Claude Smart Reports Beta: What Enterprise Admins Actually Get

Coffee Summary

  • FACT: Smart reports (beta) launched for Claude Enterprise on 10 Sep 2026 — analyze team usage: work done, cost, friction, and patterns worth packaging as shared skills.
  • FACT: During beta, each org can run up to 10 reports per month free; the limit resets on the first calendar day of each month.
  • FACT: Unavailable for orgs using CMEK, HIPAA configurations, or Access Transparency; also unavailable for Claude Code on Enterprise with zero data retention.
  • FACT: Org-level "Allow attribution to individual users" is off by default; unveiling names/session IDs is logged when enabled.
  • OPINION (AIImpish): Treat Smart Reports as an adoption/ROI tool — Anthropic says not for individual performance or employment decisions.

What happened

Claude Help Center docs describe Smart reports (beta) for Claude Enterprise: Claude samples transcripts in a scoped window, clusters workstreams and deliverable types, attaches spend, and writes a drillable report admins can download as HTML.

An Owner or Primary Owner enables Organization settings → Capabilities → Analytics → "Smart reports (beta)" (off by default). Then Primary Owners, Owners, Admins, and custom roles with analytics view can create and view reports.

Why it matters

Enterprise rollouts stall on "what are we getting for this spend?" Smart Reports answer with workstream-level cost, friction categories (missing connectors, approval gating, tool failures, rework), and suggested reusable skills — not just seat-level tokens. For enablement leads, that becomes a checklist: scope a team, fix top friction, package a repeated workflow, bring evidence to renewals.

What changed

Report contents (FACT)

Creators choose: team/scope, time range up to the last 28 days, products (chat, Claude Code, Claude Cowork), focus areas, optional files, share list, template/custom questions, then Create. Generation takes a few hours; email notifies when ready.

Sections include workstreams (sessions and spend), deliverables produced, cost per session by output type, task outcomes, common frictions, inefficiencies (unusable or personal/off-topic — aggregate count/cost only), reusable skills with suggested prompts, most expensive sessions, complex autonomous work, and custom-question answers. Charts are interactive; download HTML keeps drilldowns.

Privacy and attribution (FACT)

Personal conversations appear only in aggregate. Org toggle "Allow attribution to individual users" defaults off (drilldowns show "User", hide session IDs). When on, viewers can reveal names/IDs; each unveil is logged. Per-share attributed view only appears if the org toggle is on.

Enablement checklist (FACT)

1. Confirm Claude Enterprise and org is not on CMEK, HIPAA, Access Transparency, or Claude Code ZDR. 2. Owner/Primary Owner enables Smart reports under Capabilities → Analytics. 3. Decide attribution policy before flipping the individual-attribution toggle. 4. Scope via RBAC Groups or SCIM department / costCenter (urn:ietf:params:scim:schemas:extension:enterprise:2.0:User:department and …:costCenter). 5. Create a named report (team, ≤28-day range, products, questions) → Create. 6. Budget the beta 10 reports/month; request more via account team or in-product after the cap. 7. Archive → Delete to retire reports; do not use outputs for HR actions.

Who should care

Enterprise Owners/Admins owning ROI narratives; enablement teams packaging skills; security/privacy reviewers on attribution and CMEK/HIPAA/ZDR exclusions; IdP admins who must pass department/cost center for useful filters.

Limitations

Beta quotas can change. CMEK/HIPAA/Access Transparency/ZDR exclusions block many regulated orgs. Reports sample transcripts — directional, not a full ledger. Docs forbid use for individual performance or employment decisions. Team filters need RBAC groups or SCIM fields.

What to do next

1. Pilot one function (e.g., sales or eng) over 14–28 days with attribution off. 2. Fix the top friction category before buying more seats. 3. Turn one "reusable skills" card into an org skill. 4. Push SCIM department/cost center with human-readable names if needed. 5. Write a rule: adoption planning only — never performance management.

AIImpish Take

Smart Reports are Anthropic's Enterprise answer to "show the work, the waste, and the skill opportunities." The win is real if you respect exclusions, the 10/month beta cap, and attribution defaults. The failure mode is treating transcript sampling like an HR scorecard. Leave attribution off until legal signs off, and use the first report to kill friction, not to name names.