AI and accountability:

AI and accountability: what UK procurement law expects when automation meets public money

The principle at the heart of UK public procurement has always been clear: public money must be spent transparently, and the people responsible for spending it must be able to explain and defend their decisions. In England, Wales and Northern Ireland, the Procurement Act 2023 has raised the bar on what that transparency looks like in practice, from mandatory KPI reporting on contracts exceeding £5 million to quarterly payment disclosures for payments above £30,000. In Scotland, the Procurement Reform (Scotland) Act 2014 and the Public Contracts (Scotland) Regulations 2015 impose their own accountability framework, requiring contracting authorities to publish procurement strategies, report annually on performance and embed Fair Work First principles across their supply chains.

At the same time, AI adoption across public sector procurement is accelerating. From local authorities and NHS trusts to universities and blue light services, contracting authorities across the UK are using it to draft specifications, summarise tender responses and categorise spend data. Suppliers are using it to write bids. And in both cases, the governance frameworks that should sit around this technology have not kept pace with how quickly people have started relying on it.

This creates a tension that every public sector procurement team needs to address, regardless of which legislative regime they operate under. Both the Procurement Act and Scotland’s procurement framework require you to demonstrate how and why decisions were made. AI, by its nature, can make that harder to do.

The accountability question that procurement law forces you to answer

Under both the Procurement Act and Scotland’s procurement legislation, contracting authorities must be able to explain the rationale behind procurement decisions. That includes how evaluation criteria were applied, why a particular supplier was selected and, if challenged, what evidence supports each score. In Scotland, the Procurement Reform Act’s emphasis on sustainable procurement and community benefit adds a further layer of accountability that procurement teams must be able to demonstrate.

When a procurement professional reads a tender response, forms a judgement and writes up their assessment, the accountability chain is clear. The individual can explain their reasoning. The moderation panel can interrogate it. The audit trail shows who assessed what and why.

When AI is introduced into that process, the chain becomes less visible. If a procurement officer uses an AI tool to summarise a lengthy tender response before scoring it, the question becomes: did they evaluate the supplier’s actual submission, or an AI-generated interpretation of it? If the AI missed a nuance, omitted a qualification or emphasised one section over another, the evaluator may never know, and the resulting score may not reflect the bid that was submitted.

This is not a hypothetical concern. Ward Hadaway’s analysis of AI and the Procurement Act notes that contracting authorities should document the operation and rationale of any AI tool used in procurement, to uphold the principle of accountability and ensure that suppliers have access to meaningful explanations where AI has influenced the outcome. That documentation obligation sits alongside the Act’s broader transparency requirements and is one that many organisations have not yet built into their processes.

Where the compliance exposure sits

Three areas of the procurement lifecycle carry particular risk when AI is involved.

Evaluation and scoring. Both the Procurement Act and Scotland’s procurement regulations require contracting authorities to assess tenders against published criteria and to provide meaningful feedback to unsuccessful suppliers. When AI tools are used to assist with evaluation, whether by summarising responses, suggesting scores or flagging areas of concern, the authority must still be able to demonstrate that a qualified human made the final judgement. Neither regime prohibits the use of AI in evaluation. But both require the contracting authority to own the decision, explain it and defend it if challenged. A score that was influenced by an AI summary the evaluator did not verify against the original submission is difficult to defend.

Standstill and feedback. Under both procurement regimes, unsuccessful suppliers are entitled to feedback that explains what they scored and why. This feedback must be specific, fair and traceable to the evaluation criteria. AI can draft these letters quickly, but a generic output that does not reflect the panel’s actual reasoning is a compliance risk. If a supplier challenges the decision and the feedback letter does not align with the moderated scores, the authority’s position is weakened. Human review of every piece of formal feedback is not optional under either regime.

Data confidentiality. PPN 017, updated in February 2025, is explicit that contracting authorities should put proportionate controls in place to prevent confidential information being used as training data for AI systems. While PPN 017 applies directly to central government, its principles are relevant across the UK public sector. In Scotland, the Scottish Government’s AI Strategy 2026-2031 and the Scottish AI Register reinforce the expectation that public bodies will use AI transparently and ethically, with the Digital Public Services Delivery Plan setting a clear direction for responsible AI adoption across devolved public services. When procurement professionals paste tender documents, pricing schedules or internal strategy papers into publicly available AI tools, that data may be ingested by the model. Most contracting authorities, whether in England, Wales, Northern Ireland or Scotland, do not yet have clear policies on which AI tools are approved for procurement use and what categories of data can be shared with them. This is a governance gap that needs closing.

PPN 017 and Scotland’s AI framework: what they require and where the gaps are

In England, Wales and Northern Ireland, the Government’s Procurement Policy Note on AI transparency (PPN 017) provides the closest thing to official guidance on this issue. It applies directly to central government departments but sets the direction of travel for the wider public sector.

