Maximizing Construction IQ and Machine Learning for Predictive Analytics

DEEPAK MAINI: Good morning, good afternoon, good evening, wherever you guys are. Thanks for joining us in this AU 2023 class about maximizing Construction IQ and machine learning for predictive analytics. A quick introduction about myself. My name is Deepak Maini based out of Sydney, Australia. I'm a principal business consultant with Autodesk. With over 24 years experience in the industry, I've helped our customers implement technology on projects ranging from a few million to several billion dollars.

It's my 12th year as an AU speaker and also eighth year as an AU mentor. I'm also a guest lecturer at various universities locally within Australia and overseas. Written a few books on some design and construction software such as Advanced Steel, Navisworks, and so on. And this is my aim. Next, I want to introduce my colleague and good friend Steven Bloomer. Steven?

STEVEN BLOOMER: Thank you, Deepak. My name is Steven Bloomer. And I'm a regional BIM lead at GHD. Our background is building engineering but, most recently I've been leading major infrastructure projects. Day to day, I lead a highly motivated team of BIM professionals. I'm ISO 19650 management professional and currently pursuing my master's in BIM and integrated project delivery at Bond University. Thanks, Deepak.

DEEPAK MAINI: Brilliant, Steve. Now before I start this class, I want to say a massive, massive thank you to a few people. Speakers such as Steve and myself, we can come in here and just present without any worry in the world because of all the hard work that these people put in the background, including our recording partner Mediasite. Thanks, everyone, for giving us this opportunity and making it so much easier for us.

All right. These are the learning objectives of this class. And this is the description. I'm sure you've gone through this when you signed up for this class.

I want to start this class with an interesting concept. This is the definition of artificial intelligence. It's the theory and development of computer systems that are able to perform tasks that normally require human intelligence such as speech recognition, decision making, and so on. And this is the definition of hallucination. It's an experience involving the apparent perception of something that is not present.

Now you would be wondering, why am I talking about artificial intelligence and hallucination? That's because AI hallucination is not that uncommon. So what is AI hallucination? As per zapier.com, AI hallucination is where AI models generate incorrect information but presents it as if it were a fact. And as per techopedia.com, AI hallucination is where the large language model generates false or misleading information but presents it as a confident response.

So confident response. Makes things up and presents it as a confident response. So why does AI hallucinate? Again, as per zapier.com, it's because of the insufficient, outdated, or low quality training data. A term that we use quite often is overfitting where AI models are trained on limited data. But one of my favorites here is where it says AI hallucination occurs when AI goes off the rails trying to please you. Like seriously, trying to please me? Stop pleasing me and give me the correct information. That's what I want.

Now I'm sure you would have seen some examples of AI hallucinations but I've got a couple of examples here for you. The first one is from New York Times. Now a question was asked of ChatGPT as to when did New York Times first report on artificial intelligence. So ChatGPT came up with this answer about this conference at Dartmouth College on July 10, 1956.

Now did that conference happen? Yes. At this college, there was a conference. But did New York Times actually report on this? The answer is no, they did not. So ChatGPT took some information that was available and then made things up to bring you this information which was actually not right.

Another example-- and this is really funny for me but it may not be funny for the actual lawyer-- where a lawyer actually used ChatGPT to provide some information that he included in his 10 page brief citing more than half a dozen relevant court decisions. There was one little problem though. The problem was that none of those actual court decisions happened. Nobody could find any information about those including the judge himself. So the lawyer actually had to apologize to the court.

And I'm sure you must have seen this image about chihuahua or blueberry muffins. For humans, it's very easy to understand which one's blueberry muffin and which one's chihuahua. But for some of these AI models, it's not easy. And again, as I was telling my friend Steve the other day, if we can't find out the visual difference, I'm pretty sure we can taste the difference. Right, Steve? Not that I would want to ever do that.

All right. So you would be wondering, why am I talking about AI hallucination? That's because it's really important for us in this class. Why is it important? Just hold on to that. We're going to come back to that very soon.

Now let's start some numbers. Let's start talking with some numbers here. According to Construction Dive Report, bad project data cost contractors worldwide in 2020 only over $1.8 trillion. So $1.8 trillion was the cost to the contractors just in 2020 because of bad project data. Out of that, 14% of all rework in construction globally was caused by bad data. Now if you put a dollar value to it, it's close to about $280 billion that the construction industry had to pay because of rework cost because of bad data.

