How Artificial Intelligence is Reshaping Public Procurement
- Fanny Ganti
- 6 days ago
- 5 min read
Updated: 2 days ago
A deep dive into the OECD's 2025 landmark report on AI adoption across government contracting, from fraud detection to automated bidding, and the challenges that remain.
13%
of GDP in OECD countries
54%
of AI solutions target pre-tendering phase
€4.15B
audited
by Brazil's AI in 2023
8 days
average audit time (was 400 days)
Public procurement is one of the largest and most complex functions of governments, accounting for roughly 13% of GDP in OECD countries. Amid persistent challenges around inefficiency, fragmentation, and corruption, governments are increasingly exploring how Artificial Intelligence (AI) can enhance transparency, value for money, and strategic decision-making in procurement.
The OECD's 2025 report "Governing with Artificial Intelligence" dedicates a landmark chapter to how AI is currently used in public contracting, illustrating both demonstrable gains and important risks that remain..
01 The State of AI in Procurement
Governments are turning to AI in procurement primarily to enhance operational efficiency, reduce administrative workload, support data-driven decision-making, address workforce constraints, and improve risk detection and oversight. AI is being used across the whole procurement lifecycle (from planning to contract management), but adoption remains uneven. Notably, a majority of solutions are concentrated in the early stages of the procurement process:
54% of AI tools focus on pre-tendering and planning activities,
31% on tendering,
just 4% on post-award supply operations, and
only 11% of solutions support the full lifecycle end-to-end.
Common AI applications include:
advanced spend classification and analysis,
supplier scouting and market mapping,
automated risk detection and fraud flagging,
chatbots and digital assistants for real time support and,
negotiation support and optimisation through predictive analytics.
From Ukraine's ProZorro system generating EUR 250 million in savings to Brazil's Alice AI conducting audits in 8 days instead of 400, the evidence is mounting that thoughtful AI adoption can deliver real gains.
The technologies in play range from machine learning and robotic process automation (RPA) to large language models (LLMs) and natural language processing (NLP), each suited to different tasks across the procurement cycle.
02 Case Studies in Action
Real-world implementations making an impact
Across four continents, governments are already deploying AI procurement tools with measurable results. Here are four of the most compelling examples from the OECD report.
UKRAINE
Prozorro e-Procurement System
Launched in 2014 by a collaboration between civil society and government, Prozorro uses ML to classify procurement codes (CPV), detect irregularities, and support evidence-based decision-making. Over 30 000 users analyse procurement data annually across 49 real-time dashboards.
Euro 250M
savings since 2021 from policy changes driven by its analytics
BRAZIL
Alice - AI Procurement Oversight
Development by the Comptroller General's Office, Alice uses AI and RPA to continuously monitor procurement activities across all federal agencies, flagging fraud and irregularities in real time across 40 predefined risk typologies.
8 days
average audit time, down from 400 days before Alice
CHILE
ChileCompra & Ethical Algorithms
Chile's central purchasing body uses LLMs in its Public Contracting Observatory to detect irregularities. Its Ethical Algorithms initiative introduced standardised AI procurement templates requiring transparency, privacy, non-discrimination, and explainability.
28%
price reductions achieved through redesigned competitive frameworks
UNITED STATES
GovAI Coalition
Founded in 2023 by the city of San José, this multi-agency coalition has expanded to local, state, and federal levels. In February 2025, it launched the AI Contract Hub, a shared repository of contract templates and best practices for responsible AI procurement.
$3.3B
in federal AI-related contracts in 2022 alone
AI Across Public Procurement Lifecycle
PHASE | STAGE | AI APPLICATIONS | BIG DATA & ANALYTICS |
PRE-TENDERING | Needs Assessment | Forecasting, automated data analysis, risk & opportunity identification | Market trend analysis, demand assessment, supplier mapping |
Planning & Budgeting | Optimising budgets, predicting costs, analysing scenarios | Data-driven budget decisions, cost-saving identification | |
Specifications | Automation of specification development via document analysis | Determining effective specs based on past procurement & market analysis | |
Choosing Procedure | Optimised selection, reduced decision-making time | Analysing past outcomes to identify most effective procedure | |
TENDERING | RFP / Bid | Automating RFP drafting by analysing past documents | X |
Bid Evaluation | Automating the evaluation process | Data-driven decisions, pattern & anomaly detection | |
Contract Award | X | X | |
POST-AWARD | Contract Management | Reduced manual oversight, automated compliance checking | X |
Order & Payment | Automated order processing and payment verification | X | |
Reporting | Accurate reporting on supplier performance & procurement outcomes | Insights into performance, compliance, efficiency |
03 Risks & Implementation Challenges
Despite clear benefits, the OECD emphasizes that poorly designed AI systems can do harm at a scale and speed no human reviewer could match, making the stakes of getting it wrong considerably higher than in traditional procurement.
