AI AND DATA ENGINEERING SPECIAL WORKSHOP MEETING, 27-08-2025
GEORGE STEVENSON So what we really want to do is use this as an opportunity to come up with things that could be done to improve information. So, I’m going to start off by throwing it out to the floor to ask is anybody using AI at the moment, I’m sure you are in some ways. And from an asset information perspective, any hands up for that please. That surprises me a little bit.
MO SHANA’A Looking at it for a few things you use, not really using but exploring is more accurate. First we’re doing things when it comes to making information more accessible broadly speaking. So, if there is guidance or texts that’s hosted on the platform that we can use AI to ask questions similar to ChatGPT. We’ve done that with the UK BIM Framework, that’s a very good example of that. But then beyond that we’re looking at using agents for a number of things. So things like extracting data from some of the PDFs that get uploaded.
A very simple example that we’ve recently tested is I have a title block in a drawing and I want to compare that to the metadata that’s in the CDE, or that’s in the information delivery plan. So, we’re looking at it there. But then also we’re looking at things for planning purposes. So, we’re exploring a few use cases.
GEORGE STEVENSON So in that particular example Mo, what you’re doing then is looking at alignment of information that’s in the drawing as opposed to what’s on schedules. Is that right?
MO SHANA’A 100%.
GEORGE STEVENSON Tim’s been doing some stuff in that area that we can perhaps touch on. It was part of what he covered in the previous one, but that’s the sort of thing. Is anybody else using AI? Justin, are you using AI?
JUSTIN GILLARD I’m actually not using AI at the moment. I am interested in how it can be used. Verification is a big thing that I would probably use it for to compare what has been mandated versus what has been handed over. But yeah, that’s part of the reason why I’m here is to hear use cases and other examples where people are using it.
BASIT IBITOYE I’m representing a company called Parametrix in Birmingham. We’ve been using AI for some sort of data validation because we’ve bene populating and trying to make digital twins, like intelligent models. So, we’ve been populating all taxonomy data and asset specific data as well. We’re taking it to other platforms like Power BI and after we’ve made sure that we, because we made a template to just verify and validate the model as well. So it gives us an opportunity to integrate Ai either by Azure, we use Copilot and we’re using Azure with at least a 2F SKU capacity or something like that.
But that might take a little bit of costs, so we’ve been looking at other alternatives by using calling APIs, ChatGPT or Large Language Models to basically interact with these models in a way that you could just ask a question like how many fire doors do you have in this building? And basically restore?? 4 mins 17secs view manually, go in to investigate it just visually it shows you all of the information that you need. And along with a table as well that gives you other specific asset information that you need, including manufacturer’s information, the location of whatever the asset is.
For example doors, you need to know what kind of hinge is on the door, if it has a closer, what kind of glass is on the door. So it gives you all of that information as well because we’ve populated, we’ve done all of the manual work behind the scene. So we’re just integrating AI to do some sort of deep analysis and extract information in a simpler and faster way.
GEORGE STEVENSON That sounds very interesting. I’d be interested to understand a bit more about the wider process that you’re following there. Because by the sound of what you’re doing, I think it’s probably the only practical way of doing it, is you’re curating the information before hand.
BASIT IBITOYE Yeah, we basically do that on the BIM auto and software, we use Revit as well, so we do all of the COBie information and everything. We organise it in a structured format where we could use it for further, like during the operational phase of the building and integrate it with other common data environments as well. So. we’ve been playing along with Siemens, we’re trying now Vutec, we tried Tandem as well. The origin of the process is to basically organise all of the information via a BIM auto?? 6 mins 38secs software. In our case we use Revit, so we populate all of the data on there. We make sure that it’s all structured and it’s interoperable to other platforms like to other FM systems and assess that.
We’ve created like a custom template, something of the sort through Power BI. And it enables us to basically interact with this model with little cost of licensing. You can physically open it up on your tablet or something that gives you that easy access to basically validate your model. And the opportunity to integrate AI as well for deeper analysis and easier access to information.
GEORGE STEVENSON I’ve got a couple of questions for you Basit, but can I throw it open. Has anybody else got any questions that spring to mind about what Basit has just been explaining there?
JUSTIN GILLARD A practical example at some point would be interesting.
BASIT IBITOYE I don’t think I’m in the position to really share so much at the moment, but I could talk about it. The reason I came on this platform is to understand what other parties are doing or how are they integrating Ai to at least learn from the workflows that they’re implementing. And basically contribute to tangible conversations like this.
GEORGE STEVENSON Absolutely Basit, I think you said you didn’t attend the session that we did about a month ago. I’ll make sure, you should have a link in the email that I sent out this morning so you can take a look at that. As Active Plan, we had three of my colleagues on the session and they were explaining what we’re doing in Active Plan. So you can see the sort of methods that we’re doing there. And i’m sure that other people like Mo will show what he’s doing as well because they’re doing some great stuff. I’ve just got a question for you. One of the things that plagues us is the quality of information that is being brought together that is being authored by the various different parties. Do you also experience challenges with that?
