BIM in Asset Management Monthly Forum 10-06-2026

BIM IN ASSET MANAGEMENT MONTHLY FORUM 10-06-2026

GEORGE STEVENSON We started 6 years ago as BIM4Housing and then 2 years ago we extended the work that we’d been doing into asset management. And part of the work that we’ve been doing over the last 4 years is to really look at things from a point of view of the golden thread side of things. We were part of the team that was looking at how we could define what the golden thread was, this is back 6 years ago, and we were responsible for asset information and survey information management.  And we ran lots of workshops where we got input from subject matter experts.

Here for example, all of these people are all manufacturers, these people are all designers, these are operations people. And what we’ve managed to do over the period is get people who are competitors to actually sit around a table and collaborate and share their knowledge, which I think is really critical. We’ve got a range of things that we’re addressing, in particular statutory things, but also looking at things from a regulatory point of view so that we can then act as a conduit to feed information back to the regulators.

And one of the big challenges that we’ve got is getting senior directors to understand their duties under the new regime. And what we’ve discovered, this is from our polls of members that we’ve done over the last year, we found that people are not confident about what information they’re receiving, in particular on fire safety which is pretty shocking. But also that 89% don’t believe that their senior directors are aware of their duties under the Building Safety Act which is pretty serious.

Because what is now happening, we’re coming up to Gateway 2, this is on the Building Safety Act for residential, where people can’t actually occupy the buildings because they’re refusing that. There’s also big changes that have come through from a legal perspective because this is all retrospective now. And we’ve got a major legal action where Mace are suing Mitie for 5 million pounds for fire dampers. And Anthony Taylor that we’ve been working with from the Construction Leadership Council shared with me that the laws are changing, their tightening up in a much more stringent way.

And they’re removing some of the things that have historically been protections in terms of 12 months, say, has now gone to 10 years. So all of that is a bit negative, but hopefully is causing people to take the responsibility for providing the right sort of information seriously. I also noticed this week that Margaret from UCL has pointed me at this where the responsibility for building control, that looks as though it’s in lux. But that might be seen all negative, what we’re actually trying to do though is look at the positive side of better information, so the golden thread information for example.

And we produced this white paper earlier in the year where we’re actually looking at the activities that the golden thread supports and also what the outcomes are. So being able to have much better information, better life cycle information, encouraging closer enablement of supply chains. So these are all things that we’ve identified in the white paper. The point behind it is that a lot of this information, people are already paying for it to be collected because it’s part of the standard information process. And therefore we should be able to then demonstrate tangible value based savings, not just a threat of litigation.

So what we want to do is use this community to try and get to some of these major financial cost savings. So how do we actually go about that. Within BIM In Asset Management and BIM4Housing we’ve got these working groups and if you’re interested in any of them go to the BIMinAM website and you can see what we’ve got going on. And also we’ve got individual workstreams which are task and finish workstreams. We record every meeting and also produced a podcast for it. In this particular case with AI we’ve had a very active group working on AI for the last 6 months or so.

And what we’re also doing is then using that experience to actually look at how we can form relationships between the members of BIM In Asset Management. People who maybe can work more collaboratively together to provide tested solutions so that those can be merged with the information needs of clients. And the by doing that we can then end up with much better information that is tied around standards, standardisation and standard libraries. By doing that it means that any of the software applications that are being used can then be able to share information. So that’s part of the how.

And we’ve got a large community of software companies involved as well to share their information. And also we’ve got a lot of people in organisations that are in that as well. But this then brings me to the critical question of why do we need integration? Why do we need better information, rather than just information for information’s case. So one of the workstreams that we’ve got which Ian Upton is leading is looking at hazards and HAZOPS. So in other words looking at things from a safety management perspective and then saying what documentation and information is actually needed.

So for example for building safety cases, what information is needed and then also what information is needed for as-built and compliance. And to be able to gather that in a way that is part of a standard construction delivery process, but looking at it from the perspective of risk management. This is something that I’d like you to think about from a scenarios perspective, this is answering the question why. The principle here is that rather than look at just what documents we have been collecting, actually look at this from a point of view of what’s the problem.

So we’re creating scenarios that we can then look at to say what information is needed. So here it’s a fire damper failure, here we’ve got breach in a compartment wall, here we’ve got smoke extraction systems. And all of those are then informing what type of documentation is needed and where is it and do we have it and how should it be connected together? So my request for the group is really to have people that are knowledgeable about what’s needed who can then come and share this work with us.

So to be able to look at things from a point of view of how different parts of the building work together. In this case we’ve got a scenario where fire breaks out in a kitchen. So what we’re trying to do is stop smoke from getting in to that stairwell and there is a range of different measures that need to be in place to ensure, for example, that the compartment works. But also then to be able to make sure that smoke control detection systems and also the other things that are needed to make all of that work.

And the way that we go about doing that is to make sure that we’ve actually got the right information about each individual asset in a machine-readable format. I’m just going to finish on this bit where what we’re doing here is taking the fire strategy plan and then overlaying it to basically a 2D BIM model. So we’re federating information here that’s coming in from various different sources and then looking at it from the perspective of what’s the risk and what are the measures in place to deal with that risk. And then to be able to overlay simple information like floor plans against that.

And then to be able to see that in some sort of more joined up way where fire doors are things like that and spaces they are protecting are being covered. I’m going to jump on to, this is one of the areas that we’re working on. This is being able to take information that is being captured for fire stopping, but that’s being delivered in a PDF document. So what we’re doing is extracting information from the PDF which has got literally hundreds of images. And we’re then identifying the locations of those individual penetration seals and then writing them to a database so that we can then have that glued together.

So what it means is that we’re federating information that’s coming from various different software packages, documents and reports, into something that can then be used from an operational perspective to reduce risk. So that’s broadly the work that we’re doing around from different software companies, different people providing information and then federating it so that can all then work together.

CHRIS ALDRIDGE What George has said, I’m going to go over a lot of the same material.

GEORGE STEVENSON The difference is that we did that retrospectively, so that’s an important point. What you’re doing is on a future perspective, so rather than us having to tie those things together manually which is what we had to do, what I think you’re doing is doing it progressively.

CHRIS ALDRIDGE Yeah, George put up a screen with many more players on that this, this is just a handful and there are more. These systems are very good at what they do and they’re picked by the clients for any number of reasons, best of breed or it could be cost or historic, they’ve always used that one. So there’s all of these systems out there being used and collecting this really useful information, but it is still a bit siloed, it’s still a bit on it’s own and you need to know where to go and get it in any particular instance.

What I’m working on at the moment is putting that stuff together and federating it in a way where we can start overlapping and showing this thing in a more federated way. There’s many more than this, we’re going through and talking to these systems and we’re looking at ways of getting data from them. APIs are obviously better because we can automate even more the process, but in some instances you can get off to a quick start by the supplier files. And this doesn’t need to be all of the data, we just need to know what data is there and how it’s held and enough to link back.

