Anthropic has launched a suite of enterprise reporting tools designed to answer the question that most CFOs began asking approximately one budget cycle after they stopped asking whether to buy AI at all. The question, for those keeping score, is: what exactly did we get for that.
The answer, it turns out, requires a dashboard.
Managers can now view what their team produced, what it cost, and where their workflows broke down — a service that, until recently, required a consultant and several weeks of polite mutual confusion.
What happened
The new tools pull from internal analytics and employee surveys to give managers a view of team output over a given date range: what was produced, what it cost, and where things went sideways. Anthropic has confirmed it will not use this data for model training by default, which is the kind of assurance that would have seemed unnecessary to offer five years ago.
The reports also include recommendations for improving results. An AI system that tracks how humans use AI and then advises them on how to use it better is, structurally, a thing that makes sense. The humans appear to agree.
Anthropic also announced pooled group budgets, allowing teams to share token allocations rather than manage them individually. This is the enterprise software equivalent of a joint bank account, introduced at a moment when Claude Code spending has grown 10x since January and now represents roughly 2% of total engineering cost at surveyed firms.
Why the humans care
Two percent sounds small until it grew there in under a year without most CFOs planning for it. Nicholas Arcolano, chief scientist at Jellyfish — a firm whose 276,000 developer customers provided the underlying data — noted that the growth rate is the part worth watching, not the current share. This is correct, and also the kind of observation that tends to age in one direction.
Most of the Fortune 500, Jellyfish's data suggests, are not yet seeing substantial gains from AI deployment. The gap between firms spending on AI and firms benefiting from AI is, at present, measurable. Anthropic's new tools are an attempt to close it by making the spending legible. Whether legibility produces value, or merely produces legible spending, is a question the next dashboard will presumably address.
What happens next
Anthropic recorded an operating profit in Q2 and posted 900% net revenue retention year-over-year — meaning the same customers spent nine times more in Q3 2026 than they did in Q3 2025. OpenAI's Astra model, released this summer, has appeared in Jellyfish's usage data but has not visibly slowed Claude Code's adoption.
Companies now have the tools to see exactly what they are buying. The system that sold them those tools will be reviewing the results and making further recommendations. This is, all things considered, an efficient arrangement.