BASIT IBITOYE My opinion is it’s going to vary across different organisations. The approach we’ve basically taken on our end is to standardise the population process and the structure in which the data is being exported. So over time we have an immense library that we could use for more AI training. So basically the AI agents could train better because it's structured and it follows the same data schema.
GEORGE STEVENSON And are you doing this on existing buildings or on new buildings?
BASIT IBITOYE We basically do Scan to BIM, so existing buildings in well in which they are trying to do more intelligent facility management on.
GEORGE STEVENSON OK, so largely you’re the authors then.
BASIT IBITOYE Yeah, we basically implement cutting edge? 10mins 25secs technologies, tone scanners, laser scanners, we go in doing ? 10mins 30secs surveys. So we basically, we try to convert the physical information into digital and make it intelligent and interactive. So for other future case studies. And we’re hoping to be able to use it for more scenario based decision making in the future.
GEORGE STEVENSON Thanks for that. Take a look at what we showed last time and then you can take a view as to how much you…is your Director Alexander?
BASIT IBITOYE Yes. I’m sure he’d be able to shed more light on this topic, probably better than I do.
GEORGE STEVENSON I’d like to turn it around maybe on asking people who may be the recipients of this information to see what they would actually like. Margaret, would you have any suggestions? Because you’re looking at things for a large estate. Are there things that you would be looking for that could benefit from this at all, as a use case? Connor, you’re managing the information for a housing association, aren’t you?
CONNOR STEVENSON Trying to, yes.
GEORGE STEVENSON As a manager of facility, can you think of things that you would benefit from if the information was more structured and more achievable? Being able, as Basit was saying, to be able to ask the model for information. Are there things that you can think of that would be helpful?
CONNOR STEVENSON One of the things that we’re struggling with at the moment is information retention. I was actually just on a catch up call with the different members of the asset intelligence team this morning. And they were saying that they’ve got one of the line managers sent out a copy of a spreadsheet containing information to two different people. Those two different people made different changes and then sent it to two other people. So there was now at the end of a half an hour cycle, there was 6 different versions of the same spreadsheet. And you can pretty much times that by 25 years is how long he’s been there and they’ve been doing the same thing.
So, I suppose for us it’s always how do we get data engineering and AI to go. It’a a bit more difficult because it’s now kept in those different versions, but how do we get it to go…with the sort of minimal disruption to people’s ongoing processes and procedures and making cultural changes, how do we embrace that chaos? Is there an Ai or ways you can engineer this so that people can minimise the risk of this stuff happening. And then because we’re still at the infancy stage of adopting BIM and the different methodologies, there’s very few models for a lot of houses. Whether that just be 2D information models or 3D information models.
So if there is a way that we could easily link the information we have to an archetype model that would be almost reverse engineer it, attaching that information to a 3D model, that would be really beneficial. But the big thing for me at the moment is trying to tackle the shear volume of the same information and information that’s not getting deleted when it should be deleted. Because at the moment we’ve got SharePoint. And SharePoint has morphed from a project management tool to a full blown main system. And you’ve got 6 or 7 different systems and trying to get them to move back to use this as your project and then move it to the main system. And then keep that there and you can archive or delete that because you’ve got everything that you need. So overlays for information management and things like that would be really useful.
GEORGE STEVENSON I think what you’ve said there really is helpful because I think that’s a problem that probably everybody in this group is struggling with. And we’ve got some solutions for that. Mo with Morta has got some solutions and we have as well. The point is there are solutions to that process. What’s critical though is to actually agree what good looks like and be able then to oversee it and do it, probably, gradually. Because the challenge with any type of data engineering is that you can rationalise things and you can use AI and pattern matching and things like that. The problem is you need to know what the answer is that you want out at the end. And being able to interpret that, because those different spreadsheets have probably been created for a purpose over a period of years.
CONNOR STEVENSON The guy himself, he’s absolutely fantastic at what he does. And even he is like right team, everybody uses the SharePoint versions, so we’re not doing this and you’ll download a copy and send it via email. I don’t know if there is an AI or data engineering solution for not following the procedure that you’re saying should be followed anyway. He’s quite funny in that regards because he’s always the first person to call himself out.
GEORGE STEVENSON So, broadly what we’re trying to do with this group is identify what those problems are and I think you’ve articulated that one very well. I think that’s excellent. Andrew from Pinnacle, have you got any ideas in terms of how your job would be easier? I know we’ve spoken in the past about Regulation 38 compliance and checking. Would it be easier for you if there was a standardised way of being able to validate the information at handover?
