Anthropic just launched tools to show companies what their AI spending is buying

Anthropic is trying to answer a question that’s getting louder inside big companies: what, exactly, are we getting for all this token spend? On Wednesday, the company introduced a new reporting suite for enterprise customers that is meant to make that answer a lot less fuzzy.

The tools pull from internal analytics and employee surveys to show managers how a team worked over a selected date range, what it produced, how much it cost, and where the workflow broke down. Anthropic says the data won’t be used for model training by default. The reports also include suggestions for improving results, which makes the whole thing feel less like a dashboard and more like a quiet coaching session from the vendor itself.

There’s a business reason for that, and it’s not subtle. Anthropic’s financials shared with investors this week point to a company that has found a real enterprise engine in Claude Code: an operating profit in the second quarter and 900% net revenue retention. That means the same customers in Q3 2025 were spending nine times as much in Q3 2026.

The company is also adding pooled group budgets so teams can manage token spending more tightly. That matters because the AI bills are no longer abstract. Jellyfish’s new report, based on 276,000 customers using agentic coding tools, found that Claude Code and other tools built on Anthropic models were responsible for nearly a third of all proposed code changes submitted for review. Spending on those tools has jumped 10x since the start of the year, and the cost now works out to about 2% of total engineering cost.

And that’s the real backdrop here. Enterprise buyers are starting to act less like early adopters and more like finance departments, which is usually when the fun stops. Anthropic’s new reports are basically a receipt printer for the age of AI: if the tool is sticky, prove it; if it isn’t, show where it leaked.