ISO 19650:2026: what the revision changes and who should care

Somebody has probably already forwarded you the slide. "ISO 19650:2026" across the top, a stock photograph of a glass atrium, and a bullet promising that everything you set up in 2019 will need redoing. Trade bodies circulate committee drafts. Appointments acquire a nervous appendix, usually drafted by someone who has not read the draft either. Meanwhile the published Parts 1 and 2 sit exactly where they have always sat, and eighteen months after what ISO 19650 actually requires of a common data environment, clause by clause they have still not turned into folklore. They remain the text a document controller can test a live container against this afternoon.

A systematic review of a series this age typically does not invent a new information-management religion. It restates, aligns and tightens the seams practitioners have been living in since 2018: level of information need written as a slogan; CDE states implemented as four folders; the asset information model treated as a rename of the project information model. Those seams are already in ISO 19650-1:2018, ISO 19650-2:2018 and ISO 19650-3:2020. A later cover does not create them. What it removes is the excuse that they were ever optional.

So if you are holding a committee text, the useful thing to do with it is map it against those three seams. If the map comes back empty, what you are watching is a year on a cover.

Reading a draft without rewriting your appointment

Take each seam in turn and write two columns. On the left, what the 2018 or 2020 part already requires on a live appointment. On the right, whether a revision of the ordinary kind — one that aligns vocabulary and closes a practice gap — would change that answer. Where the left column already fails, you do not have a 2026 problem. You have a 2018 gap wearing a later year.

What you should not do is rewrite a live information protocol from a rumour, or tell a supply chain that "ISO 19650:2026" is now the appointment. It happens. An information manager reissues a protocol on the strength of a conference slide, then spends a fortnight explaining to every task team why the reference has quietly gone back again. The published parts remain the parts you can be held to until a later edition is issued and the appointment cites it.

What a revision of this series typically does

Editions of a mature ISO series usually do four kinds of work, and usually refuse a fifth.

They align the early parts with the later ones. Part 4 (information exchange, 2022) and Part 5 (security-minded approach, 2020) arrived after Parts 1 and 2, and Part 3 (operational phase, 2020) has had years of thin implementation. A revision of the 2018 concepts document typically borrows the vocabulary those later parts already use, so the series stops sounding like two generations talking over each other.

They close a practice gap. The industry implemented Figure 10 as four folders, level of information need as LOD 300, and the AIM as a folder named As-built. Revisions of this type tend to convert the argument everyone is already having into clearer recommendations, and sometimes into shall language in the process part. They do not invent the argument. They arrive late to it.

They point at methods published after the first edition. Level of information need in Part 1:2018 is a definition and a purpose — stop delivering too much. The later European and international method for specifying that need (purpose, geometrical information, alphanumerical information, documentation) is a different text. A 19650 revision typically notices that method rather than reprinting it clause for clause.

They leave the cascade alone. Organisational, asset, project and exchange information requirements; the project information model and the asset information model; the four container states; the distinction between a CDE workflow and a CDE solution. Those objects are the series. An edition that discarded them would be a different standard carrying the same number.

They do not certify software, bless four folder names, or replace the appointment. If a slide in your inbox suggests otherwise, someone is selling a licence.

Information need

ISO 19650-1:2018 defines level of information need as the framework that determines the extent and granularity of information, and that definition exists to prevent the delivery of too much. Clause 11.2 puts the framework inside collaborative production: you agree what is enough before you produce it. Part 2 then carries the idea into the project's information standard, the exchange information requirements, the task information delivery plans, and the reviews that authorise an information model. A number written in a footer cannot do any of those jobs.

The industry writes LOD 300 anyway. That is a geometry slogan with three separate ancestries — level of detail, level of development, level of definition — and none of them is the 2018 concept. Level of information need: specifying enough, not everything works through the method. The revision-watch point is narrower.

A typical later edition of Part 1 or Part 2 tightens the join between the 2018 definition and the later specifying method. When published text does arrive, expect it to be less hospitable to a whole-model number and more insistent on purpose, objects, and the three content types — geometrical, alphanumerical, documentation — against a named milestone and a named actor. That is an expectation rather than a quotation, and it is no reason to wait. The 2018 definition already rules out "everything the authoring tool can store", so the rows can be written this week.