So we get those data through. More importantly nearly all of these applications work off the principle of dropping pins down on a floor plan. So again, using floor plans that another company has also used for the same buildings for the same clients. So once we’ve got that step of getting the data we can then pull the floor plans and we have the next step where we go through a process where we align all of those plans together, so we have a consolidated coordinate space that we work within.

And we can then show where things are in relation to other things, even though they’ve been captured in a completely different way on a completely different scale using different methods. And so we go through that process of lining them up. Once we’ve done all of that we’re then on to the actual presenting that back in a useful way. And rather than having these reports in a PDF where we’ve got one per page, we can start showing them all on the same plan. Not only for the same supplier of data, but for all of the suppliers at the same time, or in any combination that makes useful sense, because obviously we don’t want to swamp the user.

And what we show on that plan can either be data that we’ve pulled and then on what we’ve worked out with that particular application over here. Or it can be quite simply real top level results which when you click on we can jump back to that system. But they key point is all of that is pulled together in this federated plan and then you can see how that information relates. So for any given space you might see 4 different suppliers supplying information that relates to that space. And on top of that you can the overload these plans and from the clients that contain fire compartmental walls and all of those extra information.

GEORGE STEVENSON My request to the group, what we’ve now got is a situation where pretty well all of the major software applications that you might use have got the ability to take to each other. That’s that term APIs. The thing is though with an API you need to define what you actually want to do. You need to know exactly what information you want from that other system and how you want to use it. And that’s where the technology people like Chris, they need input from people who are on the ground who understand what their needs are.

So going back to what I was showing you earlier where we talked about the scenarios. That’s the approach that we’re taking with Tower Hamlets at the moment, a council, but it’s a housing department, they’ve got 78 tower blocks. And they’ve now got some really rich asset data. They’ve got a data model for each of their properties with 360 photos, rich asset data. So the next question is how are they going to use it? How are they going to use that to support their safety management system, for example? Or their inspection regimes, or their reactive maintenance.

And these are the sorts of things that we need feedback from people. So if you think you would be able to contribute to that type of scenario thinking, if you could put ‘scenario’ in the chat we can engage with you on that. And we’ll share with you the work that we’re already doing. But it’s an opportunity for us to understand from people how they actually want to work.

MATTHEW A lot of the talk that we’re having in these meetings is obviously information delivery through the construction side. Now obviously we’re looking at how in the operations side and a big thing we’re looking at is building up structured data to be able to pull in to platforms to be able to do things. So user tools and combining everything from existing CAFM etc because unfortunately a lot of things that happen in operational is literally just the CAFM pretty much, and trying to combine that to make it a lot more useable.

So the issue that we’re finding a lot is to get organisations to look at API, to open up those APIs and frankly to pay those systems to implement them, they find difficult because….proposition yet before they’ve even done it. So we’re looking at a lot of keeping it simple, as you’ve got there with the files Excel element. Because a big thing for us is getting things structured, so a lot of the O&M data in operations at the moment is just unstructured or based on a CAFM structure. so we’re trying to align that through the COBie route within an existing estate.

But our hope then is as soon as you can inform through manual routines like sale input export etc to the end users, the building operators etc how important it is then to open up the Ai, it gives that value proposition to then go to the vendors and say right, let’s pay for that. Because I think that’s the issue that we’re finding on the existing estates is we don’t know whether we should pay for it and what we’re getting out of it.

FRANCO PITTONI I have a question on step 3, align the plans. The question is how do you deal with what I think realistically may happen which is a misalignment between plans coming from different sources. The second question is about step, the federation of data. I wanted to ask you if it makes sense to make an analogy between the very old concept of layers in CAD where you place certain types of information on one layer, other types of information on another layer.

CHRIS ALDRIDGE All that I can say on the alignment of plans so far is all of the ones that we’ve tried with we’ve managed to align. So we go through a process whereby we scale and ropate and basically the plans that we’re working through, it’s probably because we’re all being fed by the same plans ultimately at some point, they’re being given to us and being given to Bolster to go off and do their thing. So I think quite often there is a common source for them and then they diverge and we’re pulling them back together. But we haven’t found ones that we can’t align yet.

GEORGE STEVENSON The key thing here is in most of these situations they’re not plans, they’re just images. So what we’re doing Franco is we’ve got a one to one model, that’s what is in the master database. So therefore what we’re doing is basically taking the image of a floor plan which is what typically these systems are using and we’re just scaling it up and positioning it in the one to one environment, and therefore translating the coordinates from that to the database. So that’s the way that we’re doing it.

FRANCO PITTONI OK, I understand it, but I was just injecting a little bit of realism. I have seen many situations in which somebody is working on revision 7 and another one is working on revision 9 of the same floor plan. and then when you put them on top of each other you discover there are some misalignments.And my question was about if you discover that one plan comes in whatever form, PDF, from somebody you should go back to that author and say look, your plan is wrong, it’s not aligned with the other plan. And somebody has to do something in order to achieve this total alignment.

GEORGE STEVENSON Yeah, it’s not as sophisticated as that.  When people are doing inspections they’re probably working with a floor plan that maybe 5 or 10 years old. So they don’t normally have access to the original authors, So what we’re doing really is we’re using the coordinate points that they’ve placed. And the fundamental difference of this is that we’re using the coordinates in a world coordinate basis so that we can then align information from any source so that the locations of pins are relative to the whole building, not simply the image.

CHRIS ALDRIDGE? 27 mins 53 secs Your idea of layers on Level 4, that could well play a part of how this develops, or it could be more related to George’s scenario point of view. So when I’m in this position in this scenario in this space, what are the systems that I should be looking at? And then we show them like that, as opposed to showing them all in one go. It’s going to very much come back to usage and how people want to use this thing, which again talks to what George was saying about understanding what all this stuff presented in a particular way is useful.

TIM AIKIN I just wanted to clarify that most of the systems that you say that’s under section 1 usual tools, they’re mostly using an image. So their position of their pins are too that image, or PDF, and it’s then a proportional coordinate X and Y. And that’s great because in our system it’s a one to one so we just lay the backdrop in because it’s an image, work out what the transformation matrix is and then that is used to translate whatever PIN they place on that plan is then translated into the real world coordinate.

And I don’t think we’ve come across anybody on the surveying who actually does a vector based system because for them to do that is a workflow that they get no value out of. Jus the nature of the business, this is the way that it’s come up. But for us with operations it makes sense to have it vector based. Now the other thing is in terms of drawings, with legacy buildings what we have found on one particular building that we looked at was that the original drawings don’t actually align either.

So the floor above doesn’t quite align with the floor below, but that’s because those drawings done on a drawing board 30-odd years ago or whatever, but it is what it is. So you go right Ok, do you go with it? You could correct it later, it doesn’t matter. I just wanted to clear that up because there are challenges, but we’ll sort it out at that point.

