One identity boundary
Microsoft Entra ID provides single sign-on, while server-side RBAC decides what each user can do in each product.
One implementation brief for a multi-tenant university ecosystem spanning an AI campus experience, cross-tenant communications, and a central administration control plane.
Start here
Three applications share identity, tenant context and a common Azure SQL data foundation. Intelli Campus adds the AI and retrieval layer; Norvia Backoffice governs provisioning.
Use this diagram as the target-state reference throughout all nine phases. Each phase introduces or strengthens part of this final platform.
Open full size ↗Microsoft Entra ID provides single sign-on, while server-side RBAC decides what each user can do in each product.
Every business record is resolved through trusted tenant context. Tenant identity is never accepted from browser-controlled input.
Azure SQL connects the products, but each application owns its workflows and accesses only the data its contract allows.
Azure AI Search indexes approved sources and Azure OpenAI answers through RAG and bounded agents.
End-to-end journey
From university onboarding to a student receiving a trusted answer, each system has a deliberate responsibility.
Norvia Operations creates the tenant, products, roles and tenant administrator in Backoffice.
Norvia BackofficeProduct mappings are written to shared SQL and platform accounts are provisioned through governed APIs.
Backoffice + Azure SQLThe university user authenticates with Entra ID. Tenant and role are resolved on the server.
Entra ID + RBACStudents search grounded knowledge, ask the advisor agent and raise a support ticket when needed.
Intelli CampusUsers receive their university announcements plus global messages, isolated by trusted tenant context.
Intelli BroadcastProduct landscape
The course moves from a focused campus application to a connected, multi-product SaaS architecture.
AI-powered university support, knowledge discovery and ticket resolution with a human-in-the-loop advisor agent.
A deliberately focused announcement product that proves cross-product identity, shared data and tenant isolation.
The technology-neutral control plane for products, product roles, tenants, administrators and provisioning.
Progressive delivery
Every phase is a releasable learning milestone. Later phases extend the same system; they do not discard the earlier work.
Deliver the complete core product journey in OutSystems ODC: students can register, sign in, raise and track support tickets, while staff manage the queue and approved knowledge sources.
Move the knowledge pipeline onto university-managed Azure resources without changing the student or support experience.
Give every new ticket an immediate acknowledgement and an AI-assisted resolution by email, with a clear path back to a human support engineer.
Refactor the growing application into maintainable UI, core and integration responsibilities and remove secrets from business logic.
Replace one-shot question answering with a bounded ReAct agent that selects tools, observes results and requests confirmation before consequential actions.
Evolve Intelli Campus from one institution into a SaaS product where many universities share the platform but never each other's data.
Move core business data to a platform-neutral Azure SQL model so OutSystems and future standalone applications can use governed shared records.
Remove platform administration from Intelli Campus and deliver a dedicated .NET application for Norvia operations.
Add a second customer-facing .NET product and prove the final architecture with shared Entra identity, product access mappings and tenant-safe announcements.
Applies to every phase
The platform is complete only when its cross-cutting requirements are treated as product features, not cleanup work.
HTTPS, secure sessions, secret management, server-side RBAC, input validation and least-privilege service access.
Trusted tenant resolution, mandatory query scoping, cross-tenant negative tests and no client-supplied authorization context.
Structured logs, correlation identifiers, safe integration failure capture and health checks without sensitive data.
Semantic UI, labelled controls, keyboard navigation, visible focus, usable contrast and responsive layouts.
Idempotent webhooks, stable pagination, retry-safe provisioning, graceful fallbacks and documented recovery paths.
Clear ownership, referential integrity, audit fields, minimal retention and approved knowledge sources for AI.
Your build starts here