Anthropic is giving the companies that buy Claude a clearer view of what their token bills pay for. On October 28 it said every Claude Enterprise customer can now use three new administration tools: smart reports, pooled group budgets and in-app surveys. TechCrunch, which reported the launch the same day, described the package as an effort to show large customers a return on their spending, a question many of them are still trying to answer.
Smart reports are the core of the release. An administrator, or a team lead given permission by an owner, picks a group, a period of up to 28 days and the products to cover, which can include chat, Claude Code and Claude Cowork. Claude then reads a sample of that team's conversations, sorts them into workstreams and types of output, attaches spend to each, and lists recurring sticking points such as a connector nobody enabled or a tool call that keeps failing. Each report ends with recommendations, for example repeated workflows worth packaging as skills, and admins can add custom questions and save them as templates. Someone with billing rights can raise a spend limit directly from the report.
Anthropic sets limits on how the reports can be used. Its help pages say they are meant for understanding adoption and planning investment, not for judging individual performance or making employment decisions. Sessions that Claude flags as touching restricted topics, including health, union activity, protected characteristics, political or religious views and personal life, count only toward aggregate figures and get no summary, and admins cannot write custom questions about those subjects. The feature is off by default and is unavailable to organizations using customer-managed encryption keys, a HIPAA configuration or hybrid deployments; where Claude Code runs with zero data retention, reports leave it out. TechCrunch adds that Anthropic says the data will not be used for model training by default.
The surveys collect the employee side of the story. By default they ask what the most valuable recent use of Claude was, how it changed the time a task took, which use a colleague should copy, and what would break first if access disappeared. Admins can rewrite or drop questions, choose who sees them and when, and export answers to a spreadsheet. Respondents get a brief prompt card inside whichever tool they are working in, shown once Claude has answered.
Pooled budgets address a more mundane problem. Instead of every member receiving an identical monthly cap, a group can share one pot, so heavy users are less likely to be cut off early while colleagues leave allowance unused. Individual limits remain as a ceiling, every request counts against both, and anyone who runs out can ask admins for more from inside Claude. Billing admins are notified when the pool reaches 50%, 75%, 95% and its full amount. According to the help center, the option covers usage-based Enterprise plans and seat plans with usage credits, and it cannot yet be managed through the Admin API.
The timing follows Anthropic's IPO filing. TechCrunch notes that figures shared with investors this week pointed to a second-quarter operating profit and net revenue retention of 900%. It also cites Jellyfish, a developer analytics firm, which says tools built on Anthropic models account for nearly a third of code changes proposed by developers in its data and that spending on agentic coding tools overall has grown tenfold since January.
There are open questions. Anthropic's own pooled-budget help page still labels the feature beta even though the announcement calls all three tools generally available. The value of smart reports also depends on how representative the transcript sample is, and the customer endorsements on Anthropic's page are selected by the company.