Claude Enterprise now shows what each team's spend buys
If you run Claude for a company, you can now see spend per workstream and pool budgets per team. Here is what is on by default, and what is not.
Anthropic has made three admin features generally available for Claude Enterprise: smart reports, in-app surveys and pooled group budgets [1]. Smart reports read a sample of a team’s Claude transcripts and tell an admin what that team worked on, what it cost at list price, where sessions ran into trouble and which repeated tasks are worth turning into shared skills [1][2]. Pooled budgets let a team share one monthly budget instead of each person running into a separate limit [1]. For a company paying for Claude, the reports aim at the question a finance team asks every quarter: what is this spend buying?
Claude Enterprise admins can now run reports on what each team does with Claude and what it costs, and give each team one shared monthly budget.
for you
If you run Claude for a team, turn smart reports on for one functional group first and leave individual attribution off unless you have a clear reason.
What was announced
All three features are generally available for Claude Enterprise [1]. Anthropic frames the release around spreading work that already succeeds in one team to the rest of a company [1].
Smart reports. An admin picks a team, a date range and the products to include, and Claude analyses a sample of that team’s transcripts [1]. It groups the work into workstreams and shows the spend for each one, the deliverables produced and the most common points of friction, such as a missing connector or a failed tool call [1]. The Help Center lists the sections each report contains: workstreams by sessions and by spend, deliverables produced, cost per session by type of output, task outcomes, the most common pain points, inefficiencies, complex autonomous work, work new to the organisation, people automating their own work, the most expensive sessions, answers to custom questions, custom insights and a single list of recommendations [2]. Those recommendations include recurring workflows to turn into skills and rules to add to a code repository [1][2].
A report can cover a time range of up to the last 28 days, and the products it can include are chat, Claude Code, Claude Cowork and the unified chat and Cowork experience [2]. Reports usually take a few hours to generate and can take longer at busy times; you get an email when one is ready [2]. Each organisation can run up to 10 reports a month at no charge [2]. Queued and in-progress reports count toward that limit and failed ones do not, and the limit resets on the first day of the next calendar month [2]. If you need more, the Help Center says to contact your account team [2]. The price of extra reports is not public.
Admins can save their questions and custom insights as templates and reuse them [1]. If a report shows that a spend limit is holding someone back, an admin with billing permission can raise that limit from inside the report [1]. Owners can also let team leads who are not admins run reports, but only for the groups, departments or cost centres assigned to them [1][2].
In-app surveys. These ask members about the value of their work with Claude. The default questions cover the most valuable thing someone has done with Claude recently, how Claude changed the time that work took, which use of Claude a teammate should copy, and what would break or slow down first if the team lost access tomorrow [1]. Admins can change or remove questions and choose who gets a survey, when and where; members see it as a short card after a reply in the tool they are using [1]. Responses can be exported to a spreadsheet [1]. Anthropic says survey responses are treated as the customer’s data and that it accesses them only as needed to run the service [1].
Pooled group budgets. A group gets one shared monthly budget, so heavy users are less likely to hit an individual limit early in the month while others leave theirs unused [1]. Individual limits still apply. Every request counts against both the pool and the member’s own limit, and a member stops at whichever runs out first [1]. A member who hits a limit can ask for more from inside Claude, and the request goes to admins [1]. Setup suggests an amount based on the group’s recent usage, and admins with billing permissions get email alerts at 50%, 75%, 95% and 100% of a budget [1].
What changed, and for whom
The useful change is that cost now sits next to the work. A smart report attaches spend to each workstream and each type of output, and lists the most expensive sessions [2]. That turns “we spend this much on Claude” into “this team spends most of it on this kind of task”.
The limits matter as much as the features. Smart reports are off by default; an Owner or Primary Owner has to turn them on in the organisation’s settings [2]. Organisations that use customer-managed encryption keys, a HIPAA configuration, or a hybrid deployment that runs inference on their own cloud provider cannot use smart reports at all [2]. If an organisation uses zero data retention for Claude Code, reports cover chat and Claude Cowork and leave Claude Code out [2]. So the companies with the strictest data rules are the ones that cannot use the feature yet.
Privacy is handled by default settings, not by trust. Member names are hidden unless an Owner turns on a separate setting that allows attribution, which is also off by default [2]. Even with it on, drilldowns show “User” until the viewer chooses to reveal names, and every reveal is logged [2]. Two guardrails are always on: admins cannot ask custom questions about restricted topics, and sessions Claude identifies as touching restricted topics are used only in aggregate, with no summary [2]. The restricted list covers protected characteristics, health, union activity, harassment or whistleblower reports, political or religious views, personal life and feelings about a named person [2]. Anthropic says smart reports are not meant for evaluating individual performance or making employment decisions [1][2]. According to TechCrunch, Anthropic says it will not use the report data for model training by default [3].
There is a commercial reason behind all this. TechCrunch reads the release as Anthropic’s attempt to show customers the return on their spending, a worry for many large buyers [3]. It describes the reports as a lighter form of the forward-deployed engineering teams that AI labs place inside customer companies [3].
Turn smart reports on for one functional team, such as sales or engineering, and read the cost-by-workstream view before you renew seats or raise limits. The Help Center says team-level reports are more useful than whole-company ones.
See Claude's fact panel →Ask your vendor whether it can show spend by type of work, not only by seat or by user. If it cannot, that is a gap to raise at renewal.
Compare the three →Pooled budgets plus member limits give you two controls instead of one. Decide the pool from the suggested amount, then keep member limits as a guard against one runaway user.
Read the spend-governance guide →Who it is for — and not
This is for admins and finance owners at companies on Claude Enterprise who already have several teams using Claude and need to decide where to spend more. If you are on Claude Pro, Max or Team, the announcement does not cover your plan; it lists all three features for Claude Enterprise [1]. If your organisation runs Claude with customer-managed keys, a HIPAA configuration or a hybrid deployment, smart reports are not available to you [2], though in-app surveys can now be enabled with customer-managed keys [1].
It is also not a tool for judging people. The settings and the Help Center steer reports toward teams, not individuals, and say so in plain terms [2]. If a manager wants to use a report to rank employees, the vendor’s own guidance says not to [1][2]. The person who turns on individual attribution should be able to explain why it is needed, because every name reveal leaves a log entry [2].
This page will be updated if Anthropic publishes a price for reports beyond the free monthly allowance, extends smart reports to organisations with customer-managed keys or HIPAA configurations, or brings any of the three features to plans other than Enterprise.
∴ Claude Enterprise admins can now see spend per workstream; turn attribution on only if you have a reason to name people.