Who moves if that tightening lands: appointing parties whose EIRs still say LOD; information managers who waved through a BEP that copied the slogan; estimators who have never once been handed a row they could put a price against. Who is already fine: anyone whose exchanges name purpose, objects, fields, documents, and what is deliberately out of scope. A cover year does not reward them — it removes the cover from everybody else.

CDE states

Part 1 draws four states as a workflow: work in progress, shared, published, archive. A state is a claim attached to an information container: who may write to it, what everyone else is authorised to do with it, which gate moved it here. Status, or suitability, is a different attribute — what the container may be used for. Revision is a third. Vendor pages collapse all three into a single badge, and the four CDE states refuses that collapse at length.

Part 2 requires the project CDE to enable unique identification, metadata that travels with the container, recorded transitions, and access control on the container rather than only on its directory. Which is why a hard-copy cabinet with a genuinely managed process around it can still be a CDE solution, while a platform with four named folders and no gates is not.

A typical revision of the 2018 CDE chapter does three things practitioners already argue about among themselves. It restates, more firmly, that states are workflow claims recorded as metadata rather than a directory tree. It takes the distributed CDE — the appointing party's project environment alongside a delivery team's connected environment — out of the note most mobilisations skip and into a sentence a test can fail. And it treats the information container as a persistent, retrievable set that might be a model, a table or a subset of a file, not only a PDF. Part 4 already pushed the series towards the exchange of structured information, so a later Part 1 that noticed the push would be alignment, not a surprise.

The four-folder implementation remains the industry's favourite misreading, and a revision that restates Figure 10 in harder language hands nobody a new CDE. Ask a document controller whether a live container can occupy two states at once without anyone noticing: a WIP copy on somebody's laptop, a Shared copy in the portal, a plotted sheet being treated as Published. Where the answer is yes, the 2018 text was already being failed, and the cover year is decoration.

The AIM

Part 1 names two information models. The project information model is what delivery produces and uses; the asset information model is what the operator is supposed to be able to query. Organisational information requirements sit with the asset-owning organisation. Asset information requirements say what the AIM must contain. Project and exchange information requirements say what the PIM must produce so the AIM can be populated. Part 3:2020 is the operational-phase process for that second model, and national annex status codes already distinguish authorisation of the PIM from authorisation of the AIM. Two codes exist because there are two authorisations.

Practice treats the cut as a rename: a federation labelled as-built, a CDE archive published to the client, a COBie sheet assembled in the last month. From PIM to AIM: the transition nobody owns describes that vacuum from the site end. Seen from the series it looks identical.

A typical later edition of the early parts, written after years of Part 3 being treated as optional, tightens the cut. Expect firmer language: that an AIM is accepted against asset information requirements rather than inferred from a folder name; that OIR and AIR are organisational documents outliving any single appointment; that authorising the AIM is a distinct act from authorising the PIM. Nothing in that list would be a new object. Those sentences already sit in Part 1 clause 5 and in Part 3 — said, in a later edition, in a way a project close-out could not dodge by reprinting the delivery model.

Who moves: asset owners who never wrote an AIR a supply chain could price; appointing parties who assumed "BIM" had bought them an AIM; information managers who scheduled the cut for some time after practical completion. Who is already fine: anyone who named a person to accept the AIM before the delivery team demobilised, and who can show five maintainable assets with identity, as-installed fact, a bound certificate, and a reject rule. That labour was in the 2020 part all along.

Who should care

The honest test is not how you feel about a committee. It is whether your current appointment would still stand if a later edition only restated the 2018 and 2020 seams more firmly.

Appointing parties should care first. They write organisational, project and exchange information requirements, and they accept information models. A revision that tightens information need, or the AIM cut, lands on their templates, their information protocol, and the rows a tenderer can price. If those templates are still slogans, the later year is a deadline that has already gone past.

Information managers should care if they mapped 2018 onto a vendor default: four folders, a status list copied out of an old National Annex PDF, a BEP that says "LOD 300" and "AIM at handover". Mobilisation against Part 2 is the test, not the splash screen. A later edition that is more explicit about metadata on the container, and about a distributed CDE, will fail the same mobilisation that was skipped the first time.

