1.2.2. Platform and Infrastructure Layers
💡 First Principle: When productivity AI isn't enough—when you need custom workflows, specialized knowledge, or unique AI models—you move down to platform (Copilot Studio) or infrastructure (Foundry) layers. Each step down increases control but also increases complexity and cost. The transformation leader's job is to find the minimum layer that meets the business need.
The Platform Layer (Copilot Studio) is appropriate when:
- You need custom agents for specific workflows (HR bot, IT helpdesk)
- Standard Copilot capabilities need to be extended with custom actions
- You want to ground AI on specialized knowledge bases
- Low-code/no-code development is preferred
The Infrastructure Layer (Microsoft Foundry) is appropriate when:
- You're building proprietary AI products
- You need to train or fine-tune custom models
- You require specific models from the catalog (over 10,000 options)
- Enterprise-grade AI governance and security are required
The infrastructure layer also includes Azure AI Services (now Foundry Tools): prebuilt vision, speech, language and document APIs that developers embed in custom apps. They are building blocks for developers, not end-user productivity tools (see 3.3.2).
| Layer | Control | Complexity | When to Use |
|---|---|---|---|
| Productivity (Microsoft Copilot) | Low | Low | Standard productivity enhancement |
| Platform (Copilot Studio) | Medium | Medium | Custom agents, extensions |
| Infrastructure (Foundry) | High | High | Custom AI products, unique models |
⚠️ Exam Trap: The exam may present a simple business need and offer Foundry as an option. Foundry is typically wrong for simple scenarios—it's overkill. Look for answers that match complexity to need.
Reflection Question: A company wants a chatbot to answer HR policy questions. Should they use Microsoft Copilot, Copilot Studio, or Foundry? Why?
