Construction AI: the cladding code gets a rewrite for consistency, and the buyer now opens ChatGPT before Google

On 15 September the British Standards Institution published PAS 9980:2026, the revised code of practice for fire risk appraisals of external walls, or FRAEWs, on existing blocks of flats. It's the first update since the code was first written in 2022. The Construction Leadership Council flagged it the following day, along with a set of supporting guidance from the Ministry of Housing, Communities and Local Government.

Read the small print before you read the headlines. The revisions are designed to add structure to assessments, improve clarity and drive consistency in how the methodology gets applied. They are not meant to change the purpose of the code or the way risk is judged, and BSI has been explicit that a FRAEW competently produced under PAS 9980:2022 remains valid and can still be relied upon. So this is a change to how you do the work, not a reason to re-commission your building stock. If you were bracing for another round of everyone starting again, you can put that down.

What the rewrite is actually chasing is the thing that's quietly caused a lot of the misery: inconsistency. Two competent assessors could look at the same wall and reach genuinely different conclusions, and that spread is a big part of why leaseholders have been stuck with flats they can't sell or remortgage, and why responsible persons have struggled to know whether the report in front of them is any good. The code still takes a holistic, risk-based view rather than a tick-box compliance one, and it keeps pushing proportionate remediation, decisions matched to the actual risk, rather than a blanket rip-it-all-off approach. I'd argue the proportionality point is the one the sector has been slowest to trust, and it's the one that saves the most money on buildings that don't need the full works.

The bit I'd not skip is the MHCLG guidance sitting alongside the code. It's written for the non-technical people who commission these appraisals and have to live with the results, and it includes a guide called "Recognising quality in an FRAEW" that sets out what good looks like. What that does on the ground is hand the leaseholder and the responsible person a checklist they can hold a report against. Use it when you receive a FRAEW, not only when you produce one. The record is the point here, same as it is at Gateway 2. A consistent, well-evidenced report is what lets a resident get on with their life, and that's who this is for.

Here's the quieter story of the week, and it's one every supplier and software buyer should sit with. Writing in the construction trade press on 18 September, Paul Houston of Catalyst Marketing Agency argued that the approved-supplier list, the thing a building-product manufacturer or specialist subcontractor spent decades working its way onto, is being rebuilt from scratch by AI-assisted research. The reason is plain. Quantity surveyors, procurement managers and specifiers increasingly open ChatGPT before they open Google, type something like "suggest five UK manufacturers of fire-rated cladding who are UKCA marked and have high-rise residential experience," and get an instant shortlist. No phone call, no site visit, and often no click through to anyone's website at all.

Now the uncomfortable number. Houston cites roughly 1 per cent internal AI adoption inside construction firms against 37 per cent in professional and technical small businesses, and separate UK benchmarks this year put construction as the lowest-adopting sector overall, somewhere around 13 per cent. I'm not sure the exact 1 per cent survives contact with every firm on site, but the direction is right and it's the gap that matters: the people searching for you are well ahead of the people being searched for. A machine that has never met your sales team, and doesn't care how good your relationship with the outgoing buyer was, is writing the list.

This is where the challenger case gets literal. Challenger construction software, the broad AI-first platforms taking on the big incumbents, has always argued that being simple to find and simple to buy is a feature, not a nicety. When the shortlist is written by an AI, that stops being a marketing preference and becomes the whole game. The model cites what it can find, understand and trust: structured certifications, current case studies, clear published detail, and third-party mentions in the trade press and industry bodies. The incumbent's old moat was the framework and the long relationship. Neither of those is visible to a language model. What wins now is the credential a machine can read and a reputation it can verify from the outside, which is exactly the ground the challengers were already fighting on.

So the practical bit closes on a person, not a platform. Picture the QS at nine in the evening, programme slipping, compliance box unticked, asking the machine for a name. If your firm's proof isn't written down where that machine can reach it, you're not in the room. Tidy, current, structured credentials are the new way onto the list. That's what it's about.

Look at the two items together and they're the same point wearing different clothes. The cladding code is trying to make one assessor's report read the same as another's, because an inconsistent record leaves a resident stranded. The supplier-list piece is a warning that your credentials now have to read cleanly to a machine, because an unreadable record leaves you off the shortlist. In both cases the thing being judged isn't the conversation or the relationship, it's the document, and whether it holds up when someone, or something, reads it cold.

That's not a new discipline, it's the old one with higher stakes. The firms that keep their evidence tidy, current and structured, whether that's a FRAEW written to the new code or a case study written so an AI can cite it, are the ones that come out ahead. The ones treating the paperwork as an afterthought are about to find out that both a human regulator and a language model have stopped giving them the benefit of the doubt.

Today's action: Pick one document that represents your firm to the outside world, a recent FRAEW or your top product's credentials page, and read it as a stranger would. If it isn't clear, consistent and current, fix that one first.