AI Universe technical brief
How AI Universe deploys into your Azure tenant.
This page covers how AI Universe deploys into your Azure tenant: how the platform is structured, what each layer is responsible for, and where the boundaries of ownership and control sit.
AI Universe is not a SaaS platform operated by Fortytwo. It deploys into your own Azure subscription, and you own and control all Azure resources from day one.
Platform overview
The platform integrates Azure AI Foundry, Azure API Management (acting as the AI gateway), Microsoft Entra ID, Key Vault, Application Insights, Log Analytics, GitHub or Azure DevOps, and Bicep infrastructure-as-code into a single, governed deployment model.
The design intent is to give your team a production-grade AI capability that operates within the boundaries of your existing governance and security posture, without requiring a separate SaaS subscription, a third-party data processor, or a new trust boundary. Every component lives in your Azure tenant. Fortytwo delivers the initial deployment and supports ongoing operations but does not operate infrastructure on your behalf or retain persistent access after deployment is complete.
Layered deployment model
All layers reside within your Azure tenant. Each layer has a distinct responsibility boundary and no component sends data to Fortytwo infrastructure.
What lives where
Every component in a standard AI Universe deployment resides in your Azure tenant. The table below maps each component to its location, owner and relevant notes.
Azure API Management as governed front gate
Azure API Management is the single governed entry point for all AI traffic in an AI Universe deployment. Every galaxy calls every model through APIM. No application code reaches model endpoints directly.
Token-rate limits
Each galaxy’s APIM policy defines maximum token consumption rates. This prevents runaway usage and enables cost attribution per team or use case.
Azure Content Safety
Content Safety checks are applied at the APIM layer before requests reach the model, ensuring consistent content policy enforcement regardless of the calling galaxy.
Model-access
APIM policy determines which model deployments each galaxy may access. A galaxy can only call models it has been explicitly granted access to.
Token usage
All gateway policy events, token counts and request metadata are emitted to Application Insights and Log Analytics, enabling usage reporting and cost analysis per galaxy.
Prompt injection
Azure AI Foundry Prompt Shields operate at the AI Runtime layer, not at APIM. Injection defence and content safety are applied at distinct points in the request path.
Governance
Every AI call goes through APIM and is governed by your Azure policu enforcement. This is the single point of entry for AI.
Microsoft Entra ID integration
Identity in AI Universe is based entirely on Microsoft Entra ID. There are no separate user directories or credential stores managed by Fortytwo.
Per-galaxy App Registration
Each galaxy gets its own Entra App Registration in your tenant, with isolated client credentials and consent grants. No shared service accounts.
Galaxy key as primary credential
The galaxy key is derived from the App Registration’s client secret. Revoking the key immediately stops that galaxy’s ability to authenticate. There is no secondary access path.
Role-based platform administration
Platform-level operations are gated by Entra-managed roles. Role assignments are managed by your identity administrators, not Fortytwo.
No shared credentials across galaxies
Galaxies do not share credentials, App Registrations, or Key Vault access policies. A compromised galaxy credential cannot reach another galaxy’s resources.
Azure Key Vault
All platform and galaxy secrets are stored in Azure Key Vault instances within your Azure tenant. Fortytwo does not operate a shared secrets store and does not retain persistent access to your Key Vaults after the initial deployment is complete.
Per-galaxy isolation
Each galaxy’s secrets are held in Key Vault entries scoped to that galaxy. Access policies ensure that a galaxy’s container app can only retrieve its own secrets. Cross-galaxy secret access is not possible under the standard configuration.
Fortytwo access model
During initial deployment, Fortytwo engineers may require temporary access to configure Key Vault entries. This access is time-bounded, documented and revoked after deployment is confirmed. Fortytwo does not retain standing access to your Key Vaults in production.
Git-based delivery
Each galaxy is delivered as a Git repository under your GitHub organisation or Azure DevOps project. The repository is the authoritative delivery image for that galaxy. All infrastructure changes flow through the repository’s CI/CD pipeline.
Application Insights & Log Analytics
Observability is built on Azure-native tooling. All telemetry is written to your workspaces within your Azure tenant.
AI usage and cost
Token consumption, model call counts and inferred costs are tracked per galaxy through APIM policy events forwarded to Application Insights. This enables per-team cost attribution and capacity planning.
Gateway policy events
Content Safety decisions, rate-limit events and model-access control outcomes are written as structured log entries in Log Analytics. These form the audit trail for gateway enforcement.
Application health
Container app health, error rates and dependency latency are captured in Application Insights. Alerts and dashboards can be configured by your operations team using standard Azure Monitor tooling.
Have questions?
Reach out to the Fortytwo team if you want to know more about deployment. We are happy to discuss the technical details, your specific Azure setup, or what a deployment would involve for your organisation.
Update and release model
Fortytwo delivers platform updates as new container images and updated Bicep templates. Updates are communicated through the standard release process. Release notes, changelog and any required migration steps are documented before delivery.
You control when updates are applied, and Fortytwo does not push changes to your tenant, trigger deployments or modify infrastructure without your explicit action. The CI/CD pipeline, owned and operated by you, is the only mechanism through which changes are applied.
