AI Takes on the Built World

AI is rebuilding the field of Architecture, Engineering, and Construction (AEC). The combination of AI algorithms with sensors and cameras changes the roles of engineers by making it easier to understand project limitations and find solutions to problems in real time.

The global construction market is projected to reach nearly $22.9 trillion by 2026, according to a July 2022 report by Research and Markets. Yet the construction industry has been frustrated by new concerns, including COVID-19-related supply chain disruptions and labor shortages. AI may be critical to increase the speed to execute capital projects and ensure work and structures are sustainable. AI can also help AEC companies overcome other challenges, like rising materials costs related to geopolitical conflicts.

On a more basic level, AI can also help project teams align, reduce workloads, and free up time to manage processes more successfully, enabling project managers to deliver better quality results on time and within budget.

A recent example of AI in AEC is the use of deep learning to generate efficient floor plans and develop heat maps of spaces informed by video footage. Although past projects have focused on how outdoor built environments were used after they were constructed, future uses of AI could predict public use of spaces and structures before construction. This gives architects and contractors better estimates of what structures to build and what features are necessary to maximize use.

Another example comes from a Tel Aviv-based software company called Buildots, which seeks to filter out opinions and assumptions from construction progress reports. One of Buildots’ primary concerns is that work done in the field often does not sync with numbers from a build’s plan and budget. The company’s first step to address the issue was to equip workers with helmet-mounted 360-degree cameras.

The cameras share images with Buildots’ software, which uses AI to extract data from the shots. The company says its program can generate accurate, objective reports based on the data and images.

“Managers can plan the work for next week and receive an automatic report showing which parts of that plan were completed and how that compares to previous weeks, be alerted about work that was left incomplete, and more. Buildots capitalizes on the advantages of AI… by tracking tens or even hundreds of thousands of elements on each construction project,” explained Aviv Leibovici, co-founder and chief product officer of Buildots, in an interview with engineering.com.

And then there’s generative AI art software like Midjourney and DALL-E, which use neural networks to transform users’ text inputs into AI-generated images. The text-to-image software can quickly imagine an architect’s ideas using a few key phrases like “metal futuristic house” and “in the style of Le Corbusier.” Le Corbusier was an influential French architect and planner who emphasized access to shared greenspace.

Architects benefit by using generative AI tools because they can easily add and subtract phrases to alter their concept. They can also quickly alter a rendering by telling the software to add details like rooms and gardens. But there’s a catch, too.

The data that AI algorithms generate are only as good as the algorithms themselves. As engineers retrieve data from images such as photos and video, they can review how well the algorithms are performing. They need to determine whether the information they obtain helps them guide projects to better outcomes.

For example, AI tools like Midjourney and DALL-E could replicate an existing architectural plan, plagiarizing a real architect’s work. In addition, the software has a tendency to bias. Its responsiveness to user input may result in over-representing certain architectural styles. AI also has a tendency of utilizing more digitally generated images than real photos.

Software developers can address the concerns regarding biases by providing the program with more data to avoid focusing on particular styles and images. Developing an image-to-image process, wherein architects could share different styles for an image, could help lessen biases. Models for software could vary according to country, culture, and region.

AI can help AEC companies estimate the amounts and timings of cost overruns, but it cannot change certain facts about the industry itself. Demand is cyclical, many projects are complex, and hundreds of aspects of projects are extensively regulated. As an example, across the globe, contractors face long time frames to obtain building permits. In France, the processing time for a permit application for a house is two months. In 2020, it took an average of 213 days to obtain a construction permit.

As of 2020, in many countries, the number of interactions needed to obtain a construction permit were high. China required an average of 18 interactions, while the U.S. required 16. In addition, the construction industry faces a graying workforce, a shortage of workers, and a lack of experienced people to train apprentices and new hires.

AI has become valuable because it allows the AEC industry to do more with less. It also offers the potential to use data and solutions from past projects to guide future builds and workers. In McKinsey’s 2020 report on the construction industry, which was partially entitled “the next normal,” the consultancy group stated that digital technologies like AI have the power to enhance collaboration and pull contractors toward more data-driven decision making. The group’s prediction that innovations like AI will change how companies engage with partners has already come true.

When Leibovici and his co-founders began Buildots in 2018, they were astounded at how little data project managers typically had access to—and impressed at how they succeeded in spite of it. They knew there was a better way.