All product news
Product previewIn review

Meet Noa Agent: workspace questions in one conversation

A workspace-aware assistant for streamed hiring Q&A, private saved chats, and tightly scoped changes when Write access is explicitly enabled.

Noa Agent preview showing saved chats, a streamed interview-preparation answer, and Write access turned off
Noa Agent in a sample workspace. The conversation and workspace records shown are synthetic data.

Noa Agent brings a focused conversational workspace into the signed-in NoaMetrics app. It is designed to help hiring teams ask about the records in their selected workspace, follow an answer as it arrives, and move from reading to a small set of explicit actions without bypassing existing controls.

The current implementation is being reviewed in pull request #240. The experience shown here is a product preview, not a release announcement.

Ask with workspace context

Agent can use the active workspace context to answer supported questions about roles, candidates, interview scenarios, and available reports. Replies stream into the conversation, so the answer can be read while it is being assembled.

Chats are saved privately to the signed-in account and selected workspace. Starting a new chat gives each line of inquiry its own thread instead of mixing unrelated hiring work together.

Read first. Write deliberately.

Every new conversation begins in read-only mode. Turning on Write access is an explicit choice, and it applies only to the supported change requested in the conversation. Existing membership, validation, workflow, and quota rules continue to apply.

  • Read-only questions remain available without enabling Write access.
  • Supported writes are limited to defined actions, such as creating or updating specific hiring records; the switch does not grant unrestricted access.
  • If a write outcome cannot be confirmed, Agent stops instead of replaying the change automatically and asks the user to verify the workspace first.

What this preview does not claim

An open pull request shows the implementation under review; it does not prove production readiness. Deployment, live service configuration, and production acceptance remain separate steps.