Claude Enterprise adds in-app surveys and pooled group budgets, and smart reports leave beta

On October 28, 2026 Anthropic released three admin features for Claude Enterprise, all generally available:

1. Smart reports leave beta. Claude reads a sample of a team's transcripts and reports what the work was, what it cost, and where it got stuck.
2. In-app surveys ask your people, inside Claude, what it did for their work.
3. Pooled group budgets give a group one shared monthly spend limit instead of only per-person limits.

Together they cover the three things a rollout owner gets asked for: what people are doing with it, whether they think it helps, and how to keep the bill predictable without blocking the heaviest users. Availability can vary for organizations with HIPAA configurations or customer-managed encryption keys, so check the Help Center if you have either.

Smart reports, now GA

We covered smart reports when they entered beta on September 10 in smart reports: what your team actually uses Claude for. What changed at GA:

• Admins can raise a spend limit from inside a report. If the report shows someone hitting a limit that is blocking real work, an admin with billing permission fixes it from that screen.
• Saved templates. Custom questions and insights can be saved and reused, so a monthly report asks the same questions every month and you can compare them.
• Delegation to team leads. Owners can let team leads who are not admins run reports for their own groups.
• Findings can be shared as artifacts, so a team lead can send one finding to their team without forwarding the whole report.

Anthropic states that smart reports are not for evaluating individual performance or making employment decisions. If you delegate reports to team leads, tell them that in writing before they run the first one.

In-app surveys

A survey shows up as a short card after one of Claude's replies, in whichever Claude product the person is already using. There is no email and no separate form.

The default questions ask about value:

• What was the most valuable thing you used Claude for recently?
• How did Claude change the time that work took?
• Which of your uses should a teammate copy?
• What would slow down first if your team lost access?

Admins can edit or remove questions, choose who gets the survey, set when it runs, and choose where it appears. Responses arrive on the survey page live and export to a spreadsheet. Survey data is treated as customer data, and in organizations with a single retention period, responses are deleted on that schedule.

Getting answers you can use

Surveys and transcripts measure different things. A smart report tells you what people did. A survey tells you what they think it was worth. Use both, and use them for different decisions.

• Keep the "lost access" question. "What would slow down first" names the workflow people depend on. That is a better input for an ROI case than a satisfaction score, because you can go and measure that workflow in the system it runs in. Measuring the ROI of Claude covers how.
• Keep "which use should a teammate copy." Those answers are a ready-made list of skills to build and share.
• Survey one group at a time. Run it for a single department for two weeks, read the answers, then widen it. A company-wide survey on day one produces a pile of answers no one has time to read.
• Run it after the first month, not the first week. In week one, people describe what they tried. After a month, they describe what they kept.
• Pair it with the smart report for the same group and dates. Where the survey says a workflow saves hours and the report shows the same workstream with high friction, that is the place to fix a connector or write a skill.

Pooled group budgets

Per-person spend limits have one known problem: the heaviest users run out in the second week, while light users leave most of their allowance unspent. A pooled budget gives a group, such as a department, one monthly amount that everyone in it draws from.

How the two limits interact:

• Individual limits still apply. Every request counts against both the person's limit and the group budget, and a person stops at whichever runs out first.
• When someone hits a limit, they can request more from inside Claude. The request goes to admins.
• Admins with billing permission get email alerts at 50%, 75%, 95%, and 100% of the budget.
• When you create a pool, the setup flow suggests an amount based on the group's recent usage.

Setting the numbers

1. Start from the suggested amount, then add 10 to 20%. The suggestion reflects past use. If you are still rolling out to the group, use will grow.
2. Raise individual limits when you add a pool. The pool is there so heavy users can draw on what others leave unused. If individual limits stay where they were, the pool changes nothing. Set the individual limit as a ceiling on unusual spend, for example two to three times the group's median monthly use, rather than as everyone's allowance.
3. Decide who answers the 75% alert before it fires. Name the admin, and decide whether the answer is to top up or to slow down. An alert nobody owns is an alert nobody reads.
4. Read the requests for more. A cluster of requests from one team is a demand signal. Look at that team in the next smart report before deciding whether to raise their pool.

Pools sit alongside the tools in Claude Enterprise cost controls, which covers the analytics views and spend alerts that shipped in July. For programmatic cost reporting, the Usage and Cost Analytics API that shipped the next day reports cost by person and by group.

A first-week checklist

1. Run one smart report for your most active group and save its custom questions as a template.
2. Create a pooled budget for that same group, with individual limits raised to a ceiling rather than an allowance.
3. Name the person who owns budget alerts.
4. Schedule a two-week survey for that group, starting a month after their rollout.
5. Put the smart report, survey export, and budget usage for that group side by side at the end of the month.

The point of the comparison is a short list: the workflows people depend on, what each one costs, and where they get stuck. ROI reporting for agent operators covers how to turn that list into something leadership will read.