The Use of AI in Public Procurement

The public procurement process has been widely criticised for its complexity and resource demands, even since the introduction of the Procurement Act 2023 (the “Act”) in the UK. As governments race to keep up with the pace of AI development, public procurement has emerged as a key focus in shaping how AI is deployed and governed. The Government’s Department for Science, Innovation & Technology (DSIT) recently published its AI Opportunities Action Plan (the “Action Plan”), which outlines an ambitious role for public procurement in developing a thriving AI ecosystem and improving public services, including the use of a digital government to create new opportunities for innovators, publishing information on an “AI Knowledge Hub” and identifying “quick win” opportunities.

This marks a broader shift in public policy; governments are not just seeking to regulate AI, but also to harness it directly, embedding it into core systems through public tenders, partnerships, and targeted infrastructure spending. According to DSIT, over GBP14 billion of investment has already been committed under the Action Plan.

As AI becomes increasingly pervasive in the public sector, public procurement is emerging as a key area of transformation. While the use of AI offers significant efficiency gains, it also raises complex legal issues, particularly around transparency, accountability and fairness. For litigators and procurement professionals alike, understanding how AI is influencing procurement decisions is becoming increasingly important.

The National AI Tender Framework

Currently, AI is being used in the public sector in a small scale and in silos. Under the newly introduced Action Plan, a “Scan > Pilot > Scale” framework has been developed to scale up AI use within the sector, with a focus on supporting the domestic start-up and innovation ecosystem.

In the context of public procurement, the government is encouraging AI companies to collaborate with them by way of a “National AI Tender Approach”. The National AI Tender Approach follows the “Scan > Pilot > Scale” framework under the Action Plan and is aimed at culminating in national-level procurements to deliver a unified, citizen-centric digital government platform powered by progressive AI solutions.

The Scan > Pilot > Scale framework approach, utilised under the National AI Tender Framework, can be broken down as follows.

Scan

“Scanning” involves investment in building an in-depth, dynamic understanding of AI capabilities and their highest-impact challenges and opportunities. This requires:

• appointing an AI lead for each mission to help identify where there is potential for an AI solution within the mission setting; this involves consideration of user needs from the outset;
• a cross-government, technical horizon-scanning and market-intelligence capability that understands AI capabilities and use-cases as they evolve to work closely with mission leads; and
• two-way partnerships with AI vendors and start-ups to anticipate future AI developments and signal public sector demand.

Under the National AI Tender Approach, the scan phase comprises two-way partnerships with AI vendors to understand current AI technologies and future developments. The Gov.UK App team is simultaneously exploring how to simplify navigation of government services and tasks across departmental boundaries.

Pilot

This means developing prototypes or using light-touch procurement to create pilots in high-impact areas, to carry out robust evaluation and publish results. This requires:

• continual use of a framework on how to source AI; such a framework needs to evolve over time to accommodate given data, capability, industry contexts and evaluation of successes;
• a quick prototyping capability that can be drawn on for key projects where required, including leveraging in-house AI expertise and technical resources to build and test concepts, together with specialists in design and user experience;
• specific support to hire external AI talent;
• a data-rich experimentation environment including a streamlined approach to accessing data sets, plus access to language models and the necessary infrastructure; and
• a faster, multi-stage scaling AI procurement process that enables easy and quick access to small-scale funding for pilots and that only layers bureaucratic controls as the investment size gets larger; the Action Plan encourages compensation of start-ups for each round that they get through in respect of multi-staged “Competitive Flexible Procedures”.

In respect of the National AI Tender Approach, the pilot phase (due to take place over 2025/26) is expected to focus on testing value, technical feasibility and deliverability of AI powered system use for “key life transitions” (eg, education pathways, apprenticeship opportunities and career guidance). 

Scale

“Scaling” involves the identification of successful pilots that are capable of application in different settings to support citizens, and rolling these out beyond organisational boundaries. Scaling is pivotal if AI is to have a meaningful impact on effectiveness, efficiency and productivity, as well as maximising government spending power. This requires:

• a scaling service for successful pilots with senior support and central funding resources;
• mission-focused National AI Tenders to support rapid adoption across decentralised systems led by “mission delivery boards”;
• development or procurement of scalable AI technology that supports the use of specialist narrow and large language models for tens or hundreds of millions of citizen interactions across the UK; and
• mandating infrastructure interoperability, code reusability and open sourcing.

The scale phase under the National AI Tender Approach, intended for 2026/27, will focus on rolling out a fully integrated, agentic AI-powered Gov.UK Chat service, with the ambition of offering a unified, national-level service. This phase has intentionally been kept flexible to allow for modifications required arising from the pilot phase.

The Impact of AI Use Within Public Procurement

Benefits

The adoption of AI in the context of public procurement brings a host of potential uses, and as a result, it is expected that the use of AI in public tendering opportunities will become standard practice.

Reduced administrative burden

AI can automate repetitive tasks such as processing invoices and purchase orders, reducing time spent on some procurement processes by up to 60% (it should be noted, this is when utilised to its maximum operating capacity). Utilising AI to carry out procedural tasks can streamline the procurement process, potentially increasing accuracy and reliability and allowing those behind the procurement process to focus on higher-value tasks. By automating simple tasks, such as technical support, processing costs within the procurement procedure can be reduced by up to 40%.