DAN Basically what you just said there, what we’re doing in GS1 is lining aligning what we call the digital asset identifier at design stage to the space that it occupies. So therefore as you go through the project the space may change, but that relationship stays the same. And then when the project is actually delivered you match in the project data sheet or in the future the product passport to the digital asset identifier, so you achieve the golden thread end to end.

RICHARD FREER We are launching a new initiative, the whole of BIM In Asset Management is obviously around collaboration, but this is a specific initiative which is being headed up by Tom Roche. Tom, do you want to tell us about that?

TOM ROCHE I’ve been involved with the Golden Thread Taskforce now for around 6 months. My background, I work for Galliford Try and for Symetri. Symetri, you may think an Autodesk reseller but they are also involved in many different software applications and also support live projects from a BIM and information management consultancy side. So that dual perspective, the software and the delivery, is the view…into the industry and also the problems that I’ve seen also.

So what I’ve seen onsite is again you could have all of the requirements defined, systems in place, models verified, data captured, but the question I kept asking myself was are the facility managers actually using this information? Is what we’re producing actually built for the team who are operating the buildings? Or is it just a tick box exercise?

And that gap between what’s captured and useable, that’s what has brought me along here. And as we know technology is moving rapidly at the moment and what I want to try and help to do is to bring people together and also ensure that the data that we use is effective on projects and also to help the team who are managing and operating the buildings.

RICHARD FREER So you’re saying what you want to do, how are you going to do it?

TOM ROCHE The way we need to do this is that collaborative approach. I think we’ve got systems connected to systems and again we’ve got over 100 people on this call. But I think the way we need to do this again is to try and speak between us, speak with the subgroups that are on this forum. And again, if it’s me speaking to people one to one or as a team really.

RICHARD FREER So do you want to hold a meeting? Should we call out to get people to come to a meeting with you, is that what you’re looking for?

TOM ROCHE I think the best way, I know normally you hold one to one meetings, so I think if we could take that approach for now.

RICHARD FREER If anybody is interested in getting involved in that and learning more from Tom, just put ‘Tom’ in the chat.

FRANCO PITTONI I think that if Tom is interested you may join the so-called task exercise which is exactly about analysing what is being done during the stage when data and information is produced, and when the information is consumed. So in the near future we are going to start this exercise where we will analyse exactly what you’ve said. So if you’re interest you may join this so-called ‘tasks’ workstream.

GEORGE STEVENSON Anybody that’s interested engaging in that collaborative working side of things, maybe you’ve got a problem or you feel as though you’ve got something you can contribute as a solution, either from service or software, put ‘Tom’ in the chat and we can connect you up. Paul is up next, the progressive O&M.

PAUL OAKLEY iI’m really going to be stating the obvious because basically this is what the perception was that the industry should have been doing for the last 20 years and it really came out of what the BIM Level 2 requirements are. But we know in terms of the way that asset information is procured that quite often it is something that basically happens at the end of a construction project and is a paid for thing.

So the challenges are basically that things like O&M compilation and other information is generally assembled at the end of a project. But actually a lot of the information could be provided as we go through. We get issues with the so-called as-built information, but actually a lot of that can be through design models and sometimes design intent models and actually doesn’t reflect what was built.

We also get issues in terms of information bloat. I worked on a project recently where we had 20,000 documents of which 6,500 were actually unique. So we had multiple revisions and versions of the same documents which then obviously causes problems because actually then which one do you trust of what’s actually been delivered? And it’s that last minute chasing information, quite often when the FM teams have already started on a project, they’re trying to actually maintain and look after assets and make sure that things are there.

And they’re still chasing whether things have been tested, the records, all the compliance evidence, whether the contractor has carried out their maintenance requirements for the last 9 months before its been handed over etc. So there’s all of these consequences which rally fit in to the impact of not having the information at the appropriate stage. And the reality is there is no golden thread. All of the projects that we’ve been looking at on these things, they all talk about this, but actually they can’t show compliance and they can’t basically provide that golden thread that they’re supposed to have in place.

So the whole thing is we really need to look at the whole workflow as we go through the stages. So stage 1, early on within the project you can define the asset information requirements. You can define the COBie structures, you can deal with things like the product libraries of the products that you know about. As you get into concept design you know the asset classes, they types of objects that you’re going to have within those systems, you know about the general spaces. So all of those room data sheets, all of the other information can actually be sorted out and then you can define who is supposed to be providing the right information.

And then as you move through the design development you’ll get the generic asset types, you’ll know what all of the types are. You can produce a basic COBie file which gives you the basic structure of all of the types. You can have the nomenclature, the naming conventions etc. Each task came under 19650 or 1192 was produced in a task information delivery plan. That identifies what information is required. Now we talk about statuses, but actually purposes was not built within that.

But one of those purposes is what information is appropriate for asset management? What should get handed over from design construction into AIM? So that could be quite easily be identified as one of the purposes, but also compliance against regulations, so what’s been submitted for building regs etc, what’s been submitted for the fire safety requirements and that. All of that lot could be part of the TIBP process which basically pulls on the back of that.

Previously you’ve heard me talk about technical submissions, PTSs etc, but actually defining what the requirements are for products against those asset types. A lot of that can come out of the generic asset types identifying what PTSs that you need. And then as you actually identify the particular manufacturers within that process you can then build on those and start to provide the data that is then needed to facilitate the asset management further on.

Ad the actually as we go into construction the particular decisions are made and we can deal with those. But again, things like digital asset capture can then, a lot of this can be automated, as Tim has been showing in the AI group, if you basically know what is likely to go in to a particular space, you know they types and you can carry out those. And then basically you can actually identify what’s actually been built in the first place. That can then support the commissioning and assurance processes, how we deal with the regulation checking and that lot.

And then actually when you come over to the handover of operation a lot of the information is pre-populated ready to go. And we can actually show that whole golden thread process that at present is not being shown. So there’s these sort of key outcomes from this process. At the end of it we’ll have a proper information model, you can show the golden thread. We have a level of verification through checking your models, digital asset capture etc. So we can basically deal with the compliance evidence and basically build up that life cycle platform that we’ve been talking about.

So there’s some core principles in terms of how this needs to happen and basically it’s don’t wait for the end, make sure that information must be progressive throughout the process. Identify who is responsible for delivering the information. And it’s a continuous validation, don’t wait until the end to try and validate the data. Carry out that validation throughout the whole process while going through it. The information must be linked together so we now what asset types and which systems and all of those sorts of things to pull that information out.

And make sure we capture what has really been built, so the as-built. And then hopefully we’ll have that whole gateway and golden thread alignment that we’ve been talking about throughout these things. So obviously the whole aim is at the end we’ll actually receive a proper asset information model, based on what’s been built. So the task information delivery plans can basically identify the documentation requirements in terms of from those what has actually been provided into AIM and what the purposes of the documents have also been in terms of what’s regulatory requirement, what’s been submitted for planning, building regs, other requirements on that.

