Databricks, a data and AI company, has released its latest "State of AI Agents" report, revealing that enterprises are rapidly transitioning from simple chatbots to collaborative, multi-agent AI systems. Analyzing data from over 20,000 Databricks customers worldwide, the report demonstrates that AI agents are playing a central role in key areas such as enterprise workflows, database construction, and the commercialization of generative AI.
According to the report, multi-agent system usage has increased by 327% in just four months, with 96% of all AI requests now being processed in real time. In the Asia-Pacific region, 82% of these requests are now operating in real time, supporting core tasks such as co-piloting, customer support, and personalized services.
AI agents are also playing a leading role in core engineering areas such as database construction and test environment design, with 80% of all databases and 97% of test environments operating on AI agents. Furthermore, companies are actively leveraging AI to automate essential, repetitive tasks, improve customer experiences, and analyze market strategies. Seventy-eight percent of organizations utilize two or more AI model suites, and 60% utilize three or more.
AI evaluation and governance were also identified as key factors in commercialization success. Companies using professional evaluation tools were approximately six times more likely to transition AI projects to commercialization, and those with a well-established governance system achieved commercialization success rates more than 12 times greater. Consequently, AI-related investment increased sevenfold in just nine months.
“Only organizations with governance and evaluation systems in place can reliably apply AI from pilots to actual business operations,” said Nick Ayres, Vice President of Field Engineering for Databricks APAC and Japan. “Successful companies rapidly scale proven practices based on a strong data and AI foundation and clear accountability structures.”
- See more related articles
You must be logged in to post a comment.