Now these are all scary numbers but the question is, how do I keep on top of my project data and analytics? A phrase that's become very common these days is analysis paralysis. So there's so much information available, there's so much analytics available, but it's not actionable. There's not much we can do other than just by looking at that. But how do we actually make any sense of it? That's where it comes into picture Construction IQ. Construction IQ is for projects hosted on Autodesk Construction Cloud.

So what is Construction IQ? It's a machine learning and artificial intelligence-based capability that delivers risk analysis of quality and safety data from your projects. The key is it applies analytical techniques to help transform the data into simple but actionable insights. So the key here is actionable insights. Now these Construction IQ analytics are displayed particularly for risk, design, quality, and safety features of your project. So in simple terms, Construction IQ is machine learning and artificial based algorithms that review the project data and identify high risk areas of your projects. This then provides actionable insights to help you address those risks before they become a major drama, a major impact on your project.

Now coming back to Construction IQ, it is important for us to note that, currently, Construction IQ machine algorithms are optimized only for commercial, health care, institutional, and residential project types. So when you're setting up your projects, make sure it is one of these four project types. Now if you did not set up your project as one of these four project types at the start, you can always come back and edit it to make sure it's one of these types. I will show you a demonstration of this very soon.

So I've also been asked by several customers when we go to our insight module, we have the risk option missing. So if you have the risk option missing, all you need to do is send an email to IQactivations@autodesk.com. Make sure you include your account name and account ID, which can be found here on the Account Admin Settings page.

Now going back to Construction IQ, you would be wondering at the start of this session I mentioned AI hallucinates. So how does Construction IQ mitigate this hallucination risk? Now this is the important bit. Construction IQ and machine learning has been training our AI programs using years of data from a broad set of leading construction firms. Now these are being trained using independent as well as nested AI models.

Now when I say AI models, these are not building information models. These are programs that analyze data sets to find patterns and then provide predictions to you. These are also trained by construction subject matter experts specifically focused using Autodesk data sets and some additional data sets that are provided to us by customers and they have provided us approvals to work with their data sets.

The outputs are reviewed and validated regularly by construction SMEs to ensure that we have got the right information. And if we do not agree with the information, we can always go back and update that. And we can further train the programs. In fact, Steven's going to talk about that very soon as well.

All right. So for the next part of this class, I'm going to hand over to my good friend Steven Bloomer who's going to talk about some GHD case studies. Steven?

STEVEN BLOOMER: Thank you, Deepak, for that in-depth introduction to construction IQ and the power of AI. But how can you use this tool tomorrow? As mentioned, I'm Steven Bloomer. And I've been wrestling with how I can leverage AI on my projects. We all hear about the great benefits AI can bring but where do we start? Who else feels like that? So when I stumbled upon Construction IQ, I knew I'd found that solution. Over the next 10 minutes, I'd like to share with you some real life project examples so you can leverage the same game-changing tools within your Autodesk Construction Cloud.

To give you a bit of context, I work for GHD. We're a global network of multidisciplinary professionals providing our clients with integrated solutions through engineering, environmental, design, and construction expertise. The work I do focuses on transportation, environmental, the water, energy and resource, and built environment engineering. So the two examples I'll be providing today are specific to these sectors but remember these can be utilized by the general contractor and subcontractor markets.

Let me first address the question why. Why do we use Construction IQ on our projects? When I heard about Construction IQ, my first question was, what can Construction IQ do that Power BI dashboards can't do? I imagine you're wondering this as well.

Well, one significant advantage Construction IQ provides over traditional dashboarding-- don't get me wrong. I use Power BI. My colleagues use Power BI. Like most organizations, GHD uses Power BI to showcase a lot of information such as numbers of issues, who they're assigned to, and the priority of those issues. But Construction IQ takes that one step further. It will provide you invaluable insights.

Construction IQ leads your information that your team are putting into issues, RFIs, and forms, and through advanced analytical platform using artificial intelligence and machine learning provides you quick insights into your construction project's performance, risk management, and decision making. It acts as a digital brain, constantly analyzing vast amounts of data your team are pushing to your projects to provide predictive insights that with the likes of traditional dashboarding would not be possible.

