billion in 2023. More than 80 procurement studies have informed regulatory improvements, including EUR 250 million in savings since 2021 from policy changes related to tender corrections. Additionally, AI-driven risk assessments help oversight bodies detect irregularities, while procurement entities use data-driven insights to optimise purchasing strategies. The system’s collaborative governance model, integrating government, civil society and private sector stakeholders, promotes continuous innovation and responsiveness to emerging challenges.
Source: https://prozorro.gov.ua.
AI can help governments to enhance processes by simplifying and rapidly streamlining highly rule-driven, end-to-end workflows. Simplifying and streamlining highly rule-driven, end-to-end workflows in public procurement involves reducing complexity and automating tasks, such as with the introduction of RPA or LLMs to enhance efficiency. This can be achieved by integrating digital tools and technologies to minimise errors and accelerate the procurement process. For example, in Chile, ChileCompra has transformed how public procurement is performed and has evolved over time to use novel technologies to support government teams (Box 5.24).
In Chile, public procurement has been transformed through ChileCompra, the country’s central purchasing body established in 2000. ChileCompra operates Mercado Público, an electronic platform that centralises and streamlines the procurement of goods and services for public entities. The platform supports framework agreements, which allow multiple suppliers to provide products under standardised terms, fostering inclusiveness and competition. Over time, ChileCompra has become the largest virtual store in Chile, promoting transparency and efficiency in public procurement. Additionally, it has introduced innovative tools like ChileCompra Express, an online marketplace enabling direct purchases from pre-approved suppliers without additional tendering processes.
ChileCompra has continually evolved to address challenges, such as uneven supplier participation and inefficiencies in framework agreements. By 2014, over 850 public entities were using the platform, generating approximately 810 000 purchase orders worth USD 1.8 billion annually. Between 2010 and 2015, the number of transacting suppliers increased by nearly 180%, creating opportunities for small and medium-sized enterprises (SMEs). However, the system faced operational challenges, including a high concentration of revenues among a small number of suppliers and limited second-stage competition in framework agreements. To address these issues, ChileCompra redesigned its framework agreements by standardising product categories and introducing competitive mechanisms that reduced prices by up to 28% compared to market rates.
In recent years, ChileCompra has integrated AI to modernise procurement practices further. It introduced standardised bidding templates for AI and data science projects as part of its Ethical Algorithms initiative in collaboration with Universidad Adolfo Ibáñez and BID Lab. These templates include requirements for transparency, privacy, non-discrimination and explainability to ensure responsible AI use in government contracts. Additionally, ChileCompra’s Public Contracting Observatory uses AI tools such as LLMs to analyse procurement data for irregularities and improve compliance monitoring. These advancements have enabled more efficient oversight while promoting ethical standards in public procurement.
Source: https://www.chilecompra.cl.
AI can help streamline efforts to manage public procurement legal and regulatory frameworks. For example, in Romania, the National Public Procurement Agency (ANAP) developed a tool to improve its ability to screen new legislation, including retrieving documents in real time from public institutions' websites and converting scanned documents into searchable text (The World Bank, 2023[68]).
Enhancing procurer-supplier relations and public servant capacities
AI provide real-time communication and support through chatbots, which can answer queries, provide updates and facilitate smoother interactions. This technology helps streamline processes, reduce response times and improve overall satisfaction for both parties. For instance, in the United States, the El Paso City Council Purchasing and Strategic Sourcing department (PSS) integrated a chatbot solution, called Ask Laura, on its webpage. Ask Laura uses open-source algorithms to interpret and deal with questions and get information about potential suppliers based on business profiles and the questions asked (Collins, 2020[69]).
AI tools can facilitate real-time support and guidance for procurement professionals and suppliers, streamlining communication, and expediting query resolution. This can help enhance collaboration by offering timely advice, auto-populating forms, and creating personalised dashboards. For example, in the United States, the North Carolina Department of Information Technology (NCDIT) has introduced an AI-powered chatbot to assist state agency staff with IT procurement processes. Available 24/7, the chatbot provides instant answers to common queries, such as accessing procurement forms, submitting exception requests, and understanding procurement timelines, reducing wait times and enhancing efficiency (NCDIT, 2024[70]).
