Anthropic Shows Companies Where Their Claude Money Goes

Anthropic, the company behind the Claude models, has released new tools for enterprise customers. The tools show managers where their Claude spending actually goes. There is analytics broken down by team, employee surveys about results, and a shared token budget. That last part is a single spending cap for the whole team, instead of a cap per person. Anthropic says this data is not used to train its models by default.

The company's own numbers explain the move. Spending by the same customers grew ninefold over the past year. Spending on coding agents grew tenfold. These are AI tools that write and fix code on their own. They now account for about 2% of all engineering spending. A separate survey by Jellyfish, a company that makes analytics tools for software development, found that Claude Code and other Anthropic-based tools are behind almost a third of all code changes submitted for review. Even at that scale of use, spending is still only 2% of engineering budgets. That is less than you might guess from the growth numbers alone.