So if you're anything like me and don't want to hear about the hypothetical and want to see real life examples, the first case study I'll look at today is how we use Construction IQ in the safety in design process. Before I go into how we use construction IQ in the safety in design process, I'd like to highlight some numbers. Between 2000 and 2002, 44% of all work related fatalities on a construction industry were related to design in Australia. These numbers are Australian but I'll put my money on it they'll be similar around the world.

In Australia, it is compulsory that we carry out safety in design activities on all design projects. These come in many forms depending on the sector you work in. But the outcome should be the same-- to eliminate hazards and risks during the design stage.

Research has shown that fall from heights, struck by objects, electrocution, and caught in or between machines, rotating equipment, or collapsing structures lead to the majority of all fatalities on a construction site. Sometimes these are referred to as a fatal fall so it's critical that we mitigate these during the safety in design activities.

So how do we use Construction IQ in the safety in design process? First, we use the model to assist in this process. If you're not using the model for activities such as safety in design or model coordination, I can't recommend it enough. Model coordination will first federate all discipline models into one single view. We typically use this in parallel with design collaboration to ensure that only models suitable for sharing between the design teams are federated and used in the review process.

From here, in a safety in design workshop, we will thoroughly walk through the site in a predetermined method so not to miss any aspect of the design. As we navigate the project, issues are created with an ACC using templated types.

This example, we've identified a safety risk at the end of the egress stair at the end of the platform. Within model coordination, an issue is created. It is important to ensure the correct issue type has been selected because Construction IQ will use the issue type to group these issues. As you complete the issue, a description shall be entered and a root cause assigned. It is recommended that you customize these for your project and organization as at GHD we've taken this one step further by adding likeliness, impact, mitigation, and priority.

From here, once we've carried out the review and identified significant safety issues, this is where the magic happens. Construction IQ will now review your issues, the descriptions you have entered, and the due dates captured. Within the Insights panel, high risk safety issues will be grouped to assist you and your team in the quick resolution of these issues. As previously mentioned, the fatal four make up the majority of all construction fatalities. Construction IQ cards have grouped these particular safety risks along with fire to highlight them for actions by the design team.

Through the use of AI, Construction IQ has identified the risks and hazards within each issue description that you have entered to do the thinking for you. Imagine you're on a project with 100-plus safety issues identified. Where do you start to provide mitigation or potential control measures? If it was me, before I started using Construction IQ, I'll start at issue one and work my way through the list. This is not the most effective solution. Construction IQ allows you to review issues in order of priority to make your life easier and ultimately reduce the risk to your projects.

Now we've carried out the safety in design review using the federated model as a centerpiece for the process, we've created issues in ACC, we've leveraged Construction IQ to provide and prioritize these issues, we reviewed and provided mitigations or potential control measures within ACC, issued comments, what next? Sometimes we need to provide evidence of our safety in design process. Through the use of compiling data connector extracts and exporting issues lists, a report can be automated to meet your organization's requirements.

The second case study I'd like to share today is how we use Construction IQ to facilitate the process of site inspections. In the past, we would travel to a site with a clipboard and complete either paper form or checklist. Photos would be taken. When back in the office, a report would be drafted, photos will be linked, and a report submitted to our clients. As you can imagine, this is a laborious process, time-consuming, and error prone with photos inevitably linked to incorrect sites and noted as different locations.

With the adoption of iPads and forms with Autodesk Construction Cloud Mobile app, this has revolutionized the process. Checklists, daily reports, safety inspections, site walkthroughs, asset assessments, and certification, to name a few, can now be created as forms. I recommend creating templates for your organization to assist in this process.

Using Autodesk Construction Cloud app allows your teams to complete site inspections via the build forms in on site and in an intuitive manner. From within the form on the mobile device, issues, photos, videos, and notes can be added. One example where our teams use forms is for site inspections of clients' assets. We created a form template within ACC with predetermined questions to allow the engineers to answer. In this example, as the engineer is answering the fields by either typing or by the microphone, the fields are completed. It is worth noting these fields can also be dropdowns.

As the inspector answers the questions within the forms, photos and issues can be added to support the findings. If there's a quality concern, safety issue, or design item identified on site, these will be created as ACC issues. These issues are assigned to actual parties and tracked within ACC. Tracking items through ACC allows all parties to contribute to the closeout and resolution. Once the form is completed, the engineer inspecting the site signs the form and submits it through the system.

