From De Facto Standard on Construction Sites to an Industry-Wide AI Platform: A Video Platformer's Evolution Toward Physical AI

Executive Summary

Envalith had the opportunity to attend the company's "Construction Domain Strategy Briefing," featuring the CTO, the head of the business unit, and customer companies (Shimizu Corporation and Iwata Chizaki). In the construction domain, Safie is at an inflection point, shifting the axis of its value proposition from "viewing video" — driven by the expanding rollout of cloud cameras — to "leveraging video" through AI and agents. 49 of the top 50 construction companies by revenue have adopted the company's services, and the installed base in the construction domain has reached roughly 50,000 units, placing its position within the industry firmly at de facto standard level. What management prioritizes most is not simply supplying AI, but building a structure in which the technology is embedded in customers' business processes and data value compounds through feedback loops. Through co-creation with general contractors, the company aims to codify the tacit knowledge of construction sites into explicit knowledge via video and audio, with a longer-term vision of taking end-to-end responsibility for the real-world deployment of physical AI, including robotics.

Message from the Company

Key Discussion Points from the Interview

• Establishing and Deepening a De Facto Position in the Construction Domain

  The company has been adopted by 49 of the top 50 construction firms by revenue, with several companies standardizing on it across all sites — evidence of advancing de facto status. Management attributes adoption decisions to ease of use in the field, security standards as a cloud camera provider, and support from customers' top management for the company's growth potential and scalability. A structure in which sales, planning, and development staff all visit sites to grasp the essence of customer problems is positioned as the source of competitive advantage.

• From "Viewing" to "Leveraging": Rapid Rollout of AI Solutions

  Safie AI Studio, released in February 2026, provides a platform on which simply registering an AI application immediately initiates video analysis on cameras already in operation. Within six months of launch, 13 solutions have been delivered to customers, with development speed lifted substantially through both in-house development and partner collaboration. In addition, the deployment of "Ailytics," an AI solution for detecting unsafe behavior, has increased installed units, accumulating training data that in turn generates new solutions — a virtuous cycle now occurring at the individual site level.

Envalith's Perspectives

• 01.

  Management Sees "Embedding into Business Processes" as the Essence of Competitive Advantage

• 02.

  Customer Co-Creation Accelerates Development Speed

• 03.

  Expanding Market Size Through Multi-Track Uplift in Per-Site Revenue

• 04.

  Challenges and Countermeasures

• 05.

  Risk Factors

Q&A Highlights

• Q

  If competitors enter the AI agent domain directly, does being a video-centric integrated platform constitute a competitive advantage?

  A

  Solutions or data alone are meaningless. The greatest competitive advantage comes from being deeply embedded in customers' business processes and, through continued operation, capturing additional data and feeding it back into the system. Video data is not an easily obtainable data source, so integrating it, embedding it into workflows, and circulating feedback dramatically amplifies its value. We see that as the definitive path to winning.

• Q

  Where do the barriers to entry lie on the business side?

  A

  On construction sites, a single day of the schedule is critical, so when a camera fails, you need a structure in which a partner is on site the next day for maintenance or delivers a replacement unit. That last-mile execution capability is a strength of the company, which has already built out a partner ecosystem, and is not something a typical AI vendor can easily replicate.

• Q

  Where are the concrete benefits of adoption showing up?

  A

  At large-scale building sites, site patrols that took roughly two hours per staff member can be replaced with video checks and voice dialogue via camera, freeing that time for documentation and coordination work. By combining solutions, there are cases where site trips that previously had to be made multiple times for certain processes were reduced to zero. Automatic tracking of process progress, linkage with 3D data, and automation extending to materials ordering are also being implemented on site.

• Q

  How will the tacit knowledge of the field be turned into an asset?

  A

  The starting point is recording veteran staff's intuition, experience, and judgment criteria via video and audio, leaving a data record of what they looked at and how they judged it. AI agents are implemented on site early, and the know-how of retired veteran managers is fed back to improve accuracy. Once AI matures to a practical level, device deployment expands and whole-site visualization advances, with both AI and real-world spatial data accumulating on the company's platform — a powerful asset for the real-world deployment of physical AI.