Japanese firms are moving generative AI into real business use. Seven & I cut outsourcing costs 84%.
Sumitomo saves ¥1.2 billion yearly.
Companies must keep updating LLM choices.
What happened
A five-part special report examined how advanced Japanese companies are adopting generative AI, with the final installment focusing on the challenge of choosing large language models (LLMs, AI that understands and generates text). Seven & I Holdings cut external outsourcing costs by 84% using generative AI, Sumitomo Corporation saved ¥1.2 billion annually through company-wide Copilot adoption, and Mitsubishi Electric tackled the "RAG swamp" by improving data preparation for RAG (a method that lets an AI pull from company data to answer questions).
Why it matters
Companies are moving from a "try it first" phase to strategically embedding AI into business processes. Mitsui & Co. found that combining identification models with LLMs improves accuracy, building systems that do not "over-rely" on generative AI. Panasonic HD is developing company-specific models for factory use and home appliances, betting on specialized models that improve expertise and enable local use.
What to watch
The report highlights that LLM evolution is rapid and options are multiplying, urging companies not to commit to a single choice but to keep updating. The key lessons include using generative AI alongside existing technologies like machine learning and RPA (software that automates repetitive tasks), and committing to long-term development of specialized models despite the uncertainty.
Ask the AI about this article →
This special report captures a pivotal shift in how Japanese businesses view generative AI. The headline cases — Seven & I's 84% cost cut and Sumitomo's ¥1.2 billion annual savings — show that AI is no longer experimental but a proven tool for operational efficiency. The emphasis on company-wide adoption, as seen in Sumitomo's Copilot rollout backed by management, suggests that success depends less on the technology itself and more on organizational commitment.
The report's focus on "RAG swamps" and the need for proper data preparation highlights a practical reality: AI systems are only as good as the data they access. Mitsubishi Electric's workaround and Panasonic's move toward company-specific models both point to a maturing understanding that off-the-shelf LLMs need customization to deliver real value. Mitsui's "don't over-rely" philosophy further reinforces this — the best deployments integrate AI with existing tools rather than replace them entirely.
The final installment's core advice — don't commit to a single model, keep updating — reflects the fast-moving LLM landscape. For business leaders, the takeaway is that generative AI strategy must be flexible, with continuous evaluation and adaptation built into the plan.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Chinese large-model developer Z.ai says it can now support large-scale inference using roughly 100,000 domesti…
AI system scaling has pushed interconnect requirements inside data centers from chips and boards up to racks…
Analyst Ming-Chi Kuo says Nvidia has revived the Rubin CPX AI accelerator with a substantially redesigned arch…
Palantir Technologies stock has posted multi-year gains, including an 11x return over 3 years
Apple has escalated its legal battle against OpenAI, claiming in a new court filing that OpenAI is actively de…
Samsung Electronics has locked up as much as 70% of its memory production capacity under long-term supply agre…