So you and your team have been to site and completed the inspection via the form, identified issues, and created them on the fly within ACC. What now? Within the insights tool and utilizing Construction IQ cards, quality issues will be grouped based on risk factors. This will allow the design, contractor, and subcontractor teams to action high priority items first, focusing resources on what matters and returning to the other issues later. Through the use of Construction IQ and leveraging the AI capabilities, project teams have a more focused approach to issue resolution.

As I'm sure you appreciate, AI doesn't get it right 100% of the time. And we found this to be the case when using Construction IQ. Sometimes you may need to change the root cause or the risk factor if you don't think Construction IQ has got it right. Remember the algorithm is currently only optimized for commercial, health care, institutional, and residential type projects. This is my experience and I'm sure in time Autodesk will expand the functionality for infrastructure and industrial type projects.

These capabilities have augmented the way we manage our project information. Construction IQ is a tool that, if used by your teams, can provide clarity and increase productivity in what is typically a time consuming process of issue resolution. Leave here today and implement Construction IQ tomorrow and you will gain the same insights not previously possible. I'll now hand back to Deepak to showcase how to set up Construction IQ on your projects. Thank you, Deepak.

DEEPAK MAINI: Thanks, Steve. That was absolutely amazing, really insightful. But I absolutely loved when you said, this is where the magic happens. And I wonder if actually AI does do some kind of magic to provide us all these analytics.

All right. So anyone who's been to any of my sessions in the past would know I love live software demos. Unfortunately, in this case, it's not live so I've got some demos here for you because I like to show how things actually work on the software in the technology. The first one that I've got is setting up the project types.

As Steve and I mentioned at the start, currently the Construction IQ machine learning algorithms are optimized only for commercial, residential, health care, and institutional projects. So when you are setting up your projects, make sure that these are set up to one of these types. If not, you can always go into your project admin settings and you can edit this project type. All you need to do is on the Settings page, click Edit and then you can go into this dropdown list and select one of those project types from here.

So this is really important. Unless you set up these projects as one of the project types we mentioned, the Construction IQ and machine learning analytics will not be displayed. So just be a bit mindful of that. And I totally agree with Steve here. Moving forward, I reckon infrastructure project types would also be included. When, we don't know, but I'm pretty sure they will be.

All right. The next one that I've got for you is Insight Interface. Autodesk Construction Cloud has this module called insight which is where we see all these Construction IQ-powered cards. So from the product picker on the top left, I'm going to go and pick the insight module. And this is where on the left I can see all these different options. Currently, I'm on the My Dashboard page. But the Construction IQ and machine learning analytics are optimized for risk, design quality, and safety. If I go into risk, I can see all these different cards that I've got Construction IQ listed on the bottom left.

These cards will show you all these analytics that are machine learning and artificial intelligence-powered, so such as high risk issues, overdue issues, and so on. All these different types are available based on Construction IQ. Same thing under design. Same thing under quality, so quality risk factors, high risk issues, quality issues, and water risk issues, and so on. Similarly in safety as well, we've got all these Construction IQ-powered cards. So that's the insight module. The key is make sure you have a look at the cards that show Construction IQ on the bottom left.

The next one I've got for you is actionable insights. As I mentioned earlier, analysis paralysis is a phrase that we are hearing a lot more these days because there's so much information available to you. But if that information is not actionable, it's kind of useless. And this is where Construction IQ actually makes all of this information actionable. Fatal four, for example, that Steve mentioned is available on the safety option here under the subcontractor safety risk along with fire. There are some additional safety risk issues listed here as well.

Similarly, if I go under the safety risk factors, I can simply click on these high risk factors here and I can click on any of these issues. As soon as I do that, it opens the issues in a separate tab where I can now start taking actions to resolve those things. Same thing under quality. If I go to these high risk issues here, I can click on any of these issue IDs to look at them and then I can then click on View An Issue on the bottom left to open this issue in the Issues tool so I can start taking actions. This is what I mean by actionable insights. How cool is this?

The next one is adjusting AI data. As Steve mentioned earlier, AI may not get everything right in the first go. And this is where Construction IQ actually provides us opportunity and options to actually train the AI data correctly. So for example, in this case, I'm on the quality option. And if I go under my quality risk factors, it's showing me that these are the 11 high risk issues.

