Oracle Japan unveils Oracle Fusion Claw for enterprise AI

The forum's framing was less about new technology than about the gap between AI adoption and measurable payoff. Yazaki's numbers made that gap concrete: 16,000 of its 17,000 employees already use generative AI and training is in place, yet the company struggles to translate that into financial results. Its test with 2,423 past trouble reports, of which only 27 were usable, points to a cause Oracle repeatedly returned to — data that large language models simply cannot read.

That diagnosis shapes the advice Oracle gave. Rather than building a huge data lake up front, which it called unrealistic in both time and cost, it proposes small starts and checking results as you expand. It also argued that financial impact requires work spanning internal departments, which points toward top-down AI and data programs rather than bottom-up efforts led by individual teams.

The customer examples fit that pattern in different ways. Rakuten described using an "ECC AI Agent" to suggest next actions to stores and turn those conversations into reusable data, and said it chose to develop jointly with Oracle for speed because its own teams were fully occupied. SoftBank focused on AI security, where Oracle said it would support resilience through the latest security patches and data backup and DR environments, and claimed around 20–50% reduction in build and operation effort by governing AI behavior in a database.

On the product side, Oracle Fusion Claw is positioned as a way to bring OpenClaw-style autonomy into enterprises that need governance and security, and the 25 Fusion Agentic Applications would be its first broad application set. Whether that lands may hinge on how quickly customers can make their existing data readable by AI — the constraint Yazaki's example put at the center of the discussion.