<Session Report> Ryoto Ogawa, CEO of UTokyo Matsuo Lab Startup '2WINS,' Guest Speaker at JaSST'26 Tokyo, Japan's Largest Software Testing Conference

Ryoto Ogawa, Representative Director and CEO of 2WINS Inc., a Tokyo-based AI startup from the University of Tokyo, was a guest speaker at the technology session of Pole To Win Inc. during the 'Software Testing Symposium 2026 Tokyo (JaSST'26 Tokyo)' held on March 20, 2026, at Tokyo Big Sight.

### Exploring the Role of QA in the AI-Driven Development Era **Session Title: 'QA Response in the AIDD/SDD Era—What Does QA Observe as Quality?'**

The session featured Masayuki Kubo (Head of Advanced Technology Research Lab, Pole To Win), Kaori Goto (Head of Technology Promotion Office, Pole To Win), and Ryoto Ogawa (CEO, 2WINS). They discussed changes in QA from the perspective of field experts amidst the shift toward AIDD (AI-Driven Development) and SDD (Spec Driven Development). The session addressed a core question: 'What does it mean for quality to be guaranteed in an age where AI handles design and testing?'

Kaori Goto pointed out that the traditional premises of quality assurance must be reconsidered due to the spread of AI. Masayuki Kubo explained the characteristics of Large Language Models (LLMs), noting that since they generate output based on probability distributions, they do not always guarantee correct results. Therefore, quality under AI utilization must be managed with 'probabilistic fluctuations' in mind, shifting the focus from 'correctness' to the 'probability of correctness.' This moves the QA domain from pure engineering toward a more statistical approach.

### Limits of AI and the Human Role Ryoto Ogawa discussed the limitations of AI while acknowledging its progress. He noted that while AI is highly effective in downstream processes like code generation and test automation, it still faces challenges in upstream processes such as holistic system design, context understanding, non-functional requirements, and business logic validity. He emphasized the need to appropriately design areas where humans should remain involved, particularly focusing on creativity and judgment.

### Importance of Specification Management In the discussion on quality evidence, the importance of specification management was highlighted alongside traditional metrics like test case counts and coverage. As development speed increases with AI, tracking the intent and background of specifications becomes more difficult. Kubo argued that improving the accuracy of specifications and strengthening management systems are essential to ensure consistency. It was suggested that both AI-readable specs and human-readable specifications must be maintained and continuously synchronized.

### Evolution of the QA Engineer The session concluded by highlighting the changing role of QA engineers. They are now required to participate in quality design and decision-making at the upstream stages rather than focusing solely on verification. As AI takes over implementation and testing, humans will focus on the more 'human-centric' aspects of creation and judgment.

Through this session, it was demonstrated that quality evaluation axes are shifting from 'correctness' to 'probability' and from 'results' to 'processes' in the AI era. The redefinition of QA concepts and practices is already becoming apparent in professional settings.