Buying seats is the easy part of rolling out an AI assistant. Figuring out whether anyone is getting real value, and what to copy from the people who are, is where most enterprise programs stall. Anthropic announced three features for Claude Enterprise aimed squarely at that problem: smart reports, in-app surveys, and pooled group budgets. All three are generally available now.
This is admin tooling, not a model upgrade. But it matters, because it’s the layer that decides whether a company renews.
Smart reports: from usage data to a to-do list
Admins pick a team, a date range, and the products to include. Claude then analyzes a sample of that team’s transcripts and groups the work into workstreams. For each one you see spend, deliverables produced, and common friction, such as a missing connector or a failed tool call.
The useful bit is the output. Each report ends with one consolidated list of recommendations: recurring workflows worth turning into skills, and rules to add to a code repository. Findings can be shared as artifacts. Admins can ask custom questions about model usage and cost, save them as templates, and raise a person’s spend limit directly from the report if billing permissions allow.
Owners can also hand report access to team leads who aren’t admins. Anthropic is explicit that reports are for understanding adoption and planning investment, not for evaluating individual performance or making employment decisions. That caveat is worth taking seriously, since transcript analysis invites misuse.
Customer quotes come from DoorDash, Docusign, and others. One says each report draws on tens of thousands of sessions. Treat the testimonials as marketing, but the DoorDash line is telling: even someone close to the work says reports surfaced things they didn’t know.
In-app surveys add the human view
Transcripts show what people did. Surveys show what they think it was worth. The default questions ask for the most valuable recent use, how Claude changed the time involved, which use a teammate should copy, and what would break first if access vanished tomorrow. That last one is a sharp way to measure dependence.
Admins can edit or remove questions and control who sees a survey, when, and where. Members get a short card after a reply in the tool they’re already using. Responses show up live and export to a spreadsheet.
Surveys now work for organizations with customer-managed encryption keys, and they follow a single retention period if one is set. Anthropic says it accesses survey data only as needed to run the service.
Pooled budgets fix a boring, real problem
Per-person spend caps are blunt. Heavy users hit the wall early in the month while colleagues leave budget untouched. Pooled group budgets give a team one shared monthly amount. Individual limits stay as a guardrail, and a member stops at whichever limit runs out first.
People who hit a limit can request more from inside Claude, and admins review the request. Setup suggests an amount from recent usage, and billing admins get email alerts at 50%, 75%, 95%, and 100%.
Does it matter?
Yes, moderately. Competitors such as ChatGPT Enterprise and Microsoft Copilot already offer usage dashboards and analytics. What Anthropic adds here is interpretation: clustering work into workstreams and recommending skills to build. That ties neatly into its push on Claude Code and skills, and it gives admins a reason to keep spending.
Availability varies by configuration, including HIPAA setups and customer-managed keys, so check the Help Center before planning a rollout.