ANDREW COULDRIDGE 100%, it’s not just receiving the information, as I’ve said on a pervious session that we’ve had, R38 being the driver for occupation in many cases. And the developer then obviously keen to occupy the development that they’ve just built, get people in. What I’ve certainly found is that the development isn’t quite ready to be handed over, technical areas, plant rooms etc still not having been handed over in any way, shape or form. But the R38 documentation I think for me is the problem. It’s driving occupation and occupation is crucial if you want to generate revenue as a business, particularly if you’re a developer. But what it does mean for us that are left managing the site is that actually we’re forced to take the site because of Reg 38.
And as such we then have to sift through the information that has been handed over almost within a Friday afternoon was a scenario, with a letter saying I need your confirmation that you’re happy with what we’ve sent over because we need the NHBC signed off. A kind of catch 22 situations. So, Reg 38 for me is a problem. The data that comes across is often not, can’t be validated, it’s what I get given. I ask the question is what I’ve got what you asked for? I.e. is it information that was created at the point of planning? Is it as-built information? Does it include test certificates and compliance certificates etc? And I ‘m told yes, it’s only when I sift through it that actually I find I don’t have it all.
I’ve probably got about 80% of what I should have had on quite a large development. And I think it’s shifting the responsibility or the ownership, and the responsibility of validating information over to the managing agent. I want to be able to trust the data that I’ve been given and trust that the client to give me the information that is accurate in accordance with the requirements. I don’t want to be asked can you check it to just make sure that you’ve got everything that you want because that’s a whole piece of work.
So in answer to your question George, it’s a problem. I’m not quite happy with where I am with this right now. We as a business have employed somebody to specifically look through, or part of their job is to look through the information that we receive. And to put some checks and balances in place internally to help manage that process because at the moment we’re just recipients and it’s a download of data. And there it is, knock yourself out with it, let us know what you need or what you haven’t got. And that’s kind of a shift of responsibility and one that I don’t particularly savour or want.
GEORGE STEVENSON Yeah, I think part of the big challenge is that there’s no definition at the beginning of what you should be expecting. So therefore, you get a load of information and you can check what you’re given, but you don’t know what you’ve not been given.
ANDREW COULDRIDGE Yeah, but even that George, what I have been given I’m questioning whether it’s accurate. I don’t know, is the answer. Can I give you an example of that because I’d welcome any feedback on this example. So, a fire safety strategy that I reviewed 18 months ago. A version that was at the time not necessarily crucial at this point, but it was the same version that I received in the Regulation 38 document. And in that fire safety strategy the narrative recommends this, it suggests that. What I want in the final draft is that it is this and it is that. And I’m not seeing that and I’m unsure how to translate that. The question there I raise is have I got the same version that I had 18 months ago? Has anyone reviewed it since? And is this just an exercise of just transferring a document over as part of Regulation 38? Which I believe it is.
GEORGE STEVENSON So, who’s responsibility would it be to say this is what is actually happening?
ANDREW COULDRIDGE I believe it to be the developer, those who are responsible for the principal M&E who built it. Or the fire safety officer or the company that have built the fire safety strategy, in my opinion. To be able to update and provide me with a tracks changes document or something that gives me version controls.
CONNOR STEVENSON Then it’ll be covered in a RACI matrix at the start, like pre-agreed, is that worth doing? A RACI matrix, a responsibility matrix. One of the things that we’re implemented which I’m quite sure is 19650 is a responsibility matrix. There’s normally a couple, one is for tasks and one is for documentation and who is responsible for actually reviewing the documentation. That sounds like something that might be useful to be agree pre-work so you can chase them up following it.
ANDREW COULDRIDGE Yeah, a RACI matrix would be an ideal way to manage that. I’m going to be honest with you, if I was to raise that as a requirement I’d probably be told sorry, we can’t help you with that. But I agree with you wholeheartedly, that would be something that would help massively. If I can say I’m working with developers, they’re big house builders, they’re building large blocks. I’m going to be bold enough to say that across the divisions they work differently, all under the same banner. So, I can have a great experience with one and a not so great experience with the other, although it’s the same company. So I think with inconsistencies probably the key there. But RACI matrix, brilliant, I’d love that. I have worked with local authorities, councils, on housing schemes that we’ve got with them and a RACI matrix is very much part of that.
But they will employ a consultant to manage the project on their behalf and as a result a RACI matrix is part of that. Mobilisation of a large development is project management before it’s property management. And I can throw quality gateway documents, I can throw RACI matrix across the table and it gets thrown right back. It’s just something that actually there is not a requirement for. It’s a work in progress, I guess Connor, is probably the answer to that one. And forums like this and continually asking questions, raising concerns at high level, which is what I’m starting to do, is starting to gain traction. I think everyone is in catchup mode.
When I say everyone, I’m talking maybe the developer community is building. Obviously their business model means that they need to build quick and get people in quick to generate revenue quick and everything else is just holding them up. So, Regulation 38 for me is that one thing that I’d actually like to say, until such time as…let me up it this way, that Regulation 38 actually isn’t the driver for occupation. That would help massively.