All of that can actually feed from those basic lists of the documents and models provided. The PTS process that can actually provide the golden thread of who has made the decisions. so was it the owner who chose the product? Was it the QS? Was it the subcontractor? Was it the main contractor? All of that can be captured and a lot of that can be pre-populated at the beginning from different sources. We go into the compliance evidence packages in terms of how we deal with that, that will give us the golden thread that we keep talking about throughout this thing which isn’t presently available.

And then obviously we can then feed in to that operational life cycle platform. So a lot of the things in terms of the regulatory compliance, we’ve talked about previously ISO 19650, compliance with all of the classifications in the UK, use of COBie. We can define those sets of requirements quite easily, but actually CDM, the health & safety file also comes into that, compliance against things like Regulation 38 and the Building Safety Act. So this is work that the two IDPs and PTSs that we’re working with Morta on to basically help the process and then tying the things in with Active Plan to help resolve some of these things.

RICHARD FREER If you’re interested in getting involved with Paul’s group progressive OM put ‘progressive OM in the chat and we’ll be in touch. George, validation.

GEORGE STEVENSON What we’re now doing, building on top of that. Obviously what Paul has talked about there is doing it properly so that all of the information is progressively created and validated as it gos through. But all too often we get O&M information that isn’t joined up. So we’ve got a four stage asset validation process and some of this has come from what Tim is going to be talking about later which is the AI working group.

So we’ve got initially COBie validation and then we’ve got the O&M asset verification. Then using 360 photography to check what has actually gone into the building and then using the regulation checking as well. So the four step process starts with basically the information from the models being checked and validated. And then anything that isn’t correct automatically gets returned to the author. So the principle there is unit testing to be able to check that and validate the information.

What we’re then doing is then looking at the documents themselves. So probably half of the asset information that we need from an asset management point of view doesn’t get put i models because it’s installed by specialist subcontractors who aren’t modelling. But what they should be doing is providing the health & safety files and the O&M information in some sort of digital format ideally, but we then need to extract the information from that.

So what this process is doing is enabling us to check the O&Ms very quickly and it’s an AI supported process we’ve been developing over the last 2 years to be able to also check not just whether the files exist and whether they’re in zip files or whatever, actually looking at the content. So this is an example, a real job where we’ve had subject matter experts go through and do their audits of each of the work packages and identify whether they contain the information that they should have contained. And that then is providing us with the information to be able to produce a report back to each of the specialist contractors saying whether their information is adequate

And certainly in most cases we’re finding there is a lot of information that’s missing according to that particular process. But that can then get fed back to the specialist contractor to get them to correct it. What we’ve then done is taken that manual process and we’ve developed agents to be able to replicate that. But we very much still need the people that have got the knowledge and the experience who helped us on that first part of the journey to be able to oversee it, because we can’t just rely on AI to do the checking.

But what that does is it automates the process, so you end up with something where it’s actually generating an output that tells each work package contractor what they need to do to get to the level of completeness that we would expect. So having done that, what we’re then doing is also checking the contents of schedules against drawings and also looking at the integrity of the data and then looking at that from a retrospective perspective to say OK, how can that information be corrected.

And once we’ve then said that we should have a combination of the BIM data, the information from the O&Ms and that creates a baseline asset register that can then be validated against. So the next stage of the validation is then using 360 verification, so being able to use reality capture either through a process like OpenSpace or Qpix or one of the other great systems that are around, or just taking 360 photos with a normal camera.

So capturing the 360s then allowing a team of people to work remotely to then verify that what was claimed to be installed has actually been installed, but also then to be able to refer to schematics and things where there are assets that perhaps aren’t visible to the camera. So that’s the process that we’re following then to work through on that. So if anybody is interested in that automated O&M checking, because we’ve found that there are many O&M files which you only find that they are incomplete when you actually go to look at them.

So If you’re interested in that O&M verification, if you put ‘verification’ in the chat we’ll share with you the process that we’re following there. And also if you’ve got some projects that you need at that type of validation and checking on then let us know.

RICHARD FREER HAs anybody got any questions for Paul on progressive OM? Or George on verification?

FRANCO PITTONI One question is about these workstreams. I put my name for O&M in the past, I now put my name O&M ongoing. I was just wondering is there not a lot a of common topics in all of these different workstreams? I presume that this O&M will deal with some topics that also Paul will deal with in the progressive O&M workstream. I’m trying to understand where I should be because I’m interested in all of these things that have been said. Could somebody clarify the differences between these workstreams with different names?

GEORGE STEVENSON There’s clearly overlap in them and what we try and do with the workstreams is make them task and finish workstreams that can reach a conclusion within say 3 or 4 meetings. Now if we make it too broad, for example people that are interested in this asset verification, they may not be people that are involved in creating O&Ms. Whereas the people that are creating O&Ms or overseeing that, they have a different level of interest. So I agree that there is an overlap, but it’s a different topic as it were.

RICHARD FREER Well it’s not just that, the people that are coming to meetings now are not the same as people who are coming 6 months ago. We have new people coming through all of the time, so each of the workstreams needs to be discreet in that it covers the whole part of a focused topic and not worry about something that might have been mentioned 6 months ago or a year ago.

GORDON PRINGLE It is part of the process because you’ve got everybody in siloed positions in a process, but at the end of the day if the device for the constructed asset is not in that process they’ll never get that validation that they’re trying to seek. But that’s where I see that aspect of siloed mentality. Trying to understand it in 3 short sessions when actually yes, you need to understand your aspect of the data or the process, but we need to put the whole linear line of the process in play to get the full validation of it.

FRANCO PITTONI I couldn’t agree more. My question was the first initiative of this type which was called ‘O&M’ appears to be the parent of the O&M verification, O&M progressive. So my question is what is the intent of the scope of work of the O&M initiative? I need to understand where I should sit. The question I wanted to ask Paul, I am very happy to be in your process description testing and commissioning. I was wondering if there is any reason why you’re not using 1% the 8 RIBA stages which do not contain the term testing and commissioning?

PAUL OAKLEY There are multiple different stage options depending on which solution that you’re going through. So it was just trying to use more generic stage names as opposed to going back to specific RIBA. So there are multiple ways of dealing with the work stages. I think from what you were talking about before Franco, the issue is that we had a group looking at PTSs, we had a group looking at O&Ms. And I think there should have been some published outcomes from those, because that is really supporting the work that we’re now trying to take forwards on this because actually it assumes that those other things are in place for us to then move forwards.

GEORGE STEVENSON I think the important thing is that what we’re doing is bringing people together, we’re recording every session, we’re producing reports from each session, we’re producing podcasts from each session. So anybody that attended any of those original O&M meeting Franco, or people that perhaps haven’t attended those meetings, will be able to draw on those.