Improving risk management, oversight and accountability
Automated compliance checks, fraud detection algorithms and anomaly detection mechanisms bolster accountability by flagging irregularities and deviations from established procurement protocols. AI functionalities use ML techniques to automatically identify errors and fraud, and to manage risk efficiently and effectively (Guida et al., 2023[67]).
AI can also be used to identify integrity breaches in procurement. Both administrations and citizens are investigating AI’s potential this regard. As one relevant example, using publicly available procurement auction data, researchers have been able to use AI algorithms to detect collusion with an accuracy rate of 81-95%. Once the algorithms are trained, they can be automatically updated with the latest auctions, and with little effort on the user’s part to supervise their outcomes (Garcia Rodriguez et al., 2022[71]). In addition, in Hungary, a study analysed 119 000 government tenders from 2011-2020 to identify subtle, text-based strategies used by corrupt actors to favour specific bidders. Using ML approaches like Random Forests, the study found that text data improved system accuracy from 77% to 82%, demonstrating the potential of text mining to uncover corrupt behaviours and enhance anti-corruption policies (Katona and Fazekas, 2024[72]). Another relevant example is from Spain, where researchers developed an AI system to provide an “early warning system” predicting public corruption. The tool uses data on economic and political factors — such as economic growth and the length of a political party’s time in power — along with data on corruption cases to predict the risk of corruption in Spanish provinces (López-Iturriaga and Sanz, 2017[73]). Anti-corruption beyond public procurement is discussed below in the section on “AI in fighting corruption and promoting public integrity”.
Beyond identifying anomalies in existing data, AI’s predictive capacities can flag potential risks and irregularities and optimise public procurement processes. In Brazil, for example, the Comptroller General's Office developed Alice, a tool that uses AI to detect possible instances of fraud, enabling real-time risk management and oversight (Box 5.25).
In Brazil, public procurement accounts for a significant portion of government spending, making it a critical area for ensuring efficiency and transparency. To address vulnerabilities such as fraud, inefficiencies and errors in the procurement process, the Comptroller General's Office (CGU) developed Alice, an AI-powered system designed to analyse bids, contracts and public notices. Alice uses artificial intelligence and RPA to continuously monitor procurement activities across federal agencies, identifying risks and irregularities in real time. By automating these processes, Alice enables large-scale auditing and supports public officials in making informed decisions to improve oversight.
Since its implementation, Alice has delivered remarkable results. In 2023 alone, it analysed nearly 191 000 acquisitions and triggered 203 audits involving contracts worth EUR 4.15 billion (equivalent). Between 2019 and 2022, its alerts led to the suspension or cancellation of bids totalling around EUR 1.5 billion (equivalent). Additionally, the system has significantly accelerated audit processes, reducing the average time required from 400 days to just eight days. Alice leverages tailored natural language processing (NLP) algorithms capable of handling the unique complexities of Brazilian procurement data, further enhancing its effectiveness in identifying risks across approximately 40 predefined typologies.
By combining advanced technology with robust institutional frameworks, Brazil improved oversight, achieved substantial financial savings and enhanced accountability in public spending.
Oversight bodies including supreme audit institutions are key actors ensuring the legality, efficiency and integrity of procurement processes. (OECD, 2015[74]). Every year, Portugal’s court of audit (Tribunal de Contas, or TdC) conducts a significant number of reviews and audits related to public procurement processes (before, during and after), requiring extensive human and financial resources. TdC is working with the OECD to develop stronger control capabilities and more efficient allocation of resources. This includes the development and testing of data pipelines and ML systems to identify red flags, focusing on risks of irregularities in public procurement processes (OECD, 2024[75]).
