AI can help Scottish housing teams work faster. It can't help them meet the sustainable procurement duty.

AI can help Scottish housing teams work faster. It can’t help them meet the sustainable procurement duty.

I’ve noticed a shift over the past year in the conversations I’m having with Scottish housing associations about procurement. AI keeps coming up, sometimes directly, more often indirectly, as people describe how they’ve started using tools like ChatGPT to draft specs, summarise tender documents or research supplier accreditations. Nobody’s making a big deal of it. It’s just quietly becoming part of how procurement work gets done.

On the supplier side, the same thing is happening. Contractors and product suppliers are using AI to write tender responses, and for smaller firms in particular, it’s been a genuine help. Writing a detailed bid for a housing association is resource-intensive, and firms without a dedicated bid team have always struggled to compete on equal terms. AI has made it easier for them to put together structured, professional submissions, which is good for competition and good for housing associations who want a broader range of suppliers to choose from.

So far, so familiar. This is playing out across the whole UK public sector. But what makes the Scottish picture different is the procurement framework we operate under, and the specific obligations it places on housing associations that AI tools simply aren’t equipped to meet.

Scotland’s procurement framework asks questions AI can’t answer

The Procurement Reform (Scotland) Act 2014 requires contracting authorities carrying out regulated procurements to consider how they can improve the economic, social and environmental wellbeing of their area. That’s the sustainable procurement duty, and it applies to every regulated procurement, not just the big ones.

In practice, that means every time a Scottish housing association goes to market, it should be thinking about how the contract can support local employment, strengthen the local supply chain, reduce environmental impact and embed Fair Work First principles. Those aren’t abstract aspirations bolted onto the evaluation. They’re supposed to shape how the procurement is designed from the start, influencing the specification, the lot structure, the evaluation criteria and the contract management approach.

AI can help with pieces of that. It can research local labour market conditions, identify relevant environmental standards, or pull together examples of community benefit clauses from other contracts. But the judgement about what matters most for a particular housing association, in a particular part of Scotland, procuring a particular category of work, requires a depth of local knowledge and organisational context that no AI tool has.

A housing association in the Highlands procuring a planned works programme has different sustainable procurement considerations from one in Glasgow. The labour market is different, the supply chain is different, the community benefit opportunities are different, the environmental context is different. The sustainable procurement duty asks contracting authorities to think about all of this, and the thinking has to be specific, not generic. AI defaults to generic because it doesn’t know your area, your tenants or your strategic priorities.

Community benefit is where the gap shows most clearly

Sections 24 and 25 of the Act require contracting authorities to consider imposing community benefit requirements on regulated procurements with an estimated value of £4 million or more. But the statutory guidance makes clear that community benefit can and should be considered below that threshold too, through the sustainable procurement duty.

What I’m seeing in conversations with members is that community benefit is one of the areas where AI adoption could do more harm than good if it’s not handled carefully. AI can generate community benefit clauses that look perfectly reasonable on paper. It can suggest apprenticeship targets, local employment percentages and carbon reduction commitments. But those suggestions are drawn from training data, not from an understanding of what’s deliverable in your local area, what your existing suppliers can realistically commit to, or what your tenants and communities actually need.

A well-designed community benefit requirement is specific to the contract, proportionate to its value and genuinely deliverable. Getting that right takes conversation with the supply chain, understanding of the local labour market and a practical sense of what a contractor working in your area can actually achieve. In my experience managing supply chain partnerships across the UK, I’ve seen first-hand how the difference between a meaningful social commitment and a box-ticking exercise can come down to whether someone had done the groundwork to understand what was realistic. AI can’t do that groundwork for you.

Fair Work First adds another layer

Scotland’s Fair Work First policy requires contracting authorities to consider fair work practices, including payment of the real Living Wage, in all relevant procurement exercises. The Scottish Government is clear that this should be embedded in how procurement is designed, not treated as an afterthought.

