Databricks reports surge in enterprise AI agent use

Databricks has published new research that points to a sharp rise in enterprise use of coordinated, multi-agent AI systems, with reported usage up 327% over four months.

The company said organisations have started to move beyond standalone chatbots. It said multi-agent systems now run business workflows and sit deeper inside production environments.

Databricks based its findings on activity across more than 20,000 customers worldwide. It said this customer base includes more than 60% of the Fortune 500.

Industry uptake

Technology companies led the shift in the data set. Databricks said tech firms created nearly four times more multi-agent systems than any other industry.

Media and entertainment and healthcare also showed early adoption. Databricks said these sectors have started embedding AI agents into production workflows.

The report framed the change as a move from pilots into scaled use. It also pointed to a stronger focus on governed production deployments.

Real-time workloads

Databricks said real-time processing now accounts for the bulk of enterprise AI workloads in its environment. It reported that 96% of AI requests get processed in real time.

The company also broke out figures for Asia Pacific. It reported that organisations in the region process 82% of AI requests in real time.

Databricks linked real-time demand to several application patterns. It cited copilots, customer support and personalisation as examples.

Database automation

The research also described AI agents as central to database work. Databricks reported that 80% of databases are built by AI agents.

It also said AI agents now build 97% of database testing and development environments. The company said this change drives demand for what it described as an AI-ready database designed for autonomous, real-time AI workloads.

Databricks did not name specific database products in the report summary it provided. It positioned the trend as a shift in how organisations create and maintain data infrastructure.

Use cases

Databricks said most generative AI use cases focus on automating routine necessary tasks. It also said 40% of use cases relate to customer experiences.

In Asia Pacific, the company said market intelligence and strategic analytics rank as the top AI use cases in the region.

The report also highlighted what it called "model flexibility" as a common approach. Databricks said 78% of organisations use two or more AI model families, and nearly 60% use three or more.

That pattern suggests enterprises continue to spread workloads across multiple model providers and model types. It also implies a need for internal processes that manage performance and cost across different systems.

Governance focus

Databricks said evaluation and governance practices correlate with higher production deployment. It said companies that use evaluation tools get nearly six times more AI projects into production.

The company also said organisations using AI governance put over 12 times more AI projects into production. It added that AI governance has become a top investment priority and grew sevenfold in nine months.

Databricks framed these shifts as part of an operating model change around AI agents. It also described governance and evaluation as signals that organisations have moved beyond experimentation.

Nick Eayrs leads Field Engineering for Databricks in Asia Pacific and Japan.

"Across Asia, we're already seeing a decisive shift from pilots to production, with organisations embedding AI agents across critical workflows, infrastructure and databases," said Nick Eayrs, Vice President, Field Engineering, Asia Pacific and Japan, Databricks. "Governance and evaluation are emerging as early signals of those scaling with confidence. The organisations that succeed will be those with strong data and AI foundations, clear ownership, and a disciplined focus on scaling what already works," said Eayrs.

Databricks is expected to continue tracking agent adoption across its customer base as enterprises expand production deployments and tighten governance practices.