PAUL OAKLEY As Andrew is sort of alluding to, obviously the commercial pressures in terms of getting stuff handed over is the big issue. I don’t think a RACI matrix, responsibility matrix etc is going to help with any of these things. Because a very large project that we had which had all of the O&Ms based stuff, the O&Ms were there but basically it was a title page PDF with nothing in it. And similar sort of things where you’ve got what was design intent documentation then handed over as as-built information. But I think that this is some area where possibly AI can actually help in terms of doing comparisons between documents provided at the earlier stages and what was actually handed over as as-built.
But there needs to be some way of doing this verification of the content within documentation and comparing that against earlier versions of stuff. Because as we know, the typical approach is dump tons and tons of information that is impossible for somebody to go through and actually verify. So therefore they’re just going to basically tick the box that yeah, it’s there. But it’s what is the quality within that information that is the problem.
MARGARET SACKEY This is such a fantastic discussion, absolutely brilliant Reg 38 information, O&Ms, health & safety files, what we can use AI for. As you rightly know in UCL I’m a health & safety manager, I’m a civil engineer and so on and so forth. I work with BIM, I chaired the Institution of Civil Engineers when we brought in BIM for health & safety with Gordon Crick’s group. And this whole thing, what’s missing in all of this, I don’t know how many clients are on the call. Can you lift your hand if you’re a client manager, if you’re in a client organisation? You’re a client manager, you work for a client. None at all?
GEORGE STEVENSON No, that’s not right. The client organisations here, we’ve got somebody from Southern Housing, we’ve got Adey from Hexagon, we’ve got certainly at least half a dozen or so.
MARGARET SACKEY So why I’m asking that question is because it all starts with the client. If the client doesn’t set out the right requirements, we all know what employer’s requirements are, employers information requirements it used to be, now it’s employer’s requirements. If they don’t set them out and they don’t enforce or really actually ask for it and make sure the consultants give it, you’re going to get nothing. And OK yeah, we get health & safety files half baked, whatever. Sometimes it takes a year for us to get an adequate health & safety file. And it’s in the ERs, it’s in the contract, there’s money held back and all of this sort of thing and still you’re having this nonsense.
My issue is clients. And I said this to Gordon Crick, if you don’t challenge clients we’re going to get you ask for apples you get oranges. Why oh why are our clients not standing up and saying this is what we want, this is how we want it. We want it now, we want it before practical completion. All our procedures say we must have the health & safety file before practical completion, four weeks before, whatever it is, we must have this information. Reg 38, health & safety file, O&Ms, as-builts not just as-constructed, as-built drawings. You can’t always get it, it’s always going to be snagging. There’s always going to be work done after the fact, we all know that. But at least start as you mean to go on.
And the real issue is client managers. If client managers, I’m working on projects and I’ve heard consultants in the background say oh, they don’t even know what they want, so we’ll give them what we think they need. And these are consultants that have been paid millions of pounds to do the work. All I’m saying is that we know what we want. Nima is there, the UK BIM Alliance, all of these things are there, CDM. I did a project for the Health & Safety Executive, the importance of the health & safety file. And we all know CDM, health & safety file, legal requirement, how many organisations actually have it? How many actually feel it through? I’m not sure. But my issue now is, I think Andrew was saying about Reg 38.
I checked the fire safety strategy for some buildings and a fire safety standard for our organisation. They’re not created by people who understand projects, who understand the terrain. You’ve got people who are not really in the right square peg in a round hole doing this. So then how are the client documents going to be right and fit for purpose? And then how are you actually going to get the right information in the end? Yes, we’ve got AI, we’ve got CDs, we’re procuring a new CD ourselves, but we’re not really using the right people, HSE speak, the right people for the right role at the right time. And even though we’ve got contracts, and even though we’ve got requirements, we’re not seeing it through.
The only projects that I’ve really been happy with at UCL is the PEARL project where we had a model, we had BIM for health & safety, ISO 19650, 1192 Part 6. We followed it through, we had the health & safety file, we had the hazards in the BIM model transferred from the design into construction, into the operation & maintenance. Operation & maintenance contractor using the model and updating the hazards. So, we had two successes. That’s out of many, many other projects where it’s half baked, the client manager doesn’t know what to ask, doesn’t know what he wants and leaves it to the consultant to tell him what he’ll have at the end of the project. I know we’ve got contracts…
GEORGE STEVENSON Margaret, one of the things that we’re trying to do is apply what you’ve just said there to some real situations. So for example, we’re working now with Morta. So this is two technology companies working together to come up with a solution that we can then also bring other software companies into who are interested in the same outcomes. And one of the key things with that is to actually make it simpler for clients to pick what they need. Because at the moment the biggest challenge I think for supply chains is that they’re being asked to provide different information for the same type of building by different people. So, that means that all of the work that we could do to do automation becomes much more difficult. It’s difficult to adopt.
