Ask AI for help - Microsoft Foundry (original) (raw)

You can ask AI to assist you in Foundry. To start using AI to ask questions or complete tasks, select its icon located in the top right bar of the Microsoft Foundry portal. A chat window opens where you can type your questions and receive answers in real time. You can also ask the agent to run tasks for you.

Screenshot shows the Ask AI button in the top right bar of the Foundry portal.

Important

Items marked (preview) in this article are currently in public preview. This preview is provided without a service-level agreement, and we don't recommend it for production workloads. Certain features might not be supported or might have constrained capabilities. For more information, see Supplemental Terms of Use for Microsoft Azure Previews.

Prerequisites

To use Ask AI, you need:

Capabilities

What this AI can do - The Ask AI experience provides assistance by answering questions and performing tasks through specialized sub-agents:

What this AI can't do - While the agent is a powerful tool, it has some limitations and constraints:

Use the agent to make the most of the Foundry experience but keep its scope and limitations in mind when asking questions.

Actions and approvals

When you ask the Ask AI agent to perform tasks that require accessing or modifying your Azure resources, the agent proposes actions for you to review and approve before execution. This approval flow ensures you maintain oversight over what actions are performed on your behalf.

Screenshot shows the Ask AI chat on the right side of Foundry portal. The Ask AI agent is responding a user query by proposing to run an action for the user to approve.

The actions come from the tools available under the Foundry Model Context Protocol (MCP) Server.

To make this approval flow easier, you can change the approval settings to pre-approve some actions depending on their scope. Access the approval settings by selecting the settings icon in the Ask AI prompt chat box. By default, this experience is set to pre-approve System access actions. You can change these settings anytime, and they persist beyond your session.

Screenshot shows the Ask AI prompt chat box, showing the location of the approval settings.

Best practices and security guidance

The Ask AI experience relies on the Foundry Model Context Protocol (MCP) Server. By using this server, it implements the same best practices and security guidance.

Important

Ask AI is a preview feature and is available in the Foundry portal across all supported regions. Because it's in preview, it isn't covered by a service-level agreement (SLA) and isn't recommended for production workloads.

By using this preview feature, you acknowledge and consent to any cross-region processing that may occur. As an example, an EU resource accessed by a US user could be routed through US infrastructure. If your organization requires strict in-region processing, don't use Ask AI (preview) or restrict its use to scenarios that remain within your selected region.

Responsible AI FAQ for Ask AI

What is Ask AI in Foundry?

It's an AI agent that enables users of Foundry to navigate its capabilities, identify models, and understand how to use its resources to build generative AI applications. For an overview of how the agent works and a summary of its capabilities, see the overview earlier in this article.

What is the current status of the Ask AI feature?

It's available in Foundry as a preview feature.

Are the Ask AI results reliable?

The agent is designed to generate the best possible responses within the context it can access. However, like any AI system, the agent's responses aren't always perfect. Carefully test, review, and vet all of the agent's responses before using them in Foundry or for your application.

How do I provide feedback on Ask AI?

If you see a response that's inaccurate or doesn't support your needs, use the thumbs-down button to submit feedback. You can also submit feedback on your overall experience by using the Foundry feedback button on the top menu.

What should I do if I see unexpected or offensive content?

The Foundry team built the agent by following the AI principles and Responsible AI Standard. The team prioritized mitigating exposing customers to offensive content. However, you might still see unexpected results. The team is constantly working to improve the technology to prevent the output of harmful content.

If you encounter harmful or inappropriate content in the system, select the thumbs-down icon below the response to provide feedback or report a concern.

How current is the information provided by the agent?

The agent is updated daily to keep it up to date with the latest information. In most cases, the information the agent provides is up to date. However, there might be some delay between new Foundry announcements and updates to the agent.