The keys move on a Friday afternoon. By Tuesday the facilities team has its first real call-out — a circulating pump tripping on a system nobody in the room can name with any confidence — and three places to look. There is a crate of manuals in the cupboard, a viewer licence with a fortnight left on it, and a spreadsheet somebody in delivery started and did not finish. None of the three will say which unit the plate reading AHU-03 belongs to, where that pump is isolated, or which revision of the sheet the isolation was drawn on. So the supervisor does the sensible thing, which is also the expensive one: she walks the plant rooms with a phone, photographs nameplates, and starts typing a register of her own.
Practical completion had been certified. The building was occupied, warm and full of people. It was not operable, and the next twelve months went on buying information the owner had already paid for.
Occupied and operable are different facts
A taking-over certificate closes a contract. It cannot isolate a circuit, tell a stranger which damper sits in which system, or say which revision of the roof plan a residual hazard still cites. Practical completion, a certificate, a folder named Handover — those are dates delivery can put in a slide, and they are genuinely useful dates. They are simply not the same thing as the first operations hour: the first permit, the first call-out, the first tenant who wants the plant room opened.
Operable means a query returns an answer that someone who never walked the job can defend to a safety officer. Occupied means people have keys. Most projects buy the second and then perform the first as a reconstruction, usually with facilities labour that was budgeted for maintaining plant rather than for archaeology.
The version of the future worth wanting here does not invent a fifth information model or a new standard. It is the year in which the last delivery morning and the first operations morning are looking at the same record.
Five things that already exist, joined up
What it takes is not more files. Five objects are already on almost every job; the work is joining them so that a later person does not have to invent a sixth.
The asset information model is the destination — the model operations will query, not the project information model renamed “as-built” in the final fortnight. The asset tag is the join key that query will use. The O and M is the document bound to that tag: the valve, the warranty, the spare, rather than four hundred PDFs in a crate. Isolation is the first-day safety sentence — where you make the thing safe, on which issued sheet, with which residual still live. The issued set is the construction and as-installed information that sentence is allowed to cite, with its title, revision, suitability and recipients intact.
Miss one and the other four become theatre. A handsome AIM with no tags is a type library. Tags with no O and M are nameplates. O and M with no isolation is a brochure you cannot work from. Isolation drawn on a superseded sheet is a hazard wearing a reference. An issued set that never met the AIM is a delivery cupboard the operator will quietly ignore.
The asset information model is a destination, not a rename
ISO 19650 already splits the world in two. The project information model is what delivery produces and uses. The asset information model is what the operator is supposed to query: what is installed, where, how to maintain it, what residual risk remains, which warranty still runs. The industry talks as though one becomes the other at practical completion, by way of a folder called Final, a status letter and a spreadsheet at the end. The conversion is a separate piece of work, and on most jobs it has no natural owner.
A day-one-operable building is one whose AIM was accepted as the AIM before the delivery team demobilised. Acceptance is not storage. It is a named role on an appointment, with the standing to reject a row that will not answer an operator's question. If that role amounts to “document control, in the last fortnight,” the cut is unowned, and operations will invent a third model to fill the gap: a spreadsheet typed from nameplates, a maintenance system rebuilt from portable documents, a consultant paid to structure what was already bought. That third model is the twelve-month wait. It is also where the tidy phrase about lifecycle data being “lost at handover” actually leads — not into a skip, but into a vacuum nobody was appointed to fill.
A federation labelled as-built is still a project information model. A common data environment archive published to the client is still a delivery cupboard. A COBie sheet in a pack becomes an AIM only when the rows answer the operator and somebody with authority accepted them as the model. Two authorisations exist in the UK national annex because there are two models; treat the codes as a filing step and you perform the transition on the label while skipping it in the content. The cut nobody owns is the hinge, and day-one operability is what that hinge looks like once it has an owner.
Tags operations will still search
An asset tag is a stable identity for a thing operations will maintain. It survives a drawing revision and a room rename. A gridline is not one. Nor is a package reference, a modelling identifier nobody printed on the casing, or a filename inside a quality ZIP. The tag is the key the later query uses: this unit, this door, this board.
Without it the AIM holds types and no instances. The O and M has a chapter called “air-handling units” and no route to AHU-03. Isolation has a schematic and no plate on the valve. Inspection and test plans that never met a tag can still close a package — they cannot close an asset. Commissioning that never wrote the serial against the tag leaves you with a nameplate album. Quality records that stay in a trade folder never become handover records, however diligent the folder looks.
“We will map them at completion” is not day one. Day one, for the tag, is the first inspection, factory test or delivery check a later owner might ask about. A provisional tag you confirm later is still a tag; an anonymous signed sheet you guess at later is not. The ITP bound to the tag is how the register assembles itself while the installer is still standing next to the plant, and the occupation morning only looks operable if that binding already happened.
O and M that answers a question
Four hundred O and M PDFs are a pile. What a maintainer needs is a search that returns the valve, the warranty and the spare. The pile photographs well in a completion report and fails at the first call-out, because nobody with a leak in front of them is going to read two hundred pages looking for one torque figure.
On a day-one-operable building the manual stops being a parallel universe running alongside the model and becomes a set of documents bound to the tag: data sheet, certificate, as-left values, spare-part identity, warranty start. Completeness becomes countable, which is the useful part. You can list the tags with no manual, the manuals with no tag, and the warranties with no start date because nobody wrote the serial down when the crate was opened. Those three lists are your readiness review, and none of them requires a fresh compilation exercise.
