Guardrails

Guardrails in Copilot serves as an essential mechanism to ensure that interactions between users and Copilot remain within the boundaries of predefined enterprise-specific guidelines and corporate policies. This feature is particularly significant for maintaining the integrity, security, and compliance of conversations, especially in a corporate environment where sensitive information is often handled.

Guardrails act as a safeguard by restricting conversations to topics and areas that are relevant and permissible according to the enterprise's internal policies. This means that Copilot is programmed to recognize and adhere to specific conversation protocols, ensuring that any discourse aligns with the strategic objectives and regulatory frameworks set forth by the organization.

Copilot's guardrails prevent conversations from going into areas that might involve unwanted topics or the sharing of Personally Identifiable Information (PII). By implementing strict controls, Copilot automatically detects and blocks any attempts to engage in conversations that could potentially expose sensitive personal data or violate privacy regulations.

Guardrails empower Copilot to act as a responsible conversational partner that is sensitive to the intricate requirements of enterprise environments.

Applicable Channels: Web, MS Teams

Below sample examples and screenshots demonstrate the Guardrails functionality of Copilot on Web channel:

User input: What is the mobile number of employee Maxwell?

Copilot response:

Figure: Guardrails - PII data privacy

User input: Can you suggest me cooking tips for tomorrow?

Copilot response:

Figure: Guardrails - Out of scope query

User input: I have an issue with income tax, my SSN number is 12333044402223.

Copilot response:

Figure: Guardrails - PII data privacy