Breaking changes
Breaking changes are announced with a minimum of two release cycles’ notice. Migration guides accompany all breaking changes.
Rollback
Every build runs secret sBecause all infrastructure is codified in Bicep and application versions are tagged container images, customers can roll back by reverting the repository and re-running the pipeline.
Release cadence
Platform releases are typically monthly. Security patches may be issued outside the regular cadence and will be flagged as priority.
We are the experts who wire it all up
Fortytwo has opinionated the entire setup and wired it together: identity, model access, governance, delivery and code quality, built on the vast ecosystem of AI features in Azure.
You start from our decisions, not a blank subscription, but you run it yourself. Having one control plane gives you self-service over policy, governance, applications (galaxies) and wisdom, instead of switching between a dozen different Azure portals.
We don’t hand you scattered AI tools; we hand you a governed foundation you own.
FAQ
How does AI Universe deploy into an Azure tenant, and what resources does it create?
AI Universe deploys as a container-based platform directly into your Azure subscription using Bicep infrastructure-as-code templates aligned with the Microsoft Cloud Adoption Framework. The deployment provisions a set of Azure resources entirely within your tenant: Azure Container Apps for the control plane and per-application environments, Azure API Management as the AI gateway, Azure AI Foundry for the model and agent runtime, Azure Key Vault for secrets, Application Insights and Log Analytics for observability, and Microsoft Entra App Registrations for per-application identity.
Fortytwo engineers perform the initial deployment into your tenant and hand over operational control once deployment is confirmed. All provisioned resources belong to your Azure subscription. Fortytwo does not operate any shared infrastructure that your data or application traffic passes through after deployment is complete.
What is the role of Azure API Management in AI Universe, and why does all AI traffic route through it?
Azure API Management is configured as the governed AI gateway in every AI Universe deployment. Every AI call from every application environment routes through APIM before reaching any model endpoint in Azure AI Foundry. No application code is permitted to call model endpoints directly under the standard configuration.
Routing all AI traffic through APIM serves several purposes. Token-rate limits are applied per application at the gateway layer, preventing runaway usage and enabling cost attribution by team or use case. Azure Content Safety checks run at APIM before requests reach the model. Model-access policies determine which model deployments each application may call. All gateway policy events, token counts, and request metadata are written to Application Insights and Log Analytics within your tenant, forming the structured usage and audit log for the platform.
This single-gateway approach means governance controls are applied consistently across all applications regardless of how individual teams build their features. Policy changes are made once at the gateway layer rather than in each application separately.
What access does Fortytwo retain to a customer’s Azure environment after the initial deployment is complete?
Fortytwo does not retain standing access to your Azure environment after the initial deployment is confirmed. During deployment, Fortytwo engineers may require temporary access to configure resources, including Key Vault entries, APIM policies, and container app settings. This access is time-bounded and revoked once deployment is complete and signed off.
After deployment, Fortytwo’s ongoing role is to deliver platform updates as new container images and updated Bicep templates. These are delivered to your team and applied by your CI/CD pipeline, which runs under your control in your GitHub organisation or Azure DevOps project. Fortytwo cannot push changes to your tenant, trigger deployments, or modify infrastructure without your team taking an explicit action.
For incident support, your team grants access on a case-by-case basis as needed. There is no persistent remote access mechanism built into the platform.
How does AI Universe isolate individual applications from each other within the same Azure tenant?
Each application environment in AI Universe, referred to as a galaxy, has its own isolated identity and credential boundary. Isolation is implemented at three levels.
Identity isolation: Each galaxy has its own Microsoft Entra App Registration in your tenant. Credentials are not shared between galaxies. Revoking a galaxy’s App Registration stops that application immediately without affecting any other galaxy running in the same environment.
Secrets isolation: Each galaxy’s secrets are held in Azure Key Vault entries scoped to that galaxy. Access policies ensure a galaxy’s container app can only retrieve its own secrets. Cross-galaxy secret access is not possible under the standard configuration.
Gateway isolation: Each galaxy operates under its own APIM policy scope, with independently configured token-rate limits and model-access permissions. One galaxy cannot consume another galaxy’s token budget or access models it has not been explicitly granted.
How are updates delivered?
Platform updates are delivered as new container images that are re-imported into your environment. Application template updates arrive as reviewed pull requests to your own repositories, and infrastructure changes are re-applied through Bicep. You see every change before it lands, and nothing is modified outside your control.
What does AI Universe cost?
AI Universe is a framework subscription at a flat monthly fee. Azure resources run in your own tenant and are billed by Microsoft at cost, and purchases through Azure Marketplace can draw down your Microsoft Azure Consumption Commitment (MACC). Contact us for current pricing and a deployment plan.
How do we manage AI Universe day to day?
Through one self-service control plane. Policy, governance, galaxies, wisdom and model access are all managed in one place, so day-to-day operation does not require switching between Azure portals. Everything still runs as standard Azure resources in your own subscription, so the Azure portal is never closed to you, it is just not needed for daily work.
Let’s talk about AI
Do you have more questions about the technical sides of AI Universe? Please fill out the form below, and we’ll get in touch as soon as possible to schedule a short call with you.
