Red Hat Ansible 2.7 and the "trusted execution layer" frame

“Trusted Execution Layer for IT Operations in an Agentic Era”

On May 12, 2026 at Red Hat Summit, Red Hat announced Ansible Automation Platform 2.7 with a strategic positioning frame that is worth quoting and worth taking seriously: Ansible is now positioned as “the trusted execution layer for IT operations in an agentic era.” The announcement landed alongside several concrete shipping items:

• A technology preview of an AI-agent orchestration engine, capable of invoking capabilities via an integrated Model Context Protocol (MCP) server.
• A new automation-orchestrator workflow canvas that lets teams combine deterministic automation, event-driven automation, and AI-driven automation inside the same workflow, with shared data and unified governance.
• A universal AI bridge through the MCP server, simplifying AI-driven automation by removing the need for custom integrations between AI tools and Ansible.
• “Bring-your-own-knowledge” functionality for the automation intelligent assistant, letting customers generate more contextual AI responses against their own knowledge base.
• Opinionated solution guides for AIOps at scale, with named integrations against IBM Instana, ServiceNow, and Splunk.

Red Hat Ansible Automation Platform 2.7 will be available in the coming weeks; the automation orchestrator capability is on the calendar for later in 2026.

The shipping list is interesting. The positioning frame is more interesting. “Trusted execution layer for IT operations in an agentic era” is the clearest public articulation in 2026 of what an active operational layer actually is. The frame is worth unpacking, because it is exactly the structural pattern that mid-market and enterprise platform teams should be looking for — regardless of whether they buy Ansible.

Unpacking the Frame

“Trusted” — the layer carries an audit trail, a governance surface, and a capability-tier model on every action. It is not a copilot, not a dashboard, not a runbook archive. It is the layer that executes IT-operations work with provable governance.

“Execution layer” — the layer does the work, on production systems, in production environments. It does not just propose work. It does not just observe work. It does not just notify a human about work. It is the operational fabric that the work runs through.

“For IT operations” — the surface is the customer’s actual infrastructure (servers, networks, virtualization, containers, identity, configuration), not just a workflow-orchestration plane above the infrastructure. The work touches the operating environment directly.

“In an agentic era” — AI agents are the new entrants that the execution layer has to accommodate. The agents are not the layer; they are participants in the layer. The layer governs what the agents are allowed to do, what scope they execute in, and what audit trail they leave behind.

The full phrase is a structural definition of the active operational layer. Substitute “the platform team” for “trusted execution layer” and the sentence still parses: the platform team is the trusted execution layer for IT operations in an agentic era. Then ask the structural question: is the work shaped like a platform team can sustain it manually at 2026 cadence, or does the platform team need a coordinated team of specialized agents to deliver the same shape?

The 2026 answer is the latter, for any platform team smaller than the largest enterprises.

The Three Structural Choices Inside Ansible Automation Platform 2.7

Three Ansible Automation Platform 2.7 design choices are worth specific attention from cloud-infrastructure platform teams:

MCP as the universal integration surface. Red Hat’s choice to ship a built-in MCP server inside Ansible Automation Platform — rather than a custom integration API per AI vendor — is the same choice AWS made for DevOps Agent’s on-premises resource discovery, the same choice Microsoft made for Agent 365’s external-tool integration, and the same choice the IAN active operational layer made for cross-vendor agent connectivity. MCP has become the universal connective tissue of the 2026 agentic-AI ecosystem. The implication for any platform team is that “what integrates with what” is increasingly answered by “everything that speaks MCP, on both sides of the connection.” Platform-team architectural decisions in 2026 should default to MCP-first integration.

