AB-731 — frequently asked questions

Choose Azure Machine Learning for predictive analytics, classification, or regression tasks involving large datasets and model training. Choose Copilot Studio when the requirement is to build a conversational interface or chatbot for user interaction.

M365 Copilot is the assistant itself that works inside apps. Connectors are the bridges that allow Copilot to access external data sources (like ERP systems) to provide context-rich answers within those same apps.

No, Microsoft 365 Copilot Chat is included with eligible Microsoft 365 subscriptions like Business Standard, E3, and E5. It does not require the separate 'Microsoft 365 Copilot' add-on license.

It uses Microsoft Graph signals to provide context-aware help. This includes open files, emails, calendar events, and other organizational content you have permission to access, not just open files.

No. Azure AI Search indexes and retrieves data. Fine-tuning involves adjusting model parameters using training datasets, which is done outside the search service.

It uses semantic search capabilities and vector embeddings to understand intent and context, allowing users to ask questions in plain English rather than using strict keyword filters.

An AI governance council sets strategy, ensures ethical compliance, and aligns business units. IT admins handle technical deployment, configuration, and user support.

No, monitoring user behavior is an operational task for IT security or compliance teams. The council provides high-level oversight and policy guidance.

A chatbot engages in two-way conversation, asking questions and providing answers. A recommendation engine passively suggests items based on data patterns without initiating dialogue.

No. Predictive AI forecasts outcomes based on historical data. It does not have the conversational interface needed to interact with users in real-time.

No. Copilot only accesses data that the specific user has permission to view. It respects individual file and site permissions enforced by Microsoft 365.

It is the process where Copilot connects generative AI responses to the user's real-time enterprise data like emails, documents, and chats via the Microsoft Graph API.

No. Accessing organizational data on mobile devices is supported by standard Microsoft 365 apps and licenses. Copilot is an add-on feature, not a prerequisite for basic data access.

Graph itself is widely accessible, but using it to 'reason' over data via AI is a Copilot-specific capability. Therefore, the reasoning action requires the license.

No, if they have a Microsoft 365 Copilot add-on license, they can build and use internal declarative agents within the M365 ecosystem without an additional Studio user license.

Declarative agents are built using natural language instructions and are simpler, suitable for end-users. Procedural agents involve code and logic, typically requiring Copilot Studio for more complex, professional development.

Pay-as-you-go is for usage-based consumption of AI features when you don't have a license. Building agents that access work data generally requires the permissions granted by the Microsoft 365 Copilot license.

No, E5 is a prerequisite but does not include Copilot. You must purchase a separate per-user Copilot add-on license.

They are used by IT administrators for governance and deployment management, not by end-users to create AI agents.

Azure Speech converts audio to text (transcription). Azure Content Understanding analyzes the audio file itself to extract semantic insights like topics, sentiment, and summaries without needing manual post-processing of the transcript.

Document Intelligence uses AI models to understand document structure, identifying fields, tables, and key-value pairs, whereas standard OCR only extracts raw characters without structural context.