Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
E 24: How to Use AI in Construction from the Head of Planning at BAM Nuttall
71 Plays2 years ago
In this episode, Carlos & Jason are joined by Ivor Barbrook to complete the three-episode focus on AI and construction.
The three dive into how AI isn't changing the art of planning, how to implement new tech in this rapidly evolving area, and the real-life impact of AI and construction that Ivor has seen first-hand.
Recommended
Transcript
Introduction to Industry Leader
00:00:00
Speaker
Hello and welcome to episode 24 of the offsite podcast where we chat all things construction and technology. My name is Carlos Caballo. And I'm Jason Lansini. Now our guest today is someone that I've known quite a long time. He's an industry leader with regards to the adoption of technology and is the head of planning for a tier one contractor, which is BAM, UK and Ireland. There has a real focus driving that adoption of technology across the organization.
00:00:27
Speaker
Welcome to the pod. Over. How are you?
Tech Gaps in Construction
00:00:29
Speaker
I'm very well, thank you, and thank you for having me.
00:00:32
Speaker
Iver, it's a pleasure to meet you, like we've not met before, but I've heard your name and I saw your presentation that you did an event like last year that Carlos and the team hosted. And I think for context of maybe people listening that there's a lot of people in construction, on my perspective, there's a lot of people in construction that are talking about and writing about and saying we need to adopt technology to solve this problem or that problem.
00:00:59
Speaker
And only a percentage of those are actually doing something about it and taking steps. And I would say that for those listening, that I was one of the people that are definitely doing something about it. So it's a pleasure to meet you, Ivor. That's great. It was always an element of talking the talk.
00:01:17
Speaker
Are you actually walking it? And there is the difficult bit. And if you are walking it, how do you actually get everyone to follow you through to actually start to deploy these new ideas, these new technologies? Very easy to keep the status quo as you go forward. Absolutely. So, Ivor.
Understanding AI in Construction
00:01:40
Speaker
Full disclosure, over the past few weeks, we've run a bit of a mini-series on AI. We spoke to Dev, who's the CEO of Nplan, who are, I guess, a competitor to your chosen tool, which is Nodes and Links. Obviously, you use at BAM. And we've also spoken to a guy called Santosh, who's a planning professional based over in Australia, who's a bit of an AI skeptic. So we're trying to get this balanced view of a vendor.
00:02:05
Speaker
a skeptic and someone who actually embraces technology. So I guess to kick off as obviously a tier one contractor and as a leader, what is your view on AI generally? AI is definitely an emerging market and a capability. You've got to be careful you understand what AI is actually doing for you.
00:02:26
Speaker
And we tend to think of AI as a Star Trek element, and Captain Kirk asking the computer to give some information. We are a long, long way away from that. So what is AI actually doing for us? Well, we're training it to look at a particular set of data and to provide some answers back to us from it.
00:02:51
Speaker
in a particular way. And that means that we need to be clear about what is the need. What is it we're trying to actually get AI to do for us? And how is it going to provide that information? Then you need to be clear in your own mind the context in which it's providing that information. But once you've got that clear, it's very beneficial to use AI, because it's providing information very quickly for you.
00:03:21
Speaker
and in a format that's easily and readily digestible. And at the end of the day, it's about humans making informed decisions. And if you're clear where the AI is getting its data and that data is trusted data,
00:03:37
Speaker
and the information that you're getting from that data is freely derived, then it's of great benefit. And we're starting to see emerging on all sorts of platforms now, and the benefit is clear to be had at the end of the day.
00:03:54
Speaker
and whether it's coming from straightforward reporting of progress or reporting of change or anything of that nature, AI can start to look at a depth of history as well as consider the future and the potential future.
00:04:14
Speaker
But at the end of the day, it is information and it's information as in Star Trek. We need to make informed decisions on what we're going to do and how we're going to manage the situation. Yeah, I feel no matter what sort of individual's views are on the use of AI, the construction of AI, I think the unanimous agreement is it's the rapid sort of provision of useful information that can help you make better decisions. I think that's like the clear win for musicals like that.
00:04:42
Speaker
And if we think about you as an organization, obviously first went through a process of, right, this AI technology now exists in the market. Did you sort of initially push that because of like a clear use case that you were maybe sort of pitched or was there sort of this, I guess, sense of pressure from industry that this technology exists and we need to understand it and extract value from it?
