FAQ for site copilot
These frequently asked questions (FAQ) describe the AI impact of copilot feature in Power Pages.
What is copilot in Power Pages?
The copilot feature in Power Pages provides you with an easy way to configure a GPT-powered Microsoft Copilot Studio copilot for your website. The copilot enhances the interaction experience for the website users by enabling them to ask natural language (NL) questions and receive summarized responses with relevant links. This experience allows site users to get the necessary information available on the website - for example, through FAQs or knowledge articles easily using the copilot without the need to search and locate the information manually.
What are the system’s capabilities?
A copilot for sites created in Power Pages for providing generative answers to customer queries. Bing indexing is used for unauthenticated public site content, while Dataverse Copilot handles the indexing of private and authenticated user-specific content. This enables web users to ask natural language questions and receive summarized responses. You can test the copilot before publishing it to ensure that it provides appropriate summaries. Service administrators can also turn off the publishing of the copilot at the tenant level to prevent accidental or unintended exposure of AI capabilities on the public site. Additionally, you can navigate to the associated Microsoft Copilot Studio copilot for advanced configuration and customization.
What is the system’s intended use?
Copilot on Power Pages site enhances the conversational capabilities by enabling site users to ask queries in natural language and receive summarized responses. Copilot has the ability to add more complex conversational features, such as contextual understanding, entity recognition, and sentiment analysis. These capabilities allow the copilot to better understand user inputs and provide more accurate and helpful responses.
How was copilot feature evaluated? What metrics are used to measure performance?
The capability was evaluated on a collection of manually curated question-and-answer datasets, covering multiple industries.
More evaluation was performed over custom datasets for offensive and malicious prompts and responses.
What are the limitations of this feature? How can users minimize the impact of the copilot limitations when using the system?
- This feature doesn't include a mitigation for checking for correctness of responses returned by the copilot. Nonfactual responses might be generated if the URL from the information is gathered (and what the maker provided) contains incorrect information.
- GPT answers don't support websites that sell prohibitive products or services, as these terms get intentionally blocked by content moderation. An exception to this exclusion is if the content moderation slider gets set to Low.
- That the copilot sometimes returns misguiding responses for high-risk domains that include healthcare finance, communications, and legal.
What operational factors and settings allow for effective and responsible use of the system?
You're able to test the copilot before making it available to your site users. Evaluate copilot's performance—that is, the ability of copilot to return relevant, accurate, and offensive-language-free responses to site user's inquiries.
When a feature is enabled, the copilot is created with a content moderation set to High by default. This setting filters the offensive content by using Azure Open AI’s content filtering and Azure Content Moderator.