Artificial Intelligence–AI–has taken the world by storm this year. Its reference has increased by over 77% from 2022. From chatbots to entertainment recommendations, chances are you are interacting with some form of “artificial intelligence” daily. Probably, more often than that.
With all these new technologies emerging seemingly every day, I can’t help but wonder about the ramifications of AI on the construction industry. However, before we dive into the specific use cases and future prospects of AI in construction, let’s take a moment to examine the fundamental differences between AI and analytics.
What is AI?
To say that AI is a vast subject is a major understatement. Technically, some analytics you interact with are AI–sort of. Analytics involves the process of analyzing data to gain insights and inform decisions. According to McKinsey, artificial intelligence is a machine’s ability to perform the cognitive functions usually associated with human minds. The relation between the two is how AI incorporates analytics to enhance its capabilities of enabling data-driven decision-making.
In simple terms, AI is any computer making a decision. These decisions arise from historical data or some algorithm coded to make a decision looking at data, similar to a human. Where things get more complicated is within the different levels of AI. Basically, AI is an umbrella term encompassing two subsequent levels–Machine Learning (ML) and Deep Learning (DL).
While AI, ML, and DL tend to be used interchangeably, they are quite different. Whereas Machine Learning lives within AI, Deep Learning lives within ML and AI.
Machine Learning vs. Deep Learning
Machine learning uses data from previous actions and processes to predict what the potential outcome may be. All the machine needs is a large set of inputs and outputs to learn from. From there, it learns how to answer problems by running data through tens of thousands or millions of examples to determine what action would logically come next.
Machine Learning is about giving predicted outputs. Deep Learning is more concerned with figuring out commonalities between data sets. As explained by IBM, Deep Learning (DL) simulates the behavior of the human brain by creating neural networks with three or more layers and learning from large amounts of data. However, it is important to note that this is just one approach within the broader spectrum of deep learning techniques.
The integration of AI, ML, and DL is increasingly being utilized in the construction industry to tackle various challenges and improve processes. Yet, the industry needs to develop a deeper understanding of the importance of good, clean(ish) data. Only then can the epidemic of delays and overruns–two of the industry’s biggest challenges–be solved.
Large Data Sets = More Ability to Average Out Anomalies
Big data is a relatively new concept in the construction industry. After all, cloud computing and big data for construction became readily available in 2016, whereas in other industries, it has been around for well over a decade.
In order for AI, ML, and DL to transform the way the industry functions, we need big data. As mentioned above, any Machine Learning algorithm aims to predict. For example, let’s say I wanted to predict house prices based on square footage.
Now, ML does not have to be a linear regression–it could be any level of function. But, for simplicity, let’s say I wanted a linear regression: y=mx+b. From an AI standpoint, this is where the AI is trained. The machine picks a random m value and a random b value.
After that, the machine will determine how far off it is from matching the imported data. Then it adjusts the values thousands of times–or however many times you want, really. This is referred to as an epoch, which is a complete training pass over the entire training set, such that each example has been processed once. In doing so, the machine has picked the values that give the lowest amount of error, rendering a line of best fit.
The Importance of Large Data Sets in AI
However, let’s say I added some values that were way outside the line of best fit. Then, the machine would recalculate the line based on the new data points.
Why did the line shift? Because I had a minimal data set. Small data sets are erroneous in ML because one data point can significantly throw off the line. However, if I had 10,000 data points and only one was an anomaly, it would average out and become negligible. Similarly, if I had five anomalies, their impact would be so small; I would never notice it.
This is the basis of Machine Learning: it requires massive amounts of data to produce accurate results. This leads me to my next point–the importance of good, big data.
Quality Data for AI in Construction
Let’s explore ChatGPT’s Machine Learning process–the Large Language Model or LLM. OpenAI indexed billions of data points to create their product, but it’s not always right.
It’s the same for all AI. Anything AI-related will always give you the popular opinion because its output is based on what it has seen. The output may, in fact, be wrong. This is because thousands of other circumstances are unknown to the machine. It learns from what has been input, rendering the output more about popularity than correctness.