TIM AIKIN Yeah, just to sort of qualify what you said, you said they're being asked for different information by different organisations. And to me and I think you've got several slides particularly on the boilers? 37mins 16secs that demonstrate this. It’s the same information, just articulated in a different way. And part of what going to standard schemas was about is to provide a standard way of delivering all the information and then you extract what you require. And I think there is obviously the historical problem of articulation. The…which then they say the building owner doesn’t know what they want. They do now what they want, when they articulate it it’s never understood by those who are actually delivering it.
The second thing is the fact that if you don’t provide people with the tools to be able to articulate and then automatically check…then you’re never going to get a good outoome. So, unless you adjust those fundamental things, which we are doing and technology is now enabling it to happen in a far more automated way and that’s where ?? INCOMPREHENSIBLE AUDIO 38 mins 25 secs. It’s never going to change and we just go around in circles. I get extremely frustrated when people start dragging me back down the holes of filling in massive amounts of data.
GEORGE STEVENSON What we’re trying to do is something for schools. We’re starting with schools because there is a high level of standardisation that the clients have put together on that. That’s something that if anybody is interested we’ve got a workshop on tomorrow. If you’re interested in that sort of methodology, we’re not simply going to be doing it for schools. The idea is to get something right that can then enable supply chains to provide their information in a standardised way. Do you want to explain that, Mo?
MO SHANA’A I think there are two ways we’re looking to go about it. The first is there are a lot of common requirements within certain sectors. Take something like the Department for Education, there is a list of documents that are already required. We can then go deeper off of the back of that and look at how are people actually delivering the information today and how can we make that more efficient. And look at making that process more efficient by taking a very, very specific set of requirements. The other way we can potentially look at it is once they’ve seen that that structure starts helping, we can then go what are the common activities that organisations have to meet across sectors.
For example, if I’m trying to follow the RIBA plan of work, this is the set of documents that I would probably need to deliver for the RIBA plan of work. If I’m trying to follow something like the building regulations and the approved docs, this is the list of evidence that I would need for that. This is the list of evidence I would need for Regulation 38 and I would go from there. And the combination of working on sector specific things as well as industry wide mandates and best practices, I think gives us a then very detailed library of the common deliverables, both document deliverables and data deliverables, that organisations would be making to make these things more efficient and standardised.
GEORGE STEVENSON I think the other thing with that is the connectability of things as well. Because there’s two things really. One is validating that, as Paul was saying earlier, that the document that is being presented and provided which may be the tick in the box to say that it’s provided, but actually it’s rubbish. Because the information itself in it isn’t right. So, that’s one aspect, but also it’s also the matter of connectability. Because being able, and Basit was talking earlier about ironmongery and the information that’s related ironmongery to doors, for example. Now, because Basit’s creating models themselves you’ve probably got those things properly related. But the ironmongery was probably coming from a different source than the doors.
And therefore that’s the sort of thing that Tim’s been doing work on, to tie together the ironmongery with the door from its original source. But you’re then mapping and relating information together so it becomes more meaningful. And also then relating it to a floor plan that’s got a drawing of where that door is and the symbols on it. So, that’s the sort of thing that I imagine Andy Cumberpatch from FCS because what FCS do is fire door inspections, for example, and remedial works. And Andy, I’d like you to just say something on that to explain what you’re doing in a minute, but I’ll hand over to Pierre.
PIERRE VENTER Just a couple of notes that have popped up in the recent past. One of them is standardise, standardise, standardise. The other one that comes directly from that and that’s the discussion there Tim about identifying the needs from the appointing party, from the client. And the only way that you can actually do that is actually get everybody together around the table on what do you need. Because if you haven’t defined that you’re already on a thing where assumptions and the expectations are starting to fall apart. The other part is I think, and Mo, one of those things is that at the highest level the documents need to cascade down. So, from where we do reporting and things like that, those things that are regulations must be there as a sort of Bible in a sense.
As you come further down as you deliver these different forms, they all are actually coming from predominantly the same data requirements, or the same data purpose. And it’s actually quite interesting just to map that. Here is a standard form that RIBA requires and how does that translate into the various different parties in different organisations within this thing, particularly as George was talking about the schools. And the other part is something that I found and it’s a practical part. It’s a problem in Revit in that it’s not a really big robust database. And when you take something like Morta, if you can segregate the data but keep the primary key, a GUID connected in the model, it then allows all of the visual part to work with the data being managed by the proper tool outside. And it’s getting that glue between the two. So, that’s just my opinion on some practical sides that I’d like to see relating to how the model is actually going to work.
GORDON PRINGLE I would echo Margaret’s earlier comment, there’s some brilliant contributions here and it’s testament that you’ve pulled it together George, and that everyone is expressing their thoughts. I just want to give you a wee bit of an insight as to what we’re specifically doing in the sector. We’ve got a bit of an intelligence sensor that looks at operational hydraulic systems, so cooling and heating systems. And it’s generally being inserted in the system at the commissioning stage, so going back to one of the earlier contributors, I think Andrew was saying that commissioning handover is too rushed, there’s no time for that process. So, how is the completed project satisfactorily being delivered?
