As enterprises expand their use of AI, many are moving beyond experimental chatbots towards agentic systems designed to reason, plan and act across business workflows.
According to Databricks’ 2026 State of AI Agents report, which aggregates telemetry from more than 20,000 organisations, including over 60% of the Fortune 500, AI agents are increasingly being deployed in production environments. The report indicates a growing focus on multi-agent systems within enterprise AI strategies.
The move to multi-agent systems
Rather than relying on single chatbots, organisations are beginning to use multi-agent systems to orchestrate end-to-end workflows. Databricks reports a 327% increase in multi-agent usage over a four-month period in 2025. Technology companies are building these systems at nearly four times the rate of other industries, according to the research.
AI use across industries
Organisations are applying AI to address sector-specific challenges. Examples cited in the report include predictive maintenance in manufacturing, market intelligence in retail, and medical literature analysis in healthcare. Databricks also notes that 40% of leading AI use cases relate to customer support, advocacy and onboarding.
Enterprises are also using multiple large language model (LLM) families, including ChatGPT and Llama, aligning models to specific tasks. By October 2025, a majority of organisations had adopted two or more model families to maintain flexibility and optimise performance for different workloads.
Governance and evaluation
The report highlights governance and evaluation as key factors in moving AI from experimentation into production. Organisations using unified governance frameworks were found to put 12 times more AI projects into production. Those using evaluation tools deployed six times more projects, using them to measure, test and improve model quality and reliability throughout development and deployment.
Database operations and application development
Databricks reports that AI agents now create 80% of databases, while 97% of database testing and development environments are built by agents. This approach reduces the time required to clone, branch and test databases, supporting faster experimentation and deployment.
The increased use of AI agents also supports approaches such as “vibe coding”, where business users without deep technical expertise can build AI applications. Since the public preview of Databricks Apps, more than 50,000 data and AI apps have been created, reflecting broader participation in application development.
