In early 2027, in the German city of Koblenz, a construction team will begin work on a new residential “micro living” project. But some of these workers don’t wear hard hats: a part of the team is made up of four-legged, AI-powered robots designed by London-based startup All3, which is developing a system to automate large-scale construction projects.
“It looks pretty much like ‘Starship Troopers’. A shipping container arrives at the construction site, the doors open, and the robots autonomously go onto the site, getting all the instruments they need from the same container,” said co-founder and CEO Rodion Shishkov.
All3’s Mantis robot training in the company’s lab. Credit: All3
The Koblenz project will provide an early commercial test of All3’s approach. The company is aiming to prove that powerful new AI models, combined with modern robotics technology, can transform the construction industry: a sector that has been slow to digitize and automate.
“While every other industry in the world has been progressing in productivity in the past 50 years, productivity in construction has actually gone down rather than up,” Shishkov said. “The housing crisis, and unaffordable housing, are significant problems. When you dig a bit deeper, you see that high costs and low labor productivity are a big reason for these problems.”
Productivity growth: Federal Reserve Board, "Total Factor Productivity by Industry" (2023), federalreserve.gov. Industry GDP shares: U.S. Bureau of Economic Analysis, GDP by Industry accounts, ~2019.
All3 is part of a growing crop of startups using AI to address complex inefficiencies in construction. In 2025, 77% of construction tech funding went to AI-enabled solutions, compared to 35% in 2024.
But construction sites are less controlled than factories. Applying data and automation to messy building sites is a challenging task, which, in some cases, requires rethinking construction from the ground up.
Automating uniqueness
Shishkov said that construction has historically resisted mass automation partly because, unlike a car or laptop which can be made identically on a production line, every urban building project presents different problems to solve.
“Every building plot is unique. Neighboring buildings are unique. The visuality of the street we are building on is unique, which means the building has to be unique as well,” he said.
All3 is developing an integrated construction system with three core elements: an AI-led architectural design platform; factories that automate fabrication of individual building components; and its four-legged robots that put those components together on the site. Shishkov said the integrated approach is intended to keep building designs within the bounds of what today’s robotics technology can achieve.
All3’s architecture platform helps users to design buildings that fit into tightly constrained spaces. Credit: All3
“We realized we would need every building to be designed not just for the plot, not just for compliance, but also for the constraints of robots that can actually be built and be operated today,” he told The Infinite Loop.
All3’s robots aren’t yet able to automate every stage of the construction process – some human labor will still be required – but the company said that buildings it designs can cut up to 30% of costs compared with conventional construction methods. It also said project timelines can be reduced by up to 50% and that CO2 emissions can be reduced by up to a quarter.
“Beyond human capability”: Adding AI to conventional job sites
All3’s robots present a futuristic, sci-fi vision of how construction could look in the coming years, but other companies are using AI to drive efficiencies on more traditionally run building sites.
Buildots, founded in 2018 in Tel Aviv, has developed a platform that combines sensor data from lidar and cameras with computer vision models, to track construction site progress and measure progress and identify activities at risk of delay. It does this by ingesting 3D models from architectural designs and plans, and comparing installed work with the model and planned construction schedule.
Workers capture 360-degree images for Buildots using sensors like cameras mounted on hard hats. Credit: Buildots
“As a superintendent, I can have as many as 20 different trades working at the same time. Each and every one of those will employ dozens of different workers. Traditionally, I don't really have the ability to calculate the precise progress that they’re making on a daily and a weekly basis,” explained Amir Berman, the company’s VP of industry transformation.
“Without Buildots, they might think they're kind of on track, but then eventually realize something like: ‘Oh shit, I am 12 months into the job now and I’m noticing that I haven’t started my electrical work. That should have started three months ago.’”
Berman said that Buildots helps contractors to avoid costly delays that impact tight margins on construction sites, by being able to predict what jobs need to start and when. He added that keeping track of so many shifting and interacting variables is a perfect job for AI, which can spot patterns and recommend how to optimize a project better than any human project manager.
“Without the ability to comprehend the status on the job site at scale, it becomes something that is beyond human capability, no matter how good your team,” Berman said.
Laying new data foundations for construction
Buildots said it has now analyzed more than 425 million square feet of construction site data, and the company is now looking to turn this vast dataset into a platform that can inform better building projects before they’ve even begun.
“We started noticing that there are different patterns for what dictates a good project, and what dictates that a project has higher possibility of risk along the years,” Berman said. “Companies measure every project and its own original schedule, but who told them that that original schedule was good?”
Buildots is now developing a new set of industry metrics and benchmarks to help project planners optimize construction from the word go, something it believes will reduce waste and the industry’s environmental impact globally.
In comparison, All3 is still in relatively early days when it comes to building a proprietary dataset, as it prepares to begin construction work on its first project, but Shishkov said that its robots will quickly learn and improve based on the information their sensors gather.
“We have literally terabytes of data from video cameras, lidar, radar and other sensors, which all goes to the cloud to enhance the future behavior of our models,” he told The Infinite Loop.
As All3’s Mantis robots get smarter, they will begin taking on more tasks on the construction site, with Shishkov predicting that the company will be able to reduce the need for human labor by 10-12 times.
And, in an industry that is facing acute labor shortages, the efficiencies that All3 and Buildots are trying to drive are likely to be welcomed by developers around the world that are grappling to deliver projects on tight margins.