Now if I do not agree with the AI program, I can go to this particular issue. It's currently listed as a high risk issue. I can say, no, it actually is a low risk issue. I can also then provide options to the program as to why am I updating the predictions. How cool is this? As soon as I go ahead and save this, issue is now changed to a low risk issue. Similarly, if I go to design, I could do that. Same thing with the risk option.

Currently, I've got these 11 high risk issues under quality risk factors. I could go to any of these issues. Let's say in this case, I'll go and click on this issue ID188. And I can update this risk level here. And similar to what I showed you last time, I can then provide feedback to the program as to why am I updating this risk level. So we are given the opportunity to train the AI program based on our project. How cool is this?

The next one that I've got for you is creating custom dashboards. This is really important for us. Whenever we are working on a complex project, we want all our project stakeholders to actually have a consistent view of our issues and RFIs. This is where I can create custom dashboards.

To do that, on the Insight page, I'll click on Create Dashboard. I'll go ahead and name this to let's say hospital. And now I've got this option called card library where I can go and add cards I want. I'm going to start with the address and the weather cards. I can now drag and drop these cards and place them wherever I want. Next, I want to add some Construction IQ cards. In this case, I'm going to pick these two cards here and simply drag and drop them.

Going back to Card Library, I want some factors-based Construction IQ cards, in this case safety and risk, quality, and so on. Again, I'm going to simply drag and drop them to place them wherever I want, so setting up my dashboard the way I like. And lastly, I want to add few more cards here. First, I'll go and add some high risk safety and quality issue cards. And at the end, I'm going to add the fatal four that Steve mentioned. So all these cards are now added. I can save this dashboard. And it is now listed on my insight module. How cool is this?

Now once I've created these custom dashboards, I want to share these with my project members or extended team members. How do we do that? So this hospital dashboard, I can simply click on Export. And I can share this with a company, a role, or a user. So I can go and pick any of these from here. If I have some external members that are not a part of this project, I can even share these dashboards with them with some message.

The other option I've got is share where I can share this dashboard with all my project members, selected project members, or only me. Again, selected project members could be based on companies, users, roles, whatever I want. In this case, I've selected GHD Company and I'm going to go ahead and say done. All the members of the GHD Company will get a notification that there's a dashboard shared with them. They can simply click on that dashboard and then they can see these dashboards on their insight module as well. How cool is this?

So that's all the demonstrations that I've got for you in this session. I want to start talking about some key takeaways for you. So Steve and I talked about a few different things. But if I have to give you four key takeaways from this session, the first one is Construction IQ and machine learning algorithms are optimized for commercial, health care, institutional, and residential project types. So make sure your project type is set to one of these. As I showed you in the demonstration, if you did not do it while setting up the project, you can always go back and edit it.

The second one is the Construction IQ and machine learning based algorithms are currently optimized for risk, design, quality, and safety features that are listed on the insight module. The next one is Construction IQ has been working really hard in mitigating the hallucination risk by training the AI programs or AI models using years of data from a broad set of leading construction firms. But again, if we do not agree with some of that information, we can always provide feedback to the AI models using the Issues window to change the risk level. So we are also given opportunity to update and train the AI models. Again, how cool is this?

I'm going to end this class with saying artificial intelligence and machine learning are here to stay. But they're not going to take our jobs away. They're going to help us. They're going to support us, supplement us in becoming more efficient in our jobs. A very good example that I have seen is this one here from Impact DE. A really good picture showing Chat GPT is not going to take away your jobs because most of our skills-- not if not most. Majority of our skills are irreplaceable.

And that's all we have in this class. We really, really appreciate you joining us and helping us explain the concept of Construction IQ and machine learning. If you have any questions for Steve or myself, please add them in the comments box which is available on this class page. Steve and I are going to regularly look at the comments. And if there are some questions for us, we'll definitely try to answer them as soon as we can. With this, that's all we have in this session. We really appreciate your time. Steve, any last words?

STEVEN BLOOMER: I want to say thank you. Thank you, Deepak, for allowing me to co-present alongside you. And thank you for watching today's session.

DEEPAK MAINI: Thanks, Steven. And thanks, everyone. Hope you are enjoying learning using all these class pages. Thank you.