00:05:08
Speaker
I think
Evolution of Data Management
00:05:09
Speaker
it's definitely, you know, we need to do things better and we need to do things faster. And there's no doubt about it. Our world has changed in the last 10 years with technology. I've been in the industry some 40 years and I have to say that for 30 years it was rather stagnated. And then suddenly the idea of managing data and structuring data has been a big driving force in the industry over the last 10 years.
00:05:38
Speaker
And I've come back to trusted data that if we start to think about how we use or generate data and how we store that data in a format that is clear to use,
00:05:55
Speaker
And we did that for a human to understand. But that data doesn't have to be so well structured when you get to AI. And so you've got a pool of information. As long as you trust that information in terms of where it's come from, whether it's accurate or not, you can then start to think about how you can chop and dice that information.
00:06:17
Speaker
So you have to be clear about what it is that you're going to get out of it and how you're going to use it going forward.
00:06:26
Speaker
So the industry is almost naturally going forward with the technology and certainly with the advent of BIM as we used to call it, you know, that was sort of 2010-13 as that came out. That's restructured our thinking entirely about software platforms and how we use it.
00:06:50
Speaker
And we certainly moved from that computer that sat on to the desk, almost going back to the early days of mainframes where information held centrally.
00:07:01
Speaker
And we're doing more and more of that, but with platforms providers, such as a Vets and many others where we are holding our own data on trusted third party platforms. And we're starting to have that sensible discussion now of how we actually
00:07:21
Speaker
prize that data and start to use it beneficially and add to our other platforms and systems so that we're getting a mix of data rather than just pure data as an aspect of a singular topic.
00:07:35
Speaker
Yeah, the broad data set is really important. But we had a conversation with Santosh in the last call. And this was in the context of QRSA, so using AI for QRSA.
AI's Role in Risk Assessment
00:07:50
Speaker
And one of the big problems was only looking at scheduled data. So we lack a lot of context that you could get from other systems and other data sets.
00:07:59
Speaker
Yeah, it's definitely quite an important aspect to this at the moment because we need to be looking broader. We need these, yeah, you want a platform to understand everything rather than small sort of slices of data across different systems within the organization. I guess that does kind of lend itself to the vision of BIM. You've got to be careful here because you might use AI to look at a whole host of programs in the past.
00:08:26
Speaker
And to understand really why have they changed, what activities in that program have been extended, how contract completion dates changed, how plan completion dates changed. So there's a lot of things that have gone on in programs historically. And yes, you can apply AI to that.
00:08:46
Speaker
And you can say, right, OK, here's my current working program. We're just about to start this job. And what sort of time risk do I need to allocate to this program?
00:08:58
Speaker
And you can get the AI to look at past, and it will look at the descriptions, how you ordered your activities, and compare that to your new program, and generate some time allowance, because it now begins to recognize that because of the order of events in the program, we're putting CFA piles in, or we're doing some marine piling, or whatever. It understands what it is that you're doing, but it doesn't understand the context of the change.
00:09:29
Speaker
Now, it could be that the piling you had 10 weeks in there, but it took 12 weeks on average when you look at history, but it could be for a multitude of systems, anything from actually we got something like 20% more work instructed, or we found some unforeseen ground conditions, or production rate was wrong. You don't know quite what the context is with all that.
00:09:56
Speaker
So that sort of AI, how do you use that? What sort of context do I have? And as a contractor, yes, I can use that, but I've got to be careful here because it's not only got my uncertainties and my risk in there, it's got the clients' uncertainties and risks in there. It's also got third-party uncertainties and risks in there. So you go, okay, this is actually a great tool to work with a client with to understand our global risks. But for my own risks,
00:10:25
Speaker
I then need to think about using QSRA in terms of actually analyzing the uncertainty in my own program and coupling that with risk as an analytical process, then using Monte Carlo to understand how that potentially might impact my program. And that gives me a standard distribution curve, which enables me to understand my risk.
00:10:53
Speaker
my potential time risk. It's quite interesting to compare the two, because you can start talking with a client and say, well, historically, this is the time we need to allow for a program. We've done our own analysis, and we believe we have a time risk of why your x is much bigger than y.
