Anthropic 向 Claude Enterprise 推出智能報告、問卷及羣組共用預算

Anthropic made three Claude Enterprise features generally available on October 28, 2026: smart reports that analyze what teams accomplish with Claude and what that work costs, in-app surveys that gather employee feedback, and pooled group budgets that let administrators set one shared monthly spending pool per team.

Smart Reports: Workstreams, Costs, and Privacy Guardrails

An administrator starts a smart report by choosing a team, a date range covering up to the last 28 days, and the products to include (chat, Claude Code, Claude Cowork, or the combined chat and Cowork experience). According to the Claude Help Center’s setup guide, Claude then reads a sample of transcripts in that scope, groups them into workstreams and output types, attaches spend at list price to each group, and writes up the findings. Reports usually take a few hours to generate and can be downloaded as HTML with drilldowns intact.

Each report shows what the group used Claude for most, ranked by sessions and by spend; the deliverables produced, grouped by output type such as analysis, documentation, content drafts, and code; the average cost per session for each kind of output; task outcomes; and where the most expensive sessions concentrate. A pain-points section catalogs what got in the way, including connectors that were never set up, output that missed the ask, approval or sign-in gating, tool failures, and rework loops, while an inefficiencies section counts sessions that produced nothing usable, shown only as an aggregate count and cost.

A consolidated recommendations list identifies recurring workflows worth turning into shared skills and rules worth adding to a code repository, and any finding can be shared as an artifact. Reports can also answer custom questions about model usage and cost, and administrators can save questions and custom insights as reusable templates. Administrators with billing permission can raise a user’s spend limit directly from the report, and owners can let team leads who are not admins run reports for their own groups. Reports can be scoped to groups that carry role-based permissions, including groups synced from an identity provider through SCIM, or filtered by department and cost center attributes.

Each organization can run up to 10 reports per month at no charge, and the feature is off by default until a Primary Owner or Owner enables it in organization settings. Anthropic states that smart reports are meant for understanding adoption and planning investment in Claude, and that they are not designed for evaluating individual performance or making employment decisions.

Two privacy guardrails stay on permanently: custom questions cannot touch restricted topics, and sessions Claude flags as involving those topics feed only aggregate analysis without session summaries. The restricted list covers protected characteristics, health and medical matters, union activity, harassment, whistleblower and legal complaints, political or religious views, personal life, and sentiment about a named person. Attribution to individual users is off by default; when an organization enables it, viewers still see “User” in place of names until they actively reveal them, and each reveal is logged. Organizations using customer-managed encryption keys, a HIPAA configuration, or a hybrid deployment cannot use smart reports at all, and organizations with zero data retention for Claude Code get reports covering chat and Cowork only.

Anthropic’s announcement includes statements from Monday, Slalom, Wix, Deloitte, DoorDash, and Docusign. “Each smart report is based on tens of thousands of sessions, surfacing the most common pain points and the recurring workflows worth turning into skills,” said Ran Yitzhaki, a software engineer at Wix. Docusign said the reports show how its teams use Claude, from coding Docusign IAM, its AI-native contracting platform, to everyday work, and that the visibility helps it demonstrate real outcomes to leadership and fix connector bottlenecks. Vivek Kulkarni, Deloitte’s managing director for AI transformation, said early insights are helping inform decisions about deployment and governance. DoorDash AI enablement lead John Jannone said smart reports surfaced things he did not know even though he works close to how his teams use Claude.

In-App Surveys: Default Questions and Data Handling

Surveys are available to Enterprise plans, and Primary Owners, Owners, Admins, and custom roles with analytics view access can create them. The Claude Help Center’s survey guide lists four default questions: the most valuable thing the user has done with Claude recently and how they could tell the outcome was better than it would have been without Claude; what, if anything, Claude changed about the time that work took; one way of using Claude that a teammate should copy; and what would break or slow down first if the team lost access to Claude. Admins can add or remove questions, target all members or specific groups, set the survey’s start and end dates, and choose whether the card appears in chat, Cowork, or Claude Code.

Users see the survey as a small card at a natural break — after a reply in chat, between tasks in Cowork, or after Claude finishes responding in Claude Code — and a survey never interrupts a running task. The card carries the organization’s name, states that answers go to the organization’s admins rather than Anthropic, takes about 90 seconds to complete, and appears to each user only once. Results aggregate on the survey page while the survey runs, with response counts and rates, and each survey exports as a CSV containing the submission time, the user’s email, their group, the surface where they responded, and one column per question.

Survey responses are treated as the organization’s data, which Anthropic says it accesses only as necessary to operate the service. Surveys support customer-managed encryption keys and follow the organization’s retention setting, with surveys and their responses deleted on that schedule. They are not available to organizations with a HIPAA configuration, and Anthropic’s documentation instructs regulated customers not to place regulated data such as PHI into survey responses.

Pooled Group Budgets: Shared Spend, Alerts, and Limits

A pooled budget adds one shared monthly amount for a group on top of each member’s own monthly spend limit. Every request counts against both, and a member stops at whichever runs out first. The design, as Anthropic describes it, keeps people who use Claude heavily from hitting an individual limit early in the month while colleagues leave budget unspent, with the member limit remaining as a guardrail against unusually high individual use. When a member hits a limit, a “Request more” button appears inside Claude and the request routes to admins for review. Users never see the group’s name or the pooled amount.

Anthropic’s pooled-budget documentation recommends sizing the first amounts from the previous month’s actual spend rather than an estimate: a pooled budget of about three times the group’s monthly spend and a member limit of about three times the top spender’s usage. Its worked example puts a group that spent $1,450 with a $310 top spender at a $4,350 pooled budget and a member limit of roughly $950. Everyone holding the billing permission receives email alerts at 50%, 75%, 95%, and 100% of a budget, and the usage bar turns amber at 75% and red at 90%. The documentation suggests considering a 50% increase if a budget is 75% used before the 20th of the month.

Pooled budgets are available on usage-based Enterprise plans and on Standard or Premium seat plans with usage credits enabled, where they apply only to usage credits and seat-included usage does not count against the pool. When users belong to several groups, the largest pooled budget pays first by default, with an optional custom priority order. Budgets can be frozen at $0 or paused without losing the amount. Management happens only in Organization settings, the billing permission applies organization-wide, and pooled budgets are not yet available through the Admin API.

Anthropic notes that availability for the three features varies by configuration, including for organizations with a HIPAA configuration or customer-managed encryption keys, and points administrators to the Claude Help Center for eligibility and detailed setup steps.