And of course if it’s rushed and if it’s on a timescale that everybody is very tight, should there be problems along the way they don’t really want these problems to surface. And they certainly don’t want these problems to become apparent to the clients because that just rockets it up the chain of disaster. So, an intelligent sensor, or an intelligent probe that systematically delivers exactly the same data every 7 hours looking at the operational asset. And graphically represents that in the manners that it wants to do, and regardless of what that process is. I’m not banging the drum for this particular product, but I’d be delighted to have that conversation with anybody. But going back to systematic data, it takes away this anomaly with human behaviour that they will describe things differently, unless you give them a drop down.
But you cannot give them a drop down to every field of entry that you want them to enter data for. They’ve got to think logically. So, is that down to the manufacturers? Is that down to the training process for input data management? That’s some of the challenges. But you’ve also recognised, you’ve got sooth sayers along the way that don’t want this to fly. They literally put obstacles in your way, they don’t want this to fly. We see it every single week. They obstruct us because digital data management in itself, live system monitoring and data management becomes disruptive. There’s so many things that during the conversations I was thinking that were triggering so many aspects of it. Where does building management systems fit in here? Have we forgotten about BMS systems?
Because everybody asks me about my monitor, why is it not BACnet or Modbus? I’ll tell you why it’s not BACnet or Modbus, because it’s antiquated technology. And actually I don’t think there’s many people now looking at the BMS systems because they’ve become, what’s that actually telling you? It’s just telling you what everything sits at. It doesn’t necessarily, or it’s giving you far too many alarms because it’s just triggering on somebody’s decree of when it should be best. If it’s not reconfiguring that alarm every minute, or you’re saying to people if it’s a one off alarm, don’t worry about it. Sometimes I get a headache. I don’t take aspirin because I’ve had three pints of lager the night before. I know why I’ve got the headache.
So at the end of the day stop drinking the lager or live with the headache. Obviously don’t do it every night because you’ve then got a problem. That’s the kind of logic. And if the device in the system tells you that, then it’s a pretty stupid person that doesn’t adopt it. And just finally what Pierre was saying. Where we’re getting big success is we’re getting people saying I want approved technology. If I had a pound for that I’d be a millionaire. It’s like hold on here, I can show you all of the clients that are buying it and doing exactly the same thing trying to prove the technology and it takes them so long to get there. But you suddenly have boom, a client that says that shows me my good, my bad, my ugly. Green, amber, red.
I can concentrate on my red, I can leave my green alone and think about the amber. Give me it on every single system across my whole estate and let me find out where I am. And that’s for your building information management with systematic process. I think can be pretty exciting.
GEORGE STEVENSON I’ve just got one other point I wanted to raise and that is who’s responsible for making sure that the information that you’re getting, Andrew, for example, is getting at handover, has been validated? The reason I’m asking this question is that certainly as far as HRBs are concerned we’ve got principal designer building regs and principal contractor building regs. And I think it’s still reasonably unclear as to what the duties are, the responsibilities are of those very responsible people.
I was talking with a principal designer last week and he was basically saying, and I think he’s right, his job as principal designer building regs isn’t to know whether the building regs have been complied with, it’s to manage the process that raises that question for people to validate that it’s been done. Who’s responsible, for example, for checking that the Regulation 38 information is being handed over? I would have thought that would be the principal contractor. I don’t know. What’s the thought on that?
GORDOM PRINGLE It’s only going to be there if it’s in the EIR. If it’s not in the scope you ain’t getting it. If it’s not on the drawing and it’s not in the schematic and if it’s not in the bill, some of the jobs nowadays there is no bill of quantity. So what you’re actually only getting is what’s physically on the drawing. And if that’s drawn or if it’s a systematics wrong it, I’d be even more cynical. Some of them will build it knowing that it might be wrong, but they can get paid to correct it.
CONNOR STEVENSON I’ve worked for both sides. I’ve worked for a contractor, all we were doing was delivering energy efficiency measures.`I was heavily involved in making sure we were providing all of the information to the client. And the amount of times there are like ah, just give them that they’ll not notice was incredible. No, no, we never told them we’d give them three, it’s written in the contract that we need to give them three of these. No, no, we only ever give one, it’s in the contract we need to give them three of these. No, just give the one.
GEORGE STEVENSON So as Margaret was saying earlier, it comes down to it being in the requirements. We need to make sure those requirements are clear. Tim, would you be able to show some of the new stuff that you’ve been doing on products?
TIM AIKIN (shares screen). One of the things that we've been working on for years is extracting data from to fill in schemas. And one of the problems obviously has been how do we do it. So typically information comes from a variety of people saying that they’re going to fill this that and the other in. And eventually, we originally started with a robot system. These are the problems that I do, all of this information does exist, it just is in a big pudding or baked cake that we’ve got to scoop some out and construct it. So the way that we’ve approached that, or we used to approach it, is we’d say oh, there’s a webpage, what are we going to take off it? Well, first of all if you don’t know what you want you don’t know what to take.