00:11:14
Speaker
So we could work collectively together to start managing those risks. And so we are more and more getting into a cooperative approach to project delivery and our clients will get more and more involved. So if you like this is the next step of the journey as we go forward. Absolutely.
00:11:36
Speaker
So on the theme of, I guess, the journey, within the AI space, so if you think about AI applications, the pace development is really fast, right? So new
Integrating Evolving AI Tools
00:11:46
Speaker
capability set up every day, existing applications to sort of develop into bigger and better things. Most tech companies are like integrating into the ecosystem. How difficult does that make your job in terms of picking tools, piloting tools and purchasing tools, when the landscape changes at such a pace?
00:12:06
Speaker
Yeah, the landscape of the art of planning isn't changing. We still have to sit down as humans and consider the logical sequence of building something. It could be in time that AI does come in and take over that role. It will look at the 3D model and it will recognize all the items in there.
00:12:27
Speaker
Why wouldn't AI then start to consider the order it should be done, and indeed the duration that you'll need for each activity? You know, if you go into a hospital, there are surgeons that are using AI at the moment to actually consider how they're going to carry out a particular operation. And there may be some element of automation in there in order to get the accuracy.
00:12:52
Speaker
Even with the surgeon, at the end of the day, they're using message as information. They are looking at it and applying the human sensibility to it to say, is this logical? Is this really the right way to do it? And what will happen is they will tweak it to get the product or the output that they are looking for, they believe is actually necessary for a patient like himself.
00:13:14
Speaker
What you're actually doing is speeding up the process and allowing those who have the knowledge then to make the final adjustments to get what they need. But the counter to this is, okay, if we've got AI doing that and you've got a generation like myself who would say, oh, that's not quite right. We need to tweak that. What happens with future generations? I haven't got my experience. And where do they get that experience from?
00:13:42
Speaker
We'll never have AI as a 100% output for everything we do. We do have to bring people on to understand what it is that's actually physically happening out there. Be able to assimilate that sort of information in their mind. But yes, they can start to use AI to speed up that process and get a better output and probably a faster output and learning curve for themselves as they go forward.
00:14:10
Speaker
So it's a very careful balance as we go forward. Yeah, it leaves you in a pretty dangerous spot if people end up at the point with no skills. It's a bit like if you have driverless cars and no one knows how to drive the car, the machine stops. Like, how do you actually deal with that? You need to have that base. Right. Jason's never stayed quiet for 15 minutes before. I've got one more question before I hand over. From what you've implemented so far, what results have you seen if you were to sort of pin it on one particular output or achievement?
00:14:40
Speaker
Well, that's very, very difficult. We are looking for more and more information laid out about it. And as we go forward, we're looking for things to happen. Our projects are getting bigger and more complicated.
00:14:56
Speaker
And without doubt, I've been reviewing a 40,000 line program for the significant project in billions. How do you do that? And to be honest with you, you cannot get into the nitty gritty of it. It's all on a team that put this program together over several months. And suddenly they're asking me, can you go and spend a couple of weeks doing a review of this? Well, you know, you have to have an element of trust of the professionalism of the planners that are working on it.
00:15:26
Speaker
But there are some basic tests and approach that you can start to apply to it. And with that, yes, the health of the program is the first thing I go to. I'm interesting to understand the density of connectivity and all the activities, the links, the predecessors and the successes, how they've been put together.
00:15:47
Speaker
I'm interested to know how detailed is this program and what is the maximum duration built into this program for an activity and you might set yourself some rules that really anything greater than three weeks or four weeks, you don't really know what's going on in there. It needs breaking down further so you understand the program better.
00:16:09
Speaker
Obviously, the level of program you're working at is obviously going to have a big drive on that. Then you start to want to understand, okay, how many activities don't actually have any successes of just loading there. Now, I talk about that not as an AI, but as a bit of technology that can go rooting around a program and finding all that out.
00:16:29
Speaker
But what it is telling me is how robust, how healthy is that program to change if you start to restructure it, or I call it like poking a wet sponge. If the program's not well structured, you can poke it and you make a little wet pull around your finger, but the sponge doesn't change shape. And that's because it's too loose. The same thing happens with Carlos, actually, if you poke him. Right.
