Responsible AI FAQ for Creating calculation models with Copilot (preview)

Important

Some or all of this functionality is available as part of a preview release. The content and the functionality are subject to change.

Note

This feature is included in Microsoft Sustainability Manager Premium.

This FAQ provides answers to frequently asked questions about the AI technology used in Create calculation models with Copilot (preview) in Microsoft Sustainability Manager. It includes key considerations and details about how the AI is used, how it was tested and evaluated, and any specific limitations.

What does it mean to create calculation models with Copilot?

Creating calculation models with Copilot is a feature within Microsoft Sustainability Manager that uses OpenAI's GPT (Generative Pre-trained Transformer). It enables users to formulate calculation models through natural language input, simplifying the creation process.

What can creating calculation models with Copilot do?

Creating calculation models with Copilot facilitates the creation of calculation models by allowing users to articulate their requirements in natural language. This feature enhances efficiency and reduces the complexity associated with setting up calculations in Microsoft Sustainability Manager.

What is the intended use of creating calculation models with Copilot?

The primary purpose of creating calculation models with Copilot is to empower sustainability practitioners by streamlining the process of creating calculations. Its aims to enhance accuracy, efficiency, and user understanding of calculation model fundamentals.

How was Copilot evaluated? What metrics are used to measure performance?

Copilot was evaluated through a thorough red teaming process, testing for common risks such as violence, self-harm, sexual content, and hate and fairness. In addition, Copilot’s performance was measured using extensive grounded response testing. It tests for coverage of intended uses to assess accuracy and the extent to which responses are grounded in reference data.

What are the limitations of creating calculation models with Copilot? How can users minimize the impact of Copilot’s limitations when using the system?

Copilot's training might not cover all sustainability scenarios. Currently, it processes prompts exclusively in English and is focused on carbon emissions calculations. Users can only use Copilot for the initial model; refining through a conversational experience isn't supported.

To mitigate these limitations, users are encouraged to explore known sustainability scenarios, consider language compatibility, and primarily use Copilot for the initial model, with subsequent modifications made through the calculation editor.

What operational factors and settings allow for effective and responsible use of creating calculation models with Copilot?

Effective and responsible use of creating calculation models with Copilot requires system administrators to opt in and enable the feature in Microsoft Sustainability Manager settings. Prior to activation, administrators should review cross-geographical and AI disclosure considerations to ensure compliance within their organization.