Obayashi Corporation, a Japanese construction and engineering firm, cut the time spent researching regulations and organising requirements by 67% during a trial of Tektome’s construction-specific AI platform.
The contractor found that using Tektome reduced the work to an average of 1.4 hours per person per week, down from 4.3 hours previously. The figures came from testing with practitioners, as Obayashi examined generative AI in the early stages of building design.
Tektome, a developer of AI solutions for the construction sector, provides the AI agent platform. According to Tektome, Obayashi’s Architectural Design Legal Department initially reviewed the agent’s accuracy and usability. Currently, designers are employing it on multiple projects during the trial phase.
An internal survey of Obayashi’s design departments found that researching laws, ordinances and standards was the most time-consuming task.
AI takes on early-stage regulatory research
Building projects in Japan require designers to check legislation, including the Building Standards Act, the Fire Service Act and energy conservation rules, as well as regional ordinances and official notifications.
During the initial planning and design stages, information needed for these checks can be distributed among different parties. Designers must determine which requirements apply while many aspects of a project are still being worked out.
Obayashi uses generative AI to extract and organise relevant legislation and design requirements, search for related information, and highlight matters requiring attention.
The system combines regulatory data with Obayashi’s knowledge, including previous interpretations of legislation and design verification experience. Supporting material allows designers to check the sources behind the results.
Tektome said its regulations research agent reaches a conclusion and cites the relevant statutory provisions. If the agent cannot resolve an issue, it marks the matter as requiring confirmation rather than giving a definitive answer. Final judgement remains with people.
Obayashi records 67% reduction in research time
Obayashi tested the system with practitioners conducting regulatory research during early design.
Average time spent on regulatory research and requirements organisation fell from 4.3 hours per person per week to 1.4 hours with Tektome, a reduction of about 67%.
Obayashi also reported that 100% of respondents felt the system effectively organised issues and clarified which points needed checking during legal verification. The company did not disclose the number of respondents.
Among the reported responses, 66% cited using the system as a sounding board before consultation, and the same proportion cited identifying potentially relevant laws and ordinances. Half cited reviewing previous considerations and identifying statutory provisions or supporting grounds, while 16% cited identifying and organising issues.
Regulatory checks move earlier in the design process
Obayashi’s Architectural Design Legal Department handles designer consultations and reviews design content from a legal perspective. In the pilot, the AI agent sits between designers’ preliminary research and the department’s specialist review.
Designers can question the agent as issues arise, organise relevant legislation, and identify points that still need confirmation. Before approaching the legal department, they can also use it to check whether they have overlooked anything.
Obayashi aims to make regulatory checks more continuous as designs develop, rather than concentrating them at individual review stages. It also expects earlier identification of regulatory issues to help uncover risks sooner and reduce later rework.
The agent can present information in tables and diagrams for internal use. Its answers include references so designers can check the underlying provisions rather than rely solely on the AI response.
Obayashi has previously applied AI to design work
The Tektome pilot is not Obayashi’s first foray into generative AI. In its 2024 corporate report, the contractor said it had released O-Bot Assistant, known as Chat O, an internal ChatGPT environment powered by Microsoft’s Azure OpenAI Service. Obayashi also provided employees with generative AI guidelines, information on how to use the technology, and internal training.
In June 2025, Obayashi announced an AI programme that automates structural section design by drawing on knowledge accumulated by its structural engineers. The company said the system could reduce the section-design process, which had previously taken about a week, to one day.
Obayashi plans to build on the findings of the Tektome trial as it develops its use of generative AI in the design process. Tektome said it would continue to evaluate the pilot with Obayashi and consider future uses of the technology.
