Visible & Invisible AI: Why We Need Both in Construction
The best technology often feels like magic. You drag a finger on a touchscreen and the desired result happens instantaneously. With a minimum amount of physical and mental effort, you’re productive.
AI has the potential to make most technology you use feel like magic. We’ve already said 2023 will be a tipping point for construction technology adoption thanks to AI, but are builders aware of how much AI they are already using?
It’s critical to understand the invisible AI you’re accessing daily, not just the visible AI that’s getting all the credit. Invisible AI technology has improved construction management for decades, and it’s only going to get better.
Where is AI in construction most visible and most understood?
Did you know some of the earliest AI software development for construction emerged in the 1980s and 1990s? Systems were developed to simulate human decision-making and workflow expertise for processes like structural design and project scheduling. While the capabilities were limited, they were artificial intelligence nonetheless.
Most of those original systems have long been shut down or replaced by modern machine learning algorithms and databases. Today, we see and understand AI in several ways throughout a construction project:
Reality Capture
Drone-based photogrammetry and 3D laser scanning equipment, like OpenSpace and Reconstruct, use AI to capture and process massive datasets of imagery. AI then converts these datasets into 3D models that can be searched, analyzed and monitored for construction progress. Anyone who used a camera in the 90s can understand the difference in outputs between a camcorder and a drone-mounted camera equipped with AI.
Schedule Optimization
In construction scheduling software, like InEight and ALICE Technologies, users can clearly see AI in action when asking the software to analyze varying schedules for optimal resource availability, sequencing and variability based on historical data. What once took digging into files of old project schedules, AI now analyzes and deconflicts on-screen in real-time.
Generative Design
BIM was probably the first construction software to make contractors truly appreciate AI, thanks to companies like Autodesk and Vectorworks. The way generative design AI works through BIM to create and iterate building and component designs is like watching a magic show from backstage. You may not see every sleight of hand but you see enough to appreciate the expertise of the illusion.
Construction Robots
Autonomous equipment, like SafeAI, and robots, like MULE and SAM, may be the most tangible examples of AI on construction projects. They use computer vision, machine learning, and other AI to automate physical tasks, like bricklaying, allowing us to very literally see AI at work.
Digital Twins
To put it simply, digital twins are digital representations of a real-world asset. For example, BIM is full of digital twins of new builds and renovations. Digital twins, like those managed in Trimble Connect and Tekla, give stakeholders the ability to analyze, manipulate and iterate on decisions and changes to simulate outcomes before taking action on the real-life asset.
Safety Monitoring
AI, especially through video monitoring and analysis, allows contractors to analyze large amounts of video feeds and detect unsafe conditions, PPE compliance, and other hazards. With solutions like Twenty20 and Newmetrix, users can clearly see the power of AI searching multiple live video feeds and flagging problems in real-time (or even just before they happen), quicker than any human viewer could.
Predictive Maintenance
Smart tools have grown in popularity as builders understand the power of tracking equipment and initiating maintenance before it’s too late. AI is very tangible and visible in the alerts and tracking placed on materials used every day through solutions like BuildingsIOT and CIM.
AI capabilities are most visible where they augment or replace manual or repetitive human work.
Physically interacting with AI drives confidence and understanding, but behind-the-scenes-AI will soon drive so many construction workflows that it’s imperative it’s also understood.
Where is construction AI least visible and least understood?
Software’s goal is usually to improve the speed, efficiency and accuracy of workflows - this is often done through automation and analysis, and these often depend on some type of AI. So much invisible AI exists in construction software and hardware today. While you could argue it’s “good” that contractors didn’t know - the solutions were that seamless and easy to use - it has also hurt the industry’s collective confidence and trust in AI solutions. Let’s discuss where invisible AI often hides to start building clarity and confidence for all users.
Natural Language Processing
When your contracts, project plans, or emails start providing suggestions and flags, that’s AI working behind the scenes to generate insights based on the text or images scanned. This type of AI is purposefully subtle so it doesn’t feel invasive, but understanding it can help you apply it in more areas.
Predictive Analytics
The goal of so many AI solutions today is to process the massive amounts of dormant data from the last few decades into predictions for the next few decades. The data analysis and machine learning that identifies project cost, schedule, design and subcontractor risks is complex and opaque to most. The results feel like magic because the algorithms behind the scenes aren’t showing us their math.
Compliance Tracking
AI validation of documents and designs for regulatory compliance lacks transparency. Variation in document types and legal language make AI work extra hard in the background to identify and track risks.
Automated Traffic Monitoring & Material Tracking
AI that tracks site vehicle movements and material transport works seamlessly in the background, only flagging project managers of changes or issues. Just like it’s hard to understand how Amazon delivers our packages in two days, algorithms that govern geographical movement on construction sites are complex. Users aren’t shown how things arrive, just that they’ve arrived.
Chatbots
You could argue that AI chatbots are one of the most user-friendly, understandable AI user experiences. What’s less clear, is the natural language processing and the data source behind every answer a chatbot gives. Even ChatGPT has come under pressure for not releasing more details about where they pull their data from in their latest version.
Fraud Detection
AI pattern recognition and machine learning algorithms for fraud detection are based on large historical datasets of fraudulent activities. Since those datasets aren’t visible, it’s hard to understand exactly how and why certain activities are flagged. Users may not care in the case of obvious fraud, but when the flagged activity isn’t clearly a fraudulent activity, what’s happening with the AI in the background matters.
AI applications that operate behind the scenes on vast amounts of data without direct human interaction or prompts tend to be less visible and understood. Improving transparency into these "black box" AI systems will build more trust and adoption.
Humans want to use technology they understand in order to trust it when things go wrong.
Invisible and visible AI are critical to construction projects and will continue to be. The burden is on AI startups and technology providers like us to better explain and sell any invisible AI we offer. This kind of AI can have huge impacts on winning bids, project fees, team productivity and so much more - but it’s harder to grasp for those without a tech background. Is there an artificial intelligence software or technology that feels like a “black box” to you? What do you wish an AI developer could explain more clearly to you? Leave us a comment or write to us at blog@documentcrunch.com.