Japanese companies are leveraging artificial intelligence (AI) to enhance software development speed. As productivity becomes a top priority, businesses are adopting generative AI as a means of operational innovation. However, the increasing costs associated with high-performance AI models present new challenges in measuring their effectiveness and managing expenses.
On June 11, the Nihon Keizai Shimbun reported that the American AI startup Anthropic held a developer event titled "Code with Claude" in Tokyo the previous day, marking its entry into the Japanese market. This year's event, the third following those in San Francisco and London, attracted around 500 engineers from Japanese companies. Anthropic plans to establish its first Asian office in Japan by October 2025 and has formed partnerships with NEC and Hitachi.
During the event, Anthropic announced the public release of its new AI model, "Claude Fable 5." This model utilizes the same foundational technology as the previously limited-release high-performance AI "Claude Mythos," but has enhanced safety measures to refuse responses to potentially harmful instructions, such as those related to cyberattacks. The Nikkei evaluated Fable 5 as capable of autonomously working for extended periods in complex fields like programming, mathematics, and finance.
Caitlin Lesh, Anthropic's head of platform engineering, stated in her keynote address, "The performance of AI models like Fable is improving exponentially, but their roles in business do not yet match that progress. We want to bridge that gap with Claude." One of the tools aimed at closing this gap is "Claude Code," introduced in 2025, which automates programming tasks and leads to the development of AI agents capable of handling complex assignments autonomously. Anthropic continues to invest in high-performance AI development based on usage fees from businesses and institutions.
Japanese companies are already intensifying their use of AI. According to the Nikkei, Mercari, a leading second-hand trading platform, reported a 90% increase in engineering productivity due to AI implementation. An AI representative from Rakuten Group explained that the release cycle for major features has been shortened from every three months to every two weeks. Fujitsu partnered with Anthropic in May to enable approximately 100,000 employees to utilize Claude in their daily tasks, aiming to transition to an AI-driven development system over the next decade, with plans to incorporate AI in over 90% of all projects. They anticipate productivity will more than double compared to 2025 levels.
The spread of AI is prompting changes in the business structure of Japan's system development industry. Traditionally, compensation has been based on the number of developers and hours worked. If AI reduces development time, this model could lead to decreased revenues for development firms. Fujitsu's Takahito Tokita stated that without changing the business model, "significant growth is unlikely."
Cost is a pressing issue. The rise of "vibe coding," where programming tasks are delegated to AI, has led to a rapid increase in the usage of "tokens," which represent the amount of data exchanged between users and AI models. Anthropic employs a pay-per-use model based on token consumption for businesses and developers, with fees for the new high-performance model set at double that of its previous flagship model, "Opus."
As AI usage costs rise, companies are now faced with the need to evaluate cost-effectiveness. The Nikkei reported that Uber Technologies in the U.S. has already exhausted its annual AI budget, imposing limits on employee AI usage. In Japan, many companies tout their AI adoption rates as a measure of success. However, increased usage does not necessarily translate to higher revenues, shorter development times, or cost savings. The Nikkei highlighted the lack of appropriate metrics to measure actual effectiveness.
Shinpei Miyoshi, an executive at PwC Consulting, noted, "Many Japanese companies are using generative AI adoption rates as evaluation metrics, failing to design proper performance indicators linked to results. There are numerous cases where usage increases without effective measurement, leading to rising costs."
As AI establishes itself as a tool for enhancing workplace productivity, companies' concerns are evolving. They are moving beyond the decision of whether to adopt AI to determining which tasks to assign to AI and how to justify costs based on performance metrics. The competition among Japanese companies is shifting from the speed of technology adoption to the precision of managing cost-effectiveness.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.