Red Hat Ansible targets agentic AI as execution layer

Red Hat is positioning Ansible as the execution layer enterprises can use to turn agentic AI outputs into governed IT actions. Sathish Balakrishnan, vice president and general manager of Ansible at Red Hat, said the goal is to bridge AI-generated insights and reliable production operations without letting AI agents act unpredictably on critical systems.

Balakrishnan made the comments during Red Hat Summit 2026 in a conversation with theCUBE’s Rob Strechay and Rebecca Knight on SiliconANGLE Media’s livestreaming studio. He said the industry is moving from AI experimentation toward real-world agentic operations, but many organizations still do not have the execution infrastructure needed to control what those agents actually do.

According to Balakrishnan, automation is the base layer for both AI and IT operations. He said AI can unlock many possibilities, but it is probabilistic, meaning it can produce different answers or approaches to the same problem. “You don’t want AI agents mucking up your production servers, ever,” he said.

The issue, as Red Hat frames it, is that production systems need deterministic action, not repeated guessing. Balakrishnan said it makes little sense to use AI to patch servers when the steps are already known and should be carried out the same way every time. He argued that using AI to generate multiple ways of doing the same job can waste expensive tokens and GPU cycles while adding complexity to operations.

Red Hat’s answer is the Ansible Automation Platform 2.7, which includes a new automation orchestrator. The company said the orchestrator unifies three modes of automation - task-based deterministic, event-driven, and AI-driven - under one governance plane. That setup is meant to let enterprises keep control over automated actions while still using AI where it adds value.

The release also makes Ansible’s Model Context Protocol server generally available. Red Hat said this allows any AI tool of choice to call Ansible playbooks, giving AI systems a controlled path into approved automation workflows.

Balakrishnan said agents should only act within their domain of knowledge. He said AI should be used when a team encounters a new problem it has not yet discovered, while known issues should be resolved and moved into a deterministic catalog of actions.

Red Hat also said Cisco Systems is now reselling the Ansible Automation Platform, which Balakrishnan described as validation that Ansible has become the de facto standard for network automation. The comments came as Red Hat used the summit to present Ansible as the layer that can make agentic AI usable in production IT environments.