Anthropic Launches Claude Smart Reports to Track AI Costs at Work
Anthropic's new Smart Reports feature shows enterprise customers how teams use Claude and what it costs, but it also raises questions about employee monitoring.

Anthropic has introduced "Smart Reports", a Claude feature that lets authorized managers see how their teams use the AI assistant and what that usage costs. In theory, it also lets them keep an eye on individual employees. The company's own help page says the tool was not designed to evaluate individual performance or to inform employment decisions.
Companies are surprised by AI bills
The launch comes as many businesses struggle with the price of AI. An AWS study found that only 15 percent of German companies use AI extensively enough to create new products or processes, down from 21 percent a year earlier. In a Bitkom survey, a third of German firms said the high cost of their AI use had caught them off guard. Apollo Global Management chief economist Torsten Slok has argued that companies outside the tech sector are not yet seeing measurable benefits and increasingly treat AI as a cost that squeezes margins.
Some firms respond with "modelmaxxing": rather than defaulting to the most expensive frontier model, they assign different models to different tasks. First "model routing" startups already sell platforms that match each task to the most cost-efficient model. Others simply track token consumption.
What Smart Reports show
According to Anthropic, the reports reveal how a team uses Claude and which tasks it handles. They also show costs at list price, where friction occurs and which recurring patterns could be bundled more efficiently.
The feature is free for Enterprise customers but switched off by default. Each organization can generate up to ten reports per month. Only authorized managers have access, and they define the scope, such as a time period or a team. Claude then reads a sample of transcripts in that scope, groups them by workflow and output type, assigns costs to each group and summarizes the findings. Sessions Claude classifies as confidential feed only into the aggregate analysis and carry no personal details.
Anthropic stresses that the insights are not meant for assessing individuals or teams, yet they inevitably say something about efficiency. Media reports suggest enterprise customers account for roughly 80 percent of Anthropic's revenue, so features aimed at cost control could matter ahead of its planned IPO.
Apps worth a look
Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.
Talk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.
Open EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.
Pick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.
More from this day
See all →
CrowdStrike to Buy Fiddler AI and Dreadnode for $495 Million

Reddit Plans to Require Login Even for Reading Posts

GitHub Copilot Fusion launches, picking the best AI model per task




