Skip to main content

Episode 4: Disrupting Capital Projects with AI

A Deloitte Real Assets Advisory podcast

Host Eoin Ó Murchú is joined by Deloitte Real Assets Advisory partner Hugh Dullage and the CEO of nPlan Dev Amratia as they discuss how AI and machine learning can optimise the delivery of Capital Projects. nPlan’s machine learning technology accurately forecasts the duration of construction projects and significantly supports the de-risking of major projects. This has earned nPlan recognition as a finalist in the recent Royal Academy of Engineering MacRobert award for engineering innovation.
Key questions
  • What is the difference between AI and advanced analytics?
  • What is nPlan and how does it help clients in the infrastructure and construction sectors?
  • What benefits can data analytics and machine learning bring to the industry?
  • What are the risks in utilising AI and ML?
  • What does the roadmap for AI look like in its role in delivery of capital programmes?
  • How should organisations in the capital project space incorporate AI effectively into operations?

Find out more

If you are interested in any of the topics discussed during this episode, please find useful links below:

  • Occupiers & Capital Projects
  • Real Assets Advisory Careers
  • Transcript

    Hello, my name is Eoin O’Murchu and welcome to the Deloitte’s podcast series Futureproofed. This podcast explores and challenges how technology and innovation is disrupting major capital programmes, infrastructure and real estate projects. Today’s episode is all about the role AI is playing in disrupting the delivery of complex programmes, and I’m delighted to be joined by two guests. Hugh Dullage a partner in Deloitte’s Real Assets team, Hugh is a chartered engineer by training and advises large infrastructure and capital programme clients on technology and innovation strategies, programme controls, productivity and delivery excellence across multiple sectors, including transport, water, telecoms, defence and energy and Dev Amratia the co-founder and CEO of nPlan. nPlan makes extensive use of novel techniques in machine learning, coupled with the largest data set of schedules in the world, to systematically quantify delivery risk on major capital programmes.

    Welcome to you both.

    Dev Amratia:Good to be here.

    Dev. First question to you, can you explain to our listeners what the difference between AI and advanced analytics is?

    Dev Amratia: Sure thing. AI is the field, the best way to think about it is AI is where the system that is doing the analysis is learning to become better and better at such analysis. It is therefore intelligent artificial intelligence. The word intelligence is used to describe the fact that the analytics can produce varying outputs as varying inputs constantly arrive. Just like how you and I can process information in different ways. We are intelligent beings, I would hope. These algorithms are also capable of doing such a valuable tool in the toolkit is different in that it uses set rules to come up with the set answers. So, if this, then that, is very simple analytics.

    nPlan are doing some really interesting work in that space. Maybe you can give us a big introduction to nPlan and the challenges you are helping your clients in the infrastructure and construction sectors address.

    Dev Amratia:Sure thing. nPlan forecasts the outcomes of construction projects so we can say how long a project will take and what might go wrong with it using data from historical projects or projects that have finished. The way we do that is we use advanced processing methods in machine learning to understand that historical data, and then once it’s learnt that, it can then look at a new project and say, oh, I’ve seen these things before, I’ve learnt them and therefore I believe these things might happen in the future. It’s sort of the same as what very, very experienced construction planners or construction schedulers might do when you bring them into a project for a review. They’ll say, I’ve seen these things happen before and therefore I think this is what might happen, or these sets of things might happen to this current or future project. We don’t put people in there. We have algorithms that are capable of doing this at scale.

    So, you’re constantly learning and constantly evolving from the information you have and the projects you work on.

    Dev Amratia:Yeah, we’ve now processed just over 740,000 project schedules. It represents about £1.5 trillion pounds of constructions projects that have already finished, for a sense of scale that is 15 times the size of the UK construction market year on year, which means we get data from all over the world. It is by a whole order of magnitude the largest dataset of construction outcomes on the planet.

    Well, and obviously data and technology is integral to that. Hugh, drawing on some of the points that Dev has mentioned there, what are some of the challenges that capital project organizations face when trying to compile that data?

    Hugh Dullage: Yeah, one of the challenges we often see with many clients delivering major capital projects is the data comes together from different organizations who are not used to necessarily working together and they come together to deliver the project and so the data structures or the terminology or the definitions might be different from one project to another and one of the great things about the kind of technology that Dev is talking about with nPlan is the ability to learn and to bring all those different data sets together and to analyze them in a way that means that you don’t have to have exactly the same data structure each time you use it.