00:16:57
Speaker
I think he's probably a little bit structured. So that's a hell of a hell. So what I can say about is, well, when you actually start to think about QSRA and the Monte Carlo, what you're doing is introducing change into activity, uncertainty, more time into that activity.
00:17:19
Speaker
And if it's a wet sponge, then the Monte Carlo can do all these things. And at the end of it, basically nothing's changed because you're just trying to vote this wet sponge. You need something a bit more structured.
00:17:33
Speaker
So, okay, doing a QSRA is quite difficult on a 40,000 line program. It takes
AI in Project Plan Review
00:17:40
Speaker
some weeks. I've only got a couple of weeks. So I can start thinking about using AI, knowing that actually, yes, I've got everyone's issues and risks in here.
00:17:50
Speaker
that I'm going to get a very quick output in a matter of a couple of hours. And I can start to think about how that aligns with the approach and the program, the time and risk that I've actually got in that program. So I'm already in a position to make some judgments about that program. Then after that human error comes in, you sit down with a planner for a day and you start asking them questions and how they put this together.
00:18:15
Speaker
Have I actually physically looked at the program? No. Have I actually opened the program? No. I'm using technology to inform me to ask the right question.
00:18:25
Speaker
Either that, if I was to break my silence, if no Carlos had to go at me, I was maintained my silence because I was intently listening and finding it very interesting. So that's the difference, Carlos, I guess, to other meetings we might be in. Well, I'm not normally that injured, honestly. I can think of mental topics to talk about.
00:18:54
Speaker
One of the things that we talked about in previous weeks was about use cases for emerging capabilities of artificial intelligence. And you brought up some really interesting ones that we hadn't discussed. So obviously there's the review of a schedule and having some sort of, whether it's AI or just
00:19:16
Speaker
Literally some logic go and check links and, uh, uh, more basic. You're sorry. Um, and your example of risk and considering it like a global risk from a client side is interesting as well. Are there any use cases that you're, uh, seeing as viable at the moment in terms of during the delivery of a project? Cause we posited some ideas, uh, last week, but it's, they might be a ways off.
00:19:41
Speaker
Well, yes, you know, we get a lot of technology now that can write reports for you on an automated basis. And there's a lot of platforms now which will compare programs, obviously within the construction industry, especially within the UK, we use NEC for contract where we submit a program on a monthly basis. And we also provide a commentary that goes with that.
00:20:08
Speaker
And AI is starting to come into this area. Certainly I'm having discussions with platforms now where we're asking for it. Here is the program we submitted last month. Here is the program we're submitting this month.
00:20:21
Speaker
And the software can certainly go in and detect what's changed between the two programs. It can tell you which activities have been extended, which activities have been completed, which has been completed ahead of schedule or behind, and any logic leaks that have been changed in there.
00:20:41
Speaker
That then has to be built back into some sort of commentary, which can take a fair bit of time to do. And, you know, you might spend a couple of days putting that commentary together to submit the package.
00:20:56
Speaker
But AI can provide that commentary and it comes back to the position where I was saying, having got a commentary, you can tweak it and make the changes that you think should be and the context in which it should be going over. The information is the error made. But the way it's couched or perhaps the position it's taking isn't quite right.
00:21:22
Speaker
So all the hard work of the commentary is being derived and what you're able to do now is go through that two or three page report and make the tweaks to give it the context you believe it should have to go forward to the time.
00:21:37
Speaker
the client can equally do the same and consider it from their point of view going forward. Obviously, where we get more all kind of working with the client, I would like to think that, yes, we can actually generate commentary that is neutral and we can all support in terms of its context and the information it's providing.
00:22:00
Speaker
So you have that. Then it's quite a short step to think about how do you manage change in the program and the impact of change in the program. And again, you're only comparing here is our original intended plan and changes now occur. This is the new consequence of that change. And you can start thinking about using AI to provide reports on matters such as refloats in the program that's being used up.
00:22:27
Speaker
how that's affecting your resource leveling in the project, how it might be impacting on critical dates and the end date. So, you know, it's a fairly quick way to get out a fairly instant output.
