As companies spend more on AI coding tools, a basic question becomes harder to avoid: what is that spending producing? Anthropic’s new reporting suite is designed to help enterprise customers answer it by connecting team activity, output and cost.
Reports connect usage with results
Anthropic’s reporting tools let managers review a team’s work over a chosen date range. They can see what the team produced, how much it cost and where workflows broke down, using internal analytics and employee surveys.
The reports are also intended to recommend ways to improve results. Anthropic says it will not use the data for model training by default. For managers, the combination of cost, work output and workflow problems could make it easier to identify where AI tools are helping and where teams may need to adjust how they use them.
Anthropic is also introducing pooled group budgets so teams can manage token spending. Together, the budgets and reports address two sides of the same challenge: keeping track of usage and understanding whether that usage is delivering value.
Rising costs sharpen the question
Companies are still working out how best to deploy AI. The source article describes token spending as particularly strong among venture-backed firms, while most Fortune 500 companies are not seeing major gains from AI deployments. That gap makes it important for businesses to understand both what they are paying for and what their teams are getting from it.
A report from software development tools company Jellyfish, based on use of agentic coding tools by 276,000 of its customers, offers one sign of the pressure. Claude Code and other tools built on Anthropic models were responsible for nearly a third of proposed code changes submitted for review. Spending on those tools had grown 10x since the beginning of the year, and their cost was about 2% of total engineering cost.
Nicholas Arcolano, Jellyfish’s chief scientist, told TechCrunch that this growth was not something most CFOs had planned for. His point was that it is no longer enough to spend on AI tools; companies also need to optimize how they use them. Anthropic’s reports are aimed at giving managers information to do that.
Enterprise adoption builds staying power
Anthropic’s challenge is not only to show that its tools work, but to make their value visible to the businesses using them. The article says Anthropic leads among enterprise customers, while also noting that it can be difficult to identify what makes an agentic coding tool stick.
Jellyfish’s AI product lead, Nikolas Albarran, told TechCrunch that Codex became more visible in Jellyfish’s data with Astra, OpenAI’s latest model. Yet he described Anthropic’s enterprise relationships and the effort customers had already put into standardizing workflows as reasons they might not switch. Once a company has built processes around a tool, a competing model alone may not be enough to prompt a change.
That investment can include learning how a tool handles particular workflows and adapting team practices around it. Albarran said switching has a cost, even when tools are mostly similar, because each has its own quirks. For companies, the decision is therefore about more than comparing model capabilities: it also involves the effort required to change established ways of working.
A lighter form of customer support
AI labs are investing in forward-deployed engineering teams, which place AI consultants inside customer organizations to help them get started. Anthropic’s reports can be understood as a lighter version of that support: instead of relying only on hands-on help, customers get information and recommendations they can use to improve their own workflows.
The reports do not, by themselves, establish that an AI tool is paying off. They give managers a way to examine spending, output and breakdowns together. As AI costs grow, that visibility may help businesses decide where to keep investing, where to change their processes and whether the results justify the expense.