Claude can now show what a company's AI spending is used for

Anthropic has made the Smart Reports tool generally available for Claude Enterprise, the business edition of the AI assistant Claude. This happened on October 28, 2026, according to Anthropic's own announcement. According to search results, the feature has been in beta since September 10.

The idea is to give companies an overview of what they get out of their AI spending, which according to Channel Insider has grown quickly. The outlet writes that Claude usage has increased tenfold since the beginning of 2026 and now makes up about 2% of total engineering costs. The figure comes from an analysis by Jellyfish.

How the reports work

The user selects a team, a time period and the Claude products to include. Claude then analyzes a sample of the team's conversations and groups the work into workflows, meaning types of tasks that recur.

The report shows usage per workflow, what has been delivered, and which problems keep recurring. Examples include missing integrations and failed tool calls. It also offers recommendations, such as turning recurring workflows into reusable skills. Users can also ask their own questions about model usage and costs and save them as templates.

Wix's Ran Yitzhaki says, according to Anthropic's announcement, that each report is based on tens of thousands of sessions and shows the most common problems.

Administrators with billing permissions can adjust spending limits directly from the reports. Team leads without admin permissions can run reports for their own groups.

Privacy and limitations

Anthropic writes that the reports are not meant for assessing individual performance or making hiring decisions. They are meant for understanding adoption and planning investments.

The "Allow attribution to individual users" setting is turned off by default. In that case, member names and session IDs are hidden. If it is turned on, viewers can see names and session IDs, and the action is logged. Sessions about sensitive topics such as health, harassment and political views are included only in aggregated analysis, without summaries.

The reports are based on samples, not all sessions, and typically take a few hours to generate. Each organization can run up to 10 reports a month at no charge. Organizations with customer-managed encryption keys, HIPAA configuration or hybrid deployment with customer-managed inference cannot use the feature.

Surveys and shared budgets

Two other features have also become generally available. In-app surveys are short questionnaires that appear as cards in Claude after a response. The default questions cover the most valuable outcome, time savings, recommendations to colleagues and the consequence of losing access.

Pooled group budgets give a group one shared monthly budget while the individual limits remain in place. Each request counts against both limits, and the member is stopped at whichever limit is hit first. Administrators receive email alerts at 50, 75, 95 and 100% of the budget.