Launch of AI-Driven Development Platform 'Egeria NextCode V2'

PKUTECH Inc. (Headquarters: Chiyoda-ku, Tokyo; President and CEO: Liu Shenqiu) announces the launch of 'Egeria NextCode V2,' a next-generation AI-driven development platform.

This platform converts complex system design documents into structured Markdown format using proprietary algorithms, transforming existing system design assets into a 'common language understandable by AI.' This enables a seamless transition to AI-driven development using IBM Bob—an enterprise development support partner powered by AI agents provided by IBM—not only for new development but also for existing systems. Additionally, the platform enables automated integration testing based on structured design documents. Through this integrated workflow—from design structuring to development support and test automation—the platform achieves both speed and quality in development, helping eliminate knowledge silos and improve development productivity.

Background: Four Barriers Facing IT Teams in Transitioning to AI-Driven Development

As many companies explore transitioning to AI-driven development, the following four barriers hinder on-site transformation:

① Limitations in AI Interpretation of Existing System Specifications

Design documents in Excel, filled with diagrams, business flows, and screen transition charts with complex layouts, cannot be accurately converted into Markdown by general-purpose AI.

② Limitations of Manual Integration Test Design and Execution from Basic Design Documents

Manually extracting test cases and performing execution and verification from unstructured design documents requires enormous effort and carries risks of regression and missed checks.

③ Burden of Large-Scale Test Data Creation and Masking

Significant effort is spent on masking production data for privacy protection and generating vast amounts of dummy data for comprehensive testing.

④ Risk of Leaking Confidential Information and Design Documents to Public Clouds

In enterprise environments, highly confidential design documents and business data cannot be sent to public clouds, preventing access to efficiency gains from the latest AI technologies.

Core Value: The Optimal Solution for AI-Driven Development

To fundamentally overcome these four barriers and achieve both development speed and quality, we have established five core values centered on IBM Bob, integrating IBM’s advanced AI technologies. This enables the elimination of knowledge silos and standardization of quality management, transforming IT investment into capital for advancing DX initiatives.

AI-Driven Development Workflow for Existing Systems with Egeria NextCode V2

【Core Value ①】Structuring Basic Design Documents with Proprietary Tools

Existing AI struggles to interpret logical relationships in complex diagrams and matrix structures. Our tool analyzes the logical connections of layout and symbols using proprietary algorithms, converting specifications into structured design documents that AI can interpret. These structured documents serve as the foundation for high-precision coding and automated integration testing.

Business flow diagrams and screen transition charts are also accurately converted into Markdown using proprietary technology.

【Core Value ②】AI Coding Support via IBM Bob

IBM Bob is an enterprise development support partner powered by AI agents provided by IBM. When applying to existing systems, design document interpretation becomes a barrier, which our Markdown conversion tool resolves, enabling a smooth transition to AI-driven development.

Additionally, we provide feedback to improve IBM Bob’s maturity through practical implementation.

【Core Value ③】AI-Powered Integration Test Case Creation and Execution

The platform automatically generates test cases that cover business logic from structured design documents. It standardizes test design—often dependent on individual expertise—with over 90% match accuracy to human-created test items, while also covering human oversights. The entire process—from test case creation to execution and verification—is fully automated in one flow. The reason for accurate integration test design lies in autonomously following the six steps below, faithfully reproducing the thought process of experienced human engineers.

【Test Structure Definition】

Step 1. Theme Creation: Logically define the axes of business scenarios to be verified from the Markdown-converted design document.

Step 2. Related Information Listing: Organize correlations among user permissions, events, and database information.

Step 3. Table of Contents Creation: Structure based on functional classification criteria, thoroughly eliminating AI hallucinations.

【Test and Execution】

Step 4. Integration Test Case Creation: Generate comprehensive test cases at 120% coverage compared to human-created ones from the established framework.

Step 5. Test Data Creation: AI precisely infers and generates data required to pass test cases.

Step 6. Automated Test Execution: Complete the entire process from execution to result verification in one flow.

The Six Steps of Egeria NextCode V2’s AI-Driven Testing Tool

【Core Value ④】Automated Test Data Generation via AI Inference

Using a small set of real data as seed, AI inference models automatically generate test data. Even in environments where production data cannot be used, this eliminates the need for complex data masking and massive dummy data preparation, significantly reducing test preparation effort.

【Core Value ⑤】Local LLM Support

Enables the use of local LLMs within the customer’s environment, eliminating dependency on external APIs. This reduces the risk of leaking design documents and customer data to external parties, making it suitable for highly confidential sectors such as financial institutions, government agencies, and manufacturing.

Future Outlook

The platform has already been piloted with five customers.

Going forward, we will enhance the maturity of our proprietary Markdown tool and improve matching rates with human-created test items by accumulating real-world implementation experience across various development sites. This will evolve the automation workflow from specification definition to testing into a higher-performance system, reducing maintenance costs. Ultimately, we aim to realize an environment where AI autonomously supports all development projects we assist.

Inquiries

PKUTECH Inc. (General Inquiries) Phone: 03-3526-7778 Hours: 9:00–12:00 and 13:00–18:00 (Excluding weekends, holidays, and company-designated holidays) Contact: Harada Email: [email protected] Inquiry Form: https://pkutech.co.jp/enquire/service-contact/