Claude Enterprise can now analyze employee chats in smart reports

Anthropic has made smart reports generally available for Claude Enterprise. The feature lets an administrator select a team, a period of up to 28 days and products such as chat, Claude Code or Claude Cowork. Claude then reads a sample of transcripts and groups activity into workstreams, output types, costs, pain points and sessions that produced no usable result.

The reports can identify recurring workflows that might become shared skills, recommend repository rules and answer custom questions about usage and spending. Administrators can drill into session groups, and billing staff can raise a user’s limit from the report. Anthropic also launched in-app employee surveys and pooled team budgets alongside the reporting feature.

The privacy boundary matters because the system is analyzing workplace conversations. Smart reports are off by default, and individual attribution is disabled unless the organization enables it. Names remain hidden until a viewer actively reveals them, and each reveal is logged. Anthropic says the reports are for adoption and investment planning, not employee evaluation or employment decisions.

Permanent restrictions prevent custom analysis of protected characteristics, health, union activity, harassment and whistleblower complaints, political or religious views, personal life and sentiment about named people. Sessions flagged for those topics contribute only aggregate data without summaries. Organizations using HIPAA configurations, hybrid deployments or customer-managed encryption keys cannot use smart reports, while zero-retention Claude Code sessions are excluded. The controls reduce risk, but employees should still understand that work chats may feed management analytics.