Anthropic has introduced Smart reports in beta for Claude Enterprise, adding a reporting layer designed to help organisations understand how their teams use Claude. According to the September 10 release-note entry, Smart reports analyse the work getting done, what it costs, where sessions run into friction, and which repeated patterns may be worth packaging as shared skills. The feature is available on Claude Enterprise plans.
The addition is a practical shift from simply providing access to an AI assistant toward helping administrators and programme owners assess whether that access is producing useful work. In a large deployment, activity alone is a poor measure of value. A growing number of conversations does not show whether people are completing work, encountering avoidable obstacles, or repeatedly recreating the same successful workflow. Smart reports are aimed at making those questions easier to investigate from a team-level view.
From usage figures to operational insight
Enterprise AI programmes need a way to connect adoption with practical outcomes. The release note says Smart reports analyse how a team uses Claude and report on the work getting done. That wording suggests an emphasis on the shape of activity rather than a simple count of active users. For managers, the useful question is often not whether staff opened an assistant, but whether it helped move analysis, writing, research, coding or operational tasks forward.
The cost element is equally important. AI spend can be difficult to govern when it is spread across many users, projects and workflows. A reporting feature that places costs beside observed patterns of work can give teams a better starting point for prioritising investment. It may help them identify high-value uses that deserve broader enablement, as well as activity that needs clearer guidance or a different process. Anthropic has not detailed individual report fields, export options or cost-allocation controls in the release note, so organisations should confirm those implementation details in their own Enterprise environment.
Finding friction before it becomes a habit
Smart reports also call out session friction. This is one of the more useful parts of the beta announcement because failed or awkward sessions are often invisible in aggregate usage reporting. A team may begin a task in Claude, abandon it after an unclear result, or repeatedly spend time supplying context that could have been organised once and reused. If reporting can make those patterns visible, AI leaders can turn individual frustration into a specific improvement opportunity.
The right response will vary. Some friction may be resolved through better source material, clearer access to approved tools, or a more reliable prompt pattern. Other cases may reveal that a workflow requires review steps, data controls or specialist expertise that should remain outside an assistant. Reporting should not be treated as an automatic performance score for people. Its value is in helping a team improve its working system while maintaining appropriate human judgment, privacy and governance.
Shared skills as a scaling mechanism
The release note specifically says Smart reports can highlight repeated patterns worth packaging as shared skills. This connects analytics to a tangible next step. When several people independently discover a useful way to structure a task, a shared skill can preserve that approach for the wider team. Instead of relying on informal prompt copying, teams can document the context, preferred steps, controls and expected output that make a workflow dependable.
That can reduce variation in recurring work. For example, a research group might identify a common pattern for summarising customer feedback, while a technical team might standardise how it asks Claude to explain an incident timeline. A shared skill does not remove the need to check outputs, but it can make successful methods easier to repeat and improve. Organisations should still review any shared workflow for data handling, permissions and ownership before encouraging broad use.
What Enterprise teams should do next
Because Smart reports are in beta, Claude Enterprise administrators should treat the feature as an opportunity to establish a measured feedback loop. Start by identifying a small set of questions that matter: which teams are getting sustained value, where users encounter recurring blockers, which activities consume significant resources, and which proven patterns are candidates for a shared skill. Comparing these findings with direct user feedback will help avoid drawing strong conclusions from metrics alone.
It is also sensible to set expectations about what reporting is for. A constructive rollout focuses on improving workflows and support, not monitoring individual employees. Teams should explain which data is available to administrators, how it will be used and how people can raise concerns. That approach is more likely to produce honest feedback and useful adoption signals.
Anthropic's Smart reports beta gives Claude Enterprise customers a new way to observe and refine AI-enabled work. The release is relatively focused, but it addresses a recurring operational need: converting scattered assistant sessions into evidence about cost, friction and repeatable practice. As the beta develops, the most valuable updates will be those that help organisations turn those observations into safer, clearer and more effective day-to-day workflows.