Who signs, who owns, and what's on the record

NG Bailey, the Yorkshire-based engineering and services group, turned over £706.7m in the year to 27 February 2026 and lifted pre-tax profit 47% to £26m. As part of its 2030 strategy, reported at the start of July, it's creating a chief AI officer role. Chief executive Jonathan Stockton called it an "important step" in providing "clear" leadership, so the group can identify, develop and implement AI opportunities in a structured, responsible and secure way. I named this title in passing last week, so treat this as the fuller reading rather than fresh news.

Read past the corporate phrasing and there's a genuine shift underneath. Most contractors have AI happening to them: someone in preconstruction quietly using a model to write a method statement, someone in the bid team running documents through a chatbot, and nobody's decided whether that's allowed, where the data goes, or who's accountable if it's wrong. What a chief AI officer does, at least in theory, is put one person's name against those decisions with the authority to stop a bad one. Not the acronym, the accountability.

One honest caveat. A new C-suite title can be real governance or a press release with a job description attached. The test is whether the person can veto a tool the sales team loves, and whether they own the boring stuff, data residency, model choice, the audit trail, rather than just the innovation showreel. But the direction is right, and the signal to everyone smaller is clear. If a £700m business has concluded AI adoption needs a named owner, the mid-tier firm running the same tools with nobody in charge is carrying that risk without knowing it.

For your board pack: name the person in your business who owns AI decisions, and give them the authority to say no. If you can't name them today, that's the appointment to make, and it doesn't need a new salary line, it needs a clear remit and a mandate.

For two years the data centre story in these briefs has been about demand, grid and the squeeze on trades. There's a quieter shift running in parallel, and it changes what clients will ask of the people who build and fit out these things. The Cyber Security and Resilience Bill, currently moving through Parliament with Royal Assent expected later in 2026, makes data centre services an "essential service" in their own right, inside a new data infrastructure subsector. Colocation sites above 1MW of IT load, and enterprise sites above 10MW, become Operators of Essential Services, with duties around cyber risk, operational continuity and incident reporting. Crucially, the Bill brings supply-chain suppliers into scope too.

That was reinforced by a cluster of government digital and energy policy updates published on 7 July, covering cyber resilience, the governance of the Digital Markets Unit and the Great Britain Corporate Power Purchase Agreement market. Taken together, they point at a more regulated, more closely scrutinised digital infrastructure economy. A data centre is no longer just a power-hungry, MEP-heavy building. It's a power-secured, cyber-sensitive, digitally regulated asset, and that reframes what a client expects from the contractor, the designer and the specialist subcontractor.

So what does it mean on the ground? If you build, fit out, commission or supply data centres, the due diligence you face is about to widen. Expect questions about your own cyber posture and, more awkwardly, your subcontractors', because the regulation follows the chain. The comparison only goes so far, but it rhymes with what CDM did years ago: a duty landed at the top and the evidence requirement rippled all the way down to the smallest supplier. The firms that treat this as an IT problem for someone else will find it's a procurement problem for them.

The procurement filter: if data centre work is in your pipeline, ask your commercial lead one question this week. When a client's next tender asks us to evidence supply-chain cyber controls, can we answer, and can our key subcontractors? If the answer is a shrug, that's the gap to close before the tender lands, not after.

On 13 July, Newforma introduced a Microsoft Teams connector for Newforma Konekt, the platform it markets as the golden thread of AEC information management. On the surface it's a tidy convenience: link a Teams channel to a project and the messages flow into the project record without anyone changing how they work. Underneath, it's doing more than convenience. It syncs public and shared Teams channels to their matching project, capturing messages, replies, reactions and attached files, and it keeps the version history. Edit a message and Konekt holds both the current and the prior version with timestamps. Delete a message in Teams and it stays in the Konekt record, clearly marked as deleted.

That last detail is the one worth pausing on. Most of the important decisions on a project aren't made in a formal RFI or a signed instruction, they're made in a chat thread, half in shorthand, and then they scatter when the project closes and people move on. Two years later a Building Safety Act matter or an adjudication turns up and the question is always the same, who knew what and when, and did anyone change the story afterwards. A record that preserves edits and retains deleted messages answers exactly that question. Newforma is an American AECO software house, not a UK compliance vendor, so this isn't pitched at the golden thread in the Building Safety Act sense. But it lands there anyway, because a durable, searchable, tamper-evident communication trail is precisely what the dutyholder regime asks you to be able to produce.

