Claude now shows bosses how staff use AI and what it costs

Anthropic has added a tool that lets companies look closely at how their employees use Claude, its AI chatbot, and how much that use costs. The company calls the feature "smart reports." It is available to businesses on Claude Enterprise, the version built for large organizations. Business Insider described it as something like an AI report card.

What the reports do

A manager picks a team, a time range of up to the last 28 days, and which Claude products to include. They can also add focus areas and their own questions. Claude then reads a sample of that team's chat transcripts. It groups the work into themes and types of output, attaches the spending to each group, and writes up what it found.

What gets flagged

The reports can also point out "inefficiencies." According to a Claude support page, that includes sessions that produced "no work value" or were "personal or off-topic." In plain terms, an employer could see that some of the money spent on AI went to chats that had nothing to do with the job.

Limits Anthropic describes

• Each company can run up to 10 reports a month at no charge.
• Anthropic says the reports should not be used to evaluate individual performance or make employment decisions.
• Chats on restricted topics, such as health, legal issues, union activity, race, ethnicity, or political and religious views, are used only in combined totals and get no summaries.
• Personal conversations do not include individual sessions or names.

Anthropic says the reports are meant to help companies understand adoption and plan their investment in Claude.

Why companies want this

Many businesses are trying to get a handle on what AI costs and whether it makes staff more productive. Some are also deciding how widely to let people use it. Bloomberg reported that Goldman Sachs recently limited access to Anthropic's Claude Code and Microsoft's GitHub coding tools to engineers only, after weighing productivity and risk concerns. The move reportedly annoyed some traders who had been using the tools.

What is not clear

The material does not say how large the transcript sample is or how Claude decides that a chat has "no work value." It also does not say how companies will use the findings, whatever Anthropic's guidance says. Workers on Claude Enterprise should assume that summaries of how their team uses the tool may reach their employer.