---
type: analysis
slug: claude-can-write-the-sql-but-teams-still-own-the-metric
title: "Claude can write the SQL, but teams still own the metric"
description: "Dashboards moves Claude into connected analytics: it can query business data and expose the SQL and refresh time behind a chart. That makes exploration easier to inspect, not automatically safe to trust as routine reporting."
published_at: 2026-10-09T11:00:00.000Z
topics: ["claude"]
source_urls: ["https://www.testingcatalog.com/claude-launches-dashboards-and-motion-in-beta","https://testingcatalog.com/claude-launches-dashboards-and-motion-in-beta","https://claude.com/resources/articles/dashboards-and-motion"]
author: "Andrey Reshetnikov"
author_profile: https://newruntime.com/owner-profile.md
---
# Claude can write the SQL, but teams still own the metric

Dashboards moves Claude into connected analytics: it can query business data and expose the SQL and refresh time behind a chart. That makes exploration easier to inspect, not automatically safe to trust as routine reporting.

## Summary

Anthropic has launched Claude Dashboards in beta on paid plans, connecting to sources including Redshift, BigQuery, ClickHouse, Databricks and Snowflake. Users can ask questions in natural language and receive dashboards with chart refresh times and underlying queries exposed. Claude Motion, which produces editable code-based animation with MP4 export, is in beta for Team and Enterprise. The report establishes availability and features, not accuracy or production reliability.

## What happened

The important shift is from self-contained artifacts to a workflow that reaches into connected data. According to the supplied report, Claude retrieves answers from supported data systems and builds dashboards; viewers can inspect the SQL behind charts and see when they were refreshed. Query visibility helps a reviewer investigate how a result was produced, but it does not establish that the chosen tables, joins, filters or business definitions were right. Motion is a separate artifact: code-authored animation rather than video-model output, with editable code and MP4 export. Enterprise administrators must enable Dashboards and Motion, which are off by default.

## What the archive adds

According to the dossier, Anthropic’s documentation describes Dashboards as generating and running SQL against connected sources (d2). Anthropic’s earlier March 2026 launch of interactive charts and persistent artifacts provides product context for this extension (d3). The dossier also reports an internal analytics workflow finding: accuracy was 21% without curated skills and a human-owned semantic layer, versus above 95% with them. That is Anthropic’s internal-workflow finding, not a benchmark of Claude Dashboards; it nonetheless illustrates how much definitions and context can matter when asking questions of business data.

## Trend signal

AI products are moving from drafting standalone content toward answering questions over live organizational context. That raises the bar: the output is not just a chart but an implicit claim about what a metric means and which records count. Exposing SQL and freshness gives analysts concrete inspection points, while leaving data modeling and organizational definitions in human hands. The likely near-term role is therefore faster exploration atop existing analytics practice, rather than an automatic replacement for governed reporting.

## Why now

According to the dossier, the move extends Anthropic’s earlier work on interactive charts and persistent artifacts (d3). Connected sources and generated SQL make conversational exploration a practical next step beyond static visuals (d2). The primary report’s plan segmentation and administrator controls show that Anthropic is widening access while leaving enterprise enablement to organizations.

## Connections to other events

The useful boundary is between seeing a query and validating an answer. A readable SQL statement can still join the wrong tables, omit a relevant population or encode a definition that differs from the company’s canonical metric. Refresh timestamps answer when data was refreshed, not whether the underlying result is correct or authorized for every viewer. Our read is that Dashboards can shorten the path from a question to a testable analysis, provided someone who knows the data checks the query and result before it becomes a decision input.

## Why it matters

Builders should prepare a small set of canonical datasets and definitions before inviting broad use of natural-language analytics. Buyers can pilot with questions whose correct answers are already known, then compare the generated SQL and output with the trusted source. That makes failures legible and helps identify whether the blocker is a bad query, an ambiguous metric or a permissions issue. The tiered rollout adds a deployment decision: enterprise administrators need to choose whether to enable these beta features and for whom.

## What happens next

- **Anthropic will publish additional guidance or product documentation clarifying how Dashboards handles query review, data permissions or metric definitions.** (6 months, 65%). Watch for: A Dashboards documentation or release update describes one or more of these controls or validation practices.
- **At least one publicly described enterprise pilot will position Dashboards as exploratory analysis alongside existing reporting, rather than as the authoritative reporting layer.** (12 months, 60%). Watch for: A customer account or case study explicitly describes analyst validation or comparison with an existing trusted report.

## Limits and uncertainty

The primary report covers product features and availability; it does not establish accuracy, refresh behavior under real workloads, adoption or customer outcomes. The dossier’s 21% versus above-95% internal-workflow finding is attributed to Anthropic and is not a Dashboards evaluation. The evidence also does not settle how complex joins, conflicting metric definitions or permission edge cases behave. This take may be too cautious if pilots show reliable results with little review, or too optimistic if connecting and modeling enterprise data proves cumbersome.

## Skill to build now

Practice validating generated SQL. This week, take one business question with a known answer, ask Claude Dashboards to answer it, then inspect the tables, joins, filters and metric definition against the canonical report; record what differed and why.

## Reader takeaway

This week, choose one recurring question with a trusted answer and test it in Claude Dashboards. Check that the generated SQL uses the intended source, joins, filters and metric definition, then compare its result with the trusted answer and note any discrepancy. Treat the exercise as a query-validation pilot, not approval to replace routine reporting.

## Open question

Will visible SQL and refresh times be enough for analysts to trust conversational dashboards, or will organizations need a governed semantic layer before these outputs can shape routine decisions?

## Sources

- [ANTHROPIC 🔥 : Claude Dashboards is now available on all paid plans; Claude Motion is now available on Team and Enterprise plans.](https://www.testingcatalog.com/claude-launches-dashboards-and-motion-in-beta)

- [testingcatalog.com](https://testingcatalog.com/claude-launches-dashboards-and-motion-in-beta)

- [Build live dashboards and animate explainers with Claude | Claude by Anthropic](https://claude.com/resources/articles/dashboards-and-motion)

- [Get started with Claude Dashboards](https://support.claude.com/en/articles/17454700-get-started-with-claude-dashboards)

- [Claude now creates interactive charts, diagrams and visualizations](https://claude.com/blog/claude-builds-visuals)

- [How Anthropic enables self-service data analytics with Claude](https://claude.com/resources/articles/how-anthropic-enables-self-service-data-analytics-with-claude?trk=public_post_comment-text)

- [Conversational Analytics in Looker overview](https://docs.cloud.google.com/looker/docs/conversational-analytics-overview?hl=en)

- [Analytics APIs](https://platform.claude.com/docs/en/manage-claude/analytics-api)

- [Jamf Claude Enterprise case study](https://claude.com/customers/jamf)

Author: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json