Empowering external actors and strengthening trust in government
AI can also empower external actors, such as citizens and civil society organisations, to conduct third-party assessments of public procurement programmes and government spending (Santiso, 2022[76]). Governments have developed platforms that enable stakeholders to access open public procurement data, facilitating a transparent exchange of information (Attard et al., 2015[77]). These portals publish both structured and unstructured data covering various phases of the contracting processes. To enhance the presentation of this data, some government agencies have integrated AI-driven dashboards into these platforms that display statistics and indicators relevant to the contracting processes (Ansari, Barati and Martin, 2022[78]). Countries including Colombia (2024[79]), Chile (2024[80]) and Mexico (2024[81]) have adopted such initiatives.
AI technologies can enable stakeholders to monitor procurement activities and manage risks with unprecedented granularity by providing real-time access to procurement data, audit trails and performance metrics. Moreover, AI-powered transparency initiatives can promote greater public trust and confidence in government procurement practices, fostering a culture of integrity and accountability. AI tools can help governments to anticipate demand, identify potential risks and optimise their procurement processes. In Brazil, for example, the Labcontas project (GLOBO, 2018[82]), which brought together 96 databases with information relevant to the work of the Brazilian Federal Court of Audit (TCU), has enabled automated checks of public tenders posing a risk of potential corruption (EC, 2020[66]).
Managing risks and challenges
A variety of potential risks and challenges are areas of concern for the use of AI in government (Andersson, Arbin and Rosenqvist, 2025[83]; Shark, 2024[84]). The issues most seen in OECD work and the evaluation of AI used cases are discussed below.
Associated risks
Inadequate or skewed data in AI systems
Lack of transparency and explainability
AI systems with inadequate or skewed training data can present concerns, as systems used to evaluate bids might favour certain bidders due to skewed training data, leading to unfair procurement decisions. Moreover, the ability of public procurement teams to understand the accurate functioning of algorithmic systems is constrained by human biases in perception, multi-layered complexities and other factors (Hickok, 2022[63]). These types of AI systems can have harmful outcomes that might impact a larger portion of society. In procurement, a human public employee reviews bids and decides one by one, whereas an AI system can review many bids and make decisions in a matter of seconds and minutes. Therefore, if bias exists in the AI system, the speed and scale of harm will also exceed that of a human review (Hickok, 2022[63]). To prevent these outcomes, procurers need to ensure that the AI has been trained on representative datasets. In addition, the AI systems should be designed with fairness in mind, considering factors beyond cost and efficiency. Finally, AI performance should be monitored in real-world scenarios to detect and address any emerging harms.
Concerns also arise regarding AI’s lack of algorithmic transparency. When an AI system is deployed within the public administration, governments should make and honour commitments to principles of fairness, accountability and transparency. The system should be sufficiently explainable; the contracting authority should obtain sufficient information on how the system works and the data it has been trained on to derive its conclusions. Otherwise, public actors will have embedded systems without an independent capability to maintain and monitor these systems. Without these capabilities, alternative oversight and accountability mechanisms will also not be available due to the initial lack of transparency or subcontracting arrangements (Hickok, 2022[63]). In the United Kingdom, the Office for AI (OAI) and the Government Digital Service (GDS) produced a guidance in partnership with The Alan Turing Institute to safeguard public trust in the use of AI in procurement through the use of the FAST Track Principles: fairness, accountability, sustainability and transparency (GOV.UK, 2019[85]).
Implementation challenges
Inflexible or outdated legal and regulatory environments
Lack of high-quality data and the ability to share it
Skills gaps
Risk aversion
Data and vendor lock-in
Many jurisdictions lack regulations and formal guidance on AI use, leading to legal ambiguities and potential challenges from unsuccessful bidders questioning the fairness of the process. Given these regulatory gaps, there is a growing need for regulatory frameworks and guidelines to ensure clear guidance for AI in public procurement, promoting transparency and fairness, reducing legal ambiguities and minimising challenges from unsuccessful bidders. Box 5.26 illustrates how some governments are coming together to overcome this challenge.
