The perils of over-reliance on AI in construction procurement

Everyone is talking about AI, and the area I work in – public sector construction and housing – is no different. Over the past 18 months its use has exploded, with people integrating AI into their everyday working routines and increasingly building it into the systems and products our industry relies on.

But there is an issue here. I’m in the procurement side of construction, supporting local authorities and social landlords to buy goods and services needed to build new homes and improve existing ones. As contracting authorities, our sector is bound by the Procurement Act 2023 – a raft of important rules that ensure public money is spent transparently, fairly and in a way that delivers value for money.

That principle sits at the heart of everything we do, and it’s exactly where the rapid rise of AI starts to create both opportunities and challenges.

Yes, artificial intelligence can help us work faster and smarter, but over reliance risks undermining the safeguards we have in place to protect the public purse.

Let’s begin with the tender process. We are seeing an increased use of AI from contractors to support bid writing, something that’s particularly helpful to SMEs with limited resources. This adoption is understandable, but it can present problems for building firms and buyers if it’s used without guardrails.

For example, AI‑generated bids that aren’t properly checked can oversell and overcommit, leaving companies unable to meet the contractual obligations the system has written for them.

On the other side, procurement teams are facing many more generic, template-perfect and sometimes unsubstantiated responses and they’re having to spend extra time differentiating, spotting AI embellishment and drilling deep into claims.

There’s also a real risk that many suppliers score very highly on quality, pushing competitions into a price‑only race that rarely delivers good outcomes.

Another issue is an over-reliance on AI in areas where human judgement and context are critical. Tender evaluations, for instance, still require individuals with operational knowledge to assess bids. Relying solely on AI could result in inaccurate scoring and, potentially, a contract being awarded based on a generic response that scores well on AI is opposed to one specifically tailored to a housing organisation’s operational needs.

The same goes for offer letters. Once a contract is awarded, feedback is sent to unsuccessful building firms, outlining what they scored and why. Writing these letters is time‑consuming and it’s tempting to turn to automated tools. However, it is vital that all feedback directly relates to the scoring criteria and outlines clearly the evaluators’ rationale for a moderated score. Error in these letters could increase exposure to challenge and that’s why human understanding and oversight are essential here.

Despite these issues, it’s worth recognising that AI offers big opportunities for contracting authorities when applied with human knowledge, capability and clear controls.

For instance, it can be trained to sense check your bid assessments and offer letters, providing an impartial view on whether your scores and feedback are fair based on tender criteria, weighting and supplier responses. Rather than taking the lead or influencing, AI’s role here is to moderate and benchmark.

That ability to process, check and organise information is equally valuable at the front end of procurement. AI can act as a drafting assistant, shaping early requirements, proposing structure and identifying key elements like the latest accreditations a supplier must have, before a technical expert writes the full specification.

AI can also significantly reduce the administrative burden associated with ongoing contract management. In the construction sector, teams often manage large volumes of disparate spend data, particularly for building materials, where categorising thousands of individual product lines can be both complex and time-consuming.

By leveraging AI, organisations can automate data classification and analysis, enabling rapid identification of off-contract purchasing and instances where higher-priced alternatives are being used. This provides faster, more accurate spend visibility, supports robust benchmarking, and helps organisations assess whether value for money is being achieved while identifying opportunities for cost savings and contract compliance improvements. 

With a brand new public procurement regime going live just over a year ago, many buyers are still getting to grips with the new rules. Artificial intelligence can support with navigation and interpretation, such as clarifying next steps around a particular procedure. However, to avoid inaccuracies, always ask your AI system to provide sources and links to back up what it’s telling you.

 It feels like we are at the beginning of this AI journey in public sector construction procurement. Although the government produced a relevant procurement policy notice last year and the Local Government Association released a guide to responsibly buying AI (which also explores how procurement teams are using AI tools), the commentary on this issue is noticeably quiet.

To confidently harness AI without being either complacent or afraid of challenge we need an honest and practical discussion about the opportunities, risks and governance required going forward. Only then will we be able to strike the right balance between automation and human judgement.

Rob Peck is chief commercial officer at Procurement for Housing (PfH)