Subscribe
Share
Share
Embed
Reimagining AEC is Dr Sepehr Abrishami's independent podcast on how the built environment handles information. Each episode takes one subject from digital construction, from the ISO 19650 series and the common data environment to the golden thread, soft landings and the arrival of artificial intelligence in design and surveying practice, and tells it as a single argument built on the published standards, guidance and evidence. Sepehr is a researcher and former architect who leads the Centre for AI in Construction and teaches building information modelling at master's level. The episodes are written for postgraduate students and practitioners who want the reasoning behind the rules, not a summary of them.
An industry that wants to know how digital it has become can inspect its projects or ask its people. Since who asks matters as much as who answers, I should say that I am Dr Sepehr Abrishami, a researcher with more than thirty peer-reviewed journal articles in my field. The podcast is Reimagining AEC, which I run independently, and this episode reads a survey that took the second route, the NBS Digital Construction Report 2025. NBS, a specification platform that is now part of Hubexo, ran the survey in 2025.
Of the professionals who answered, 72.3 per cent say they have adopted building information modelling, or BIM. A figure like that is soon quoted in bids, dissertations and board papers. Suppose a client's board paper turns it into one sentence. Almost three quarters of the UK industry now uses BIM, so we can require it of every bidder. I wrote that sentence myself as a teaching exercise, and no client has put it to a board. Still, a board asked to approve it would be making a procurement decision on the strength of one survey figure.
So the question this episode answers is what a survey of this kind can carry as evidence, and what it cannot. The aim is that you can quote any figure from the report with its question, its sample and its limits attached. It should also let you interpret its findings and rewrite that board paper sentence so that the evidence supports it. I start with who answered. Then I move through BIM and the information inside models, through cloud working and artificial intelligence, and through environmental measurement and the industry's view of itself, before returning to the sentence.
Who answered comes first, because every percentage in the report is a share of a particular group of people answering a particular question. NBS distributed the survey electronically, under the code of conduct of the Market Research Society. It received 559 responses, and only 358 of them were complete. A further 114 were partial and 87 were disqualified. So 559 counts the people who started, not the people whose answers lie behind every result. The report never prints the base for each question, meaning the number of people who answered it. Its appendix is uneven too. It says that 559 professionals completed the survey, yet lists 358 complete responses, and its shares by location and by role imply different bases.
Two thirds of respondents, 66.5 per cent or 315 people, are consultants, designers or specifiers. Contractors make up only 12 per cent, and clients fewer than 7 per cent. The report itself warns that its perspectives may be weighted more heavily towards design professionals. And 27.7 per cent, more than a quarter, are based outside the UK, with Europe the largest group. So the phrase the UK industry, in our board paper, already reaches beyond the evidence.
Those facts point to three limits. The first concerns who answered. People chose to take part, which makes this a self-selected sample, and those engaged with digital practice are likely to be over-represented. The report describes no weighting to correct for that, so its results describe its respondents rather than the industry. The second concerns coverage. More than a quarter of the sample is outside the UK, no result carries its base, and at least one, on classification, sets aside those who were not sure. The third concerns what the answers are. A yes to a question about BIM records a respondent's own view, not an audit of projects. And NBS is not a neutral observer. The UK guidance on concepts, published in 2019 by the UK BIM Alliance with the Centre for Digital Built Britain and the British Standards Institution, records that NBS authors Uniclass 2015. That is the UK classification system. The report also carries advertisements for NBS specification and product platforms. None of this makes the survey useless. Counting its predecessor, the NBS National BIM Report, the series spans fifteen years, which gives it real value for showing direction. The headline figure on BIM is the first to test against these limits.
The BIM question opened with the words based on your previous answer. Respondents had just described what BIM meant to them, so each one judged adoption against a personal definition. The answer was 72.3 per cent, against 70 per cent in 2023 and 71 per cent in 2018. For seven years the figure has barely moved, and those with no plans to adopt fell from 12 to 8.4 per cent. NBS reads the flat line as BIM finding its natural market penetration level. A careful reader adds a caution. A flat line in a self-selected panel led by designers may show saturation among the kind of people who answer this survey, rather than across the industry. Now return to the board paper. Three quarters of respondents, most of them designers, is not three quarters of the contractors who would bid. Nor does the figure say whether those who have adopted BIM work to the ISO 19650 series, the international standards for managing information using BIM.
The report comes closest to that second question through how respondents define BIM. Describing it as 3D models with data fell from 26 to 14.6 per cent. Describing it as following the BS EN ISO 19650 standards rose from 22 to 30 per cent, now the most common answer. NBS calls this a maturing understanding of BIM. Two cautions apply. The 2023 shares add up to 108 per cent, while the 2025 shares add up to 100, so the question or its options probably changed, and the comparison is only indicative. And the standards themselves have not moved. BS EN ISO 19650-1 and BS EN ISO 19650-2 remain in force in their 2018 editions, with the UK National Annex of 2021, which adapts Part 2 for the UK. Revised drafts of Parts 1 and 2 went out for UK public comment, as nima, the industry body formerly called the UK BIM Alliance, announced in March 2026. No new edition has been published. For the board paper, only 30 per cent of respondents tie BIM to ISO 19650. Using BIM is also not the same as meeting a client's exchange information requirements, the requirements a client sets for the information each appointed party must deliver. Adoption among these respondents has levelled, then, while their understanding has moved towards the standard.
