THE ESSENTIALS IN 30 SECONDS
- Kajima has deployed FieldAI-powered robots at more than 10 of its construction sites in Japan within about a year of a 2025 enterprise service agreement.
- The robots have completed over 2,000 autonomous missions across different robot types, tasks and sites.
- Uses include site rounds, 3D scanning, progress tracking, off-hours security patrols and first response after earthquakes or heavy rain.
Japanese construction group Kajima announced on November 5 that it has begun full-scale deployment of robots driven by autonomy software from FieldAI, a California-based physical AI company. According to Kajima, the robots are now working at more than 10 of its construction sites in Japan, and have completed over 2,000 autonomous missions across a range of robot types, tasks and sites. The contractor stresses that this is continuous use on active sites whose layout changes daily, rather than demonstrations under controlled conditions.
From pilots to an enterprise agreement
The two companies have worked together since 2023, testing the technology on real job sites. In 2025 they signed an enterprise-level service agreement that shifted the partnership from one-off projects to wider deployment, and Kajima reached the 10-site mark roughly a year later. No contract value has been disclosed, and neither the release nor the coverage names the specific robot models in use.
The system is built around an autonomy unit that combines cameras, LiDAR and an onboard edge AI computer, and that Kajima says can be mounted on different robots. Processing happens on the robot rather than in the cloud, which lets it pick a safe path in real time and operate indoors or underground where satellite positioning is unavailable. The software relies on FieldAI's Field Foundation Models, which the companies describe as mixing data-driven learning with physics-based reasoning and estimates of uncertainty.
What the robots do
On first start-up, a robot explores an unfamiliar area and builds a 3D map, which it then refreshes on later rounds while avoiding obstacles, taking photos and capturing point clouds. Data is uploaded at the charging station and made available on the web, and staff can send robots to a destination from a tablet or PC while watching live video. Kajima uses the 360-degree images and colour point clouds to track construction progress. Robots also patrol sites at night, early in the morning and on non-working days, switching on bright LEDs and an audio alert if they detect an anomaly. After earthquakes or heavy rain, they can enter a site before it has been declared safe to collect images and video.
Kajima presents the rollout as a response to labour shortages in construction, freeing engineers from routine checks so they can focus on safety, quality and scheduling. It plans to extend the software to more sites and to other technologies it is developing, including construction machinery. Independent data on time saved or error rates has not been published.