00:22:42
Speaker
on the impact of change. Where does a human come into that? Well, the human has to add the bars in the program to reflect the change. Obviously, those become the bars, which are the new bars, which is picking up on. And then the report is talking about those new bars and how it's impacting the program.
00:23:02
Speaker
Yeah, I think that's a really great example. I think a lot of people when we think about this immediately go to grand outcomes of like big brother, automated scheduling, those sorts of things. But AI has proven already that summarization is like a really good power that it has kind of in a capability today. And there's a lot of admin busy work time spent across lots of parts of different businesses and in construction and in planning and managing change.
00:23:31
Speaker
I think that's a really great example. I was probably, when we started this series of conversations, probably sitting more on the skeptic fence and I've kind of moved the needle the other way.
Tech Adoption Analogy
00:23:43
Speaker
I'll tell you a little story. It goes back in history. When we used to thrash the grain, it used to be physically done by hand. The crop would be gathered from the fields and it would be physically thrashed by hand using sticks and twigs and things like that and then tossed in the air to get the seed out.
00:24:06
Speaker
I'm not going to make any comment about Carlos again now either. Well the dressing machine then arrived and of course that put a lot of people out of work and they were landowners who were being killed over it. There was quite a rebellion over it. You come to today you wouldn't entertain harvesting your fields without a combined harvester, the technology. So what's changed? Yes there's not so many people involved anymore
00:24:36
Speaker
and the skill sets that have been deployed are much higher than they used to be but the yield we get out of it allows humans to go and do something else so this is just technology where things are going to happen quickly you may not need such a big team you need to actually embrace change and know that the job you're doing today
00:24:56
Speaker
is not necessarily the job you're going to be doing tomorrow. With the advent of technology is opportunity and we couldn't say that because of the thrash machine and we stopped beating the seed out that that's been detrimental to human society, human race. We adapted in the end and we are what we are today.
00:25:18
Speaker
And really that analogy needs to be thought about as we go forward. Things will get better. There'll be different opportunities. Opportunities in jobs that we haven't even thought about just yet. But that will provide a quality of life. We cannot really believe we should go back to beating the seed out of the stalks to get our oats or wheat or whatever it is.
00:25:42
Speaker
On that topic, I might change pace slightly. I know that listening to this podcast are a bunch of people, I'm thinking of a couple that are in heads of planning roles at sort of tier one contractors in different places around the world. And lots of them are either going through or thinking about going through some sort of process to like build consensus or implement some sort of technology, whether it's AI or any other thing.
Choosing Valuable Tech Solutions
00:26:12
Speaker
I guess I'd be super interested to understand, one, how do you separate fluff from things that are valuable when you're getting a million different technology vendors shouting at you to adopt them versus another thing? How do you build momentum or consensus within an organization? And then maybe what is the process that you think about? Is it like pilots? Is it like a bottom-up approach? Or are you getting top-down consensus? That's my mega question to finish over.
00:26:39
Speaker
Well, you've covered about four or five different areas there. So that is a mega question, probably take me 20 minutes to answer it. First of all, you've got to actually understand what is the need in a business? What do you really need? There's a lot of technology out there and very shiny things. But you've got to remember, I mean, I've got a community of some 200 planners, and they're very busy people.
00:27:04
Speaker
And the last thing they want is to play around with toys which they can see no value for. It doesn't give them any benefit at all. That doesn't necessarily mean that that's going to take you anywhere if you just give them something that they can immediately use. So you've got to think about technology and approach that you're taking and where you would like to be in four years' time, five years' time, and how you're going to implement that.
00:27:33
Speaker
And when you implement it, you've got to implement it in a way that gives benefit to the end user. They can actually see that they're going to get something out of this new bit of technology. It's going to make their life a little bit easier. And then you need to think about this is a lot of change for individuals and focus how you introduce that change. There's a lot of sheet dip approach to change that everyone gets to know every single at once. And that frightens people.
00:28:03
Speaker
If you can actually target change individual groups so they know that this new bit of technology is coming out and they're going to get this or that out of it. And the reason why we like you to do that is because we need this at the center and we've already got a plan of how we're going to couple that with other information.