The GovAI Coalition is a multi-agency initiative dedicated to promoting the responsible and ethical use of AI in government. Founded in 2023 by the City of San José, the coalition has since expanded to include local, state and federal agencies across the United States. It serves as a platform for cross-agency collaboration, knowledge sharing and AI governance best practices, helping governments simultaneously promote innovation while ensuring accountability.
GovAI Coalition members have collaborated to create a suite of public procurement templates and knowledge-sharing tools that any public agency can use to jumpstart its own AI governance programme. Among these resources, the GovAI Coalition developed the AI Contract Hub, launched in February 2025 in partnership with Pavilion. This platform streamlines AI procurement by offering a shared repository of contract templates, cooperative agreements and best practices. The hub aims to reduce procurement costs and timelines, improve contract transparency and expand access to AI vendors. As AI adoption in government grows — reaching USD 3.3 billion in federal AI-related contracts in 2022 — the GovAI Coalition helps ensure that public agencies have the tools to procure AI solutions efficiently while upholding public values.
AI can significantly enhance the entire public procurement cycle. However, its impact is limited without standardised and accessible data, which necessitates a coherent, government-wide data governance strategy. Moreover, government agencies often implement various AI systems without unified standards, resulting in incompatible systems, fragmented data and inefficiencies in aggregating procurement information. Restrictions on data sharing contribute to these challenges and make it challenging to get the most out of AI systems by broadening the scope of system training and analysis (Andersson, Arbin and Rosenqvist, 2025[83]). Although OECD countries have made considerable progress, data containing all relevant public procurement information is still largely unavailable, and very little exists as open data (defined as data reusable in an accessible format) in most evaluated countries (da Rosa, 2023[86]).
Digital skills gaps — as well as a lack of understanding of AI’s potential — are relevant hurdles to the successful deployment of AI in procurement processes (Guida et al., 2023[67]). The power of AI alone is not enough for its successful adoption of advanced procurement platforms. Data management, cultural change and skills development are fundamental (Handfield, Jeong and Choi, 2019[87]). If such gaps and limited understanding persist, public entities may struggle to effectively implement and manage AI systems or mitigate their risks, leading to inefficient and potentially improper procurement processes.
Another interesting finding is that procurement managers are mostly sceptical of AI, believing that the typical skills of the human buyer are strictly related to negotiation and that this knowledge, often tacit and not formalised, cannot be transferred to autonomous agents or systems (Guida et al., 2023[67]). Consequently, they feel that this knowledge cannot be effectively transferred to autonomous agents or systems. This risk aversion highlights the importance of digital readiness and the level of digital skills possessed by public officers, as these factors are crucial for the successful integration of AI in procurement processes.
Poorly designed or restrictive data licensing agreements can create data lock-in, preventing the contracting authority from sharing the necessary data with the AI developer, thus limiting the AI system's effectiveness. And there is a risk of vendor lock-in, which makes the contracting authority heavily reliant on the AI vendor's proprietary technology and data formats.
Untapped potential and way forward
OECD findings and external research identify little dedicated research on AI for public procurement, and a fairly low level of AI maturity in public procurement entities (Andersson, Arbin and Rosenqvist, 2025[83]). To help public procurement organisations assess their own maturity and identify factors needed for growth, the IBM Centre for The Business of Government has developed an AI maturity model for public procurement that may be a useful reference (2023[88]).
If successfully adopted in the field, AI’s potential in public procurement includes automated supplier evaluation, predictive systems to anticipate events — like product shortages and optimal timing for securing best pricing — detection of potential influences from broader economic and geopolitical issues, and the creation of smart bidding platforms that automatically match procurement needs with the most appropriate bidders (Shark, 2024[84]). LLMs could be used promote integrity in public spending. For instance, models such as those that power ChatGPT can support public procurement officials in analysing large amounts of data on a company and potential contractor to screen for fraud or corruption risks (Ugale and Hall, 2024[89]).