    That’s brilliant. Yeah, so it’s the data is underpinning everything you’re doing and it’s almost driving that forward and we’re all used to AI you touched upon earlier on and it’s fast becoming part of our lives from day to day. You know, like we all we’ve all come across it. Even when you go online to shop, for example, or even when people are trying to create their images online. Dev, what type of benefits, particularly in productivity and risk management can data analytics and machine learning bring to the industry?

    Dev Amratia:Certainly, I can illustrate that with an example. One of the largest rail projects in the UK uses and nPlan to help them understand risk and uncertainty. Effectively they’ve got a two parted problem. The first part is they need assurance of the direction of travel they’re going in and they need to share that with their stakeholders. That could be the Department for Transport, their stakeholders could be their contractors, their projects that will interface with this other project and so there are many stakeholders in that need to understand when things will finish or when things will happen. Um, so they use nPlan to forecast parts of the project or the entire project so that they can give a fair and you know, they’re not marking their own homework when they share what nPlan says as the outcome of the project. So, they used us to justify their outline business case as the first problem space. The second problem space is how do you drive performance on the project. So typically, or at least - my background is I was a project manager with Shell in my past life I’ve managed very large projects in my career and I always felt like I was fighting fires, not little ones, but, you know things would keep happening to me as we were delivering the project and my team and I and we constantly felt the stress of having to deal with these fires. So, the problem there was we had no way of foreseeing the fire before they came and whacked us and you know put us, made us and put is in a very uncomfortable position. So, what we do with this rail project is we give them these insights about here are the things that you now need to think about and the impact that those things could have should you choose to do nothing about them.

    That’s a huge mindset shift though for the client.

    Dev Amratia:Oh yes, huge.

    They’re putting huge amount of trust in you here.

    Dev Amratia: Yes. Because all of a sudden, what was either reserved for the most experienced people on the planet to give to a project team is now at the fingertips of hundreds of project professionals. Anyone can just login and have a look. Not that they should just have a browse, but they should actually do something with it. So, trust is a huge challenge that we go through with our clients on how do you trust that this machine, which his telling you something about the future, is correct? How did this machine understand my special rail project or any project frankly? Right, it’s really complicated challenge that we’ve had to navigate through. But suffice to say we’re making lots of progress in this space in getting through it. And a lot of that comes from showcasing how well these systems can work in parallel examples and also giving them the evidence underneath it to say, here's why this algorithm thinks this so that they can the person can introspect themselves and feel comfortable with taking action and then reminding them that this is the opinion of the data. This is no one is attacking you. No, we’re not trying to make you look good or bad. The data says this, so, you know, it’s probably worth listening to at least to a part of it.

    Yeah. Hugh you must have similar experience working with clients where they’ve had to embrace technology in a meaningful way to be successful?

    Hugh Dullage: Yeah, absolutely. I think it’s fascinating what Dev’s just alluding to there. So, it’s not just the technology challenge of getting the algorithm right and being able to produce the tool to do the analysis in the first place. There’s also the challenge of the adoption and convincing people that actually, you know, this bit of technology you don’t necessarily fully understand is going to give you an answer that’s reliable and it’s interesting in the capital projects context because obviously you know where we use clever technology to help plan routes and journeys and things either in you car, or on public transport, you know if you get it wrong, it tells you something. You’re a few minutes later, early, it doesn’t really make a big difference, when you’re talking about capital projects, the scale of the investment involved and the timescales through to delivery to prove whether this forecast is right or wrong or their risk identification is helpful is obviously much bigger scale. So yeah quite, quite a different scenario.

    That’s really interesting Hugh you as well. You’ve just touched on something that everyone’s used to, which is travelling or even the weather like Devin. How have you embraced, you know, the ambiguous AI that’s out there in your world?

    Dev Amratia: Yeah, we’ve taken a lot of inspiration from tools that we are now starting to take for granted that are fascinating applications of machine learning and AI. The weather is one particular example of a forecast, right? The weatherman now tells us what the probability of rain will be in advance of time and other than my dad we all generally trust the weathermen. We think that this thing or this being is, you know a very informative source that we can rely on, and commercial weather is even more so, right. The people choose to evacuate oil and gas platforms off the back of a forecast generated by a machine. It’s not a human being deciding these things. It’s data analytics at his very finest. So, one of the couple of inspirations that we’ve taken from it and the first is be like, keep it intuitive. No one, no one enjoys changing their mind when the information they’re looking at is difficult to digest. The weather apps most of us have on our phones are really simple tools that are actually communicating incredibly complex information at us. So being keep it simple, remain intuitive. The second is demonstrated, so Hugh said something very interesting earlier. He said, how do you actually prove the efficiency of these tools when a project could take ten years from inception to completion? You know I’m not going to sit around waiting for ten years to say, oh, look, I was correct and then you have the problems that it’s not statistically valid if you just proved it once. Um, so what we do at nPlan and we do with our clients is we do redacted back tests. That means I said we have 740,000 schedules, we train the algorithm on about 80% of it and then forecast close to 100,000 projects, go to forecast. Just think about how many projects we’re saying. So, we rewind the clocks, pretend it’s 2015, 2013 and say let’s pretend we’re a sanction point for all these 100,000 projects of all types of sectors and forecast forward and then check how often did the algorithm get it right or wrong? What did it get right and wrong? And then when we talk about intelligence and learning, the algorithm is capable of taking whatever it figured, whatever it, you know, mis forecasted and learning from that right.

