Over 30% Using Generative AI: The Blind Spot in Business Process Design

Urgent Shift from “Using” to “Integrating”

According to a survey by The Yomiuri Shimbun and Teikoku Databank, over 30% of Japanese companies are already using generative AI in their operations, primarily for writing and information gathering. At first glance, this suggests steady progress in digital transformation (DX) for Japanese firms.

However, I see a hidden risk in these numbers. Many companies are stuck at the stage of “using generative AI as a tool” without moving to the phase of “redesigning business processes to leverage AI’s capabilities.”

In fact, AI use cases reported by the Reform Industry News also focus on “improving efficiency in existing tasks” like meeting minutes and plan proposals. That’s not inherently bad. But from a management perspective, the real questions should be: “Which tasks can we eliminate by using AI?” and “How can we change the quality of decision-making?”

The Reality Behind “30% Using Generative AI” and Hidden Challenges

Teikoku Databank’s survey reports that as of April 2026, 31.7% of Japanese companies use generative AI at work. By industry, information and communications leads at 61.4%, followed by manufacturing at 35.2% and services at 28.1%.

But these numbers include many cases where AI was “tried but not adopted” or “used only in specific departments.” From my consulting experience, the difference between successful and failed AI adoption isn’t tool performance—it’s whether companies have “rewritten their business processes to match AI’s characteristics.”

For example, if you delegate meeting minutes to AI, simply “recording and transcribing” has limited impact. What truly matters is “changing how meetings are run, assuming AI will auto-generate minutes.” This means enforcing rules like clearly stating speaker names, and separating decisions from action items during discussions.

Learning from the Renovation Industry: Successful “Task-Specific AI” Patterns

What’s noteworthy in the Reform Industry News article is that it goes beyond “AI use case presentations” to explore AI tailored to industry-specific workflows. The renovation industry involves multi-stage processes: client meetings, estimates, construction management, and after-sales follow-up.

The success pattern here isn’t “using generic generative AI as-is,” but “building custom AI trained on industry terms and practices.” For instance, training an estimate AI on past project data allows it to automatically suggest standard pricing and timelines. Training a meeting-minutes AI on industry-specific terms (like “wallpaper replacement” or “baseboard replacement”) ensures error-free transcripts.

The advantage of this approach is relatively low implementation costs. Adding proprietary data to a generic AI can start at a few hundred to a few thousand USD per month. Unlike custom development, you don’t need to hire specialized engineers.

Three Principles of “Business Design” That Managers Often Overlook

When integrating generative AI into operations, managers must follow three key principles.

Principle 1: Clearly Separate “Tasks for AI” from “Tasks for Humans”

AI excels at gathering, organizing, and summarizing information. Meanwhile, creative judgment, interpersonal negotiation, and ethical decisions should remain with humans. If this line is blurred, AI task quality can’t be guaranteed, leading to double work as humans recheck everything.

In my experience, a ratio of 80% AI-handled tasks to 20% “judgment-required” human tasks works well for many operations. Only by designing with this ratio in mind can companies shift from “using” to “integrating.”

Principle 2: Build a “Verification System,” Not “Use AI Output As-Is”

Since generative AI produces text probabilistically, it can output factually incorrect content (hallucinations). The key is to embed a verification mechanism into the workflow.

Specifically, create a “double-check system” where AI-generated minutes or proposals are reviewed by another AI or human. My team uses a three-step process: generate with Claude, verify with ChatGPT, and final-check by a human. This reduces hallucination risk to nearly zero.

Principle 3: Think “Process-Wise,” Not “Department-Wise”

Many companies consider AI adoption by department: “Let’s introduce AI to sales” or “to accounting.” But business processes cut across departments. For example, the order-to-delivery process involves sales, accounting, manufacturing, and logistics.

AI adoption should be considered “process-wise.” Streamlining one department won’t boost overall throughput. Identifying bottlenecks and redesigning the entire process with AI leads to true efficiency gains.

Starting “Business Redesign” from 0 a Month: Your First Step

By now, many managers might think, “I want to start, but I don’t know where to begin.” Here’s a concrete first step.

First, classify your business processes into three categories: “Information gathering/organizing,” “Judgment/creativity,” and “Execution/communication.” Next, identify tasks in the “Information gathering/organizing” category and evaluate whether AI can handle them.

For example, consider these tasks:

• Meeting minutes creation (AI tools starting at $70–$200/month)
• First-level customer inquiry email responses (ChatGPT API from a few tens of dollars/month)
• Market research report generation (AI research tools at $140–$350/month)
• Internal manual creation/updates (AI writing tools around $70/month)

These tools require no special technical knowledge. I’ve had clients—managers with little IT background—implement them independently and see results within a month.

Conclusion: AI Adoption Success Depends on “Organizational Design”

We’ve entered an era where over 30% of Japanese companies use generative AI. But don’t be satisfied with this number. What truly matters isn’t “using AI”—it’s “redesigning business processes around AI.”

Instead of spending time on tool selection, first visualize your business processes, clearly separate AI-handled tasks from human ones, and then introduce tools starting at a few hundred dollars per month based on that design. If you get this order right, generative AI will become a powerful management asset.

Are you ready to move to the next phase beyond the “30% using AI” milestone?