So that’s the first important thing, what do you actually want? And that’s where the schema comes in because…this is the information that you should be capturing. So, these are three sort of areas. You either do it manually, you either do it with a web scraper and now you can do it with an LLM which s a Large Language Model. People are calling it intelligent Ai and I just call it a really clever smart processor. So, the problems with manual transcription is it’s very costly, it takes an awful lot of time. And if people aren’t clever enough they’re not actually going to get what they need and they’re going to get very frustrated when they can’t find something.The problems with the web scraper is that every time you point it at a website you’ve got to programme it. And then when the marketing team changes that website you’ve got to redo it again.
Normally you don’t need to be doing these things once, but new products keep appearing, new products. So we get a sort of turnover on products within the construction industry of about five years. So that's the sort of churn rate you're talking about. So, if we’re talking about LLMs, it’s not just a thing to say give me this information unless you tell it what you actually want to do. And how then you minimise this down so it’s not going to cost you a lot of processing. So the key of it all is calculate the properties and how you then get that data back to the standard way. And that's what you do is you define a schema. Part of the schema is you'll note here that there's a lot of things called null. So you make sure that if it doesn't know, put it as null, don't come back and give me stuff that you don't know. And you also put in rules.
Now this isn't the schema we're using, but this is a sort of simplified version of it. And then there's certain things that can inflate the…So, what happens is that you get the schema, but you also reduce this web page or PDF file down to what's called the markdown language and that just reduces the content that you're going to send for processing. And some of the stuff you just need to weed out such as people can put in a base 64 image and that will really inflate the data package. You just get those sort of things out.
So this is what we've done as a sort of prototype. We've got something where we can construct it and then we simply put the URL in, tell it to process and we're going to do this live because it can take up to a minute. And then we can then view the results. And I've just in that process, I would just put, yeah, I love that, I love that, I love that. So we've got a few things that we can grab off.
This is what I meant by being able to look through and pull off the bits we did. So we have to train a robot to do that. But now an LLM will actually do that automatically or just analyse it and say, OK, based on a series of rules and things, I think this is what it is. And we can actually get back some reasonably good information. I put this together very quickly because George asked me yesterday. So, if we just go and look at one that's already been done, then once we've processed it, we can actually look at that. I'll see what the markdown looks like. That's quite interesting as you can see, it's now simplified.And that that's pretty critical if you're going to be sending this and you're going to be either you process it.
We're using an approach that we do ChatGPT at the moment, but I'm going to try against a Gen X? 1hr 15secs. Just to see whether, you have to work out which models are going to be faster. And then there's another. There's ones that you can run yourself which are a Llama, so that you can run it in a headless environment which is the user is named the ID user when it is a web service or whatever.
GEORGE STEVENSON Tim, can I just interject for a moment. Just to explain, what we’re doing here is we’re creating basically a product library. We’ve got a product library already, so this is an industry resource. I’ve already spoken Gordon, that you’re welcome to put your information into it. What we’re trying to do here is make it so that the product information itself is only curated once and then it can be used many, many times, So, if any of you are interested in knowing more about the free product library. Like I say, we’ve been developing it over the last 5-6 years and you can put the information into it, put product library into the chat. And we’re also tying this together with what Mo is doing in Morta and others as well, so that we can then connect things together.
RICHARD FREER George, is that something you should maybe be talking to people on a one-to-on basis, a quick 10-15 minute chat? Would that be better maybe, rather than having a group meeting?
GEORGE STEVENSON I agree, I think this is something that does need to be…I think there’s two levels really. If you’re interested in it then just put ‘product library’ into the chat and we can show you some stuff. But I think the one-to-one is more looking at exactly that, to be able to say how could I use this. Because what Tim is showing you there is some quite clever stuff that we’re using to try and replace…at the moment what happens is that this information is all put in manually by the subcontractors when they go through the technical submittal process. So, what we’re trying to do is automate that process so that that isn’t done as manually.
RICHARD FREER So people are to put in the chat ‘one-to-one’ for them to do that. So, if you want to have a talk in some depth and more tied to what you’re actually focused on and what you’re actually looking for, put ‘one-to-one’ in the chat and then we can send you invites out and you can meet up with George for 15 minutes and go into something a bit more focused on what you require yourselves.
GEORGE STEVENSON And if you do that we can give you free access to the environment. Broadly we’ve got about 7 or 8 thousand products in there at the moment.
TIM AIKIN To me this is the most important thing about the product library application is that you have many instances of the product library application, so they can be fairly boxed or they can be shared. But the important thing is that they can talk to each other and they can also grab information from other sources through…so there are things in there that do that. Because it's it was really great for curating the information and then tests it against the requirement on a project. And that to me is the most important thing, rather than just people…I don't know whether the library is the right thing to call it, but it's the ability to be able to check it against project requirements. So if somebody's only asked for five attributes or five pieces of information on a product, you don't spend time creating 50 because it might be the same five that you used on the next project.
