Introduction
Moving artificial intelligence from experimental pilots to production systems is a challenge many organizations face. The key isn't just building models—it's operating them with consistent identity, governance, and security. At Red Hat Summit 2026, Microsoft and Red Hat showcased how Microsoft Azure Red Hat OpenShift serves as a secure, scalable foundation for this journey. A standout example is Banco Bradesco, a major Latin American financial institution, which unified governance across more than 200 AI initiatives on this platform. This guide walks you through the steps to achieve similar platform modernization and production AI deployment, leveraging the same principles that earned Microsoft the Red Hat Ecosystem Innovation Award for Platform Modernization.
What You Need
- An Azure subscription with permissions to create and manage resources.
- Basic familiarity with Kubernetes and container orchestration concepts.
- Access to Red Hat OpenShift – either via Azure Red Hat OpenShift (ARO) or a self-managed cluster.
- Azure Active Directory (Entra ID) configured for identity management.
- Existing AI models or pilot projects that need operationalization.
- Governance and security policies defined for your organization (e.g., compliance requirements like PCI-DSS for financial services).
- Development and operations teams who can collaborate on CI/CD pipelines.
Step-by-Step Guide
Step 1: Assess Your Current State and Define Objectives
Before diving into technology, evaluate where you stand. Banco Bradesco started with strict regulatory and security requirements at massive scale. Map out your existing application portfolio, identify which workloads are ready for containerization, and list all AI pilots that need to move to production. Define clear goals: consistent governance, unified identity management, and scalable infrastructure. This assessment will guide every subsequent decision.
Step 2: Provision Azure Red Hat OpenShift as Your Hybrid Cloud Foundation
Azure Red Hat OpenShift (ARO) is a jointly managed Kubernetes platform that combines Red Hat OpenShift’s enterprise Kubernetes with Azure’s cloud services. Provision a cluster via the Azure portal or CLI. Ensure you enable Azure Active Directory integration for single sign-on and role-based access control. This creates a secure, scalable foundation like the one Banco Bradesco used. For production, configure high availability and disaster recovery options such as multiple availability zones.
Step 3: Integrate Azure Identity, Security, and Policy Services
The power of ARO lies in its deep integration with Azure. Connect your cluster to Azure Policy to enforce regulatory and corporate compliance across all namespaces. Use Azure Defender for Containers for threat detection. For identity, map Azure AD groups to OpenShift groups so permissions flow seamlessly. This is how Banco Bradesco unified governance across 200+ AI initiatives—by leveraging Azure’s policy and security capabilities consistently.
Step 4: Containerize and Migrate Existing Applications
With your platform ready, begin migrating existing applications. Start with a pilot application that can be containerized easily. Use tools like Azure Migrate or Red Hat Migration Toolkit for Containers. Ensure each application runs in its own namespace with appropriate resource quotas and security contexts. Monitor for performance degradation. Once verified, repeat for other workloads, prioritizing those that will interact with AI models.
Step 5: Build and Deploy AI Models with Consistent Governance
Now operationalize your AI models. Package each model as a container (e.g., using ONNX or TensorFlow Serving). Deploy them onto ARO as Kubernetes deployments. Use Azure Machine Learning workspaces integrated with OpenShift to manage model versions and track experiments. Apply the same policies from Step 3 to AI namespaces—ensuring no model can access unapproved data stores. Banco Bradesco’s platform unified governance across all AI initiatives, meaning every model adheres to the same security rules.
Step 6: Scale from Pilots to Production Systems
Move from individual deployments to a robust production system. Implement horizontal pod autoscaling based on CPU or custom metrics (e.g., inference requests). Use Azure Monitor and Red Hat Advanced Cluster Management for observability across clusters. Create CI/CD pipelines with Azure DevOps or ArgoCD to automate deployments and rollbacks. For Banco Bradesco, this meant moving beyond proof-of-concept to a production platform handling thousands of transactions per second across 200+ AI workloads.
Step 7: Establish Continuous Governance and Optimization
Production AI isn't a one-time effort. Continuously review policy compliance using Azure Policy reports. Optimize costs with Azure Cost Management rightsizing recommendations. Schedule regular security scans and update OpenShift as new patches become available. Encourage feedback loops between data scientists and operations teams to refine models and platform performance. The journey from pilot to production is iterative—the recognition Microsoft received as Platform Modernization Partner of the Year underscores the importance of ongoing partnership and optimization.
Tips for Success
- Start with a small, high-value pilot that aligns with your business goals. Banco Bradesco likely didn't migrate all 200 initiatives at once; they proved value first.
- Leverage joint support from Microsoft and Red Hat. Since ARO is a jointly supported service, use both support channels when facing issues.
- Invest in training for your teams on both Azure and OpenShift. The hybrid nature requires skills in both ecosystems.
- Automate everything—from cluster provisioning (Infrastructure as Code) to model deployment (CI/CD). Automation reduces human error and accelerates scaling.
- Measure what matters: track time to deploy, compliance violations, and inference latency to quantify modernization success.
- Engage with the ecosystem: Consider participating in the Red Hat Ecosystem Innovation Awards to benchmark your progress against peers and gain visibility.
By following these steps, you can transform how your organization operates AI—moving from isolated experiments to a unified, secure, production-grade platform. The path taken by Banco Bradesco and recognized at Red Hat Summit 2026 can be your roadmap too.