I'd temper the enthusiasm with one thing. Sweeping every message, reaction and deleted line into a permanent record is a data-governance decision, not just a features decision. It's a lot of information about your people and your project sitting in one searchable place, and on a dispute it cuts both ways. So the boring questions come first: which channels get captured, who can search the archive, where does it live, and how long do you keep it.

Worth doing: before switching a capture tool like this on, agree a simple retention and access policy with whoever owns your information management, so the audit trail helps you rather than becoming disclosure you didn't plan for.

For weeks this brief has carried the good-news side of data centres. Savills has vacancy down to 8%, Turner and Townsend rank them the UK's second most in-demand sector, Clearstone is putting £3bn into Ebbsfleet. Here's the counterweight, and it comes from an unlikely place: the Bank for International Settlements, the institution central banks themselves bank with. In its flagship annual report, given a fresh run by Bloomberg on 14 July, the BIS lays out the shape of the risk.

The five largest hyperscalers are on course to spend more than a trillion dollars on AI-related capex across 2025 and 2026. That spending is running ahead of what those firms earn and ahead of their free cash flow, so some are issuing debt to cover it. Worse, in the BIS's reading, is how it's all wired together. Hyperscalers take equity stakes in AI labs, the labs commit to multi-year purchases of chips and compute from those same hyperscalers, and the terms, in the BIS's own words, are typically poorly disclosed. It flags the risk of the same asset being pledged more than once, and reaches for canal mania in 1830s America and the dotcom bust for its comparisons, both of which ended in sharp corrections.

I'm not sure the doom scenario survives contact with how much real demand there is for compute, and the BIS's job is to worry for a living. But the direction matters for construction, because the thing being financed here is buildings. Sheds full of racks, on our land, built by our contractors. A lot of firms have quietly leaned their two-year forward book on the assumption the pipeline holds. The comparison only goes so far, but think of it like a single anchor tenant carrying a whole retail scheme. When the covenant's strong, everyone's happy not to ask questions. The moment it wobbles, you wish you'd diversified. The people this lands on first aren't the hyperscalers, they're the commercial director who committed labour and the subbie who geared up for one client.

The takeaway: if a single sector is now more than a fifth of your pipeline, treat it like client concentration risk, because that's what it is. Ask where the funding for each scheme actually sits before you commit long-lead resource to it.

The data centre pipeline is meant to keep UK contractors busy, and the caveats so far have been about power and planning. Here's a different kind of caveat, and it comes with a figure. Buildots' Intelligence Lab, the aggregated research hub it launched in late June drawing on anonymised data from projects worldwide, found a gap of 20 to 50% between the mechanical, electrical and plumbing output planned for a given week and what was actually delivered on data centre builds. That's the firm's own aggregated data, so read it as evidence from a vendor with a product to sell, not an independent audit. But the size of it is hard to wave away, and it lines up with what anyone who's run an M&E-heavy programme already suspects.

Think about where that bites. A data centre is, structurally, a shed wrapped around an enormous amount of building services. The value and the risk both live in the MEP. If a fifth to a half of the planned services work isn't landing week on week, the programme isn't slipping at the margins, it's slipping at the core, and it's doing it quietly because the superstructure looks fine from the road. That's the trap on these jobs: the bit that photographs well is on time, and the bit that decides your completion date is drifting. Buildots also pushed its progress tracking into the superstructure phase on 1 July, so it's now watching frame as well as fitout, which is the point, you can't manage a gap you can't see.

So the standing discipline for anyone resourcing against this pipeline is the same one as ever, just with sharper numbers behind it. Measure planned against delivered, at trade level, every week. Not overall percent-complete, which flatters everyone, but the specific question of whether this week's M&E actually happened. Catch a 30% delivery gap in the weekly data and it's a conversation. Catch it at handover and it's a claim, and it might just end up in the court we opened this brief with.

