327%

That’s how much multi-agent workflow adoption grew on Databricks between June and October 2025.

Just one of multiple compelling pointers in the Databricks “State of AI Agents” report, published 27 January 2026, drawn from telemetry across 20,000+ organisations, over 60% of which are Fortune 500.

Our recent post focused on the Lakehouse architecture and why Japan’s biggest enterprises are building on it. We now turn our attention to what exactly is taking shape.

AstraZeneca used Databricks Agent Bricks to parse over 400,000 clinical trial documents and extract structured data in under 60 minutes, all without writing a single line of code. Franklin Templeton built a governed fund analysis agent that reduced research from days to seconds. Production-grade AI agents that reason, plan and execute complex business tasks are emerging across every sector.

Despite this, only 19% of organisations have deployed agents at scale.

The report found that what separates the 19% from everyone else is not the model, but governance. Companies using governance tools deploy 12x more AI projects to production. Those using systematic evaluation deploy 6x more.

Because every data asset, model and tool sits inside Unity Catalog with on-behalf-of access control, agents can delegate tasks securely without the permission sprawl that kills most enterprise deployments.

Nowhere is the tension greater than in Japan.

A projected shortfall of 220,000 IT professionals by 2026. A culture that demands consensus, auditability and proof across multiple business units before anything touches production. But momentum nonetheless.

Toyota is using its Obeya initiative to build AI agents that capture the institutional knowledge of retiring engineers. And Databricks achieved ISMAP certification on 28 January 2026, clearing the path for Japan’s most security-conscious organisations.

The appetite is real. But so is the gap. AI agents built on governed platforms are not replacing that missing workforce. One engineer with the right agent architecture can now cover what used to require a team.

The enterprises set to scale agents in Japan will not be those with the best models. They will be those that can prove to every stakeholder, with documentation and audit trails, that the agent will not hallucinate customer data or breach internal boundaries.

That is where we work. As a Databricks partner, we help enterprises in Japan and Europe close the gap between pilot and production.