Practical Case Study on Integrating AI Agents into Construction Management Shared with 150 Participants at 4th Architecture AI Study Group

The 4th Architecture AI Management Study Group was held in Yaesu, Tokyo. It was the first time the event was held simultaneously online, with approximately 150 people participating.

LIFEFUND Inc. (Hamamatsu, Shizuoka; CEO: Takuma Shirato) held its 4th Architecture AI Management Study Group for construction company managers on Tuesday, June 2, 2026, in a hybrid format with a Tokyo venue and online, attracting a total of 150 participants.

The day's events included sessions on LIFEFUND's internal use of AI agents, a case study by EXCEED GROUP on the cross-sectional analysis of gross profit management data, and an AI-based quantity surveying workshop using drawings. A participant survey (n=63) recorded an overall satisfaction rate of 4.21/5, with CEO Shirato's first lecture achieving 4.65/5. This study group shares practical case studies on how to integrate AI into construction companies' management decisions, site operations, and data utilization, rather than just using it as a standalone tool.

Industry Shift: From 'How to Use AI' to 'How to Integrate into Operations'

The architectural industry faces a worsening labor shortage and high dependency on individual skills, yet the adoption of generative AI remains limited.

According to Teikoku Databank's "Survey on Bankruptcies due to Labor Shortage (FY2025)" released on April 9, 2026, bankruptcies due to labor shortages hit a record high of 441 in FY2025, with the construction industry accounting for 112 cases, or 25.4% of the total. Source: https://www.tdb.co.jp/report/economic/20260409-laborshortage-br25fy/

Furthermore, ANDPAD's "Survey on AI Utilization in the Construction Industry" released on February 12, 2026, which surveyed 2,000 construction professionals in December 2025, found that 34.8% use AI in their daily work, while 47.3% have "no plans to use" it. Source: https://andpad.jp/news/20260212

While the importance of using AI is becoming widely recognized, a major challenge for construction company management is determining 'which tasks, in what order, and to what extent to delegate to AI.' The 4th Architecture AI Management Study Group addressed this by sharing methods to use AI not as a standalone tool, but as a system that divides work within the company by connecting it with internal data, business rules, and human approval processes.

Lecture 1: Roadmap for AI Agent Management and Concrete Strategy for AI in Construction

In his keynote speech, Takuma Shirato (CEO of LIFEFUND Inc.) delivered the management message that "the phase of using AI as a tool is already over."

LIFEFUND, based in Hamamatsu, Shizuoka, operates in custom housing, renovation, real estate, and AI education. The company has an annual revenue of 2.71 billion yen and builds 115 houses per year. Through three years of continuous, company-wide AI training, they have achieved 1.6x sales growth and 1.3x employee growth, along with a 4% increase in real wages (while increasing annual holidays from 110 to 120 days).

The Main Battlefield Shifts to AI Agents

From the 'era of standalone tools' like ChatGPT, Claude, and Gemini, the shift is towards an 'era of deploying AI comprehensively' centered on foundational AI platforms and Agents. Mr. Shirato organized AI utilization maturity into five levels.

Image of the 5 Levels of AI Management Level 1: Individual Use (Individuals start using tools like ChatGPT) Level 2: Departmental Adoption Level 3: Company-wide Standardization (All employees use AI in daily work) Level 4: Proprietary Data Integration — Feeding company-specific information to AI Level 5: AI Agent Collaboration — Working with AI teams

What is Context?

Mr. Shirato defines 'context' as 'company-specific information that humans know, but the AI does not yet.' By systematizing and providing information such as reasons for past failures and successes, on-site judgment know-how, relationships with subcontractors, a sense of price negotiation history, and unspoken rules and customs, the AI becomes a practically useful entity.

"An AI without context is like a new employee with a high IQ who doesn't know the company. The more context it gets, the closer the AI becomes to a veteran who understands business flows, decision criteria, and past history," said Takuma Shirato.

Takuma Shirato's AI Management OS — 'Handle This' Sets the AI Team in Motion

A live demonstration of LIFEFUND's operational "AI Management OS" was performed. The system, where simply sending 'Handle this' on Chatwork triggers a Management Agent to pass tasks to a team of specialized agents (Estimation Agent, Schedule Management Agent, Customer Support Agent, Analysis Agent) that work autonomously, was revealed.

"The company has multiple AI agents. Each has its own responsibilities and operates daily," said Takuma Shirato.

Decomposing Construction into 117 Tasks — A Concrete Roadmap for AI in Construction

In the latter half of the lecture, Mr. Shirato applied the AI Management OS concept to construction sites. LIFEFUND inventoried its construction tasks sequentially, breaking them down into 117 tasks. These were classified by difficulty (K-1: 43, K-2: 41, K-3: 33) to clarify the priority for AI implementation and the tasks requiring human judgment. A 5-step roadmap for AI in construction (① Decompose → ② Small-scale efficiency improvements → ③ Feed data → ④ Delegate → ⑤ Improve) was also presented.

[Implications for Managers] AI in construction is not about 'cost reduction' but is a strategic weapon to increase the number of projects without increasing staff. Starting with tool selection based on 'what AI can do' leads to a standstill. The starting point is to decompose business flows and create the context for the AI to read—this change in mindset is the most necessary perspective for today's construction managers.

[Implications for the General Public] The time that construction managers spent on 'administrative work instead of site management' is being freed up by AI. A system where 'supervisors can concentrate on the site' contributes to construction quality and site safety.

Lecture 2: Revenue Structure Revealed Through AI Analysis of Data from 140 Projects

The second lecture was given by Yoshinobu Togashi, CEO of EXCEED GROUP from Yamagata Prefecture. His company is the number one fastest-growing company in Yamagata in terms of 5-year cumulative growth and achieved the number one ranking for new construction starts among locally capitalized companies in FY2025. He shared the practical process of using AI to conduct a cross-sectional analysis of accumulated project data.

Cross-sectional AI Analysis of 140 Projects — Three Trends Emerged

Cross-sectional AI analysis of data from 140 projects confirmed a deviation from planned profit in 56 cases (approx. 40%). A closer look at the data revealed three structural trends.

① Orders without approval flow: Accumulation of orders made based on on-site decisions. ② Dispersed ordering channels: Ordering channels were not unified across projects, preventing overall optimization.