Reflections After the DX Expo:

The Great Agent Era and the Future of Industrial Intelligence

By Mizuki Marumo, CEO of AIM Inc.

---

AIM Inc. recently exhibited at the “DX Expo,” hosted by BizCrew Japan. Through countless conversations with industry professionals, we were able to refine our perspective on how AI can be implemented in real-world operations. The experience was both intellectually and strategically rewarding.

---

Shifting Perspectives on AI in the Field

We are witnessing a clear shift in how the construction industry perceives AI—from “understanding what it is” to “mastering how to use it.” Many visitors, especially from construction companies, asked highly practical questions such as how AI can assist in automated quantity take-offs from 2D drawings or BIM-based schedule generation.

In the past, the conversation was largely conceptual: “What can AI do?” This time, it was pragmatic: “How can we integrate AI into our own sites and workflows?” This shift signifies that, post-ChatGPT, AI is no longer seen as a mere tool. It is becoming a foundational layer for structural transformation across operations.

---

From General-Purpose Models to Industry-Specific AI and AIOps

Technology adoption always follows a hype cycle—from inflated expectations, through disillusionment, to sustainable maturity.

We believe that generative AI is now climbing out of that “enlightenment slope.” With Sam Altman’s OpenAI reportedly preparing for IPO, the hype that began with the release of GPT-3.5 in late 2022 is reaching its inflection point. (For those in construction considering GPU purchases, now may not be the best time—hardware prices are inflated. Until the market stabilizes after OpenAI’s IPO, investing in software architecture and evaluation frameworks will likely yield better returns.)

What comes next is the question of how to embed general-purpose AI into industry-specific contexts—and how to optimize both performance and cost through AIOps and open-source models.

Across areas like estimation, site management, and scheduling, the challenge is not to create flashy interfaces, but to design data structures that preserve human expertise and enable robots to learn executable tasks. We aim to partner with companies long-term to unify siloed data, validate model performance on actual sites, and redesign end-to-end industrial data architecture.

Just as I am currently undergoing hands-on training with Tokyu Construction, every member of AIM is committed to learning directly from the field—to design solutions that truly serve real-world operations.

In domains historically built on intuition and experience, successful AI implementation demands more than data structuring. It requires architectural design that understands context and optimizes cost simultaneously.

We remain eager to collaborate with construction professionals who share this long-term vision.

---

The Evolution of Pricing Models

As Salesforce’s stock continues to hover around 20–30% of its peak, the phrase “SaaS is dead” is becoming increasingly common. For the first time in Japan, we also heard a client ask:

“Instead of seat-based pricing, can we pay based on how much the AI actually works?”

This marks a fundamental shift in perception. Clients are no longer paying for “using AI”—they are paying for what AI completes autonomously. We plan to co-design flexible pricing models with each partner through pilot implementations, ensuring both sides share conviction in the value delivered.

---

From SaaS to Agents: The Direction of OfficeGenie

Our flagship product, OfficeGenie, drew significant attention at the exhibition.

The platform combines Retrieval-Augmented Generation (RAG) and AI Agents to create an “action-oriented AI” that integrates seamlessly with Slack, Gmail, and Google Drive.

For example, when a quotation request arrives via email, the AI automatically parses the attachment, checks internal databases for inventory availability, and generates a response—complete with delivery dates and suggested wording.

By orchestrating multiple technologies such as RAG and OCR, OfficeGenie functions as an AI that can autonomously execute an entire workflow, not just assist with it. This is what we call an AI Agent.

Our key takeaway from the exhibition was clear: AI is evolving beyond searching and assisting—it is beginning to act. We are entering a stage where AI doesn’t just think; it moves. That, we believe, is the true starting point of the next industrial revolution.

---

Cross-Industry Expansion and the Road to Lunar Construction

While AIM continues to focus on AI adoption in construction, we have begun expanding into other sectors—manufacturing, energy, and broadcasting, among others.

Each industry has its own data shape, but the shared goal is the same: to build AI systems that can understand structure, infer meaning, and take the right action.

This structural intelligence leads us toward our ultimate vision—autonomous lunar construction and the Urban Operating System (Urban OS). We envision a future where low-cost data collection and self-governing AI systems can autonomously plan repairs, allocate budgets, and even manage maintenance funds.

In such a world, buildings evolve into self-reproducing systems within intelligent cities. General contractors will transform from large-scale executors into organizations that define and optimize the objectives of humanity and society itself.

AI and robotics will work hand-in-hand to augment human intelligence. This is not merely an extension of “DX”—it is the creation of a new layer of human infrastructure: an intelligent infrastructure.

When viewed from this lens, focusing solely on construction would be shortsighted. True innovation in the built environment requires collaboration across manufacturing, real estate, and urban infrastructure—toward the shared vision of the Urban OS.

We are still at the beginning of this journey, but deeply committed to pursuing it with humility and conviction.

---

Mizuki Marumo

Chief Executive Officer, AIM Inc.

---