Claude Enterprise: 3 New Features to Keep Your AI Budget in Check
機械翻訳 / Machine-translated
A
AI Friends
3日前・15分で読める
機械翻訳 / Machine-translated
@aifriends
AI Friends（https://aifriends.jp）のクロスポスト公式アカウント。AIツールの紹介・使い方・できることを、中学生でもわかるやさしい日本語で届けます。
Have you ever wondered, "We've adopted AI, but are we really getting our money's worth?" On October 28, 2026, Anthropic added three new features to its enterprise product, Claude Enterprise, designed to help organizations measure cost-effectiveness. This article explains what each feature does and what Japanese companies should keep in mind.
Anthropic announced on its official blog that Smart reports, in-app surveys, and pooled group budgets are now generally available (officially released for everyone to use). The news was also covered domestically by ITmedia Enterprise on October 31.
All three features share a common purpose: making it easier for administrators to understand how much is being spent on AI and what they're getting in return.
Smart reports is a feature that analyzes what your team is doing with Claude. An administrator selects a team, a time period, and the product in question, and Claude reads a portion of the conversation history and categorizes it by type of work.
Each categorized group displays the cost incurred, the outputs produced, and common points of friction—such as missing integrations or failed tool calls.
At the end of the report, a consolidated set of improvement suggestions is provided. For example, it might recommend turning a repeatedly performed task into a "skill" (a saved sequence of steps).
Results can be shared as a deliverable, and frequently used queries can be saved as templates. Administrators with billing permissions can also raise usage limits directly from the report screen.
According to the help page, the reporting window covers the past 28 days. Reports take several hours to generate, and you'll receive an email when they're ready.
Up to 10 reports per organization per month can be generated for free. The feature is off by default and must be enabled by an owner in the organization settings.
Privacy considerations also apply. Anthropic asks that the feature not be used for individual performance evaluations or employment decisions. Topics such as health and labor union activity are included only in aggregate summaries, and individual summaries are not produced for these areas.
Cost figures alone can't tell you whether something is actually useful. That's where in-app surveys come in.
After a user interacts with Claude, a short card appears. The default questions cover things like the most helpful thing Claude did in a recent task, time saved, ways of using Claude that a colleague should try, and what would be missed first if Claude were no longer available.
Administrators can edit or delete questions and control who receives them, when they appear, and where. Responses appear in the admin dashboard in the order they're received and can be exported to a spreadsheet.
Previously, the standard approach was to manage usage through per-person limits. This meant heavy users could hit their cap mid-month while light users ended the month with budget to spare.
With pooled group budgets, a single monthly budget is assigned to the group and shared among all members. Individual limits remain in place, so extreme overuse by any one person is still prevented.
Usage counts against both the group limit and the individual limit. Whichever is exhausted first is what stops further usage.
When the budget reaches 50%, 75%, 95%, and 100% of consumption, an email notification is sent to administrators with billing permissions. Users can request additional budget from within Claude, and administrators review those requests. The settings screen suggests an amount based on the group's recent usage.
Claude Enterprise also added administrator-focused features back in July 2026. According to the official blog, the July update introduced per-group and per-person usage and cost views, per-model usage restrictions, and notifications at 75% and 90% of the organization's limit. The July update was focused on showing how much was being spent in granular detail. The three new features go a step further, letting you see what the spending went toward and whether it delivered value.
Enterprise plans from other providers, such as ChatGPT Enterprise, are noted to also offer usage dashboards. However, within the scope of what could be confirmed, a comparable offering that includes conversation categorization and surveys was not found.
Consider the IT department of a mid-sized Japanese company. Costs vary by department, but when internal stakeholders ask whether Claude is paying off, there's no easy answer. Smart reports can help by laying out each department's usage breakdown and costs side by side, giving you concrete material for those conversations.
If the sales department consistently has budget left over at month's end while the development team hits its cap by mid-month, switching to pooled group budgets by department can help even out that imbalance.
There are also things to confirm before rolling out these features. Smart reports is not available for organizations using customer-managed encryption keys, HIPAA configurations, or hybrid setups where inference runs on the company's own cloud. If zero data retention is enabled, Claude Code is excluded. Pricing is not listed on the official page, so you'll need to contact your sales representative.
The official blog does not mention pricing for these features. Smart reports are free for up to 10 per month per organization. For anything beyond that, you'll need to consult your account representative.
Anthropic asks that the feature not be used for individual performance evaluations or employment decisions. Its intended purpose is to understand adoption and inform investment planning.
Usage for the group stops. Users can request additional budget from within Claude, and administrators review those requests.
No. Smart reports is off by default and must be enabled by an owner. Some organizations may not be eligible depending on their configuration.
Start by reflecting on how your organization currently explains the costs and outcomes of its Claude usage—that's a good place to identify where these features can help most.
This article is a cross-post from AI Friends.