Combined deterministic, event-driven, and AI-driven automation on a single canvas. The automation-orchestrator capability lets teams design a single workflow that mixes deterministic steps (do X, then Y, then Z), event-driven steps (when this event fires, do Q), and AI-driven steps (use the agent to decide what to do given this context). The structural insight is that 2026 IT-operations work is not purely deterministic and not purely AI-driven — it is a mix, and the mix lives inside the same workflow. Platform teams that have separate deterministic automation surfaces (CI/CD pipelines, Terraform), separate event-driven automation surfaces (event bridges, Lambda), and separate AI-driven automation surfaces (agents on a copilot platform) are paying for tool sprawl that the 2026 unified-canvas pattern collapses.

BYOK on the intelligent assistant’s knowledge base. “Bring-your-own-knowledge” is the Ansible frame for the same structural choice that BYOK-on-model-keys answers in the agentic-AI category. The customer’s knowledge — runbooks, operational policies, internal documentation, post-incident review records — is the asset. The agent vendor is the operating surface. The agent vendor should not own the customer’s knowledge; the customer should bring it in and the vendor should consume it under explicit governance. The structural pattern repeats: in agentic AI, the customer’s operational knowledge is the customer’s asset, not the vendor’s.

See the IAN team run on your cloud. We connect to your AWS account via a scoped read-only role, run the Observe-tier agents, and leave you with a concrete audit report — cost waste, security exposure, compliance gaps, and a labor-offset estimate. You keep the findings regardless of next steps. Get a free infrastructure audit →

Where the Frame Leaves Gaps for Cloud-Infrastructure Platform Teams

Ansible Automation Platform 2.7 is the most articulate public expression in 2026 of what an active operational layer is. It is also a specific product, optimized for a specific customer profile (Red Hat customers running Linux servers, network devices, Windows estates, and Kubernetes via OpenShift). The frame leaves three structural gaps for cloud-infrastructure platform teams whose work centers on AWS / GCP / Azure / Kubernetes:

The execution-layer assumes Ansible-native targets. Ansible’s strength is configuration management against servers, network devices, and standard enterprise infrastructure. Ansible’s coverage of cloud-native operational primitives — rightsizing an AWS Auto Scaling Group, rotating a Secrets Manager credential, applying a Terraform plan against an EKS cluster, draining a Kubernetes pod, auditing an AWS Security Group — is via integration modules and community-contributed roles. The coverage is real; it is also second-class to the Ansible-native server-management work. Cloud-infrastructure platform teams need an execution layer where AWS / GCP / Azure primitives are first-class, not bolted on.

The orchestration shape is workflow-canvas-centric, not agent-team-centric. The automation-orchestrator surface in 2.7 is a workflow canvas where teams compose deterministic, event-driven, and AI-driven steps. The shape is excellent for IT-operations workflows with a clear sequential or event-driven structure. The shape is less excellent for the cross-pillar agent-team work that cloud-infrastructure platform teams do — where the cost-agent’s rightsizing recommendation is an input to the security-agent’s exposure scoring, and the SRE-agent’s incident investigation reads from the deployment-agent’s release log. The agent-team shape is graph-shaped, not workflow-shaped.

The customer profile is enterprise IT, not cloud-native platform engineering. Ansible’s customer base is enterprise IT operations. The 2026 cloud-native platform-engineering team is a different shape — smaller, cloud-native, container-native, frequently MCP-native from the start. The execution-layer pattern translates; the specific Ansible-Automation-Platform product is heavier than the cloud-native platform-team needs.

How IAN Helps: The Active Operational Layer for Cloud Infrastructure

IAN is the AI DevOps team for cloud infrastructure, delivered as a coordinated team of specialized agents on the active operational layer. IAN’s shape is the same “trusted execution layer for IT operations in an agentic era” frame, applied to the cloud-native infrastructure surface that the Ansible-centric expression does not directly cover:

