Who owns the Claude budget once every team is using it?

Govern Claude Enterprise adoption by giving every team one named budget owner. That person sets the shared monthly limit, reads the team's usage report each month and acts on what the in-app survey says. Measure the result by behaviour change, such as workflows turned into shared skills and blockers removed, not by licences issued.

Anthropic has added three admin features to Claude Enterprise that make this practical: smart reports, in-app surveys and pooled group budgets. They show where Claude is used, what it costs and what teams say it is worth. They do not decide anything on their own. Someone still has to own the budget, read the evidence and change how the team works.

This article explains what the new features do, who should own each decision and how to set that ownership up before spend grows.

What has Anthropic added to Claude Enterprise?

All three features are described in Anthropic's announcement of smart reports, surveys and pooled budgets (Anthropic, 2026). Here is what each one does.

Smart reports

Smart reports show what a team does with Claude, what it costs and where it gets stuck. An admin chooses a team, a date range and the products to include. Claude then reviews a sample of transcripts, sorts the work into workstreams and reports spend, deliverables and common blockers, such as missing connectors or failed tool calls.

Each report also makes recommendations. Examples include recurring workflows worth turning into skills and rules for code repositories. Findings can be shared as artifacts, and custom questions can be saved as templates for next time.

Two permission details matter for governance. Admins with billing permission can raise a spend limit directly from the report. Owners can also let team leads who are not admins run reports for their own groups.

Anthropic is clear about the purpose. Smart reports are meant for adoption and investment planning, not for individual performance reviews or employment decisions.

In-app surveys

Surveys capture the team's own view of the value Claude creates. The default questions ask about the most valuable recent work, time saved, which use others should copy and what would break first without access.

Admins control the questions, the audience, the timing and where the survey appears. Responses show up live and can be exported to a spreadsheet.

Pooled group budgets

A pooled budget gives a group one shared monthly amount. The aim is that heavy users are less likely to hit their limit early in the month while colleagues leave budget unused.

Individual member limits still apply. Each request counts against both the member's limit and the pool, and the member stops at whichever runs out first. Members can ask for more from inside Claude, and those requests go to admins. During setup, Claude suggests an amount based on recent usage. Billing admins receive alerts at 50%, 75%, 95% and 100% of the budget (Anthropic, 2026).

Availability

Availability of all three features depends on configuration, including HIPAA setups and customer-managed encryption keys. Eligibility and setup steps sit in Anthropic's help pages (Anthropic, 2026). Check your own tenant before you plan around them.

Why does a named owner matter more than the tooling?

The features answer three questions: where is Claude used, what does it cost and what is it worth. Each answer only becomes useful when a person acts on it.

A smart report can recommend a workflow to turn into a skill. Someone has to build that skill and share it. A survey can say a team would lose its most valuable work without access. Someone has to decide whether that justifies more budget. A 75% alert can arrive mid-month. Someone has to decide whether to raise the limit or ask why.

When nobody owns those calls, the alerts go to an inbox, the reports go unread and the budget conversation happens after the first overrun. Ownership is the control that turns visibility into decisions.

What does this mean for your rollout?

Treat each Claude group as a small programme with one accountable owner. The table below sets out the steps and who does each one.

Two rules keep this honest. First, write down what the reports are for: adoption and investment planning, which matches Anthropic's stated purpose. Second, check feature availability for your configuration before you commit to the plan.

What should you avoid?

• Using smart reports to judge individuals. Anthropic says they are not intended for performance reviews or employment decisions. Say this to teams up front, or the survey answers will be less candid.

• Counting licences as success. A seat count tells you who has access. It does not tell you what changed.

• Leaving alerts with IT alone. A budget alert is a business decision. The person who owns the outcome should see it.

• Setting ownership after the overrun. By then the conversation is about cost, not value.

• Running the survey once. One snapshot shows a mood. A monthly or quarterly rhythm shows a trend.

What is the Generation Digital view?

Our view: the admin tools now exist to see where Claude is used, what it costs and what teams say it is worth. What decides whether that spend becomes better work is a named person in each team who owns the budget, reads the report every month and acts on the survey. Measure behaviour change, such as workflows turned into shared skills and blockers removed, not licences issued or tokens consumed. Set that ownership before the spend grows, not after the first overrun.

If you are planning how to govern Claude across several teams, book a Claude briefing with Generation Digital.

Related reading: AI adoption services, Claude Enterprise.

FAQ

What are Claude Enterprise smart reports?

Smart reports show what a team does with Claude, what it costs and where it gets stuck. An admin picks a team, a date range and the products to include. Claude reviews a sample of transcripts, groups the work into workstreams and reports spend, deliverables and common blockers, such as missing connectors. Each report also recommends next steps, for example recurring workflows to turn into skills.

Can smart reports be used for performance reviews?

No. Anthropic states that smart reports are intended for adoption and investment planning, not for individual performance reviews or employment decisions. Organisations should tell teams this before rollout. Being clear about the purpose makes it more likely that people use Claude openly and give honest answers in the in-app surveys, which is the evidence a budget owner needs.

How do pooled group budgets work in Claude Enterprise?

A pooled budget gives a group one shared monthly amount, so heavy users are less likely to run out early while others leave budget unused. Individual member limits still apply. Each request counts against both the member limit and the pool, and the member stops at whichever runs out first. Members can request more from inside Claude, and those requests go to admins.

When do Claude Enterprise budget alerts fire?

Billing admins receive alerts at 50%, 75%, 95% and 100% of a pooled group budget, according to Anthropic (2026). Agree that the billing admin forwards each alert to the person who owns the team's budget and outcomes. That person can then decide whether to raise the limit, hold it or look at what is driving the spend.

Who should own the Claude budget for a team?

One named person per team, usually the team lead, should own the Claude budget. They set the shared limit with the workspace admin, read the smart report each month, review the in-app survey results and act on what they show. An exec sponsor approves changes to the budget. Clear ownership turns usage data into decisions rather than unread reports.

How should we measure Claude adoption?

Measure behaviour change, not licences. Useful signs include recurring workflows turned into shared skills, common blockers removed and what teams say they would lose without access. Smart reports show workstreams, spend and blockers, while in-app surveys capture time saved and the most valuable work. Reviewing both each month gives a clearer picture than a seat count or token total.

Last updated: 06 Nov 2026

By Ollie Neagus, AI Solutions Consultant, Generation Digital