    Correcting its homework.

    Dev Amratia: And it corrects itself. It’s like, oh, I made a mistake there. I’m going to go back and not make that mistake again and it does that 24 hours a day, seven days a week. Right. You know, we would get pretty bored of remaining in such a vicious learning cycle. But machines, don’t. They keep going and they keep getting better and better.

    So that’s a huge opportunity is that one of the biggest opportunities you see for AI and ML in the capital project space?

    Dev Amratia: It is one, but no, it is not. It’s not the only one. I think of course it’s the biggest because it’s the one I’ve sunk my entire career on but, I think that the plethora of solutions out there are all marching in the right direction, which is how do you take, for example, the other example that I’m seeing a lot more coming, coming much more into the front line is measurement of progress, right? It’s a very – not a lot of people get thanks for walking their site with a notepad and then coming back into the office and then, you know, translating that into their schedule file. Instead, you can just chuck a camera out and the camera gets loaded with a BIM model and the BIM model against computer vision can track how well you’re doing on site at a millimeter accuracy. Right. It spots more things than any human being could ever spot and does it at a fidelity that you know we just don’t have time for. Those types of examples are amazing, and they really do complement it sort of become creates a suite of products that are all in totality very valuable.

    Yeah. Hugh, what Dev is describing is, in my view, is that nPlan is part of a wider ecosystem of solutions. Is that what you are seeing?

    Hugh Dullage: Yeah. I completely agree. I completely agree. I mean, I think the power of nPlan and some of the other tools that can help with schedule optimization and risk management and so on is enormous. But in my mind, they’re all part of a wider opportunity for productivity improvement, and part of that is through risk management but there are other things out there. I was down at the with one of our alliance partners the other day, at their innovation lab and looking at some of the technology they’re starting to use and one of the things they were demonstrating was NLP on natural language processing technology to identify risks very early through unstructured data, through emails, through instant messaging and so on and being able to identify them early means obviously you can get on to managing them earlier. So yeah, different kind of technology but in a similar sort of space. And the other area that we’re looking at is around productivity and improvement productivity and in particular around wearable technology. So, understanding what people are actually doing on site, what is driving any inefficiency that we’re seeing in the productivity of people on site and various surveys have looked at this and found that 40-50% of site time is not fully productive. So, there’s a huge opportunity there if we can use some of this technology to identify where the lack of productivity is coming from, whether it’s training or access or whatever it might be.

    There really is a lot of data out there people if they’re willing to share the people movements and so it’s a really good example. People have their phone and their watches which can track their movements, but also just huge health and safety element to that as well of, you know, being able to intervene if and when things go wrong and these are some of the positives and we’ve talked a lot about the positives, but what are some of the maybe risks or other things to consider? And when we talk about AI and ML, Dev keen to get your view.

    Dev Amratia: Yeah. I think there are two that I would I highlight the first a bit more comical than the second is the risk of ignoring it or doing nothing about it and saying that us you know let these let the boffins deal with that. That is a real risk to a, an individual’s career and b, the project’s outcome, do not just think that you can do this on your own. Think of these tools as being productive adders to the capability of team. The second risk is from the technology itself. So, the technology is moving at a pace that honestly even for me as someone within the sector and using the technology every day and developing it, it is moving at breakneck speed and the speed at which it’s moving means that for the organizations that are using the technology we have to take a lot of we have to invest a lot into ensuring that what we’re doing is legitimate, correct and can be back traced to being that, right? We talked about how some of the forecasts we’re producing at nPlan could be worth million, hundreds of million of pounds and dollars when implemented or not implemented. That’s a lot of weight to put onto our shoulders and if you sort of don’t keep the right checks and balances in place within the organization on like how can we demonstrate the efficacy of these things? Is it traceable? Is it, you know, is it ethical to produce a forecast without the homework at the back being done? I think that’s a real risk inside all of the tooling that we are starting to see is that if the tools are developed without that rigor behind them, you can very quickly. It’ll take us years to build trust, but days to lose it through these through an incident that may occur.

