What moved
Somebody will have forwarded you the chart by now. Published labour-exposure work, taken as a class, puts construction near the bottom on observed AI coverage, with office and knowledge occupations higher up the same tables. That ranking is what travelled in May: not a paper anyone had read end to end, but a bar at the reassuring end of a chart with a sentence underneath about the trades being safe.
The method behind that class of research is not mysterious. An occupation is scored by how much of its task mix overlaps with what current systems can already do, or by how much AI is already observed in the work. Desk labour that lives in text, spreadsheets, mail and software scores high. Site labour that lives in weather, plant, craft skill and a particular piece of ground scores low. Construction, as a sector average, is dragged down by the physical majority — which is an accurate description of the work, and neither a rounding error nor a shield.
Consolation is the sentence that survives into the slides. The ranking is low, so the industry is safe; the trades will not be replaced; the cabin can wait, information discipline can wait, the unofficial pack can wait. That is weather. The interesting part is what the ranking cannot see.
A sector average hides the people whose working day already looks like the occupations at the top of the same tables. The document controller. The planner. The quantity surveyor. The permit coordinator. The person who recuts one issued set three times because each client's portal wants it differently. Their week is text, revision, suitability and a pack that has to be accepted. It is not a trowel. And if that slice is badly instrumented — a folder with a login, a USB in a glove box, a cabin PDF that is current because someone said so on a call — then an exposure method has almost nothing to observe in the first place. Low coverage, in that case, is a measurement fact about the record rather than a safety fact about the job.
Being near the bottom is therefore a position rather than a verdict. It means the published container is still, on ordinary evidence, a clerical object rather than something anyone retrieves from under pressure. It means there is still time to name the issued set — these sheets, this model state, this revision, released to these companies, for this purpose — before an assistant gets parked on a shared drive and called intelligence. Office work has already bought that pattern, mostly by accident. Construction has not, at scale, because the working set is still a plotter and a bag. That is not immunity. It is unused room.
When coverage does arrive on a live job it will not arrive as a machine that lays brick. It will arrive as checking, as register reconciliation, as a summary of last week's model, as a permit pack that claims to have been read. Clerical labour does not evaporate because a ranking put the sector low; it stays with whoever is asked, at four on a Friday, to prove which revision is current. A job that treats the ranking as a holiday from that proof meets the same accident as everyone else — a superseded sheet in a gloved hand, because the office believed it had time.
What changed in the standards
Not much, and I would rather say so than dress it up.
No new part of ISO 19650 was published in May. The UK BIM Framework did not issue a replacement text to mark a research week. The Building Safety Act's information duties in England were already in force long before the chart circulated; they did not arrive because a table put construction near the bottom, and they are not satisfied by an exposure score. They still want an identifiable, available, change-controlled golden thread, which is a harder thing to produce than a reassuring paragraph in a deck.
The texts your appointments point at are the parts already in circulation. They were not new this month. They were in use this month, which is a different fact and the one that actually governs your week. They assume an appointing party, appointed parties, an information protocol, a common data environment and a published container. None of them assumes that a sector ranking retires the need to name those objects, and none of them treats physical work as a reason to leave the issued set unpublished.
The revision conversation around the series stayed where it was, which is to say a drafting conversation. Working groups collect comments, national bodies circulate drafts internally, and none of that yields a clause identifier you can print on a cover sheet or a commencement date you can write into an appointment. Live appointments claiming alignment to the series were still aligning, if they aligned to anything, to the published language. Reading a research week as though the parts had moved is how a job buys a slogan and keeps the same register.
Gulf permit-stage BIM requirements remain a class of public-record rule where the text is public, and I have no May circular worth citing. If an annex, a circular or a draft ballot did move, it will turn up in a later issue with a reference attached.
A pitch that says the job is ready for AI because construction is less exposed is not a clause. Openness that a party outside the authoring licence cannot actually exercise is still a brochure word, whatever the ranking said about the sector. Labour-exposure season is not a standards event.
One thing to watch
Two weeks after a ranking has done the rounds, the question is not whether anyone can still recite it. Everybody can. The question is whether the document controllers are still arguing about what it named — immunity for the trades, or a record so unpublished that a study method could barely see it. If that argument is still running, it belongs in the ordinary weekly business of the job, next to the unofficial site pack and the ordinary accident of building from a superseded sheet, rather than in a research recap.
Watch the sentence that treats physical work as a shield. Physical work is the majority of this sector and nobody sensible is dismissing it. It is still not a reason to leave the information slice unnamed. If the only answer a job has to a question about exposure is that the ranking said no, then the question actually asked — can you produce the issued set — has gone unanswered.
Watch the word safe, too, because it is doing two jobs at once. If it means the published container can be extracted, named and handed to every appointed party, good: say so and move on. If it means nobody has bought an assistant yet because a table put the sector low, that is a rumour rather than a control. And if three applications now claim to be the environment while the plotter still takes a USB, May moved nothing that matters here.
So, two things worth doing before June. Ask your appointing party, in writing, which container is the published one on this appointment and who can extract it without holding the authoring licence. And when the next briefing arrives with a chart in it, ask whoever brought it whether the path they are proposing changes a published state or only a slide. Neither request costs anything. Both tend to produce a long pause, which is usually the most informative part of the meeting.