Biased or Skewed Training Data Leading to Unfair Bid Evaluation
AI systems trained on unrepresentative data may systematically favour certain bidders, amplifying inequities at machine speed across thousands of decisions.
Lack of Algorithmic Transparency and Explainability
Contracting authorities that cannot explain how their AI works cannot properly oversee it. The UK's FAST Track Principles (fairness, accountability, sustainability, transparency) offer a governance model.
Outdated Legal & Regulatory Frameworks
Many jurisdictions have no formal guidance on AI use in procurement, creating legal ambiguity and risk challenge from unsuccessful bidders.
Data Lock-in & Vendor Dependency on proprietary formats
Restrictive licensing arrangements can trap governments in dependency on proprietary vendor formats, limiting long-term flexibility and oversight capacity.
Digital Skills Gaps and AI literacy & Risk Aversion
Procurement managers remain sceptical, believing negotiation expertise is uniquely human and cannot be transferred to AI systems, a cultural barrier as much as a technical one.
Fragmented Data Governance
Without governance-wide data standards, agencies build incompatible systems and siloed datasets, limiting the scope and accuracy of AI-driven analysis.
04 The Way Forward
What the OECD recommends
The OECD outlines a clear framework for governments seeking to harness AI responsibly in procurement. Success, it argues, depends not just on technology but on governance, culture, and collaboration.
1 Invest in Data Governance Infrastructure
Standardise and open up procurement data government-wide. AI is only good as the data it is trained on. Fragmented, inaccessible data is the single greatest limiter to effective AI adoption in procurement.
2 Build Procurement Capacity & Digital Skills
Fund training programmes that equip procurement professionals to use, manage, question and oversee AI tools responsibly, not just operate them.
3 Establish Clear Regulatory Frameworks
Develop AI-specific procurement obligations and documentation requirements applicable to both external vendors and in-house development. Apply the OECD AI Principles as a baseline.
4 Prioritise Transparency & Explainability
Contracting authorities must be able to understand and explain how their AI systems work. Commit to fairness, accountability, and transparency as non-negotiable design requirements. Adopt the FAST Track Principles as a baseline for all AI procurement systems.
5 Foster Cross-Sector Collaboration
Engage government, private sector, academia, and civil society to share best practices, accelerate adoption, and validate AI appropriateness for each use case. Collaborative models like Ukraine's ProZorra and the US GovAI Coalition demonstrate the power of shared governance.
6 Monitor, Evaluate & Iterate Continuously
Subject all deployed AI systems to ongoing performance monitoring, real-world impact assessments, and stakeholder feedback loops, rather than deployed and forgotten.
Core Topics
Public Procurement | Artificial Intelligence | Algorithmic Decision-Making | e-Procurement | GovTech |
Technologies
Machine Learning (ML) | Large Language Models | Natural Language Processing | Robotic Process Automation | Predictive Analytics |
Governance & Ethics
Transparency | Accountability | Algorithmic Fairness | Data Governance |
Explainability | FAST Track Principles | OECD AI Principles | Vendor Lock-In |
Applications & Functions
Fraud Detection | Spend Analysis | Supplier Scouting | Bid Evaluation | Contract Management |
Risk Management | Compliance Monitoring | Collusion Detection | Anomaly Detection |
Key Systems & Initiatives
ProZorro (Ukraine) | Alice (Brazil) | ChileCompra (Chile) | GovAI Coalition (USA) |
Challenges & Barriers
Data Quality | Skills Gaps | Risk Aversion | Regulatory Gaps |
AI offers compelling opportunities to transform public procurement by enhancing efficiency, transparency and strategic value. However benefits are not automatic, they depend on robust data governance, skilled professionals and ongoing oversight.
In the words of OECD: AI can make public procurement more dynamic and responsive — but only within an ecosystem of governance, capacity, and collaboration.
Sources:
Subscribe to Transformative Procurement Newsletter for the latest insights, strategies, and updates straight to your inbox.
Comments