Launch of "zenshot AI," a Physical AI Agent for the Construction Sector

Zen Intelligence Inc. has launched "zenshot AI," a physical AI agent designed for the construction industry to address labor shortages. By simply taking photos while walking on-site, the AI understands the environment and automates tasks like safety checks, quality control, and progress tracking, aiming to standardize management quality. The system is powered by a specialized Vision-Language Model (VLM) developed through Japan's GENIAC project, enabling it to process spatial and temporal data beyond simple image recognition.
product_launchNQ 100/100出典:PR Times

📋 Article Processing Timeline

  • 📰 Published: April 14, 2026 at 21:00
  • 🔍 Collected: April 14, 2026 at 12:31
  • 🤖 AI Analyzed: April 14, 2026 at 18:11 (5h 39m after Collected)
Zen Intelligence Inc. (Headquarters: Chuo-ku, Tokyo; CEO: Daiki Nozaki; hereinafter "Zen Intelligence") has announced the launch of "zenshot AI," a physical AI agent for the construction sector. zenshot AI is an AI Agent that understands site conditions and automates parts of construction management tasks, including safety, quality, and process control, simply by walking around a construction site and taking photos with a camera. This service implements a Vision-Language Model (VLM) specialized for construction sites, developed under the "GENIAC" project, a generative AI foundational model development support initiative promoted by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). Its key feature is the ability to understand the construction site as a spatial environment by considering not only image recognition but also location information, correspondence with drawings, and continuity with past recordings. Under the vision of "Increasing Supply Capacity with Physical AI Agents," our company will promote business transformation through AI in industries involving physical spaces, starting with construction. **Labor Shortages in the Construction Industry and AI Utilization** In recent years, the advancement of generative AI has rapidly transformed operations in white-collar fields, targeting tasks like document creation, information retrieval, and customer service. However, in physical domains like construction, AI application remains limited because the starting point of work is the physical space itself. Within the construction industry, the shortage of site supervisors and skilled workers has become a serious issue. Construction management involves multifaceted checks and judgments regarding safety, quality, and processes, and requires appropriate instructions and record-keeping based on the specific conditions of each site. Consequently, differences in experience tend to manifest as differences in work quality. As the labor shortage progresses, maintaining construction quality and securing supply capacity have become critical management challenges for construction companies. What is needed to address these challenges is not an AI that only handles already digitized information like text and forms, but an AI that can grasp what is where and in what condition on-site, and based on that understanding, make judgments, give instructions, and create records. Our company defines an AI that understands the site starting from such physical spaces to support and automate practical work as a "Physical AI Agent." In the construction domain, it is crucial to understand the site not just from image data, but based on data that captures the site space over time, considering positional relationships, spatial structures, correspondence with drawings, and continuity with the past. **"zenshot AI" Overview** **Automating the Construction Management Work of Site Supervisors Just by Walking with a Camera** zenshot AI is a physical AI agent specialized for the construction sector. It can be used simply by walking around the site and taking photos with a camera. Based on the acquired data, the AI grasps the situation of the construction site and automates parts of the construction management work. Traditionally, these tasks were performed by site supervisors who patrolled the site, making checks, judgments, and records based on their experience and knowledge. zenshot AI aims to reduce the workload of site supervisors and standardize management quality by having the AI take on a portion of these construction management tasks. **Equipped with a Construction-Specific VLM Developed in GENIAC** At the core of zenshot AI is the construction-specific VLM developed in GENIAC. This model is based on multimodal site data that integrates spatial, object, and semantic information of construction sites over time, and was developed with the aim of understanding the situation and context of construction sites with high precision. This enables information processing that leads to judgments necessary for construction management, not just by recognizing objects in an image, but by capturing the entire site as a space. **Main Features** zenshot AI supports and automates the following construction management tasks: * **Safety Hazard Identification:** The AI checks the entire site's condition and generates指摘 for hazardous areas and inadequate safety measures. * **Process Progress:** The AI understands the chronological changes of the site, allowing for the management of process progress. * **Quality Inspection:** The AI extracts inspection points from design drawings and automates inspections by collating inspection standards with on-site data. * **Construction Management Records:** The AI reads the daily construction status of the entire site and the design drawings to extract construction management record photos. This allows the AI to take on a portion of the confirmation, judgment, and recording tasks that previously depended on the knowledge of experienced site supervisors. We aim to create an environment where even less experienced site supervisors can perform construction management tasks to a certain standard. **Future Development** In the future, our company will work on further automating construction management tasks through zenshot AI. Based on the construction-specific VLM developed in GENIAC, we will improve the accuracy of site understanding and expand the scope of its application, thereby contributing to the improvement of supply capacity in the construction industry. **Inquiries about the Service** For inquiries about zenshot, please use the site below. zenshot service site: https://zenshot.ai **About Zen Intelligence Inc.** Zen Intelligence is a Physical AI startup that develops Spatial Intelligence, which perceives, reasons, and acts within the 3D space of physical sites and its changes over time. Since its founding, the company has consistently focused on developing and providing AI and robotics technologies based on site data. Currently, it offers "zenshot," a construction AI product for the construction industry, centered on 3D Vision and foundational models. Through Physical AI, the company aims for "Re-Industrialization," which involves reconstructing the very nature of industry into an AI-native form. **** Company Name: Zen Intelligence Inc. (formerly SoftRoid Inc.) Location: Sumitomo Fudosan Yaesu-dori Bldg. 6F, 2-14-1 Hatchobori, Chuo-ku, Tokyo Representative: Daiki Nozaki, Representative Director Established: July 21, 2020 URL: https://zen-intelligence.ai/ **** * Selected for METI/NEDO GENIAC 3rd Term: Development of an AI foundational model to automate construction site management (2025) * Selected for Toyo Keizai's "100 Amazing Ventures" (2024) * Selected for Forbes JAPAN's "100 Notable Japanese Startups of 2024" (2024) * Selected for Nikkei Architecture's "100 Ventures Updating Architecture" (2022) * Selected for IPA Mitou Advanced Project: Realization of a robot service that patrols construction sites to collect and analyze data (2021) * Selected for the Ministry of Land, Infrastructure, Transport and Tourism, Kanto Regional Development Bureau's project on the development, introduction, and utilization of unmanned and labor-saving technologies at construction sites (2020)