Again, AI can help with background research. It can identify current Living Wage rates, flag relevant Scottish Procurement Policy Notes, and suggest fair work criteria wording. But deciding how to weigh fair work practices in a particular evaluation, how to assess whether a supplier’s fair work commitments are genuine rather than performative, and how to monitor delivery through the life of the contract, these are judgement calls that require experience, sector knowledge and an understanding of how suppliers actually operate.

For smaller Scottish housing associations where procurement often falls to the asset team or finance director rather than a dedicated procurement function, the temptation to lean more heavily on AI is understandable. There’s not enough time or resource to do everything from scratch, and AI seems to offer a shortcut. But the areas where Scotland’s procurement framework demands the most careful thinking are precisely the areas where AI shortcuts carry the most risk.

Scotland has its own AI guidance, and it’s worth knowing about

One thing that’s often missed in the broader conversation about AI in procurement is that the Scottish Government has published its own guidance, separate from the UK Government’s PPN 017. Public Contracts Scotland and the Procurement Journey website both carry guidance for buyers and suppliers on the responsible use of AI in Scottish public sector procurement. It’s still early-stage and fairly high-level, but it establishes an important principle: AI use is permitted in Scottish procurement, by both buyers and suppliers, but it must be used responsibly and in accordance with each public body’s own policies.

For housing associations that don’t yet have an internal policy on AI use in procurement, that last point is significant. The guidance assumes you have one. If you don’t, you’re in a position where your team may be using AI tools without any agreed boundaries about what data can be entered, what types of content need human sign-off, and how AI-assisted decisions should be recorded in your procurement documentation.

Given that your annual procurement report needs to demonstrate how your procurement activity has delivered against your published procurement strategy, and your Annual Assurance Statement to the Scottish Housing Regulator requires you to confirm compliance with all regulatory requirements, having no documented position on AI use in your procurement function is becoming harder to justify.

Where AI genuinely helps, and where it doesn’t

I don’t want to give the impression that Scottish housing associations should be avoiding AI. Used in the right way, it’s a genuinely useful tool for teams that are stretched thin, which is most of them.

It’s helpful for the research phase of a procurement, pulling together market intelligence, identifying relevant standards and accreditations, and structuring early-stage thinking about what a specification might need to cover. It’s helpful for data work, categorising spend, identifying off-contract purchasing, benchmarking prices. And it’s helpful as a sense-checker, reviewing your evaluation reasoning against the published criteria to make sure you haven’t drifted.

Where it falls short is anywhere the Procurement Reform Act asks you to exercise judgement about how your procurement can improve outcomes for your area. The sustainable procurement duty, community benefit requirements, Fair Work First, these all require you to make decisions that are specific to your organisation, your tenants and your local context. AI can inform those decisions. It can’t make them for you, and if your procurement documentation suggests it has, that’s a problem when it comes time to report on what your procurement actually delivered.

The responsible approach is to be clear with your team about where AI sits in your process, use it where it adds value, keep human expertise at the centre of the decisions that matter, and make sure you can evidence that in your annual procurement report and your Assurance Statement.

If you’re working through any of this in your organisation, we’re having these conversations with members across Scotland and I’m always happy to talk it through.

Kathryn Irons is Head of Key Accounts at Procurement for Housing Scotland (PfH Scotland)

This article is based on an opinion piece published in Housing Digital.

How PfH Scotland can help

Quantum, our spend and contract management platform, gives you the data visibility you need to track supplier performance, monitor community benefit delivery and produce the evidence your annual procurement report and Assurance Statement require.

Procurement consultancy including support with specification development, sustainable procurement duty compliance, community benefit design and evaluation, helping you get the human judgement right on the things that matter most.

Frameworks and DPS solutions designed for Scotland’s procurement legislation, giving your team compliant routes to market with suppliers who’ve already been through our vetting process.

---

---

Sources:

Scottish Government, Procurement Reform (Scotland) Act 2014: statutory guidance, October 2022

Scottish Government, Community benefit requirements in procurement

Public Contracts Scotland, Guidance on the use of Artificial Intelligence (AI), September 2024

Scottish Government, Fair Work First in procurement guidance

Scottish Government, Scottish Procurement Policy Notes (SPPNs)