“The hare of science and technology lurches ahead. The tortoise of the law ambles slowly behind.” - Justice Michael Kirby,

By Claire Tucker, Head of Public Law and Regulatory and Lebohang Molapi, Candidate Legal Practitioner, Bowmans

The recent publication of the Draft General Public Procurement Regulations, 2026, under the pending Public Procurement Act, No. 28 of 2024 has signalled the end of business-as-usual for state purchasing.

Regulation 4 envisages “a structured, systematic and data-driven approach” to procurement planning, management and supply-base development through the analysis of expenditure patterns, internal requirements and external market information.

The Draft Regulations require procuring institutions to use this analysis to identify categories of goods, services, infrastructure and capital assets, thereby enabling the development of strategic procurement plans aimed at achieving value for money while balancing the fundamental procurement principles of fairness, competitiveness, transparency and equitability.

Regulation 5 further provides for the establishment and maintenance of a prospective supplier database by the Public Procurement Office to support and enhance procurement processes.

The Draft Regulations clearly require a move away from manual compliance towards intelligent automation.

Global perspective

Globally, the use of AI in public procurement is already well established. According to the Organisation for Economic Co-operation and Development, the public procurement agency in Chile, ChileCompra, uses AI to process PDF documents and extract key information for the purpose of monitoring procurement processes, significantly reducing administrative burdens and manual workloads. Beyond document extraction, AI is also being used to classify procurement data at scale.

Similarly, in Ukraine, the e-procurement system Prozorro utilises machine-learning technology to classify procurement items using Common Procurement Vocabulary codes based on tender and item descriptions. This has improved classification accuracy, facilitated the analysis of sector-specific market trends and enabled suppliers to identify relevant tender opportunities more effectively, thereby promoting competition and value for money.

In New South Wales, Australia, comparable technologies are used to automatically classify procurement expenditure across large datasets, allowing for more sophisticated analysis of government spending patterns.

Further examples demonstrate the expanding role of AI throughout the procurement cycle. Municipalities such as Tempe and Murray City in the United States have adopted large language models to assist with the drafting of requests for proposals.

In San Antonio, a pilot AI system analyses procurement contracts, extracts key contractual information such as renewal dates and payment terms, and assesses compliance with internal policies.

In Finland, Palkeet, the Financial and Human Resources Service Centre for the Finnish Government, has deployed a combination of robotic process automation and machine-learning technologies to automate procurement-related administrative functions. One automated system validates incoming invoices for completeness and formatting requirements, while another maintains and updates the supplier register following the conclusion of contractual arrangements.

One of the most sophisticated examples of AI-assisted procurement oversight can be found in South Korea. The South Korean Fair Trade Commission developed the Bid-Rigging Indicator Analysis System (BRIAS), which analyses approximately 60 000 procurement cases annually using data obtained from more than 720 public entities. Through the application of advanced AI models, BRIAS identifies bidding patterns that may indicate collusive conduct. This enables regulators to prioritise investigations and allocate resources more effectively.

Reportedly, the system has generated penalties amounting to almost forty times its maintenance costs over a seven-year period. South Korean procuring entities have also embraced AI-driven tools to predict procurement demand, recommend opportunities to suppliers and anticipate bidding congestion, illustrating how AI can be utilised not merely to automate tasks but to understand and respond to market dynamics.

AI is also increasingly being used to lower barriers to participation in procurement systems. In the United States, the Department of Defence utilises AI-based tools to match small businesses with procurement opportunities, thereby improving access to public contracts. Similarly, in Paraguay, AI tools developed with support from the Open Contracting Partnership identify tenders likely to be suitable for small and medium-sized enterprises ("SMEs") based on historical procurement data, promoting broader participation and competition.

Challenges

AI undoubtedly presents significant opportunities to improve efficiency, oversight and strategic planning in public procurement, but its successful implementation depends upon robust governance frameworks, reliable data systems and regulatory safeguards.

Despite these opportunities, significant challenges remain. Data quality continues to be a substantial obstacle to the successful implementation of AI. Poor-quality, incomplete or biased data results in flawed outcomes and, in certain circumstances, may exacerbate existing inequalities. For example, an AI system designed to recommend contract opportunities may disproportionately overlook women-owned businesses if the data on which it was trained does not adequately reflect their participation in the market.

The use of AI in procurement also raises important concerns regarding transparency, accountability, explainability and procedural fairness, particularly where automated systems are used to influence sensitive decisions such as supplier selection or contract awards.

Many AI initiatives also remain confined to pilot phases and fail to achieve lasting operational impact. This is often because of institutional and regulatory constraints, insufficient technical capacity, uncertainty regarding the integration of AI into decision-making processes, and deficiencies in underlying data infrastructure.

South Africa

While the use of AI is expected to enhance efficiency, transparency, supplier engagement, market analysis and fraud detection within public procurement systems, its introduction into procurement law also gives rise to practical and legal challenges.

South Africa currently lacks a comprehensive legislative framework specifically regulating the development, deployment and use of AI. As a result, there are no clear legal parameters governing accountability, transparency, explainability, liability and human oversight in AI-assisted decision-making processes.

Specific legal considerations in the South African administrative law context are the extent to which procuring entities may rely on automated systems when exercising statutory powers and applying discretionary considerations.

As such, while one could foresee AI being used without challenge for functions such as contract management and demand forecasting, the use of AI in activities such as supplier selection and bid evaluation would raise questions regarding procedural fairness, transparency, accountability and the constitutional principles underpinning South Africa's public procurement framework. The challenge for South African lawmakers and procurement practitioners, therefore, will be to ensure that technological innovation advances the objectives of section 217 of the Constitution.

True winners

As the use of AI continues to evolve, procurement law must be adapted to ensure technological progress remains aligned with the principles of constitutional and administrative justice. However, the goal is not for the tortoise of the law to outrun the hare of progress, but rather to build effective barriers that keep the race on course. The true measure of the successful use of AI in public procurement will not be how fast it can process a tender but how best it serves the public interest and upholds constitutional integrity.