What happened
On October 28, 2026, Anthropic said smart reports, in-app surveys, and pooled group budgets are generally available for Claude Enterprise. Business Insider and Futurism covered the feature in the days after, mostly from the angle of what it means for employees. According to Anthropic's Help Center, an admin picks a team, a date range of up to 28 days, and the Claude products to include. Claude then reads a sample of the transcripts in that scope, groups them into workstreams and types of output, attaches spend to each group, and writes a report. The report covers things like deliverables produced, cost per session by type of output, common pain points (a missing connector, output that did not match the request, rework loops), and the most expensive sessions. A few details matter. Smart reports are available on Enterprise plans and are off by default. An owner has to turn them on. Anthropic's documentation says each organization can run up to 10 reports a month at no charge. By default, drilldowns show "User" instead of a name. An owner can separately turn on an option that lets a viewer reveal names and session IDs, and Anthropic says each reveal is logged. Anthropic also says the reports "aren't designed and should not be used for evaluating individual performance or making employment decisions." It lists sessions about topics such as personal conversations as restricted, so they are used only for aggregate analysis, with no summary. These are Anthropic's descriptions of its own product. I have not tested it.
Why it matters for business owners
Most owners who pay for AI cannot answer two plain questions. What is my team actually using it for? And what does each kind of work cost? Most owners cannot answer those from the invoice. A bill shows usage totals. It does not show that one team spends most of its budget on first drafts nobody uses, or that another team is quietly saving hours on a task you never knew about. That is the useful side of this feature, and it applies beyond Claude. Any business paying for AI seats or usage will eventually want a view of work, cost, and friction. Anthropic's own list of use cases includes finding an integration nobody enabled and turning repeated work into a shared routine. The other side is just as real. A tool that summarizes what staff typed into an AI assistant is also a monitoring tool, whatever the label says. Futurism and Business Insider both framed it that way, and the framing is fair even if Anthropic's intent is adoption and planning.
What owners should not misunderstand
Do not assume this is a productivity score. The report estimates cost at list price and groups sampled sessions. It does not know whether a session saved an hour, closed a deal, or avoided a mistake. A cheap session is not a good session, and an expensive one is not a wasteful one. Do not assume it reads everything. Anthropic says Claude reads a sample of transcripts. A sample can miss things, and a team of a few people can produce a thin or easy-to-guess picture. Do not assume the privacy guardrails are a substitute for your own policy. They depend on Claude correctly classifying which sessions are sensitive, and I did not find a published error rate. Anthropic's guardrails are a vendor setting. Your commitments to your staff are yours. Do not assume this is only for companies using Claude. Other AI vendors have admin dashboards, and the same questions apply to them. If you use a different tool, the useful move is to ask what its admin view shows and what it can expose. And do not assume small businesses need an Enterprise plan to ask these questions. Smart reports are an Enterprise feature. The questions about work, cost, and trust apply on any plan.
The operational lesson
AI spend becomes manageable when you can tie it to a kind of work. Total spend tells you almost nothing. Spend per workflow tells you where to look. But how you collect that view decides whether people keep using the tool honestly. If staff learn that managers can read summaries of what they asked an AI assistant, some will stop using it for the useful, messy, exploratory work, or move to a personal account. That second outcome is worse for you: the work now happens outside any control you have. So the goal is a measurement approach people can accept. Measure at the team and workflow level. Say in advance what you will and will not use it for. Keep it away from performance reviews. That is a management decision, and a tool setting cannot make it for you.
What a serious business should do next
Before turning on any AI usage reporting, write down what question you want answered. Good ones are: which workflows use most of our AI budget, where do people get stuck, and which repeated tasks should become a standard process. Decide who can see the data and at what level. Team-level and workflow-level views are enough for most budget decisions. Think hard before enabling anything that reveals names. Tell employees first, in plain words. What is collected, who can see it, what it will be used for, and what it will not be used for. If you cannot say that out loud to your team, do not switch it on. Treat the numbers as questions, not verdicts. If a team's spend looks high, ask what they are doing before concluding anything. Check the legal side for your location and industry with your own advisor. Rules on monitoring employees vary, and this is not legal advice. If you do not use an enterprise plan, you can still get most of the value manually. Ask each team to list their three most common AI tasks, how often they happen, and what goes wrong. It is slower, but it also builds the trust that makes the numbers honest.
The Atlacis view
Atlacis helps owners slow down, understand the workflow, map the risk, and choose the right tool or system before spending. Usage data is only useful when it answers a real business question and your people trust how it is gathered. If you are paying for AI and cannot say what it is buying, a short conversation can help you map the workflows, the costs, and the ground rules before you add more tools or more reporting.
The short version
- Anthropic says smart reports are generally available for Claude Enterprise as of October 28, 2026. They are off by default and need an owner to enable them.
- Per Anthropic, reports sample a team's transcripts, group them into workstreams, and attach list-price cost to each kind of work.
- The useful part for owners is spend per workflow and where people get stuck. Totals on an invoice cannot show that.
- Anthropic says the reports should not be used for evaluating individual performance or making employment decisions. That rule is yours to set and keep.
- Names and session IDs are hidden by default, but an owner can allow a viewer to reveal them. Each reveal is logged, per Anthropic.
- Decide the question, the viewing level, and what you will tell employees before you measure anything.
Where ATLACIS can help
Sources
- Anthropic Help Center: Get started with smart reports
- Anthropic: Smart reports, surveys, and pooled group budgets for Claude Enterprise (October 28, 2026)
- Business Insider: Claude's new AI report card tool flags 'inefficiencies' (November 2, 2026)
- Futurism: Anthropic debuts tool for your boss to spy on your Claude chats (November 3, 2026)