FRANCO PITTONI I understand that, are you referring to my first question or to the last question? Because the last question if you want me to be very specific, as you know I am working very hard towards launching the tasks project. One of the fields of this database is called work stage, and I need to decide, we need to decide which language we use to attach this specific metadata to each task.

EMMA HOOPER I just wondered if the group would like us to present on a project that we’ve got called Project Colin which won the information management award at the Digital Construction Awards this year. It’s about looking at O&Ms, health & safety files, fire & emergency files and actually breaking the information down so that it’s only specified and delivered once, but then it’s grouped for many different purposes within the CDE.I just thought that might be useful given some of these projects that you’re talking about at the moment.

PAUL OAKLEY One of the questions was about standard O&M structure, which actually that was part of the work that was carried out by the O&M group and was published.

RICHARD FREER Yeah, so I think part of the lesson here is look on the website, it’s a resource. Ok, now the breakout star of 2026 is the leader of our AI workstream. Tim, keep us up to date.

TIM AIKIN I’m just starting presenting, I think generally Ai is just everywhere, but I’ll just give a recap on where we started and how we’ve got to where we are. So when we started…the working groups first sessions were about sharing experience with the new thing which was the LLMs, the Chat GPTs and the Claudes and the Geminis and Copilot and that sort of stuff. And obviously people were sharing their experiences of identifying the risk. So I decided that we should really do a sort of survey, so we couched that around O&M production.

And what was highlighted was really the things about trust and accuracy. The key insights really were those two things because there was a lot in the headlines about how people are using these chatbots as boyfriends and girlfriends and getting in to all sorts of issues. And also businesses letting agents run riot and ending up either with large bills or suddenly having their whole development platform deleted because it thought I’ll start with a clean sheet. So it was quite important to understand that you needed to have explained as to why something reaches the output that it does.

And so the survey that we did, these were the headline percentage results. We had 28 responses, 96% using it, and it was interesting about the ones who weren’t, there were a few and I don’t think that is the state today. And I think that when we do this survey again we will couch it in slightly different terms so we can cover more aspects. So what we did, as we were progressing we could see that there were quite a lot of gaps in people’s understanding. There were people at one level and people coming into it sort of fresh, so we needed to address those things and needed to explain.

So we did some slide decks and presentations so that people would understand what was meant by these different terminologies. One person commented I haven’t got a clue what MCPD is which is fair-dos and that needed clarifying. And again, the same with the RAG and the vector databases, people don’t understand what they are. So we did some slide decks that put them into the context of how they are used in the real world.

And we didn’t actually cover this as a slide deck, but it’s quite important because it’s been going on for years which is the object recognition which is used in medical imaging, it’s used in auto-driving cars and things like that. And we also got people to start sharing and showing how they’d used it in the real world. And in fact I’ve got one guy who’s done an MCP connector on to Autodesk Civils and he’ll give a presentation of what they did with that.

The other thing was about the guidance, there was very little guidance and I think back in March RICS had just issued their guidance. There’s already been guidance for people in the electrical industry, but that was more aimed at manufacturers. And the important things are to also recognise that AI is sort of the umbrella of it all, but there are nuances and its critical to understand these differences. So when we talk about object recognition, there’s two things: there’s detection and recognition.

And if I just couch it like this is that if you’re looking and trying to detect something you're saying is that a person? Or is it a face? But if you’re trying to do recognition, is that Tim’s face or is it George’s face or Christine’s face? So that’s the sort of difference between the two things and it’s pretty important to understand. The other is context. If you don’t put things in context, so people are just generally uploading to an LLM a big document and asking something, which includes regulations, you could get the regulations from anywhere in the world unless you actually put the guard rails into it.

GEORGE STEVENSON What I was going to get Alan (who has not joined the meeting) to do was just explain what we’ve been doing with the Construction Leadership Council around determining, specifying what as-built means. Now this may seem a bit crazy, what the Construction Leadership Council have identified that there’s a real challenge, particularly I think because of the way the regulations have caused people to think twice about.

So as far as the as-built group is concerned, we’ve been running workshops with the Construction Leadership Council over the last month and we’ve got another one this Friday. And part of that is to determine, to get an explicit probably a contractual definition of what as-built means. Now what we’re doing with that is making it so that it then ideally is going to be put into building regs as some sort of definition. But certainly the Construction Leadership Council are looking to put it into contracts.

And part of the reason for this is to try…Anthony, Alan Binns who was hoping to be here isn’t here, so I’m going off the cuff, so could you just explain what we’re trying to do together?

ANTHONY My background is in the occupied sector and for many years I’ve been given suboptimal information when I’m in occupation. Whether that’s from a new building or whether that’s somebody moving a building between managing agents, facilities managers, the quality of the information has been bad. So we’ve raised the issue and I also lead a portfolio for of the CLC which s looking at buildings in occupation.

And that group wanted to define what as-built information actually means because when we ask it meant different things to different people, or let’s say different definitions coming through different disciplines throughout the process. So what we’ve done is set out a) let’s define what we want in occupation, what should we expect? And we spent the last month working out what people think is reasonable and should be expected. That is not to say that we think it can happen now.

What we’re looking for is what should be being produced somehow. And then we’ve established a definition, we’ve gone through now to expand on that definition to say OK, if you’re talking about verifying or validating something that is the minimum level that is acceptable to validate or to verify something. And next Friday we’re going to be talking about those minimum levels. Th next stage then is to look at how this information should be collected.

And Emma Hooper who was speaking a little bit earlier was talking about the same sort of thing, the point being is we are basically working in the same direction. What we’re looking for is the far end, what those in occupation want. And we’re talking about the occupying of any building, we’re not talking about HRBs or residential or anything in particular. We’re talking about any building that is handed over and any project. So if you’ve got a commercial building and you’re moving a boiler from one floor to another, it is a project and you should have information that’s given to you.

So the plan is to cover all of that and then to work with Emma and others in the industry to say OK, if that’s the quality and the detail of the information that we want how do we collect it along the way. So it should all stitch together in the long run. That’s the plan of action.

GEORGE STEVENSON If anybody is interested in understanding a bit more about what the group has decided is important from an as-built perspective, you’re welcome to join on one condition and that is that you take a look at what we’ve done before. Because it’s all there, there’s probably a 15-minute podcast there which summarises things. We’ve got the actual meeting itself and also a transcript. So if you’re going to come along, the next meeting that we’ve got is on Friday, please put ‘as-built’ into the chat.

And if I could just say that Anthony has been chairing this very effectively, it’s been like herding cats with us all. But the really powerful thing about it is that he’s kept us focused on not so much dealing with the reasons you can’t do it, but focusing on what is desirable so that we can then work together to actually make it work.

