According to a new report, enterprises are quickly moving from generative AI tools underpinned by a single LLM to multi-agent systems that can autonomously orchestrate workflows and take actions across business processes. Databricks’ 2026 State of AI Agents report found use of these systems grew sharply over the latter half of the data period, with multi-agent workflows up 327 percent over the latter half of 2025.
“AI agents mark the next phase of enterprise AI adoption as organizations transition beyond pilots and chatbots toward agentic systems that can reason and orchestrate workflows,” researchers wrote in the report. “Enterprises that are succeeding are aligning AI use cases to business goals, developing focused agents that complete necessary operational tasks specific to their industry and organization. In the data layer, AI agents require an infrastructure shift that occurs only once every decade. “
The research, based on based on anonymized activity data from more than 20,000 global Databricks customers, shows a significant shift in how organizations are using artificial intelligence agents, moving from isolated experiments to more complex, system-level deployment.
According to the report, most AI use cases are focused on automating routine necessary tasks like producing market intelligence, customer support classification and routing, claim processing, and customer onboarding, among others. Forty percent of AI-automated workflows are related to customer experiences, the Databricks research found.
AI agents are also playing a growing role in core infrastructure work. The report highlights that 80 percent of databases are built by AI agents and that nearly all database testing and development environments are now created by agent-driven automation, trends that are driving demand for new database architectures suited to autonomous workloads.
In addition to adoption metrics, the report emphasizes the importance of governance and evaluation practices. Organizations that deploy evaluation tools and governance frameworks are seeing materially higher production flow for AI projects, suggesting that operational discipline remains a key enabler of scaling agentic AI.
The report does acknowledge possible negative effects of the unprecedented automation acceleration on job loss.
“If AI agents gained the capacity to act with greater autonomy across domains, within only a few years it would likely accelerate labor market disruption,” the report’s authors wrote.
For technology providers and end users in process automation, the report points to a broader transition from simple generative AI use cases toward integrated, multi-agent automation that intersects with data, workflow, and operational systems.