The next part looks inside the model, beginning with manufacturers' digital objects, the modelled versions of real products that designers place in their models. Use runs ahead of supply. Among consultants, 88.3 per cent use these objects regularly or occasionally. Among suppliers, 73.4 per cent provide them for at least some products, and NBS infers growing pressure on those who do not. It also claims a jump of more than 20 percentage points in four years. That claim compares 73.4 per cent providing objects for some or all products in 2025 with 56 per cent providing them for most or all products in 2023, and 52 per cent in 2021. Those are different categories, and the 2025 survey offers no most or all option. Its nearest figure, 22.4 per cent for all products, is narrower still, so the published numbers cannot show the size of the change. The appendix also counts only 65 suppliers in the whole sample. And an object is useful only when it carries the information a requirement asks for. That is why level of information need matters, the method for defining the quality, quantity and granularity of information required, now set out in ISO 7817-1 of 2024.
Classification matters because of the common data environment, or CDE, the agreed source of information through which a project collects, manages and shares its information containers. An information container is any named, persistent set of information, such as a model, a drawing or a schedule. ISO 19650-2, in clause 5.1.7, requires each container in the CDE to carry classification metadata, and ISO 19650-1 recommends that classification follows the framework in ISO 12006-2 of 2015. The UK BIM Framework's guidance on the CDE, written by John Ford and published in 2020, names Uniclass 2015 as the preferred UK system and notes that the UK National Annex references it. So the rise of Uniclass, to 58.9 per cent from 46 per cent in 2023, fits the standards. The Common Arrangement of Work Sections, the older system, fell from 33 to 14.6 per cent, although NBS notes that many practices still use it for specification. A project may therefore use one system in its model and another in its specification. Three cautions follow. Just under a fifth of respondents are not sure which system their organisation uses, which is a gap in governance. The 2025 shares exclude those answers, and the report does not say whether the 2023 figure did. And NBS, as I said, authors Uniclass. If our client's requirements named Uniclass, and bidders resembled these respondents, almost one in five could not say which system their own organisation uses.
Behind the objects sits product data. A product information management system, or PIM, keeps a manufacturer's product data in one governed place. Among suppliers, 28.6 per cent use one, little changed since 2023, and 18.4 per cent do not know whether their organisation does. Then there is the digital product passport, a machine-readable record of a construction product's information across its life cycle. The European Union's new Construction Products Regulation, EU Regulation 2024 number 3110, entered into force on 7 January 2025. The European Commission gives 8 January 2026 as the date from which its main part applies. Fewer than half of respondents, 46.1 per cent, were aware of the regulation and the passports, and of those, 83 per cent support adoption in the UK. The report's conclusion later calls this 83 per cent backing from UK respondents, yet nothing in the report limits the question to respondents in the UK. It is support within an informed minority. A Director of BIM Services, writing as an information manager, put the condition plainly. In that respondent's words, the regulations will only impact if clients believe there is a benefit. Such product data ends up in the asset information model, the owner's record of an asset in use, which the episode on information management in the operational phase takes further.
The tools of daily work come next, beginning with cloud computing. Cloud use rose from 80 to 86.3 per cent in two years. The report projects use above 95 per cent by 2027, but that rests on stated intentions, and even if every one held, use would reach 94.5 per cent. The sharpest change is in using the cloud for digital signatures, contracts and forms of appointment, which almost doubled, from 32 to 61.7 per cent. The report gives collaboration an 11 per cent increase without saying whether it means percentage points, so no earlier value for collaboration can be stated.
Storing files in the cloud is not the same as running a common data environment. Under ISO 19650, the CDE, that agreed source of information, is a workflow as well as a technology. Each information container moves through four states in turn. It starts in work in progress, where its authors develop it. It passes to shared once it has been checked and reviewed, so that other teams can coordinate with it. It moves to published once it has been authorised for use and, where the client receives it, accepted. Finally it goes to archived, which keeps a record of every version. The order matters, because each move is a point where someone answers for the information. The survey asks about technology, not workflow, so it cannot say how many respondents run a compliant CDE. A board paper that treated bidders' cloud platforms as proof of a working CDE would claim more than the survey supports. The episode on the common data environment follows that workflow in detail.
Artificial intelligence, or AI, is the fastest moving finding in the report. In 2020, 9 per cent of respondents said their organisation used AI. By 2023 it was 22 per cent, and in 2025 it is 42.5 per cent, with a further 37.7 per cent planning to start within five years. Those with no plans fell from 43 per cent to 8.4 per cent. Now read the question. It asks whether, within your organisation, you use artificial intelligence and machine learning. One colleague using a chatbot can turn that answer into a yes, so the figure measures presence, not depth, quality or governance. The direction has support elsewhere. A contribution to the report from the Royal Institute of British Architects, or RIBA, cites the RIBA AI Report 2025, which RIBA bases on a survey of its members. In that survey, 59 per cent of respondents used AI on at least the occasional project, up from 41 per cent in 2024. RIBA's report page does not give the sample size. Agreement between independent samples strengthens a claim about direction.