ANTHONY Indeed, we are focusing on what should be delivered, what’s reasonable. Not just yet how do we get there because that is accepted that the procedures are not in place at the moment at any level in any size of these projects, being a giant one or a tidily one. But what we’re saying is let’s start with what we want and then work our way back to how we get there and that will take a bit of time.

TIM AIKIN If I just get to the slide, we’re looking at the guidance which I think is quite important because a lot of people are sort of in the dark about where to go with that. So what’s come out of this? Obviously the theme that runs through all of these workstreams is the golden thread and one of the problems I and lots of other people came across is about receiving photographs and not having a clue where they belong. And so I looked at this and thought what can we actually do.

The problem comes from the fact that people…this actually goes to what Gordon was saying about the digital passport and things. Photographs are like artefacts, but they’re useless without a manifest at the moment, but they don’t have to be. And this is how they’re stored on an Android and this is how they’re stored on an iPhone. And so they automatically embed what’s called metadata in a format called EXIF. So they have a timestamp, they tell you what the exposure was (f2 or whatever).

They give a geolocation, so you now from aGPS satellite where it is. It won’t tell you where it is in the building, but it will give you the location of the building. It will tell you what the camera was, what phone it was and it will then give you a profile about other aspects of that phone. But there is nothing else on it, so I looked and thought surely you should be able to shove something more on to it. So out of that I developed an idea of actually embedding the data in.

So what we wanted to do though is actually make it really easy to do because if you have to type things in on a phone and you are on a site and it’s chucking it down with rain, busy working environments they are difficult. So can we actually use voice to do it? Because voice recognition is already built in to all of the phones and the fact it’s actually an AI model. And you can convert that to text. so the next challenge was actually getting a context model, so this is to do with what’s called natural language processing. And to use something lightweight that takes what you’ve just said and says yeah, you’re on this floor, you’re in this space and it’s this type of asset. Those were the sort of challenges.

And then write it into the artefact rather than not just into a manifest. When I say manifest think of a little database and when I say artefact think of a file. So we’re writing it on to the file and it’s then travelling with it. I did this this morning and I screen transmitted it to my computer, there’s no sound on this but I am actually saying this is the ground floor dining room window. So we take the photograph, it then says are you OK with it and then it moves on to listening, have you got anything to say?

So it listens to what you’re saying and it then takes that on the phone, so no internet connection and says oh right, you’re saying ground floor dining room window. And you say happy with that, go add details and immediately it tags it up. So then that’s added that to that. When you finish those it actually just runs you start back to take the next photograph, I’ve just skipped it to cut that bit out to go to the gallery. So we can go back to that and then we’ve also got a GDPR. We can say can I do a GDPR check?

So this is scan and detect any faces and if it doesn’t find it you can say yeah, I’ve done the GDPR check. And then this is what gets stored inside the file, this is the metadata that’s actually built in. So when you pass that on to somebody else they’ll have this information. So that’s where we’ve go to with this and there is a desktop application as well and that’s quite cool because you can manage thousands of photographs across hundreds of sites.

Just on the GDPR, we’ve done some work for blurring faces anyway on 360 images, I thought well, we should be able to do this easily in the form of phone. And this is actually something nobody wants to address and particularly it’s probably not as relevant if you’re onsite in construction, but it is if you’re doing on premises surveys and things like that, because the premises are being used. So you might accidentally catch a photograph of somebody, or even a photograph on a wall or on a desk. So you certainly don’t want to get in trouble with that.

So again, we can do it at two points. We can either do it on the phone or you can do it on the desktop. And although on Android it’s actually built in to Google machine language tool kit it didn’t work very well, so I pushed in another one which does work. And I’m expecting a better result on Apple because it’s probably more advanced technology and the chips are certainly far more capable. So this is really a summary of where we’ve got to really.

There is one other thing, if you’re interested I’ve actually published some publicly available source code which is this one here, the MPC server for Revit, is a fork of somebody else’s repo (code repository) which is based using OJS. And I thought we’re on C sharp stack and I just want an executable, so this is my version but it just uses a C Sharp executable and I’ve added a few more tools into it as well.

The second one came out of because we were going to get involved in a project which had IC files it was to do with ICs, so I made this one public and I’ve built this from scratch with Claude, the coding agent that I’ve used to help me build it and you can access that one. Again, it’s an EFT, it’s a C Sharp executable, but you can connect that to an agent and the agent can then open up an IFC file and then you can interrogate it. So it’s like a no graphical interface.

GEORGE STEVENSON I think the really important point here, Richard, how many people have we got in the AI working group?

RICHARD FREER 70 or 80.

GEORGE STEVENSON The important thing here is that Tim hasn’t done all of this on his own, he’s had input and contributions from people that are part of that group and he’s been doing a lot of contributing as well to them. So it’s a collaborative effort and I’m really inspired by it because we’re getting people that are experimenting with AI and we’ve got some people that are profound experts in it.

RICHARD FREER Can I just correct us, there are 234 in the AI group. The next session is the 23rd of June, so if you’re interested in getting involved in that put ‘AI’ in the chat.

TIM AIKIN Can I just ask, I’ve got one guy who said he would do a presentation. If anybody else has got a presentation or something that they’d like to share then please let me know and we’ll try and slot that in. Because shared experience is invaluable, what works and doesn’t work.

GEORGE STEVENSON There’s a lot of people, like me for example, who doesn’t have the technical skills to actually do it but I know what I need. So if you know what you need in terms of information and you think that AI could help with that you are equally valuable to the group because it’s understanding what the use cases are that we can then apply and work together on.

TIM AIKIN Yeah, I think also what I found quite amazing is that over the last couple of years is how AI is helping me build deterministic programmes. That means that they are programmes that you could write manually, but they just speed up the process. In fact virtually all of thread lens is being built like that and the new one that I’ve got is Thread 360 which is running the detection. Again, the interface and everything is being done using a series of AI agents.

MARGARET SACKEY I’m just putting something in the chat for you. On the 23rd of June there is a really big IOSH BSA event 12-5, so I’m just hoping that’s a session on the 23rd, if it can be in the morning it would be really helpful.

DEXTER DASH One of the things I wanted to ask, with regard to the presentation with the image and the metadata being assigned by voice recognition, have you guys looked at how the information populates whatever space it’s brought in to? I know you mentioned a like to Revit.

TIM AIKIN OK, just to clarify, the image metadata has got nothing to do with Revit. What it does is you can build a tree of all of the spaces, or you can import it. And then it will use natural language to work out what space you are actually in based upon the tree that you’ve got. And then you confirm that that is the right one. The other thing that we’re going to put on to it is if you’ve got a vector spacial model as well you could also click that.

In terms of the asset types, you have a list of asset types so it uses natural language to work out which one that you actually need. Now then if your location tree, if we think of the building as a location tree buildings, floor, space, or building, floor, zone space, because what we do is we use zones for things like apartments so that you’ve got context there. Then that information is embedded into the image metadata using the UID (Unique Identifier) that you’ve got back in your host system that it came off. And the actual code that you’ve got, so it’s like belt and braces.

