Building an Automation Center of Excellence with Ansible Automation Platform 2.7

This post summarizes Building an Automation Center of Excellence with Red Hat Ansible Automation Platform 2.7, the deck behind Level Up CTO Daniel Goosen’s session at Red Hat Summit: Connect 2026 in San Jose.

The through-line is simple: an Automation CoE is not a team that writes all the playbooks. It is the control plane that lets the rest of the organization create, consume, govern, and measure automation at scale, especially as AI generates more of it.

Download the slides (PDF)

Why now

Automation is still how organizations become cloud-native. AI has not reduced the need for governance; it has increased the amount of automation that has to be trusted.

The session framed that as a compounding loop: more generated automation → more operational surface area → more need for standards and a CoE. DevOps maturity also shows up in the AI numbers: high-maturity teams are far more likely to embed AI into the SDLC than low-maturity ones, and most estates are already hybrid.

The zero-to-one problem

The hard problem is rarely “can we write a playbook?” It is turning individual wins into an organizational capability:

Scripts (one person) → playbooks (one team) → platform (across teams) → CoE (repeatable at enterprise scale).

Organizations usually stall for operating-model reasons, not syntax: hero engineers, duplicated content, credentials that don’t scale, no product owner for the platform, no value story for leadership, and training that starts after go-live.

The innovation paradox

If you last saw Ansible Automation Platform more than a year ago, 2.7 is a different conversation. Platform engineering, governance, and AI-driven automation now sit in the same product story. Read the features as CoE primitives:

• Self-service portal — how do we democratize execution without giving up control?
• Visual execution-environment builder and content discovery — how do we standardize runtime?
• Browser-based development workspaces — how do we standardize how automation is built?
• OIDC with HashiCorp Vault and short-lived, job-scoped credentials — how do we reduce credential risk?
• Automation dashboard (technology preview) — how do we prove value?
• BYOK assistant and AAP MCP server (technology preview) — how do AI agents invoke approved automations safely?

Five stages of maturity

Maturity is not playbook count. It is the ability to create, govern, consume, and measure automation:

1. Individual
2. Team
3. Platform (AAP + governed execution)
4. CoE (enablement, standards, metrics)
5. AI-native (agents invoke trusted automation)

A short diagnostic before you argue about catalog size: who creates automation, who can safely consume it, how runtime consistency is enforced, how business value is measured, and whether AI can invoke approved automations safely.

What a CoE actually does

Four verbs: build, govern, enable, measure.

Centralize the platform and guardrails. Federate domain expertise and delivery. The KPI is more teams automating independently without sacrificing safety or consistency.

The reference architecture is a paved road from intent to trusted execution: consumers → experience (portal, API, AI/MCP) → control (controller, EDA, RBAC) → supply chain (Git, private hub, execution environments) → targets, with identity, secrets, observability, audit, lifecycle, and metrics cutting across. Treat automation like a product: build, validate, package, publish, run, measure.

Deep specialization from a partner is not a failure of self-sufficiency. Software plus implementation plus training is often the shortest credible path to a customer team that owns the platform.

Lead with business value

Do not start with “we need an automation platform.” Start with outcomes leadership already owns: faster (lead time, MTTR), safer (change failure, audit), cheaper (hours saved, cost avoidance), scalable (active users, reuse). Define the measurement model before the first production sprint so ROI is part of the platform, not an annual archaeology project.

A practical first 90 days: assess and pick 2–3 use cases; build guardrails, golden paths, and champions; ship production work and publish metrics. Exit criteria is a repeatable operating model, not an impressive demo.

The goal of an Automation CoE is not to centralize automation. It is to decentralize automation safely.

Session announcement: Building an Automation Center of Excellence with Red Hat Ansible — Summit Connect San Jose