A late ZIP named 08_O+M_FINAL fails in the same way as a health and safety file compiled as a binder — a snapshot assembled for the certificate, already out of date for the stranger who will use it. The facts were free while the room was still full of the people who knew them. After the room empties, every field becomes an inference, and enough inference turns an O and M into fiction.
Isolation you can find before the first permit
The first operations question is almost never “show me the federated model.” It is “how do I make this safe?” Isolation points — valves, boards, fire-alarm interfaces, the lock-out that still matters long after the temporary works have gone — are not a nice-to-have field on a COBie row. They decide whether you have a building people can work on or a building people can only sit in.
An AIM worth the name can answer, for a named asset or a named space: where the isolation is, which issued title and revision it sits on, what residual hazard remains if you get it wrong, and how you physically reach it. If getting those answers means opening a crate and guessing at a filename, occupation has happened and operability has not. A contents page is not a query. A schematic taped up in a van is an unofficial set with a following.
Isolation that cites “see drawing 123” with no revision is worse than a gap, because it looks like an answer. The sheet moved; the residual did not. The health and safety file is the safety-shaped version of the same test: residual hazards, significant services, access, the structural principles nobody may cut through, each tied to a current identity. CDM assumes that pointer already exists and says nothing about manufacturing it from unmarked sheets after demobilisation. The golden thread, where it applies, assumes isolation points as basic building information. A pack that cannot produce them on the morning of occupation has not met the duty, whatever the spine of the folder says. You cannot claim day-one operability if the first permit has to invent the isolation schedule.
The issued set operations inherit
An issued set is a dated, named, transmittable package: the drawings, models and specifications released for a stated purpose, to named organisations, at a stated revision and suitability. It is not a feeling, and it is not whatever happens to be in the latest folder. Delivery knows this sentence, or performs knowing it. Operations inherit either the last true version of it or the unofficial one.
On a day-one-operable building the operator's copy, the contract administrator's folder and the sheet stuck to the plant-room wall agree with each other. Where a print is on a wall, it matches the published container; where it does not, that is a recorded exception rather than a way of life. As-installed information updates the AIM instead of starting a second register in a facilities spreadsheet. A superseded roof plan gets withdrawn from the live file. Residual risk that no longer exists leaves the live AIM the way a superseded sheet leaves the live issued set — still findable as history, no longer readable as instruction.
ISO 19650's published state exists because “the file is on the disk” and “the file is authorised for this use” are two different facts. Day-one operability adds no fifth state to the four already written down. It just refuses to let operations float free of them. Ownership of the record is not a question of who bought the common data environment; it is a question of who can still say, months later, which container was released, to which organisation, at which suitability, and which tag that container describes.
The same failure, seen from the other end
Owners wait a year for data they already paid for because the facts were never written against tags and issued containers while the installer was on site. The circulated “thirty per cent” of effort lost at handover stays a slogan until you name the moments it hides: commissioning not captured, warranties left in email, COBie applied once at the end, no owner for the AIM. Progressive handover means validating asset data while the trade is still in the building, and end-loaded COBie is how an owner buys the same building twice. Asset information requirements written so a supply chain can price them are how the cut acquires an invoice rather than a hope. A register earns its keep by holding the fields facilities have actually queried, not the metadata delivery found convenient. Warranties and spares stay boring right up until the first failure outside cover. An owner's acceptance checklist is only leverage if it can demand the issued set, the register and the residual risks on the same day, from the same source.
Those are not eleven separate problems with eleven separate fixes. They are one informational claim, restated until somebody closes it.
The Tuesday that would actually matter
The industry's favourite picture of the operable future is a feed. Sensors, a twin that updates, a chatbot that “knows the building.” It is an attractive picture and it dodges the question, because it never establishes whether the facilities supervisor, the permit issuer and the residual-risk sentence are looking at the same issued fact.
The Tuesday worth aiming at is duller and stricter than that. Three people who do not share an employer can name, for the plant in front of them, the tag, the isolation, the O and M row, the warranty that still runs, and the issued revision those sentences sit on. The site copy, if one still exists, agrees with the operator's AIM. A completion walk, if there is one, produces a residual list rather than the opening entry of the asset register.
If you are the appointing party, that Tuesday is bought at three points and nowhere else. At tender, ask for asset information requirements as a priced activity with a named acceptor, and say plainly which fields facilities will query — vagueness here is what you pay for later at consultancy rates. In the appointment, keep the acceptance role alive past demobilisation, tie a slice of retention to accepted rows rather than to a delivered ZIP, and require the tag to appear on inspection records from the first delivery check rather than at completion. Third, and much the cheapest of the three, put the people who will run the building in the room from the week commissioning starts, with a written right to reject a row they cannot use.
None of that is free. The acceptance role is a fee, the tag discipline slows the first month of installation records while installers learn it, and holding retention against data will be argued at every valuation. Set against a facilities team spending its first year retyping nameplates into a spreadsheet, it is the smaller invoice, and it is the only one that arrives while the people who know the answers are still on site.
Read next: Monthly Register — April 2026, A handover acceptance checklist for owners, The twelve-month handover: why owners wait a year for data they already paid for.
