Anthropic Gives Managers a Window Into Claude Use

Anthropic has introduced a tool that lets Claude Enterprise customers analyze sampled employee AI-session transcripts, putting managers in a position to track work categories, outputs and costs. The reports can also flag sessions Anthropic labels “inefficiencies,” including work it classifies as producing “no work value” or as “personal or off-topic.” Each organization can run up to 10 reports a month at no charge.

A manager’s view of the work

Managers can select a team, a period of up to 28 days, Claude products, focus areas and custom questions. The tool samples transcripts, groups sessions into workstreams and types of output, assigns spending to each group and summarizes its findings. That gives an organization a way to inspect how employees use Claude—and to attach costs and judgments to categories of work.

Anthropic says the reports are meant to help customers understand adoption and plan investments. Its support documentation also says they “aren't designed and should not be used for evaluating individual performance or making employment decisions.” The disclaimer sits beside a tool that analyzes employee sessions and labels some activity as inefficient, leaving the company’s stated purpose and the system’s workplace uses in the same frame.

The documentation says sessions involving restricted topics—including personal conversations about race, ethnicity, health, legal issues, union activity, and political or religious views—are used only for aggregate analysis and don’t include summaries. Personal conversations also don’t include individual sessions or names, according to the support page.

Names stay hidden—until they don’t

An Anthropic spokesperson told Business Insider on Monday that reports don’t show employee names by default. But an organization’s owner can enable names alongside summaries, and report viewers can choose to reveal them; each reveal is logged. Sessions identified as personal or off-topic won’t appear with names.

The spokesperson said: “A report groups a team's work with Claude into workstreams and shows what each one cost. Opening a workstream shows short session summaries, not transcripts.” The spokesperson also said an owner can turn on the name setting and each reveal is logged. So the tool’s design includes both limits on what appears and a route for an owner to make names available.

The launch arrives as companies weigh AI costs, productivity claims and how widely employees may use the tools. The report describes management’s ability to sort and review AI use; it does not describe any employee-led response, direct action or mutual-aid effort. Nor does it identify a legislative or electoral route for workers to shape these workplace systems.

The corporate appetite for control

Bloomberg reported Monday that Goldman Sachs recently restricted Anthropic’s Claude Code and Microsoft’s GitHub coding tools to engineers after evaluating productivity and risk concerns. The move reportedly frustrated some Goldman traders who had used the tools. Here, access to workplace technology is set from above, while the traders affected by the restriction appear in the report as frustrated users, not decision-makers.

Separately, Palantir reported a profit of $1.13 billion, or 44 cents a share, in its most recent quarter, compared with $475.6 million, or 18 cents a share, a year earlier. The company makes bespoke data-management and analysis software for businesses and governments. Palantir said revenue surged as businesses increasingly turned away from artificial-intelligence labs and sought greater control over their use of the technology. The report gives no further breakdown of that business mix. It does show the commercial stakes: one company sells tools for organizational control, while another gives employers a way to categorize employee AI activity and its cost.