PPN 017 establishes several important principles. Suppliers’ use of AI in bid writing is not prohibited, but contracting authorities should ask suppliers to disclose their use of AI when responding to tenders. Those disclosures should be used for information only and should not be scored as part of the evaluation. Contracting authorities should also put controls in place around their own use of AI, particularly in relation to confidential data.

In Scotland, while there is no direct equivalent of PPN 017, the Scottish Government has taken a broader approach to AI governance through its AI Strategy 2026-2031 and the Scottish AI Register, which requires public bodies to log AI systems they are using or developing. Public Contracts Scotland has also published guidance on AI use in procurement. Several Scottish councils, including Edinburgh and Glasgow, have begun embedding algorithmic impact assessments into their procurement workflows, an approach that could become standard practice as the Scottish Government’s strategy matures.

What neither PPN 017 nor Scotland’s framework currently does is tell contracting authorities how to build an internal governance framework specifically for AI use in procurement. Both set principles but leave the practical application to each organisation. For many public sector teams, whether in local government, education, healthcare or emergency services, that is where the gap sits. They understand the risks in broad terms but do not have a documented policy that sets out which tools are approved, what oversight is required at each stage of the procurement lifecycle and how AI-assisted decisions should be recorded.

What a practical governance framework looks like

The framework does not need to be lengthy, but it does need to be specific enough to give procurement teams clear guidance on day-to-day decisions.

Approved tools. Identify which AI tools are sanctioned for use within the procurement function and which are not. Public, consumer-grade tools that may use input data for training purposes should be treated differently from enterprise tools with contractual data protection guarantees.

Data boundaries. Define what types of information can be entered into AI systems and what cannot. At a minimum, live tender responses, supplier pricing, commercially sensitive strategy documents and any information not already in the public domain should be excluded from public AI tools.

Mandatory human review. Identify the points in the procurement lifecycle where AI-generated content must be reviewed and signed off by a qualified individual before it is issued. Evaluation scores, standstill letters, contract award decisions and published notices are obvious candidates.

Documentation. Record where and how AI has been used in each procurement exercise. This supports the accountability requirements of the Act and provides an audit trail if a decision is challenged. It also helps the organisation learn over time which applications of AI are adding value and which are creating risk.

Supplier disclosure. Incorporate AI disclosure questions into tender documentation, following the example questions set out in PPN 017’s annexes. Make clear that disclosure is not penalised and is used for transparency purposes only.

The bigger picture: AI as a tool, not a substitute

The UK’s procurement legislation, whether the Procurement Act 2023 or Scotland’s Procurement Reform Act, was designed to bring transparency, accountability and integrity to how public money is spent. Those objectives do not change because AI is involved. If anything, they become more important.

AI can be a genuinely useful tool for procurement teams under pressure. It can accelerate drafting, improve data analysis, support consistency in evaluation and reduce administrative burden. But every one of those applications carries a corresponding accountability question: can you explain the decision? Can you trace it back to the evidence? Can you demonstrate that a human being, with the relevant knowledge and authority, made the final call?

Contracting authorities that can answer yes to those questions will find AI a valuable addition to their procurement capability. Those that cannot will find that the very technology they adopted to work faster has created compliance risks that are slower and more costly to resolve.

Rob Peck is Chief Commercial Officer at Inprova Group

This article draws on themes explored in an opinion piece published in Construction Management.

How Inprova can help

Quantum, our spend and contract management platform, provides the data visibility and audit trail that procurement legislation across the UK demands, giving your team line-level spend data, contract performance tracking and reporting capability in a single platform.

Consultancy support to help your organisation design evaluation frameworks, manage tender processes and build governance structures that keep pace with evolving procurement legislation and the role of AI. We work with customers across local government, education, healthcare and blue light services throughout the UK.

Frameworks, Dynamic Markets and DPS solutions reduce procurement workload and provide pre-tendered access to vetted suppliers across the public sector.

---

---

Sources:

Cabinet Office, PPN 017: Improving transparency of AI use in procurement, February 2025

Ward Hadaway, Procurement in a Nutshell: Procurement Act 2023 and the use of AI in public procurement, May 2025

Browne Jacobson, Procurement Act 2023: Key transparency changes in force for 2026

Brabners, Procurement Act 2023: Further 2026 duties for contracting authorities,

March 2026 Scottish Government, Scotland’s AI Strategy 2026-2031,

March 2026 Scottish Government, Public sector procurement policy and legislation

Scottish Government, Scottish Procurement Policy Notes (SPPNs)

DWF Group, Understanding the Scottish cross-border public procurement regulations,

January 2026 Local Government Association, Responsible buying: how to build equality data protection your AI commissioning