What changed
Administrators can map Agent Copilot context attributes to knowledge applicability metadata. During an interaction, Copilot resolves those values and Knowledge Fabric filters out ineligible content before retrieval and answer generation. Separately, agentic virtual agent knowledge tools can now generate answers directly from retrieved knowledge chunks with APT-2, instead of having Venba generate an answer and APT-2 reformulate it. This applies to supported Knowledge Fabric and chunk-indexed Knowledge V2 sources.
Vendor documentation (opens in a new tab)Why it matters
Copilot suggestions are only useful if they fit the customer's product, region or plan, and context filtering gives knowledge teams a way to enforce that. The direct-answer option removes a generation step from virtual agents, which should make them respond faster. Both changes depend on knowledge content having the right metadata and indexing.
CxCave practical context
Who should care
- Architects
- Administrators
- CX leadership
Implementation considerations
REVIEWAdd applicability metadata to knowledge articles for the contexts that matter, such as product or region, map it to Copilot context attributes, and test direct answer generation on one virtual agent for latency and answer quality before rolling it out widely.
CxCave recommendation, not a vendor mandate unless explicitly stated.