Enterprise AI agents are embedded in everyday business processes, especially within engineering and IT operations. Many organizations report active operational deployments and agent development ranks high on their strategic agenda.
New research results from Docker, Current state of Agentic AI reportingLet’s examine how enterprises are deploying agent systems and the challenges that arise as deployments scale.
Data shows that 60% of organizations run AI agents in production. Almost all cited architectural agents as a strategic priority.
Initial implementation will focus on internal workflows. DevOps, continuous integration (CI) and continuous delivery (CD) optimization lead the reported use cases, followed by security automation and general process automation. Code generation and reviews also rank prominently. These environments provide structured tasks and measurable outcomes, giving teams room to measure performance and manage risk.
Industry recruitment shows strong activity in telecommunications, financial services, and technology. Some organizations remain unfamiliar with the term agent AI, demonstrating uneven perception in the broader market.
Security remains a major barrier
40% of respondents cited security and compliance as the main barrier to scaling agent AI. Many people report having difficulty verifying whether a tool meets their company’s security standards.
Respondents described issues at the infrastructure, operational, and governance levels. Infrastructure teams value runtime isolation and sandboxing. Operations leaders cite exposures introduced by adjustments to models, APIs, and external systems. Governance stakeholders are calling for stronger audit mechanisms and consistent policy enforcement.
Immediate injection and tool addiction appear frequently in answers regarding risk. Vulnerability detection and mitigation is one of the most pressing technological challenges. Entitlement management and access control in distributed agent systems also require attention.
48% identify the operational complexity of coordinating multiple components as the main challenge when building an agent. Integrating models, connectors, and runtime environments increases monitoring requirements for security teams.
Multi-model architecture increases operational demands
Agent systems rely on multiple models. Nearly all organizations surveyed use multiple models within their architecture, with nearly half reporting using four to six models.
61% have a combination of cloud-hosted and locally-hosted models. Control, data privacy, and compliance drive decisions about whether to run models locally. Hybrid and multicloud deployments are widespread, with most organizations operating agents in multiple infrastructure environments.
Technical complexity is cited as the biggest barrier to scaling. Respondents stated that orchestration tools are immature for production environments. Security teams must consider the interactions between models, data sources, and connected services in different environments.
Model context protocols require scrutiny
Model Context Protocol (MCP) allows agents to connect to external tools and enterprise data sources. Awareness among practitioners surveyed was high, with many reporting active use.
Organizations cite the operational overhead of managing MCP servers and clients, along with the burden of installation and configuration. Security and compliance concerns remain critical.
Immediate injection and tool poisoning emerge as major risks in MCP-enabled systems. Managing authentication, credentials, and access control for MCP servers presents ongoing challenges.
Deploying MCP at enterprise scale requires improvements in discovery, manageability, and security governance.
Concerns about distribution and vendor dependencies
Agent sharing practices remain fragmented. Commercial marketplaces and source code repositories serve as common distribution channels. Internal documentation and informal processes continue to support collaboration within the team.
Security is the biggest barrier to seamless sharing. Respondents want signed and scannable agent packages, a centralized registry, and built-in policy enforcement. Version control and cross-environment compatibility further increase operational demands.
76% of respondents reported concerns about lock-in related to model hosting platforms, cloud providers, and monitoring layers. Organizations are diversifying their models and infrastructure environments to reduce dependencies, increasing coordination complexity.
Containers serve as a consistent operational foundation. The majority of organizations use containers for agent development or production workflows. In most cases, established cloud-native pipelines and orchestration practices will be extended to support agent systems.
According to the researchers, “The near-term value of Agentic AI is already becoming a reality in internal workflows. The next wave will depend on standardizing how agents are secured, coordinated, and shipped. Teams that invest in this layer of trust now, on top of the container foundation they already know, will begin by scaling agents from local productivity to durable enterprise-wide outcomes.”
