Now that AI adoption has become commonplace, many executives likely feel a sense of urgency that “the results aren’t as significant as expected.” The “Survey on AI Utilization and Business Application Development” released by Nikkei CrossTech quantifies this vague unease and reveals results that hit the core issue. According to the survey, many companies face a significant disconnect between “developing applications using AI” and “actually utilizing AI in daily operations.”
This “development-implementation gap” is the primary factor reducing the ROI of AI investments. The technology exists, the budget is allocated, yet it fails to permeate the front lines. Unless this structural problem is resolved, AI will remain nothing more than an “expensive toy.”
The Survey Reveals a Divide Between “Two Worlds”
The Nikkei survey presents several intriguing data points on the current state of AI adoption. Particularly noteworthy is the clear gap between companies engaged in developing AI-powered applications and those that routinely use AI in their actual business processes.
The development side chases the latest tech trends and repeats PoCs (Proofs of Concept). Meanwhile, the operations side clings to existing workflows and is hesitant to adopt new tools. The shortage of “bridging talent” who should fill this gap is becoming apparent in many organizations.
Looking back at our own company’s AI use cases (93 cases across 29 business areas), the key to success lay precisely in this “bridging.” Without personnel or systems capable of simultaneously understanding technical potential and operational needs, AI will not take root on the ground.
The “Shitatel” Approach Hints at a Solution
One model for resolving this disconnect is the approach taken by Shitatel, which recently launched a value chain transformation support service for apparel and lifestyle companies. The company specializes in specific industries, offering AI solutions based on a deep understanding of industry-specific business processes.
The crucial point is its “industry-specific” focus. Instead of providing generic AI tools, it digitizes and optimizes with AI the entire apparel industry flow from “planning → production → inventory management → sales.” Without this perspective of redesigning the business process itself, AI adoption remains shallow, merely “pasting” AI onto existing operations.
Executives should consider “the path to AI adoption specialized for their own industry and operations.” Whether utilizing external services or advancing in-house development, the degree of this “specialization” determines the depth of utilization.
The Barrier to In-House Development is Dramatically Lower
Here, the evolution of code-generating AI cannot be overlooked. The attention FIXER is garnering as a generative AI service company is rooted in this technological shift. Tools like Claude Code and GitHub Copilot are creating an environment where even non-engineers can develop practical business applications.
Within our company, we built an automated legal document checking system using Claude Code. Even without personnel possessing both specialized knowledge (legal) and AI development skills, legal staff can now directly prototype “the functions they need.” The monthly cost is only about $140 for the AI tool, achieving operational efficiency worth several million yen annually.
This “democratization of in-house development” becomes a powerful means to bridge the development-implementation gap. We have entered an era where operational stakeholders themselves can begin building tools to solve their own challenges.
Three Concrete Actions Executives Should Take
Based on the survey results, here are three actionable steps executives can start today.
1. Diagnose AI Suitability Using a “Business Process Map”
First, visualize your company’s key business processes and evaluate the “AI suitability” of each step. Use the following three criteria:
- Routine Nature: Is the task repetitive with clear decision criteria?
- Data Volume: Is there sufficient historical data for training?
- Integration Potential: Can it connect with existing systems (CRM, ERP, etc.)?
This diagnosis clarifies the priority for AI introduction. The first targets should be routine tasks with high potential for immediate impact, such as invoice processing, initial customer inquiry responses, and data entry work.
2. Build “Bridging Processes,” Not Just Cultivate Bridging Talent
Instead of nurturing one precious “operational staff member knowledgeable about AI,” design processes that allow anyone to participate in AI utilization. Specifically, follow these steps:
- Problem Discovery Workshops: Collect operational bottlenecks from each department.
- Prototype Creation Support: Hands-on training for creating simple tools using code-generating AI.
- Standardizing Effect Measurement: Share rules for visualizing time-saving effects in “hourly wage equivalents.”
Running this process fosters an organizational AI utilization foundation that doesn’t rely on a few “capable individuals.”
3. Establish Clear Criteria for Using External Services
Decide between industry-specific services like TechSuite’s “Media Launch Partners” or in-house development based on the following criteria:
- Is it a Core Competitive Area?: Prioritize in-house development for operations that are a source of differentiation.
- Data Confidentiality: Highly confidential data should原则上 be managed in-house.
- Rate of Change: Flexible in-house development is advantageous if business processes change frequently.
In many cases, a “hybrid strategy” is the practical solution: considering in-house development for core operations while utilizing external services for peripheral tasks.
The OpenAI-Google Alliance Highlights the Importance of an “Ecosystem”
The move by OpenAI, Anthropic, and Google to collaborate against AI model copying by Chinese companies offers implications beyond mere technological competition. It signifies that in AI adoption, an “open ecosystem” is becoming more important than a “closed system.”
This perspective is essential when considering your company’s AI strategy. Building systems dependent on a single vendor compromises long-term flexibility. A “multi-model strategy” that uses multiple AI models (Claude, ChatGPT, Grok, etc.) appropriately for different purposes and combines them via API integration is a sound approach that can adapt to future changes.
In practice, our company uses three AIs concurrently: Claude, which is strong with legal documents, for contract reviews; ChatGPT, which excels in creativity, for idea generation; and Grok for real-time information gathering. While monthly costs increase, this not only diversifies the risk of relying on a single model but also offers the significant benefit of selecting the optimal tool for each task.
Conclusion: The Executive’s Role in Turning Disconnect into “Connection”
The “development-implementation gap” highlighted by the Nikkei survey is not a technological issue but an organizational and procedural one. Only clear executive intent and concrete action can bridge this gap.
Start by visualizing your company’s business processes and diagnosing AI suitability. Then, determine the balance between in-house development and external utilization, and formulate a strategy conscious of an ecosystem that combines multiple AIs. These are practical steps you can start today.
AI is no longer just an IT department issue. It is a management resource for redesigning all business processes. Only by turning the development-implementation disconnect into a “connection” does AI investment begin to generate real ROI. The first step begins with executives shifting their own perspective from being “users” of AI to being “creators.”
Comments