A practical step: track weekly MEP delivered-versus-planned as a named metric on every services-heavy job, and treat any sustained gap above 10% as a programme risk, not a reporting quirk.

xAI announced Grok 4.5 on 8 July 2026, pitched squarely at coding, agents and knowledge work. The headline detail isn't the benchmark score. It's that this is the first model co-trained with Cursor, the AI code editor a lot of developers now live in. Grok 4.5 went live the same day in Cursor on every plan and in the xAI console, though not in the EU, where availability was flagged for mid-July. On the numbers, xAI described it as roughly comparable to Opus-class performance but faster, around twice as token-efficient, at roughly two dollars per million input tokens. Every one of those figures is the vendor's own, so treat them as a starting claim, not a settled fact. I'm not sure the "comparable to Opus" line survives independent testing intact, but the direction, cheaper and faster at the same rough quality, is real and relentless.

Why does a coding model matter to a construction audience? Not because you'll run Grok from the site cabin. It matters because of the shape of the deal. A frontier lab and a tool vendor tuned a model to the tool's own data, and shipped it inside the product on day one. That is exactly how the construction software you already pay for is going to work. Your estimating tool, your document platform, your compliance assistant, they all sit on top of a frontier model, and that model can be swapped underneath you between releases. GPT-5.6 shipped a fortnight ago. Grok 4.5 this week. Gemini 3.5 Pro is lined up for a 17 July release with a reported two-million-token window, enough to hold a whole contract and its appendices in one prompt, though as of early July that date and those specs were informed leaks, not a signed launch. The churn is now weekly, and most of it happens without you being told.

So the useful move isn't to chase the model of the month. The durable questions are the dull ones. Which model is under the hood of the tool I'm buying? Where does my project data travel when I use it? And when the vendor swaps the engine, does my data handling change with it? The EU availability lag on Grok 4.5 is a small reminder that where a model can legally run still matters, and for anyone holding client data under a UK or EU obligation, that's not a footnote.

The discipline: ask any AI vendor to name the underlying model, its data residency, and their notice period for changing it. If they can't answer all three in writing, you haven't finished due diligence.

A dated refinement to the Gateway 2 story this brief has tracked for three weeks. The Building Safety Regulator's figures for the 12 weeks to 28 June, published in the second week of July, show 368 decisions at a 77% approval rate, with external remediation at 85%, comfortably above the 65% year-end target, and the legacy 2024 backlog down to 14 from 42 at the start of the year. So the front door is opening. The side door still sticks: higher-risk internal works sit at a 73% approval rate with a 28-week median determination time, and 1,505 applications remain live in the system, 57% of them in London. If you've been quoting a client eight to twelve weeks for an internal higher-risk decision, that conversation is better had now than at week twenty.

For your board pack: pull your live Gateway 2 submissions and re-baseline the programme against a 28-week determination for internal higher-risk works.

Two data points from this month tell the real state of play. ServiceTitan's 2026 Commercial Specialty Contractor report, a survey of more than 1,000 construction leaders, found 38% now report a measurable business impact from AI, up from 17% a year ago, with the work showing up in cost estimation (24%) and bid management (22%). A ConstructConnect survey of Canadian construction leaders, out on 7 July, found the enthusiasm just as high but day-to-day use still stuck in single digits. Both are vendor-adjacent and neither is UK gospel, so take the direction, not the decimal. The intent is near-universal; the doing is not, and the gap is widening fastest on the estimating and bid desk, which is where a labour market short of roughly half a million people squeezes hardest.

A practical step: time your next three tender returns end to end, this week, before you change a thing. If you can't say how long a bid takes you now, no tool will ever prove it helped.

Houzz's inaugural UK State of AI in Construction and Design report, drawn from over 145 professionals and covered on 14 July, found 46% already using AI and saving an average of more than three hours a week, which the report puts at over £23,000 in annual productivity gains per business. Vendor-reported figures, so hold them at arm's length. But the single biggest concern, cited by roughly a third of users, was reliability and accuracy. That's the whole tension in one survey: people are using these tools and getting time back, and they don't fully trust the output. The answer isn't more enthusiasm, it's the boring governance NG Bailey just put a name against, a person who owns the checking.

The practical bit: take the one AI task already happening in your business without permission, and give it a named owner and a verification step this week.