Document controllers should care about states and suitability, and not at all about the year on the cover. If they can already name, for a live container, the state, the writer, the authorised use and the gate that moved it here, a revision of Figure 10 is a clearer poster for the wall. If they cannot, they are the people who will be asked to invent the journal after the fact, usually on a Friday afternoon.

Lead appointed parties should care wherever the BEP, MIDP and TIDP still recite 2018 without implementing it. Cascading an EIR that cannot be answered does not get easier because the series acquired a new year.

Asset owners and facilities teams should care about the AIM seam and more or less nothing else. Part 3 has been published since 2020. A revision of Parts 1 and 2 that notices it is less a new duty than a reminder that operations inherit a vacuum whenever nobody is paid for the cut.

Who can wait. Task-team authors whose appointments have not changed. Trade contractors who deliver against a named container, a named suitability and a reject rule — they are already doing the labour a tighter edition would ask for. Site gangs, who were never appointed to interpret a standard in the first place.

Who should not panic. Anyone whose Monday tests against the published parts already pass. An edition that aligns vocabulary and withdraws excuses does not make a managed process obsolete; it makes an unmanaged one harder to dress up as compliance.

1. 01
   Appointing parties

   What to watchInformation-need and AIM-cut wording in templates

   Already fine if...EIRs already name purposeobjectsfieldsdocuments
2. 02
   Information managers

   What to watchWhether mobilisation matches Part 2not a vendor default

   Already fine if...Statessuitability and a distributed CDE were already tested
3. 03
   Document controllers

   What to watchStatewriterauthorised usegate — not the cover year

   Already fine if...All four can already be named for a live container
4. 04
   Lead appointed parties

   What to watchWhether the BEPMIDP and TIDP implement 2018or just recite it

   Already fine if...Already cascading an EIR that can be answered
5. 05
   Asset owners and facilities teams

   What to watchThe AIM seam (Part 3)

   Already fine if...An AIM acceptor was already named before demobilisation
6. 06
   Who can wait

   What to watchTask-team authorstrade contractors delivering to a named container

   Already fine if...Already doing the labour a tighter edition would ask for

A Monday watch against the published parts

You do not need a draft in your hand. One live appointment and one live container will do.

1. Find every sentence in the EIR, BEP or information standard that says LOD, "level of detail", or "level of development". If there is no purpose, object, field, document and out-of-scope line beside it, the information-need seam is already open.
2. Name the current state of one container site might build from. Name who may write to it. Name the use it authorises, in a sentence a site engineer could refuse work on. Name the gate, the role and the date. If any of those answers turns out to be a folder title, the CDE-state seam is already open.
3. Name the person — a role on the appointment, not a platform — who will accept the AIM. Produce the asset information requirements that say what "accepted" means. Pick five assets and ask whether the current PIM can feed identity, as-installed fact, and a bound certificate. If the name is "document control, at the end", the AIM seam is already open.
4. Then ask the only revision question that matters. If a later edition merely restated those three published requirements more firmly, would any of the answers above change? A no means you are watching a cover year. A yes means the gap is older than the rumour.

Five assets and one container are enough to run that. You are not trying to forecast a committee. You are asking whether 2018 and 2020 have made it into the appointment yet.

What can and cannot be said yet

None of the above reads unpublished wording, and there are no clause numbers, quotes or official circulars here for a 2026 edition. A committee text in circulation does not carry the force of an issued standard, however confidently it gets summarised on a slide. Until a later edition is published and an appointment cites it, the texts you can be held to are the ones with 2018 and 2020 on the cover — plus, where they apply, Part 4:2022 and Part 5:2020.

Appointing parties should act on the published parts now, rather than holding a diary reminder for a year that has not been issued. Replace every LOD slogan in the current EIR and BEP with a purpose, object, field, document and out-of-scope line. Name, on the appointment, the person who will accept the AIM, and produce the asset information requirements that say what accepted means — then fund that acceptance before the delivery team demobilises. That costs template time and a fee line; it does not cost a new standard.

A later edition, if and when one is published and an appointment cites it, may make those seams harder to dodge. Until then a cover year is not a duty. The 2018 and 2020 texts are. If the Monday watch above already fails, waiting for a committee will not close it.

Read next: What ISO 19650 actually requires of a common data environment, clause by clause, The four CDE states: work in progress, shared, published, archive, Beyond the CDE: why the next standard is about data, not documents.