Anthropic's Claude Smart Reports give employers a new way to analyse how staff use artificial intelligence at work, including identifying sessions classified as personal or off-topic.
Announced on 28 October 2026, the feature allows Claude Enterprise organisations to examine workplace AI activity, assess the cost of different tasks and identify recurring problems, although Anthropic says the reports should not be used to evaluate individual performance or make employment decisions.
The feature addresses a growing challenge for businesses investing in AI. Companies want to know whether employees are using these tools effectively, what they are producing and whether the expense is justified.
Anthropic's approach offers that visibility, but it also raises questions about how much insight employers should have into workers' conversations with an AI assistant.
Claude Smart Reports Analyse Workplace AI Use
Smart Reports allow authorised users in Claude Enterprise organisations to select a team, choose a reporting period of up to 28 days and specify which Claude products to analyse. These include Claude chat, Claude Code, Claude Cowork and the combined chat and Cowork experience.
Claude analyses a sample of conversation transcripts, grouping sessions into workstreams and categories of output. Reports can cover activities such as producing documentation, writing code and conducting analysis, alongside estimates of the associated spending.
🛡️ Security Teams Can Now Monitor Claude Chats, Files and AI Agent Activity | Source: https://t.co/Aei24jjiLq
— Cyber Security News (@The_Cyber_News) October 1, 2026
The Claude Compliance API gives security teams visibility into employee and AI-agent activity, integrating Claude data with existing monitoring, DLP, identity,… pic.twitter.com/Xm1hLYeE6W
Managers can use the reports to review work, track costs and spot recurring problems, such as failed tools or repeated revisions. The system can also suggest workflow improvements and let authorised administrators raise spending limits when needed.
Smart Reports became generally available to Claude Enterprise customers on 28 October, although some configurations are excluded. Organisations can generate up to 10 reports a month at no extra cost, then wait for the monthly reset or contact their account team for more.
Why Claude Smart Reports Flag Personal Conversations
Anthropic's 'inefficiencies' category counts sessions that produce no usable output or are classified as personal or off-topic, along with their associated costs.
However, conversations involving restricted topics are included only in aggregate reports, without individual summaries or session-level access.
Individual attribution is disabled by default. If enabled, names and session identifiers remain hidden until a viewer reveals them, an action Anthropic logs. The system also restricts questions about sensitive topics, including health, religion, union activity and employees' personal lives.
Anthropic says these safeguards cannot be switched off and that Smart Reports should not be used to assess individual performance or make employment decisions. 'Smart reports help you understand adoption and plan your investment in Claude,' the company says in its documentation.
However, reports still show how teams use AI, what it costs and which sessions are deemed unproductive. A failed task may reflect technical problems or an unsuccessful experiment rather than wasted time, making transparency and careful interpretation important.
Employees may also want clear explanations of how their workplace uses these reports, who can access the information and how long the data is retained. Clear policies could help workers understand the purpose of monitoring and reassure them that unsuccessful tasks will not automatically be treated as evidence of poor performance.