So that’s how it works. And including also which estate, which site etc. So there is some setup, you just don’t get it and run it and magically it will work, it has to know these things to start with. But I’m making it so that it is easy to build manually, easy to import a CSV file and easy to also connect to a system using APIs. But basically, although it will be very COBie-esque in terms of the data that comes back. Does that help, Dexter?

DEXTER DASH Yes, that was quite useful because essentially I was trying to understand what the thoughts were around pulling all of that data associated with the image into whichever application is being used.

TIM AIKIN So programmatically there are very simple ways that a programmer can have an EXIF and an EXMP reader so it looks in the file and it says what’s in the header. And it reads that data and then they just say oh right. What I have tried successfully is there’s certain existing metadata which is a schema that is part of an image that does include things like locations and comments. So we’re filling those in as well.

So you can put in your own schema that means you can put in your own data definition of how you want it structured, but also you can use what’s already available. Which means that when you then bring it up in say Windows Explorer you can list those columns out.

GEORGE STEVENSON Just on that point, Tom Roche was saying earlier about the golden thread group. Symetri are part of that and they were talking yesterday about how ACC works with being able to put pins down into a spacial model. So Dexter, there may be a connection there that we can do maybe with Symetri to tie it back into ACC, if that is the sort of thing that you’re looking for as well.

DEXTER DASH Yeah, it’s really interesting for me because I deal with, for example I recently did a project where I had to organise over 9,000 files and it was all through coding. So I am very interested in that in terms of a test case or a context in regard to how it connects to different systems.

GEORGE STEVENSON The only thing that I was going to say Richard, we had the inaugural water meeting yesterday which was very interesting and very useful. We had a really good group, there was about 15 people on it. And we had people that are looking at things as water as being a risk, whether it be for heating systems or drinking or whatever. So if anybody is interested in water and the importance of how we manage that, put ‘water’ in the chat we can include you in the next meeting.

GORDON PRINGLE Just on that, a client just mentioned to me today as a result of that that their staff or their facilities management side of things look after water quality. In her view they are completely blinded by legionella and have enough on their plate with that to worry about water systems in the sense of heating and cooling systems.

So the regulation of legionella is where the focus on water is, but to the detriment of heating and cooling systems, and they’ve got systems with failing valves and pumps failing, everything, because they can’t physically keep abreast of the data, whereas the systematic process we demonstrated was really valuable for them. And that process that we were talking about earlier on of linking that data set and API ing it to a model, all becomes interesting.

GEORGE STEVENSON I think the point is, and it goes back to what Franco was talking about earlier, because what we’re trying to do is look at each of these specific information requirements in detail, in much more detail than most people are currently doing and then to be able to bridge between them. For example, the work that we were doing yesterday on water was looking at things very much from a system perspective. But if you’re talking about legionella testing you need to know where all of the outlets are.

And if you need to know where the outlets are, that falls into the as-built working group or the building safety working group or the O&M working group. So obviously we’ve got a massive set of challenges in terms of information, but what we’re trying to do is break them in to smaller bits so that we can then connect them together. So it’s that connectivity that we’re trying to achieve through what Chris Aldridge was showing right at the beginning.

MARGARET SACKEY Did you see the information I put in the chat on ethical issues in AI? The main thing is UCL have got that presentation.

CHAT

Mark Hunter……I assume the slide deck will be shared?

Debbie Carlton….George, activities and outcomes have to include information on competency and when/how/who etc. makes decisions

Ian Upton….Mat - I have an invite to you for HMNBP  Through KBS

Gordon Pringle….Matthew I maybe able to avail .csv data on our output that may assist you to demonstrate the process

Margaret Sackey….Hi. There. Zing - can we catch up 👍. Are you attending the UCL AI Ethical Governance Workshop - at Bentham next week. Thurs 18 June at UCL

Mark O'Reilly….Hi folks. Really helpful info, I'm sure this has already been passed but will these slides be available at the end of the meeting?

Vitalij Tetervov….Do we have a standard O&M structure in the industry?

Dan O'Gorman…Progressive O&M very well done, Paul.  In fact, we are following a similar drive in Ireland, and if interested, I am at Dan.ogorman@gs1ie.org,  The reason is that we have gone a few steps beyond what you demonstrated.  We are tying in with Product Data Sheets and "Digital Product Passports" in the future, so you have an interactive link with the OEM, ODM or Offsite Modular Entity

Vitalij Tetervov….I'll try again 🙂 Do we have a standard O&M structure in the industry?

Margaret Sackey….Hi Emma. Greetings.
Well done for your presentations at Digital Construction Week 👍
Is it possible  to get presentations or view recordings anywhere .?

Emma Hooper….I think PAS 1958 and the IMI have recordings so I will send you some links.

Margaret Sackey…Hi George. Are you and your team attending this UCL workshop next week.

Ian Upton….Can you send George your slides Tim ? to present

Harris, Joanna….Would be worth looking to see if ‘as-built’ is defined in BSI standards?

Tim Aikin….OK electrical storm here  PC back up

Harris, Joanna….We are working on an update of a 2012 BSI standard on facility management information so if you want to input, I’m sure it would be welcome.

Margaret Sackey….Great update Anthony 👍. We must get more from the maintenance and operations roles involved 👷‍♀️👍.  Working with Emma and Nic on this in UCL 👍

Gordon Pringle….If this is for me Margaret then no not planned in London I'm afraid.

Margaret Sackey…The message is for George Stevenson and Team.  Anyone can register via the contact details shown……

Emeka Osaji….Please accept my apologies. I have to now leave to join another appointment. Many thanks for the great contributions. Wishing everyone a fantastic week.

Margaret Sackey….Can the 23rd June. Ai. Session be in the morning if possible - there is an important BSA IOSH Event on from 12 - 5pm - happy.

Hajar Moshfeghnia Touchaei…Can you please add me to the golden Treat group?

Carolina Motolinia….Quick plug, grateful for this group's support in completing this survey to support post grad research at the University of Cambridge. It is about Positive Energy Districts (PEDs) an increasingly recognised approach for supporting the transition to climate-neutral cities. No prior knowledge is required.https://cambridge.eu.qualtrics.com/jfe/form/SV_0uDVNQL3SgLibt4

Structured Academic Summary

BIM in Asset Management June Monthly Forum (10 June 2026)

1. Overview

The forum focused on improving asset information management across the asset lifecycle, with particular emphasis on Progressive O&M (Operations and Maintenance), asset validation, digital verification, artificial intelligence (AI), as-built information, and the development of a robust Golden Thread of information. Discussions highlighted persistent shortcomings in information handover, the need for continuous validation throughout project delivery, and emerging approaches that combine information management standards, automation, AI, and reality capture technologies to improve compliance, operational readiness, and asset performance.