Streamlined evaluation

AI can evaluate the context of a bidder’s organisation, scrutinising large amounts of data on their cost, performance and compliance very quickly and accurately. Generative AI tools can also assist in the drafting of tender documents. The Action Plan subsequently advocates for AI to increase resilience and productivity across public services. Dynamic Purchasing Systems (DPSs) offer flexibility in the procurement of AI solutions, and subsequently allow the entry of new suppliers to the market and foster innovation, essential if public procurement is going to develop in parallel with the evolving technological landscape. Utilising AI within the public procurement sector adheres to the Act’s goal of simplifying processes and maximising value for money.

Managing contracts

AI can assist in putting together contracts, checking for issues, and ensuring the key terms required for compliance are included. It is advised for contracting authorities to make use of AI in the evaluation process only with legal support and strong internal AI expertise to avoid procurement risks. Contracting authorities must articulate their rationale behind each contract award and, therefore, contracting authorities should ensure that their AI is trained on scoring methodologies and evaluation criteria.

Security

AI can detect subtle patterns of suspicious behaviour from suppliers, which could indicate that those suppliers are likely to make late payments or that there is cause to suspect fraud.

Forecasting

Predictive models will be able to provide accurate forecasting for price fluctuations and shifts in demand to help procurement professionals get ahead of shortages.

Challenges

Part of the scale of AI’s ability to effect change in public procurement comes from its potential application to several stages along the procurement life cycle, from creating a tender, to responding to it, to judging that response. However, the potential benefits bring with them several risks.

Security concerns

Although the use of AI in public procurement has the ability to reduce security risks, collaboration with information assurance teams is essential to implement risk mitigations. AI models are trained on large data sets which often include personal information. If not handled properly, there is a risk of data protection breaches, resulting in not only regulatory penalties but also reputational damage. The AI Playbook echoes the need for enforced security provisions by highlighting governance and security considerations, particularly for high-risk AI systems. Contracting authorities must balance the enhanced flexibility offered by the Act with the need for robust security, ensuring that innovation is not hindered in the process.

Capacity and skills gap

The public sector’s current capacity to procure and manage AI effectively is not up to the standard required for the use of AI in public procurement. The AI Playbook suggests upskilling teams in technical and ethical AI competencies. The Action Plan outlines how insufficient AI knowledge within the sector risks over-reliance on suppliers, ultimately blurring accountability lines. 

Assessment of internal resources

The use of AI, with its efficiencies and thus time-saving abilities, could result in increased frequency of bidding by certain suppliers. This is likely to result in an increase in compliant tenders, with no corresponding increase in evaluation capacity.

Ethical and fairness concerns

AI cannot exercise discretion and judgement and must be trained using existing data, which is often biased. How AI then applies its “knowledge” to the assessment of tenders can be unfair or discriminatory, which is particularly problematic in areas such as hiring, loan applications and social services.

Additionally, the precision required in the technical specifications of the tender may not be reliably produced by models designed to have a general capability to create large numbers of tenders over time, as there will likely be a generic quality to each set of tender documents produced. The recent introduction of the Act means that there is only limited usable data on its application from which AI models can be trained to reach decisions. As a result, at least for the time being, some information in the tender may be inaccurate or even contradictory. It could therefore be a while until AI is a reliable tool for independently creating tenders.

A “Black Box”

A core principle of UK public procurement law is transparency. This principle, reflected in the Act, requires contracting authorities to provide clear and understandable reasons for their procurement decisions. Where AI is involved, fulfilling this obligation may become quite challenging. Algorithms often operate as “black boxes”, making it difficult for unsuccessful bidders to understand the rationale behind a decision, let alone challenge it effectively. Whether or not a decision reached by AI contains any element of unfairness or bias, will be difficult for those employing these systems to discern.

In litigation, this lack of explainability can complicate matters. Claimants may argue that they have been denied effective remedies if they cannot access or interpret the basis on which they were scored. This raises broader issues under administrative law, particularly around procedural fairness and the right to a reasoned decision.

These issues will be compounded when the responses being evaluated have also been put together by AI as, if the same model is employed by multiple bidders to help produce the tender submission, their submissions may be extremely similar, making the exercise of demonstrating the reasoning behind the differences in scores awarded to each supplier even more of a challenge.

While procurement professionals may seek to use AI to manage this increased volume of submissions, evaluating tender submissions may be the most challenging area in which to attempt to incorporate AI. This is due in large part to those using it not knowing exactly what the machine has based its decisions on.

Our Approach

Procurement professionals on both the buyer and bidder sides should prepare for a wave of opportunities and complexity as AI becomes more embedded in public sector infrastructure. Success will depend on thoughtful early engagement, an understanding of the new legal framework under the Act, and a willingness to embrace innovation.

There are many questions which need to be explored as AI becomes an increasingly prominent figure in public procurement, such as:

• What are the regulatory implications of using AI in procurement?
• How do we ensure that AI systems are fair and unbiased?
• What AI technologies are available, and how do they fit into the procurement process?

In order to get ahead of the emerging challenges presented by AI’s ongoing adoption, our practice maintains a position at the forefront of innovation in the legal space, having already integrated AI directly into our working processes and being the first firm to introduce a firm-wide bonus scheme linked to AI utilisation.