Azure landing zone and Microsoft Foundry for company-wide AI use
We built an Azure foundation with clear rules, provided GPT models in the EU and trained the teams for OpenAI Codex. From the first subscription to working with AI every day.
Challenge
Give teams access to generative AI and coding agents without losing sight of data protection and cost. On a foundation that carries every workload that follows.
Initial situation
No cloud foundation
No guardrails. Azure was in use without a management group structure, central policies or a single identity. Every new workload would have been set up by hand, and differently each time.
AI outside any rules
Uncontrolled data flow. Without a sanctioned platform, there was no answer to where prompts and code go, who may use which model, or what it costs.
Licences without adoption
No productivity gain. Access to a tool does not change how work gets done. The teams had to learn to do their own work with it, and operations had to learn to run it.
Solution
A platform with clear access rights, and teams who know how to use it. Compliant from the first login.
Azure landing zone
A management group hierarchy with platform, landing zones for corporate and online workloads, and a sandbox for test and development. Azure Policy guardrails, central logging and audit from the first subscription, and documentation ready for certification, including risk and threat management.
Identity and access
Single sign-on through Microsoft Entra ID, connected to the existing identity provider. Access to AI models is granted per group, not per person.
Microsoft Foundry on EU Data Zone deployments
GPT models run through EU Data Zone deployments. Prompts and outputs stay in the company’s own tenant and are not used for training.
Budgets and traceability
Cost limits and alerts per team and application in Azure Cost Management. Every access is logged in Azure Monitor and can be audited.
OpenAI Codex training
An on-site workshop over two and a half days for users and operations. It starts with fundamentals such as GDPR and the EU AI Act, setup and business case. Then the teams work on their own tasks with Codex. The last day covers roles, quotas, monitoring and budgets for the people who run it.
