A new building off Columbus Circle is now the first project in New York City to implement artificial intelligence (AI) construction technology to significantly increase efficiency in terms of costs, labor, management time and resources.
The 26-story, 197,000 sq. ft. high-rise residential tower, located southwest of Central Park at 1841 Broadway, uses the AI construction platform Buildots and is being developed by Global Holdings Management Group.
The technology is currently being used on sites across North America, the UK, Europe and the Middle East.
Buildots CPO and co-founder Aviv Leibovici noted that construction is predominantly manual work, unlike car manufacturing and other highly automated industries that rely on machine status checks. With so many people walking around a site, often from a composite of different businesses, there are very few ways to know when and how something was done.
“There’s just a lack of credible information for them to use,” he said.
AI SET TO TRANSFORM CONSTRUCTION INDUSTRY
According to Leibovici, the tech was built to serve as a source of inspiration for project managers across the globe. With Buildots AI technology, managers can quickly go into the system and understand what exactly is going on with the project, where the gaps are and what the pace of the build looks like. Instead of spending time trying to determine and quantify various aspects of the project, Buildots frees them up to make decisions and prioritize changes.
To achieve this, the Buildots AI platform captures site data via helmet-mounted 360-degree cameras, which is automatically analyzed using propriety AI algorithms. The platform provides project management teams with accurate progress reports and visual analysis.
While such data collection processes are relatively simple, Leibovici said the new tech provides unique challenges. With an acknowledgment of his befitting use of a construction pun, Leibovici said the system requires a “concrete process” on-site without the time to change how things are done in the overall development plan.
Additionally, managers need to choose how they represent the world to the AI carefully. A sequence of activities written across the top of a page of the software needs to track the exact scope of that individual cell and define it properly.
GENERATIVE AI WILL ADD TRILLIONS TO GLOBAL ECONOMY: MCKINSEY
For example, Leibovici said to imagine a scenario where a manager labels one of the activities as “walls.” But, during the weekly subcontractor meeting, the general contractor asks whether the walls were framed in the floor. If the system has walls at 20% completion, how can the person know if the 20% refers to the percent of walls framed, or walls completely done? This parameter discrepancy highlights the importance of correctly prompting the AI to analyze aspects without room for interpretation.
This process in which the AI is expected to correctly identify various objects and their context within a construction site has incrementally improved since Buildots was founded over three years ago. In the early stages, the AI frequently forwarded object analysis decisions to a human being on the team for review. However, unlike something like language, construction materials are largely ubiquitous.
Once the algorithm was able to identify drywall, that element became identifiable on all construction sites across the globe. While each build provides its own unforeseen challenges, Leibovici said the model is now “far more effective,” which affects how long it takes to process data and helps lower the cost of running the tech.
LOWE’S FOUNDATION INVESTS $50M TO ADDRESS ‘CRITICAL’ TRADESPEOPLE WORKER SHORTAGE
A study conducted by Buildots found that, on average, only 46% of areas are utilized on a project, which means that over half of all available areas are not being worked on at any given time throughout the week. Unsurprisingly, the data found that when projects succeed in working on more areas, on average, they get more done and finish on time. Leibovici hopes that Buildots AI system and software interface can help managers understand what is blocking them from working on the maximum number of areas and help them cultivate a plan to improve.
“There’s a very interesting term that I’ve started hearing from our customers. They call it scheduling false hope, which is when, because we don’t have concrete truths, we just say it like, yeah, yeah, it will be fine. I’m going to finish on time. And I even update my schedule with a new plan of how I’m going to finish on time,” Leibovici said. “But if that new plan means I need to work three times as quickly as I have done until now. And in the first week of doing that new plan, nothing changes on site, then that’s false hope.”
With Buildots, Leibovici said builders have better operational control of the project and can immediately see if they are working at the right pace. The AI can even determine the efficiency of each portion of the supply chain and automatically see who is completing their plan and who is not.
“We have customers that have done that, and we showed how they reached very significant results, even finishing ahead of time, which is something that doesn’t happen all that often in construction, unfortunately,” he said.
On the first project that Buildots assisted with that used this methodology, Leibovici said the site saw its predicted delay shorten by two months and the percentage completion per week move from around 20% (standard in construction projects) to over 50% in just a few short weeks.
Despite the rapid advancement of AI in the last several years, Leibovici believes the gap between fully autonomous construction robots and the current use of AI to interpret large swathes of data is massive. In his opinion, this data is merely a tool for the people who manage these projects, and the AI is unlikely to decide on project logistics independently.
“It’s always going to be about professionals using information that they get from machines or AI to use that to make decisions because the decision are based on so many factors that the machine will never know,” he said. “It’s based on relationships, it’s based on who’s at risk, and we need to help them out so that this isn’t the problem here. It’s based on designs and priorities and clients and sales and, I don’t know. Maybe in 100 years, AI will also make the decisions. Who am I to say? But no time soon.”
Read the full article here