AI could also make an impact in setting requirements and specifications needed by a purchasing official, assessing bids and ensuring fair and reasonable pricing, optimising supplier selection and ensuring regulatory compliance (IBM, 2023[88]). AI could help public procurement officials to determine specifications for a purchase by presenting information about previous procurement exercises, or by dynamically identifying and presenting relevant products services on the market (IBM, 2023[88]). Improving market knowledge through AI-gathered and synthesised content could also assist procurement officials in identifying reasonable and fair prices for various products and services to help serve as a baseline for direct purchase or in considering competitive bids. Moreover, AI could help to synthesise suppliers' information for procurement decision-making. Information about suppliers can be analysed using NLP methods from a variety of sources, such as business profiles, financial data and internet reviews (Burger, Nitsche and Arlinghaus, 2023[90]).
Other potential applications can be found in the private sector’s use of AI tools. Private procurement has been the subject of significantly more research, even though companies are slow to adopt AI relative to other business functions (Andersson, Arbin and Rosenqvist, 2025[83]). These approaches may serve as inspiration for public procurement offices, which may not be able to directly replicate private sector solutions due to legal and regulatory frameworks. Examples include:
Walmart has implemented Pactum AI, a negotiation chatbot for suppliers providing "goods not for resale", which aimed to enhance payment terms, secure discounts and offer flexible contract termination notices.1
AutogenAI has developed an AI tool designed to expedite the bid-writing process for procurement. This tool assists businesses in crafting proposals more efficiently, reducing the time and effort required in responding to procurement opportunities.2
Sievo, a procurement analytics company, offers a platform to enhance procurement processes that uses AI to analyse spend data, forecast demand and optimise supplier selection, thereby improving decision-making and operational efficiency.3
DocuSign developed an AI-powered contract management tool that uses NLP to scan and interpret legal documents, identifying cost-saving opportunities and ensuring compliance.4
Contracting authorities in government need to take steps to ensure they are taking informed actions and decisions when incorporating AI into their procurement processes. They should focus on reducing both risks and risk aversion, improving skills and capacity, encouraging procurement officials to engage in dialogue with suppliers, and enhancing data collection and monitoring of results. The OECD Recommendation on Public Procurement (2015[74]) and the OECD AI Principles (2024[91]) — among other OECD and international standards — help contracting authorities navigate their efforts to implement trustworthy AI in their respective processes. Without an informed and trustworthy approach, public actors could fail to seize the benefits of AI systems. They could also end up with highly integrated yet flawed systems without an independent capability to maintain these systems, or skills to monitor their performance (Hickok, 2022[63]).
The success of AI in procurement processes hinges on both the effective implementation by the buying entity and the commitment of key stakeholders. A user-focused approach is essential for digital transformation in public procurement. Success also depends on robust data governance and infrastructure, including the standardisation, sharing and use of procurement data, as well as modern computing systems needed to efficiently host and transfer data and run AI systems.
Governance mechanisms and accountability structures remain essential. Policymakers and intergovernmental organisations are establishing regulations to govern AI use in public procurement. To ensure solid governance and accountability, AI-specific procurement obligations and documentation should apply equally to both external and in-house development (Heikkila, 2022[92]).
Governments should also invest in capacity development programmes, training initiatives and knowledge-sharing platforms to support AI adoption in public procurement. By investing in capacity building and training programmes, procurement professionals can be equipped with the necessary skills and knowledge to effectively use AI technologies.
Furthermore, collaboration between public and private sectors, academia, civil society and, in some cases, the public is essential for encouraging innovation and disseminating best practices in AI-enabled procurement systems. By promoting knowledge sharing and collaboration, governments can accelerate the adoption of AI technologies and maximise the potential benefits for society. This collaboration also helps to determine whether AI is the best solution for a given challenge relative to other approaches or technologies – an important but often overlooked step (Hickok, 2022[63]).
Public purchasing bodies should prioritise the ongoing evaluation and iteration of AI systems used in public procurement, such as monitoring the performance of AI algorithms. They should assess their impact on procurement outcomes and people and solicit feedback from stakeholders.