Claude Enterprise smart reports: see AI spend by workstream

Most teams know what they pay for AI. Few know what they get for it. On October 28 Anthropic made smart reports, in-app surveys and pooled group budgets generally available for Claude Enterprise. The pitch: show managers AI spend by workstream, next to what that spend produced. That is the right question. Do not let the vendor be the only one who answers it.

What actually happened

Anthropic's announcement lists three features, all GA for Claude Enterprise. Availability "varies by configuration," including for HIPAA and customer-managed-key setups.

Smart reports. An admin picks a team, a date range and the Claude products to include. Claude reads a sample of that team's transcripts and groups the work into workstreams. Each workstream shows spend, deliverables and friction points, such as a missing connector or a failed tool call. The report ends with recommendations, for example workflows to turn into skills or rules to add to a repo. Admins with billing permission can raise a spend limit from inside the report. Anthropic says the reports are for adoption and planning, not for rating individual employees.

In-app surveys. Short cards appear after a reply. The default questions ask about the most valuable recent use, time saved, and what would break if access went away. Admins can edit the questions and export responses to a spreadsheet.

Pooled budgets. A group shares one monthly budget, and per-member limits still apply. A request counts against both, and a member stops at whichever limit runs out first. Billing admins get email alerts at 50%, 75%, 95% and 100%.

TechCrunch adds context. A Jellyfish analysis of its customers found that spend on Claude Code and similar tools grew 10x since January. Jellyfish's Nicholas Arcolano called it "growth that most CFOs didn't plan for."

Why it matters for your business

Spend by workstream beats spend by seat. A per-user bill tells you who is busy. A workstream view tells you whether "invoice reconciliation" or "support macros" earns its tokens. Use that view to decide where AI stays and where it goes.

Watch the button next to the number. The same report that shows friction also lets an admin raise the limit. That is convenient, and it is also the vendor's growth loop. Set the rule first: a limit goes up only when a workstream shows a deliverable you can count, like tickets closed, PRs merged or hours saved.

Survey answers are not ROI. "This saved me time" is a feeling. Pair the survey with one hard metric per workstream from your own systems, such as ticket volume, cycle time or error rate.

Keep a copy of the numbers. These reports live inside Claude. If you also run OpenAI, Gemini or open models, you need one ledger across vendors. Export what you can and log usage in your own warehouse.

Sources: Anthropic, TechCrunch.