Both sides of the table are using AI now.

Both sides of the table are using AI now. Social housing procurement needs to catch up with that reality.

There’s something slightly absurd about the current state of AI in social housing procurement. On one side of the table, procurement teams are using AI to help research specifications, benchmark pricing and check their evaluation reasoning. On the other side, suppliers and contractors are using it to write tender responses, structure method statements and polish their quality submissions. Both sides know the other is probably doing it but neither side is talking about it openly.

I’ve had conversations with colleagues across the sector over the past few months, buyers and suppliers, and the picture that’s emerging is one of quiet, uncoordinated adoption. Housing associations and councils are dipping in and out, using AI for some tasks, overdoing it for others and steering well clear in certain areas because they’re worried about challenge. Suppliers are doing much the same, leaning on AI to produce well-crafted bids while hoping the buyer won’t hold it against them. There’s very little guidance that speaks specifically to social housing, and PPN 017, the government’s procurement policy note on AI transparency, is aimed at central government rather than the practical realities of housing procurement.

What concerns me is the trust gap this is creating. And if we don’t start having honest conversations about it soon, it’s going to create problems that are harder to unpick later.

What I’m seeing on the supplier side

For smaller contractors and specialist firms, AI has been genuinely useful. Writing a tender response for a housing association is time-consuming and expensive, and firms that don’t have a dedicated bid team have always been at a disadvantage. AI tools have helped level that out, giving smaller suppliers the ability to produce structured, articulate submissions that can compete on a more equal footing with larger firms. That’s a good thing for competition and for housing associations who want access to a broader supply base.

But it’s created a new problem for people in my position. When AI is involved in bid writing, the quality of the prose stops being a reliable signal of the quality of the firm behind it. I’m seeing submissions that are polished, well-structured and perfectly aligned to our evaluation criteria, but that don’t necessarily reflect what the contractor can actually deliver on site. The gap between what’s written and what’s real has widened, and evaluation panels are having to work considerably harder to tell the difference.

This is particularly tricky with quality questions. When every bid reads like it was written by the same capable hand, the differentiation that should come through in quality scoring gets compressed. Competitions start drifting towards price-only outcomes because there’s nothing meaningful separating the quality submissions. And price-only competitions, as anyone who’s managed a housing repairs contract will tell you, rarely produce good results for tenants.

What’s happening on the buyer side

Procurement teams are using AI too, and mostly for sensible things. Researching market conditions before writing a specification. Pulling together background on supplier accreditations and industry standards. Summarising lengthy tender documents to get an overview before diving into the detail. Checking whether evaluation reasoning is consistent with the published criteria.

Where it gets more complicated is in the areas that require judgement. Having worked in public procurement for many years, I’ve come to recognise that a specification isn’t just a list of requirements. It’s a set of choices about what matters most for a particular housing organisation, its stock, its tenants, its operational context. AI can help with the research that informs those choices, but someone with genuine category expertise and knowledge of the housing provider’s priorities has to make the final calls. If you hand the whole thing to AI, you end up with a specification that looks professional but doesn’t actually reflect what your organisation needs.

The same applies to evaluation. AI can help you spot inconsistencies in your scoring and test whether your written feedback holds up against the criteria. It’s useful as a sense-checker, almost like a neutral moderator that can offer an impartial perspective on your reasoning. But the judgement about whether a contractor can genuinely deliver what they’ve promised, whether their case studies are relevant, whether their resource plan is realistic for a scheme of this size, that requires human experience and operational knowledge. The skills that matter most here, contextual understanding, proportionality, the ability to read between the lines and ask the right follow-up questions, are precisely the ones AI doesn’t have.

Two areas where the stakes are especially high

Feedback letters to unsuccessful suppliers are a growing source of challenge. Under the Procurement Act, suppliers receive formal written feedback explaining their scores and the panel’s reasoning. Supplier challenges around these letters are increasing, and I can understand why a time-pressed team might reach for AI to speed up the drafting. There is a useful role here: when you’ve had operatives, tenants and service managers all contributing to a quality evaluation, each responding in different ways, AI can help analyse and consolidate those central themes. That consolidation work is genuinely helpful with the time-consuming process of offer letter writing.