    Yeah, and it’s just needs to be written off once for someone to not trust it anymore and it’s down to that fundamental thing of changing the mindsets of people Hugh I think you’ve touched upon before and we’ve spoken about this.

    Hugh Dullage: Yeah, we slightly touched earlier a bit on the point about adoption. I mean, in a similar vein in building on what you were just saying Dev. I mean, we’ve all heard about the examples of some of the new AI tools that come back with some fairly odd results from Internet searches and things that kind of thing can undermine I guess the technology and the credibility of the outcome of these sort of tools. So, I think the biggest challenge really here is in the adoption and maintaining the credibility and building the credibility.

    Yeah and it’s that, as you said, that trust not only with the stakeholders who are involved in the project, but the wider, wider community and understanding that everything is being delivered in a way that is best for the programme and trying to bring AI back into the heart of that. I think that’s I assume, would be one of your missions as a business?

    Dev Amratia: Yeah, for sure. I’ll just state what our vision is. So, our vision is to build a world no longer limited by its appetite and risk. We believe fundamentally that society is held back because we’re just so scared of doing these large projects because of how badly and how volatile they are that these like beasts that we just don’t want to go very close to and an good example of that of course, we’re all sitting here in London is Crossrail two, something that I was very much hoping would come to life for my own personal commute. But it is now exceeded our tolerance of risk and so therefore the project has been shelved for now right. Right and anything I can do to try and sort of use technology to try and help people make better decisions will take us forward in humanity. That’s I mean, that’s why I do this.

    And that’s a very noble and valid reason to, to drive and make that change and I think we touched upon trust, and I think the making AI part of the business as usual is a really important aspect of this. Maybe, Dev can you give us any examples of where you work with clients who really dove in and adopted nPlan for the betterment of the program?

    Dev Amratia: Yeah, lots. I’ll use the example of High Speed 2, which is the largest rail project in Europe?

    Dev Amratia: Thanks. Very, very complicated delivery to execute on that project but it’s a good example of a forward thinking organization that understands the limitations of their team. You know, as much as they are the largest, this is also the first time most people have built something of this scale and sort of like bringing the data forward to help them understand where they need to focus. It is such a large project. There is no way everyone can be on top of everything and certainly as you go up the management ranks, that gets and becomes a harder and harder problem to face and so it’s enlightening that you know the leadership of that project see the value of using data to check their assumptions to improve stakeholder management which you can’t even express how complicated it is for such projects to have to go through this. But it’s again, it’s very encouraging to see these teams take it up, but they’re also examples of very small projects doing it right, like High Speed 2 is the big, the big one in town but they’re also really encouraging examples inside the electricity grid as an example. Right, where there’s lots of little pockets of projects going on all day, every day that none of us really get to find out about, but are so crucial to keeping our lights on, literally. Um, and even there where it’s a ten person project team and you know 50 million and you’re done in less than a year they’re using it at that little pocket of project team, spreading it across multiple projects as a portfolio of them and then rolling it upwards to their management, right? So, thinking of portfolios of projects where there are lots of small pieces. But you know, if lots of the death by a thousand cuts problem inside portfolios is another very encouraging example we see today.

    I think one of the things that struck me there when you were speaking there Dev is the different applications of it. So, I think Hugh one of the things we see with major problems all the time is that you know they’re forecasts for a cost or time can be wrong and I think a tool and AI like such as nPlan potentially benefit in forecasting and bidding for major programmes.

    Hugh Dullage: Yeah, the opportunity here is immense. Some very interesting analysis done at the Oxford Business School recently about the number of projects that deliver the expected benefits on time and on budget. Very, very small single figure percentages and if we can improve that and significantly improve that, then the benefit to deliver to society in all aspects is, is enormous. You know it’s interesting you talking there Dev about High Speed 2. We’ve obviously done quite a lot of work with High Speed 2 over the years, they handle enormous amounts of data on their reporting each month and so there’s an awful lot to go out there in terms of what you can analyze and the performance, but the ability to better predict and better forecast and identify the risks earlier, it delivers huge possibilities for more efficient delivery and increased productivity.

    And value for the wider public.

    We’ve talked a lot about the current state and the opportunities and even some of the risks but Dev keen to get your view as to what you think that roadmap in five, ten, fifteen years time is for AI and it’s role in the delivery of major capital programmes?