• Cloud-native primitives as first-class operations. Every cost-agent action (rightsize an AWS Auto Scaling Group, downshift an EC2 instance family, schedule a GCP commitment), every security-agent action (rotate an AWS Secrets Manager credential, audit a Security Group, scan a Terraform plan), every SRE-agent action (drain a Kubernetes pod, scale a node group, restart a deployment), every deployment-agent action (apply a Terraform plan, promote a release, roll back a deploy), every resource-agent action (tag, lifecycle, retention) is first-class on the IAN active operational layer.
• Cross-pillar agent team as the unit of coordination. The cost-agent, security-agent, SRE-agent, deployment-agent, resource-operations-agent, and compliance-agent coordinate as a team on the same operational fabric. Cross-pillar context (rightsizing → security exposure → incident hypothesis → release log) is the default, not a separate workflow.
• MCP-first integration surface. Every customer-side integration — AWS, GCP, Azure, GitHub, GitLab, Datadog, internal Mattermost, Slack, Kubernetes clusters — lives behind an MCP server. The IAN platform exposes its own MCP server to Claude Code, Cursor, and any MCP-aware client the customer’s practitioners already use. MCP is the universal connective tissue, on both sides of the connection.
• Capability-tier classification on every action. Observe-tier (continuous audit, drift detection, exposure scoring) runs automatically. Operate-tier (rightsize, rotate, restart, scale, drain, apply) is pre-authorized once with explicit scope. Administer-tier (IAM changes, billing-account-level actions, approval-policy changes) always requires explicit human approval. Drift between classified tier and actual scope is detected and surfaced.
• BYOK on model keys. Customer pays inference costs directly to Anthropic / OpenAI / their model provider. IAN charges for orchestration only. Pricing is usage-based on agent actions, with a monthly minimum. The customer’s knowledge — runbooks, policies, audit history — lives in the customer’s repository and the customer’s per-tenant database.
• Immutable audit trail in the customer’s database. Every cost-agent recommendation, every security-agent finding, every SRE-agent investigation, every deployment-agent release, every resource-agent tagging change, every compliance-agent reconciliation lands in the customer’s per-tenant audit-trail store. The audit trail is the reconciliation artifact for internal audit, external auditors, and cyber insurance.

The shape is the same trusted-execution-layer pattern Red Hat articulated for IT operations. The surface is the cloud-native infrastructure surface that mid-market and enterprise platform teams actually operate.

The Three-Phase Rollout

Phase 1 — Observe the cloud-infrastructure surface as it exists today. Inventory every AWS / GCP / Azure account, every Kubernetes cluster, every existing automation workflow (CI/CD pipelines, Terraform modules, ad-hoc scripts, runbooks), every existing AI surface (Copilot in pipelines, Cursor in IDEs, Claude Code in terminals, hyperscaler-native agents in observability). Classify the existing surface against the Observe / Operate / Administer tier model. Build the per-tenant audit-trail store. Two-to-four weeks. Output: the gap report against the trusted-execution-layer pattern.

Phase 2 — Codify the cloud-infrastructure skills library. Pre-authorize the Operate-tier scopes per cloud and per pillar. Write the first 20-40 customer-specific skills against the agents — rightsize-this-class-of-workload, rotate-this-class-of-credential, drain-this-class-of-pod, apply-this-class-of-Terraform-plan, audit-this-class-of-security-group. Validate the skills, instrument the audit trail, ship them as the operational baseline. Two-to-three months.

Phase 3 — Cross the cost / security / SRE / deployment / resource / compliance agent loop. The same skills layer, the same capability-tier governance, the same audit trail across every agent in the team. The operating model becomes the operational fabric. Continuous.

Red Hat’s “trusted execution layer for IT operations in an agentic era” frame is the clearest 2026 public articulation of what an active operational layer is. The shape — capability-tier governance, MCP-first integration, BYOK on the customer’s knowledge and the customer’s model keys, an immutable audit trail, a skills layer that captures customer operational knowledge — is the structural answer to the 2026 platform-engineering question of how to absorb agentic AI without losing the audit surface. The IAN active operational layer applies the same shape to the cloud-infrastructure surface that mid-market and enterprise cloud-native platform teams actually operate.

Get a free infrastructure audit → | See pricing →