But the letter itself needs to be written by someone who understands the procurement, the criteria, the moderated scores and the specific reasons a particular supplier didn’t succeed. It’s only right that suppliers who’ve spent weeks working on a bid receive a detailed outline of why they achieved certain scores. An AI-generated draft that doesn’t precisely reflect what the panel actually decided is a real problem if the decision gets challenged. The time saved in drafting could end up costing significantly more in defending a poorly worded letter.

The Procurement Act’s competitive flexible procedure opens up another area where AI has a useful supporting role. For the first time, social housing buyers can enter negotiations with suppliers at multiple points during the process. AI can help teams prepare by extracting commercial insight from existing contracts, deconstructing pricing structures and anticipating where a supplier might push back during negotiations. But the negotiation itself, reading the room, understanding what a supplier is really telling you, knowing when to push and when to concede, that’s fundamentally human work and always will be.

The conversation the sector needs to have

PPN 017 tells contracting authorities to ask suppliers whether they’ve used AI in their bid, and to treat the disclosure as information rather than scoring it. That’s a reasonable starting point, but it’s not specific enough to social housing. Our sector operates under a level of scrutiny and accountability, where every purchase must be transparent, defensible and shown to positively impact tenants’ lives, that generic government guidance doesn’t fully address.

What I think we need is a more honest, two-way conversation. Housing associations should be clearer with their supply chains about what they expect. Tender documents could include straightforward wording reminding suppliers that they’re contractually required to deliver any commitments in their responses, whether those commitments were drafted by a person or by AI. That’s not about penalising AI use. It’s about making sure everyone understands that a well-written bid creates a real obligation, regardless of who or what produced it.

Suppliers should feel able to disclose AI use without worrying that it counts against them, because for many smaller firms, AI is the reason they can bid at all. But alongside that openness, procurement teams need to adapt how they evaluate. If the written submission alone can no longer be taken at face value, that means more clarification interviews, more probing of case studies and references, more site visits where practical. The evaluation process needs to get more investigative to compensate for the fact that the surface quality of bids no longer tells you as much as it used to.

And internally, housing associations need to be clearer with their own teams about where AI fits in their procurement process. Which tools are approved? What types of data can be entered? At what points does AI-generated content need human sign-off before it goes anywhere? Most organisations don’t have a documented position on this yet, and the absence of one is quietly creating risk.

Why this matters more in our sector than most

Social housing procurement isn’t like buying office supplies. Every contract, every supplier appointment, every pound spent, has a connection to the quality of life tenants experience. A repairs contractor who was appointed on the strength of a polished AI-written bid but can’t deliver reliable first-time fixes has a direct impact on people’s homes. A materials supplier who oversold their product range and then substitutes lower-quality alternatives affects the standard of work across an entire programme.

The core principles of public buying, integrity, value for money, fairness, accountability, competition and transparency, must always be led and controlled by human hands. AI can support that work, and it should. It can free buyers from transactional tasks so they can focus on strategic thinking, partnership development and adding genuine value to their organisations. But those human skills, the contextual understanding, the proportionality, the nuanced thinking, the ethical reasoning, the relationship-building, are what make the difference between procurement that’s compliant on paper and procurement that actually delivers for tenants.

I’d like to see more of the sector talking about this openly. We’re all figuring out the same questions, and sharing what’s working and what isn’t would help everyone move faster and with more confidence. If you’re grappling with any of this in your organisation, I’d welcome the conversation.

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

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

How PfH can help

Quantum, our spend and contract management platform, gives your team the data visibility to track supplier performance against what was promised at tender stage, not just what was written in the bid.

Procurement consultancy including support with specification development, tender evaluation and contract management, making sure human expertise drives the decisions that matter.

Frameworks and DPS solutions where suppliers have already been through our vetting and due diligence process, reducing the risk that a polished bid doesn’t reflect genuine delivery capability.

---

---

Sources:

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

Trowers & Hamlins, The future of AI in public procurement, February 2025

Burges Salmon, Improving transparency of AI use in public procurement: PPN 017 published, February 2026