Claude Enterprise finally gives founders a receipt for their AI spend
Anthropic just launched smart reports and pooled budgets to show exactly what your team is building with Claude and what it costs.
Business · Source: Claude
What happened
Anthropic made three new features generally available for Claude Enterprise on October 28, 2026. The update includes smart reports, in-app surveys, and pooled group budgets. The goal is to show companies exactly what their token spend actually buys.
Smart reports sample a team's transcripts over the last 28 days and group the work into specific streams like code, analysis, or documentation. They show the deliverables produced and attach a list-price cost to each group. The reports also highlight friction points, like missing connectors, approval gating, or failed tool calls. Organizations can run up to 10 of these reports per month at no charge. Anthropic explicitly notes these reports are for investment planning, not evaluating individual employee performance.
To protect privacy, custom questions cannot touch restricted topics like health, union activity, or personal life, and user attribution is off by default. To manage costs, admins can now set a shared monthly budget for a whole team. The system sends alerts when spending hits 50, 75, 95, and 100 percent of the pool. Meanwhile, in-app surveys pop up naturally during work to ask employees how much time Claude saved them and what would break if they lost access.
Key facts
- October 28, 2026 — General availability date for the new Enterprise features
- 28 days — Maximum date range a smart report can cover
- 10 — Maximum number of free smart reports an organization can run per month
- 50%, 75%, 95%, 100% — Alert thresholds for pooled group budgets
Why it matters
Founders and enterprise buyers are tired of paying massive AI bills without knowing the return on investment. This update changes the dynamic entirely. You no longer have to guess if your engineering team is using Claude for complex architecture or just writing emails. You get hard data on task outcomes, average cost per session, and where the most expensive sessions concentrate. It turns an opaque software expense into a clear line item with measurable output.
The second-order effect is a shift from isolated individual experimentation to systemic company workflows. Because the reports identify recurring pain points and successful patterns, leaders can codify what works. If one engineer finds a brilliant way to use Claude for debugging, the system flags it so the whole organization can adopt it. Conversely, if a team wastes money on rework loops because a tool keeps failing, leadership sees the inefficiency immediately and can fix the broken integration.
For builders
Find and fix connector bottlenecks
Smart reports explicitly flag when missing connectors or failed tool calls waste time and money. Engineering leads can use this data to prioritize internal tool integrations and stop paying for rework loops.
Turn rogue prompts into shared skills
The system identifies recurring workflows that actually produce results. You can take these proven prompts and turn them into shared skills or repository rules, scaling your best employee tactics to the whole team.
Control team spend without blocking work
Pooled budgets let you set a hard cap for a department while keeping individual limits flexible. Users can request limit increases directly in the app when they hit a wall on a high-value task.
My take
I talk to founders every week who are terrified of their blind AI spend. Anthropic is smart to build tools that prove Claude's return on investment directly inside the product. When you give the person paying the bill a clear receipt showing time saved and code shipped, they will never churn.
Original reporting: Claude. This is my rewrite and opinion.