AI Executive Office — CXO Assistant Platform · View 08 of 30 · 3 · Structure
Decisions and rationale
- Azure Container Apps rather than AKS: the workload is a handful of stateless HTTP services with scale-to-zero characteristics between briefings. AKS is the right answer at a much larger estate and is a documented exit, not a rewrite
- Durable Functions for investigations because an investigation is a long-running, resumable orchestration with checkpoints — not a long HTTP request. This is what makes the asynchronous experience honest rather than a spinner
- Every service authenticates with a managed identity. There are no connection strings in configuration anywhere in this diagram
Assumptions
- Provisioned throughput for Azure OpenAI is assumed for latency predictability under the 07:10 brief spike, with pay-as-you-go overflow. Confirm capacity availability in the target region before committing
- Fabric F64 is the starting capacity for a mid-size tenant. Sizing is a per-tenant exercise driven by refresh frequency, not user count
Deliberate omissions
- Private endpoints exist on every PaaS service here but are drawn only in views 21 and 26, where they are the subject rather than clutter
- The self-hosted integration runtime is drawn once; in practice one runs per on-premises network the tenant exposes