2. Progressive O&M and Asset Information Management

Current Challenges

Paul Oakley highlighted that asset information is frequently treated as a deliverable assembled at the end of construction projects rather than as information that should be progressively developed throughout the project lifecycle. Common challenges include:

• O&M documentation being compiled only at project completion.
• As-built information often reflecting design intent rather than what was actually constructed.
• Significant information duplication and version control problems.
• Facilities management teams receiving incomplete information while already operating assets.
• Inability to demonstrate a true Golden Thread of information despite regulatory expectations.

Progressive Information Delivery

A proposed solution is the development of information progressively throughout project stages:

• Defining Asset Information Requirements (AIRs) at project inception.
• Establishing COBie structures and product libraries early.
• Identifying asset types and information requirements during design development.
• Using Task Information Delivery Plans (TIDPs) to define responsibilities and required deliverables.
• Capturing compliance evidence continuously.
• Automating digital asset capture during construction.
• Ensuring information is substantially complete before handover.

Core Principles

The proposed approach is based on:

• Information delivery throughout the project lifecycle.
• Clear accountability for information provision.
• Continuous validation rather than end-stage verification.
• Linking information to assets, systems, and compliance requirements.
• Ensuring accurate representation of what has actually been built.

3. Asset Validation and Verification

George Stevenson outlined a four-stage asset validation process designed to improve confidence in asset information.

Four Validation Stages

1. COBie Validation
   • Automated checking of structured data.
   • Immediate feedback to information authors.
2. O&M Verification
   • AI-assisted review of O&M documentation.
   • Assessment of document completeness and content quality.
3. Reality Capture Verification
   • Use of 360° photography and digital capture technologies.
   • Confirmation that installed assets match documented records.
4. Regulatory Compliance Validation
   • Verification against regulatory requirements and compliance evidence.

AI-Supported O&M Review

The forum described the development of AI-assisted processes that:

• Review specialist contractor submissions.
• Assess content completeness.
• Generate corrective feedback.
• Support subject matter experts rather than replacing them.
• Create baseline asset registers from combined BIM and O&M information.

4. Workstream Coordination and Industry Collaboration

Discussion explored the growing number of specialist workstreams within the BIM4Housing community.

Participants acknowledged:

• Significant overlap between Progressive O&M, Verification, Product Technical Specifications (PTS), As-Built Information, and Asset Management initiatives.
• The need for focused, task-based workstreams while maintaining overall process integration.
• Risks associated with siloed thinking when information management challenges span the entire asset lifecycle.

Emma Hooper introduced Project Colin, an award-winning initiative that seeks to break down information into reusable components, allowing information to be specified once but reused across multiple compliance and operational purposes within a Common Data Environment (CDE).

5. Artificial Intelligence and Information Management

Evolution of the AI Workstream

Tim Aikin reflected on the development of the AI working group, noting that early discussions focused on:

• Understanding Large Language Models (LLMs).
• Building awareness of AI terminology and capabilities.
• Exploring trust, accuracy, and governance concerns.
• Sharing practical implementation experiences.

Key AI Themes

The primary concerns identified were:

• Trustworthiness of outputs.
• Accuracy and explainability.
• Appropriate use of context and guardrails.
• Understanding distinctions between AI technologies, including:
  • Large Language Models.
  • Retrieval Augmented Generation (RAG).
  • Vector databases.
  • Object detection and object recognition systems.

Participants emphasised that AI systems require contextual information to avoid generating misleading or inappropriate outputs, particularly when handling regulatory or compliance information.

6. Defining As-Built Information

A major workstream led in conjunction with the Construction Leadership Council (CLC) is attempting to establish a clear and consistent definition of "as-built".

Identified Problems

Participants reported:

• Different interpretations of "as-built" information across disciplines.
• Poor information quality during occupation and asset transfer.
• Lack of clear expectations regarding information deliverables.

Emerging Approach

The group is:

• Defining what information building operators should reasonably receive.
• Establishing minimum verification and validation standards.
• Developing methods for collecting and maintaining this information.
• Focusing first on desired outcomes before addressing implementation mechanisms.

7. AI-Enabled Metadata and Digital Evidence

Tim Aikin presented developments in automated image metadata capture.

Proposed Solution

The system:

• Uses voice recognition to describe photographs.
• Applies natural language processing to identify locations and asset types.
• Embeds contextual metadata directly into image files.
• Records location, asset, building, and project information.
• Includes GDPR checking and face detection capabilities.

Benefits

This approach aims to:

• Improve traceability.
• Strengthen the Golden Thread.
• Reduce manual data entry.
• Support large-scale management of photographic evidence.
• Enable integration with BIM and asset management systems.

8. Emerging Themes Across Workstreams

Several recurring themes emerged throughout discussions:

• The Golden Thread remains a central objective.
• Information quality remains a greater challenge than information quantity.
• Validation must occur continuously rather than at project completion.
• AI should augment rather than replace expert judgement.
• Asset information should support operational outcomes rather than simply satisfy handover requirements.
• Stronger integration is needed between design, construction, handover, and operational phases.
• Data interoperability and system connectivity remain critical enablers of effective asset management.

 

Key Insights & Action Points

Key Insight	Implications	Action Points
Asset information is still predominantly assembled at project completion rather than progressively developed.	Creates incomplete handovers, information gaps, and compliance risks.	Develop Progressive O&M methodologies and establish information requirements at project inception.
The industry lacks a demonstrable Golden Thread despite regulatory expectations.	Organisations struggle to prove compliance and asset history.	Embed information capture, validation, and accountability throughout project delivery.
As-built information remains inconsistently defined across the industry.	Different stakeholders have different expectations and interpretations.	Support industry efforts to establish standardised definitions and minimum verification requirements.
O&M documentation is frequently incomplete or inaccurate.	Asset operators inherit unreliable information.	Implement structured O&M verification processes and AI-assisted quality checking.
AI offers significant opportunities but introduces risks relating to trust and accuracy.	Poorly governed AI could generate misleading outputs.	Develop governance frameworks, contextual controls, and expert oversight mechanisms.
Continuous validation is more effective than end-stage verification.	Errors become increasingly expensive to correct late in projects.	Introduce validation checkpoints throughout design, construction, and handover stages.
Digital reality capture provides a practical route to validating installed assets.	Creates evidence-based asset registers and compliance records.	Expand use of 360° photography, digital twins, and automated verification workflows.
Metadata-rich digital evidence can strengthen the Golden Thread.	Improves traceability and information retrieval.	Standardise image metadata structures and integrate them with asset information models.
Workstreams are increasingly interconnected.	Siloed approaches limit information value and operational outcomes.	Strengthen coordination between O&M, verification, AI, building safety, water, and as-built initiatives.
Operational requirements should drive information management requirements.	Information is often collected without consideration of operational use.	Define desired operational outcomes first and work backwards to establish information requirements.