<Session Report> Ryotetsu Ogawa, CEO of 2WINS, an AI Startup from UTokyo's Matsuo Lab, Makes a Guest Appearance at Japan's Largest Software Testing Technology Conference 'JaSST'26 Tokyo'

Ryotetsu Ogawa, Representative Director and CEO of 2WINS Inc., an AI startup originating from the University of Tokyo (Headquarters: Bunkyo-ku, Tokyo; Representative Director and CEO: Ryotetsu Ogawa; hereinafter '2WINS'), made a guest appearance at a technology session hosted by Paul To Win Co., Ltd. (Headquarters: Nagoya City, Aichi Prefecture; Representative Director and CEO: Kazuaki Shimura; hereinafter 'Paul To Win'), a company that supports problem-solving in service lifecycles. The event took place at the 'Software Testing Symposium 2026 Tokyo (JaSST'26 Tokyo)', one of Japan's largest technology conferences in the software testing field, held at Tokyo Big Sight on Friday, March 20, 2026.

Session Exploring the Role of QA in the Era of AI-Driven Development

'QA Response in the Era of AIDD (AI-Driven Development) and SDD (Spec Driven Development) - What Does QA Observe as Quality?'

Mr. Masayuki Kubo, Head of the Advanced Technology Research Office at Paul To Win, Ms. Kaori Goto from the Technology Promotion Office at Paul To Win, and Ryotetsu Ogawa, Representative Director and CEO of 2WINS, took the stage to discuss the QA response amid the changes of the AIDD and SDD era from an on-the-ground perspective. On the day of the event, it was held in a hybrid format of online and offline, and was viewed by dozens of participants.

In this session, the main theme of discussion was 'What does it mean that 'quality is guaranteed' in an era where AI handles the design and testing processes?' Ms. Goto of Paul To Win's Technology Promotion Office pointed out that the widespread use of AI is forcing a reconsideration of the traditional premises of quality assurance, and raised the need to redefine the 'state of guaranteed quality.'

In response to this, Mr. Kubo of Paul To Win's Advanced Technology Research Office explained the characteristics of AI, particularly Large Language Models (LLMs). He pointed out, 'Because LLMs generate outputs based on probability distributions, they do not always guarantee correct results, and a certain degree of error is included even in code generation and test design. Due to this characteristic, quality under AI utilization cannot be deterministically guaranteed, and requires management premised on 'probabilistic fluctuation.' How to handle such 'probabilistic fluctuation' has become a new point of contention for QA.'

Furthermore, Mr. Kubo noted that the deterministic mindset that 'quality is guaranteed if correctly designed, implemented, and tested,' which traditional software development has relied upon, is becoming difficult to uphold as AI is integrated into the development process. Along with this, the standard for quality evaluation is shifting from 'correctness' to 'the probability of correctness'.