00:28:26
Speaker
So, you know, the prime example we have is that we now ask our planners to code their program, a separate column, which gives some basic coding about new works, existing planned works, our baseline, our start dates, all this sort of stuff. It's not huge. And I think basically it is about 10 or 12 codes. That's all we're asking them to use. And it comes as a dropdown column. They get benefit out of it because they can actually filter
00:28:55
Speaker
all this information out and they can see directly themselves to change. I get benefit at the centre because I can start to analyse programmes because I can see everyone's programmes and I can start to use technology to understand which programme is submitted to the client because those are the ones I'm interested in. Out of the hundreds of programmes that are there, the what-if programmes will have you. And that allows me to start to understand for each project the change that's going on from month to month.
00:29:26
Speaker
and allow me to actually link that then with commercial information. Now commercial will go looking at that and go, oh, actually, that's really good. I don't understand this number or that number. Why is that? You start saying, well, we need to have some structured approach on the commercial side in terms of how this information comes together.
00:29:47
Speaker
So it grows and it starts to get arms and legs and it starts to get traction. This isn't done overnight, this is done over years and that's why I say it's important not just to talk it and be far ahead of yourself but to walk it, not run it and get yourself to a position where everyone's coming on board with you as you go forward.
00:30:11
Speaker
So to either to double click on that just before I know Carlos will probably be watching the clock. But to double click on that bit is your do you find that the approach of if you're trying to validate some new technology or approach or software or whatever it is.
00:30:28
Speaker
is to start with some small pilots. Do you let projects do their own things and then bubble that information up? Or is it more like you try and think about it centrally, you get some consensus in the business, and then roll it in? Which way seems to work best for you? Well, personally, you've got to understand the need at the sharp end of the business because they are making the money for you. I'm an overhead. I have a service industry. I'm here to provide a service at the end of the day.
00:30:58
Speaker
So there's no point me inventing something that no one wants to use because basically I'll go bust or go out of business and not surprisingly. So I really need to provide something that the end user, the people at the sharp end who are paying my wages at the end of the day want to use and get value out of it.
00:31:19
Speaker
But at the same time, I've got another finance group above me who are very interested to understand how are we doing as a business and where are we going. So there's a very fine balancing act between them. And that's all about project controls at the end of the day.
00:31:34
Speaker
What do we do? Yes, we understand the need from both directions and we start to bring together approaches that will enable us to service both ends of the business that we're in. We will actually get user groups together and we'll invite people
00:31:56
Speaker
in management, but also people at the shark end, there is nothing wrong with having the side engineer or the assistant planner sitting in with all the managers and people higher up in the company.
00:32:11
Speaker
and you create time and gaps for these people they're not used to sort of sitting in such a meeting and certainly don't understand actually their opinion is incredibly valuable we really need to know and understand what you think about this and why this may or may not work from there you can start to formulate how you're going to go about this how you're going to meet the needs that they're sharp in because
00:32:37
Speaker
clearly we need them to use this, how that will furnish the output that's required further up the business and right at the board level.
00:32:47
Speaker
You then start thinking about how you would implement it.
Testing Technology with Small Projects
00:32:52
Speaker
And to be honest with you, we start off modeling this on very, very simple projects. And we have what we call in the business the bridge project, which is a very simple 50 line program, which we know absolutely inside out. And we will test the environment against that and change things on that and manipulate it.
00:33:16
Speaker
and get it into a position where we are getting the platforms of software or whatever it is that we're trying to get adopted working. Then we will get an existing project and apply that to see if we can get scalability into the product.
00:33:35
Speaker
When we feel that that is working properly, we then introduce that one to a couple of trial projects out in the business itself. This can take the best part of six months, maybe nine months to actually do. And I always keep on saying to people, look, we get it right. In two years' time, we'll look back. The fact that we were two months late, or we took three months longer than we should have,
00:34:02
Speaker
really on a matter? And the answer is no. People want things to work and they want it to work in a particular way so that they can do their job. They're doing their job is about making money at the end of the day. We're here to make money.
00:34:19
Speaker
Absolutely. Right. I'm going to have to cut everyone off first. We have got one over time. That time really did disappear quickly, which is testament to how interesting it was. Thank you very much for joining today, Eva. It's been amazing. Thank you very much, everyone, for listening. Thank you. Thank you, Ivor.