What respondents use AI for follows one principle, which is that language and information tasks lead. Searching for technical information comes first, at 71.1 per cent, followed by drafting or reviewing text at 63.8 per cent. Design generation comes last among the tasks named, at 40.8 per cent. Current users rely mainly on general-purpose assistants, led by ChatGPT at 86.8 per cent, and the 21.7 per cent using company-specific tools suggests that some organisations are building their own. In these tasks a fluent but wrong answer can pass unnoticed, and respondents raised that risk themselves. A technical product specialist warned that large language models generate inaccuracy in very credible and believable ways. A BIM professional said AI must be used with a human checker for sign off. That fits ISO 19650-2. Whatever produced a piece of information, it is still checked and reviewed before it leaves work in progress for the shared state. For the board paper, the useful question about a bidder is whether that checking happens, and no adoption figure can show it.
Attitudes to AI follow one principle, optimism about the industry with a split over jobs. Most respondents expect benefits, with 85.2 per cent agreeing that AI will have a positive impact. On jobs, a majority, 56.2 per cent, expect AI to reduce staff numbers, yet only 34.6 per cent think it threatens professions. Respondents seem to expect fewer people doing the same professional jobs. Read the source with care here too. For the statement on positive impact the report also gives a second value, 53.1 per cent, without saying what it counts. The values I have used come from the report's text, where their meaning is stated, so use the number whose meaning you can defend. These are also forecasts by people who chose to answer a digital survey.
Environmental measurement and the industry's view of itself make up the last part of the evidence. The environmental question asks whether, in the past twelve months, respondents used digital technologies to calculate environmental measures for their projects or products. The rises are among the steepest in the report. Energy and water demand went from 38 to 64.2 per cent in two years, embodied carbon from 40 to 60.3 per cent, and waste from 18 to 45.8 per cent. In 2025, 90.8 per cent agree that digital technologies are creating better buildings, and 84 to 86 per cent that they improve environmental impact and safety. The report gives those last two figures the opposite way round in two places. That is one more reason to check where a number comes from before quoting it. Now test the question. It asks whether a tool was used at least once in twelve months, not how often, or whether the result changed a design decision. And asking a digitally engaged audience whether digital technology helps the environment invites agreement. The data shows growing activity, not proven environmental outcomes. Had our board paper added that bidders using BIM would lower the project's carbon, this survey could not carry that claim either.
The industry's view of itself combines confidence and anxiety. A slim majority, 51.4 per cent, now believe construction no longer lags other sectors in digital adoption, against 27 per cent in 2023. Yet worry about being left behind also rose, from 36 to 57.6 per cent, and the report notes that this concern is sharpest in smaller organisations. The report's text says 69.3 per cent strongly disagree that digital transformation is only for large organisations, up from 49 per cent. Elsewhere it gives a different value for the same statement without saying what it counts, so rejection is the safer word. Asked to rank technologies by their potential to transform the built environment over five years, respondents put BIM first, AI second and cloud computing third. Remember that 2023 and 2025 are two separate self-selected samples. The report says their profile by organisation size is similar, which helps, but shifts in attitude can still reflect who chose to answer.
That leaves the board paper itself. NBS has recently run the survey every two years, and by the end of October 2026 it had published no later edition, so these remain the current figures.
So here is the board paper sentence rewritten. In the NBS Digital Construction Report 2025, a self-selected survey with 358 complete responses, mostly from design consultants and more than a quarter from outside the UK, 72.3 per cent said they had adopted BIM. Only 30 per cent described BIM as following BS EN ISO 19650. We will therefore test each bidder's capability against our exchange information requirements at tender. The figure stays in the paper, and the decision no longer rests on it. The episode on tendering and appointments shows how that testing works.
To draw this together, go back to the opening question. Among its own respondents, a survey like this one can carry direction over fifteen years, how they understand BIM and classification, and where their awareness is low, on classification, product information management and passports. A self-selected sample led by designers cannot carry a rate for the whole industry. Self-reported answers cannot prove practice against ISO 19650, the depth or governance of AI use, or the outcomes of digital tools.
Read every figure with its question, its base and its sample beside it. Among these respondents, BIM adoption has been flat for seven years while understanding has moved towards the ISO 19650 series, so test a claim about BIM capability against requirements rather than adoption. Cloud working, AI, product data and environmental measurement are all rising, alongside clear gaps in awareness, but the questions record presence and activity. Treat those figures as signs of direction, not evidence of good governance or better results.
The report itself repays reading, appendix first. The episodes on the common data environment, on information management in the operational phase and on tendering and appointments take up the questions it leaves open. The episodes on the RICS report on AI in construction and property, from the Royal Institution of Chartered Surveyors, and on the RIBA AI Report 2026 apply the same reading to two more surveys.
Reimagining AEC is an independent podcast, and I am Dr Sepehr Abrishami. Consider how often a decision in your own practice or research stands on what professionals have said about themselves.