    Dev Amratia: The roadmap is really simple. It becomes the norm, not the exception. Like today we’re like, oh, look at that one project over there, they’re doing really amazing things. The road map is that that’s not amazing anymore and instead we’re talking about another thing that is, you know, we can’t even imagine right now. So, ubiquity, standardization, you’d be stupid to start a project and not have it checked and not have a copilot that is an AI system at your side as a project team. So, that’s the roadmap.

    So, Hugh just picking up on what Dev said there he’s talked about that roadmap and everything becoming the norm I guess companies like nPlan will form part of that disruption that the industry is facing going forward

    Hugh Dullage: Yeah, yeah huge disruption but in a way actually that can improve the way people work from day to day. So, I mean, AI has the potential to take away a significant burden of admin in people’s day to day jobs and actually give them the opportunity to focus more on doing the interesting part of the insight and analysis and so on that they do. But, as you say Eoin the opportunity for disruption here is, is significant I think that’s been recognized recently hasn’t it Dev with the you’re a finalist in the MacRobert Award for engineering is that right?

    Dev Amratia: Yeah. Um, lots of imposter syndrome with that, um, where the MacRobert Award is the most prestigious award in engineering innovation in the UK. Previous winners are Rolls-Royce for building the Pegasus engine and the vertical takeoff and landing engine systems. McLaren for building a ventilator system for getting us through Covid. Inmarsat for building global area networks delivered by satellite. So, there are some incredible innovations out there and it was really eye opening to see that the judges at the Royal Academy saw how big the problem space is in construction. So, nPlan is the first construction technology solution that has reached this level for its innovation and impact and it’s that impact story that really gets me out of bed, right? It’s they’re saying we are gob smacked by how big the problem is, how big the opportunity is and how encouraged we are that this kind of innovation will move the needle. So, yes, we are a finalist and hopefully the winner but even getting to the finalist level was um, yeah, I am gob smacked.

    Hugh Dullage: Brilliant. Fantastic recognition.

    Amazing, amazing stuff we will all know by the time this podcast comes out, whether you win or not. So, everybody listening already knows the answer, but we don’t so we’re very excited. But I think you’re right it shines a light on a sector that has been, you know, technology poor maybe, or lagging behind some of its, you know, sister sectors over time. I guess the final question I want to ask is what advice what you give to organizations in the capital project space who are looking to incorporate AI into their BAU operations and how they could go about doing it in an effective way that will be impactful? Maybe Dev first.

    Dev Amratia: My advice is to not sit on your hands on this one. It’s the power that this technology offers is, is unparalleled. But I’ll perhaps say something that listeners might not expect, which is when people use such technology at an individual level, they become happier. The reason they become happier is because they are no longer doing the boring things that they used to do. Instead, picking up on what Hugh said, the people that used to do this really tedious analysis are now becoming negotiators and communicators with other people, which is still incredibly important and not something AI is taking away anytime soon. Not in my lifetime, I think. When we when you see that transition of moving you out of the basement and up onto the first or above floors, it just makes people happier and happier people are more productive people. More productive people gets to better outcomes, right? I didn’t write the book on that. So, you know as much as we want, we can say like there are all these business benefits that we can get and those are massive. Clearly, we’ve talked about it several points in this podcast, but the last point I’d make is that it just makes people happier and if you can do that and you get the business case outcomes as well, you’re winning. Why would you want to sit this one out?

    Hugh, Dev was saying to me it unlocks the opportunity for individuals.

    Hugh Dullage: Yeah, I completely agree. It’s a game changer. People ought to be embracing it. You don’t have to go you know, full bore into it straight away. Maybe do a pilot first, test it, but this is going to change the way the industry works there’s no doubt about that. I’m sure if we look back in ten years’ time, we will see that this is it has had a huge impact on the industry. If you went back 20 years and looked at the way the Internet has changed, the way we work in construction and all aspects of our lives, I think this is the next big thing to impact the way you would deliver major capital project

    Brilliant. I think that’s a great place to leave it. My thanks to Dev and Hugh for their insights and really interesting perspectives. AI and machine learning within capital programmes is really fascinating topic and there’s lots more to come, I’m sure Dev, so hopefully we’ll have you back at sometime with your award on the table.

    If you enjoyed this episode, please do take a moment to like share and subscribe. Keep an eye out for future episodes on topics such across digital transformation, delivery, excellence, and other exciting, interesting themes in the capital projects and real assets space.

    If you want to know more about nPlan please Google them at and likewise Deloitte if you want to know more about us, please search Deloitte Real Estate Advisory Online, where we have a host of blogs, articles, roles and other interesting topics all available for you to explore.

    So until next time, thank you all for listening.

    Did you find this useful?

    Thanks for your feedback

    If you would like to help improve further, please complete a 3-minute survey