“Analytics are only as good as the data behind them:
garbage in equals garbage out.”
Good data comes at a price, and yet all data is crucial. You just have to be sure the data sets are valid before using them to make business decisions. If your data set is riddled with issues, the algorithm will just learn to predict those issues because that is when it’ll be most “correct” when evaluating the training data set. Historically, construction (especially schedule-related) data has been ungoverned. The underlying issue is that any ungoverned data set needs to go through a cleansing process to make it analyzable.
For example, when it comes specifically to scheduling, all of the progress data and things of that nature are mainly guesses. Someone on-site looks at a task and determines it is 50% done with no rhyme or reason. Not to say the guess is wrong. But it is, in essence, the opinion of one person. After the schedule is updated with actual data, it may also present a picture that is not in line with their contractual obligations, which often results in “adjustments” being made to the schedule. Also, humans make mistakes. Computers only make mistakes when they have been taught wrong, which is why data needs to be validated.
Schedule Quality: A Crucial Factor for AI Scheduling
Why is it important to make sure you have a high-quality schedule? For machines to make decisions based on the data held within the schedule, the data must be of good enough quality for that to happen.
However, nobody will ever have a 100% clean data set. If you envision you will have a 100% clean data set, you’re living a dream. It’s just not possible, which is why data sets need to be large enough so that errors get averaged out. With data sets that are both valid and sufficiently large, AI will have a significant impact on the industry through better visibility into project outcomes than has been historically available.
AI in Construction Today
Most of the AI in the industry today focuses on data gathering, which is the best place to start. Construction has traditionally relied on manual processes, complex scheduling, and intensive labor, often encountering challenges such as delays, cost overruns, and safety concerns. However, AI creates new possibilities by leveraging advanced algorithms, Machine Learning, computer vision, and data analytics.
Just like any other industry, AI in construction will automate mundane, cookie-cutter tasks– eliminating those things you repeatedly do. With these advancements, companies can analyze vast amounts of data, gain valuable insights, and automate tasks that were once time-consuming and error-prone. Let’s take a closer look into various aspects of AI’s ability to enhance the industry today.
AI in Construction Safety
According to Forbes, AI-powered technologies like robotics, cameras, and computers reduce the risk of accidents and industries on job sites by monitoring the environment and alerting workers of safety hazards and other dangers that need improvement.
Alongside the ability to monitor worksite behavior and identify safety risks based on data from sensors, video feeds, or wearable devices, job sites have already begun to see the benefits of AI in construction.
Overcoming the Labor Challenge with AI in Construction
With the intense shortage of construction workers, AI can significantly aid in streamlining and automating certain processes, providing accurate and real-time insights for workforce management.
By eliminating the mundane process of progress tracking on-site through AI powered-systems like RFID tags or computer vision technology, up-to-date information on task progress and worker location can be obtained. Furthermore, advanced algorithms can analyze labor data, ensuring teams are assigned to the right tasks, thereby improving efficiency and optimizing labor utilization.
Optimizing Construction Scheduling with AI
From groundbreaking to closeout, the schedule contains every hiccup, RFI, change order, worker, duration, task, etc., that occurs to complete a project. With the advancements of AI in construction, the framework laid out by traditional methods can be enhanced to handle the intricacies and uncertainties inherent on the jobsite.
SmartPM is focusing on carefully and intentionally leveraging AI to strengthen our predictive analytics and empower customers to create more realistic schedules that are based on reality instead of hope.
The best is yet to come. Advancements within the umbrella of AI have already vastly improved project planning and scheduling, shortening the duration of project timelines. With the helping hand of AI, scheduling bottlenecks are minimized by the algorithm’s ability to allow real-time adjustments based on changing project conditions.
Every aspect of AI in construction–safety, labor tracking, financials, resource allocation–ties back to the schedule. As every aspect of AI in construction revolves around the schedule, its implementation brings about enhanced visibility. This visibility results in reduced disputes, elevated output quality, and more effective strategic planning and risk management.