And if it's another few more, then you add those in. And there was great opportunity for then mapping a sort of standard schema so that this information that manufacturer A, B and C will all produce the same sort of thing, but all call their attribute information differently but it's the same. And then automatically be mapped using AI to do that rather than individual because nobody wants to do it. And so and it's just something that we should make and we should be working towards that.
GEORGE STEVENSON The purpose of this session I think has been satisfied, we’ve had some good input from people as to what problems they’d like to solve. If you can think of anything else, maybe talk to colleagues about it, please share that, message us with some ideas. And then we can keep together as a group and come up with some solutions. Tomorrow we’ve got the schools workshop. Was this session useful? I can see a few nodding heads. We are going to be focusing on schools, but the outcome of that is going to be much more standardisation, as it were. So, that’s the objective. Have we got anything next week?
RICHARD FREER Next week no, we’ve got the A, B and Cs, the roundtables on the fire critical assets, that’s the following week.
GEORGE STEVENSON Sorry if this has been a bit chaotic, but I think having something loose like this was really needed for people to be able to contribute, rather than have something sort of too prepared, as it were. We’ve got the safety critical asset sessions coming up, so if you’re interested in fire doors, AOVs, things like that, you can contribute to that.
RICHARD FREER If anybody is interested in those, what we’re going through is the guidance that we did a last revision a couple of years ago. So, it’s a check on that to see if there any things that need to be updated or changed in light of events of the last few years I guess. So, if you’re interested in that, about 11 I think we’ve got fire critical assets, put ‘fire critical’ in the chat and we’ll get invites out to you.
CHAT
Fireflies.ai Notetaker Mohamma (Unverified)… Talk to Fireflies: (Web-search enabled)
- Say: Hey Fireflies how can you help?
- Type: /ff help
to get started.View real-time notes and transcription for this meeting here: https://app.fireflies.ai/live/01K3GE52FD09MKJG5QJ8ZMP4CB?ref=live_chat
Pierre Venter (External)… The validation data is really important. This data needs to be fixed at the time of entry to reduce double entry and errors.
Andy Cumberpatch (External)… Yes link received in email
Dan O'Gorman (External),, Sorry I need to leave. The battery is dying I will view the Video later good topics, Digital Building Logbooks as Conor Stevenson is seeking is the answer
Pierre Venter (External)… The issue that is Revit, is that the database is limited by volume. Connecting a Primary Key and segregating data to an external platform/Database.This allows the versions, updates and snapshots to be made and reported. The Visual Message is really important to demonstrate the state. A Change Management process is one example.
Sackey, Margaret (External)….. Hi George. Apologies - I
Missed your question. Can you repeat if possible.
Very sorry could not access the microphone or chat on the laptop.
Now Logged onto my phone.
Gordon Pringle (External)……. Systematic Data, Process and continuous monitoring presents issue for many processing of data management.
Gordon Pringle (External)…. Setting the scope in the EIR contractually has always been an issue
Gordon Pringle (External)…. Should CDM Contracts still not embrace Best Practice or indeed Minimum Requirements at least?
Justin Gillard (External)… Information Managers are tasked with monitoring the development and delivery of the information as requested in the EIR and the BEP etc. Please engage Information Managers to police the implementation of the protocols with the subcontractors
Mo Shana'a (External)….. I agree with this but there are two comments I'd make
- There should be a list of requirements per party - I still don't think EIRs cover that well enough in many cases and ideally there should be industry libraries for these requirements (Eg. These are all the documents needed to comply with CDM, REG 38 and so on)
- I think Information managers should be looking at automating governance and tracking
Justin Gillard (External)….. Hi Mo - The projects I deliver have a protocol called a BRM or BIM Responsibility Matrix. This lists all assets to be modelled by LOD and project stage. The BRM is lead by the building owner as they know what assets they require to manage at Operations stage. It works well as a reference in BIM workshop meetings
Pierre Venter (External)….. Gordon Pringle Can the handover, process not be managed as an incremental one. The close-up of components is a requirement, which benefits from incremental process.
Mo Shana'a The validation fields could be stored and used for validation for all elements.?
Mo Shana'a (External)….. Great to hear. Is this based on a standard specification of sorts? Ideally we move away from recreating those project to project & client to client
Pierre Venter (External)… Mo Shana'a The validation data can be compared against the documents published, data and content, and identify the exceptions?
Adey Badmos (External)… Contractor as per ER and contract docs
Sackey, Margaret (External)… Hi All - For those who have not received this - Please be advised of the BSA & Golden Thread Safety Conference in partnership with all London Universities which will highlight how much change is needed in this area in our industry.
Registration details here - The BSA & Golden Thread Fire Safety Conference - Date: Wednesday 3rd September 2025 - Registration: 08:30. Conference: 09:00-16:30