PlugWiki

AI for data analysts: Claude and ChatGPT for SQL, charts and dashboards

Claude and ChatGPT can write SQL, explore datasets and build charts, if you validate the numbers before you share them.

Claude has an Anthropic-built data analyst plugin with commands to explore data, write SQL, make charts, build dashboards and validate an analysis before you share it. ChatGPT Business covers spreadsheet analysis, SQL and pipeline code, and apps for warehouses and BI tools. Both work best when connected to your warehouse. Without a connection, Claude writes the SQL and you run it, then paste the results back.

Good for

  • Analysts who write SQL against a warehouse every day
  • Business analysts working in spreadsheets and BI dashboards
  • Analytics engineers maintaining dbt models and pipelines
  • Product and growth analysts answering stakeholder questions

Claude

Claude's data plugin adds commands for analysis, SQL, charts, dashboards and validation.

Plan: Pro ($17 a month billed annually, $20 monthly, as of November 2026) adds Claude Dashboards (beta). Team Standard seat ($20 per seat annually, $25 monthly) for a shared analytics team.

What it can do

  • Answer data questions end to end with /analyze, from a quick lookup to a full analysis
  • Profile a dataset for shape, nulls and odd values with /explore-data
  • Write optimized SQL for your dialect with /write-query
  • Make publication-quality charts in Python with /create-viz, or interactive HTML dashboards with /build-dashboard
  • Check an analysis for method, accuracy and bias before you share it with /validate
  • Connect Snowflake, Databricks, BigQuery, Definite, Hex and Amplitude
  • Create and edit files with code execution, on every plan including Free

Limits

  • Without a warehouse connection, Claude cannot run your queries. It writes SQL for you to run, and you paste the results back or upload CSV or Excel files.
  • Claude Dashboards is in beta and not on the Free plan.
  • Anthropic lists model training as opt-out on individual plans. Team plans do not train on your content by default.

Plugs to turn on

ChatGPT

ChatGPT Business covers spreadsheet analysis, SQL and pipeline code, and warehouse apps.

Plan: ChatGPT Business for team data. Standard seat: $20 per user per month billed annually or $25 monthly, from 2 seats, as of November 2026 (Premium seats are $100 annual or $125 monthly).

What it can do

  • Turn stakeholder questions in attached files into a proposed analytics roadmap
  • Generate SQL transformation logic and validation rules from a schema and feature definitions
  • Analyse a CSV, such as retention by cohort and plan tier, with summary charts
  • Run a drift analysis on model input features
  • Chat with spreadsheets and update workbooks through the ChatGPT for Excel add-in
  • List data-quality checks and a remediation plan for a dataset
  • Use apps for warehouses, BI and streaming alongside GPTs and shared projects

Limits

  • Business plans do not train on your data by default. On Free, Go, Plus and Pro, chats can be used for training unless you opt out.
  • OpenAI's examples on its pages are demonstrations. Check each result against the source data.

Plugs to turn on

Plugs worth turning on

Answer a stakeholder question with Claude's data plugin

Claude

  1. 1

    Install the plugin

    In Claude Code, run:

    claude plugin marketplace add anthropics/knowledge-work-plugins
    claude plugin install data@knowledge-work-plugins

    In Cowork, install it from claude.com/plugins.

  2. 2

    Connect the warehouse

    Add your warehouse server so Claude can explore schemas and run queries. Use a read-only role.

    Uses Snowflake, Databricks

  3. 3

    Ask the question

    Name the metric, the cut and the period.

    Prompt
    /analyze What was our monthly revenue trend for the past 12 months, broken down by product line? Flag any month that looks off.
  4. 4

    Validate before sharing

    Run the check and read what it flags, such as a denominator that excludes trial users.

    Prompt
    /validate

Profile and clean a CSV in ChatGPT

ChatGPT

  1. 1

    Use a Business workspace

    Sign in to ChatGPT Business if the file has customer data. Business does not train on your data by default.

  2. 2

    Upload and profile

    Attach the file and ask for a profile before any analysis.

    Prompt
    Profile this CSV: row count, column types, null rates, duplicate rows, and values that look wrong. List data-quality checks and a remediation plan.
  3. 3

    Run the analysis

    Ask for the cut you need and the code behind it.

    Prompt
    Compute 3-month rolling churn by acquisition cohort and plan tier. Chart it, and show the code you used.
  4. 4

    Rerun the numbers

    Run the code yourself or recompute two figures in your own tool.

Tune a slow query

  1. 1

    Add the SQL skill

    The SQL optimization skill covers indexing and EXPLAIN analysis.

    Uses sql-optimization-patterns

  2. 2

    Give it the query and plan

    Paste the query and its EXPLAIN output. Say which engine you use.

    Prompt
    This query takes 40 seconds on Snowflake. Here is the query and the query profile. Find the three most expensive steps and propose rewrites. Keep the result identical.
  3. 3

    Prove the rewrite

    Compare row counts and a checksum between the old and new query before you swap.

Verdict

Our advice: use Claude with the data plugin and a read-only warehouse connection for SQL-heavy work, and ChatGPT Business if your team lives in spreadsheets. Validate every number before it reaches a stakeholder.

Our pick: Snowflake

Our advice: start with one connection, read-only, and one question you can verify by hand.

Best for

  • Ad hoc SQL against a warehouse
  • Profiling and cleaning files
  • First-draft charts and dashboards

Skip it if

  • You would share a result you have not rechecked
  • The connection has write access you do not need

Questions

What is the best AI for data analysis?

It depends on where your data lives. Claude has a data analyst plugin and warehouse connectors. ChatGPT Business has spreadsheet analysis, an Excel add-in and warehouse apps. Pick the one that connects to your stack.

Can Claude query my database?

Yes, with a connector for Snowflake, Databricks, BigQuery or another SQL database. Without one, Claude writes the SQL and you run it.

Can ChatGPT analyse CSV and Excel files?

Yes. OpenAI shows ChatGPT analysing uploaded CSVs and working in Excel through an add-in.

Does Claude have a data analyst plugin?

Yes. Anthropic's data plugin adds commands to analyze, explore, write queries, create charts, build dashboards and validate an analysis.

Can AI change my data?

Some connectors can write. OpenMetadata, for example, has four tools that change assets. Use a read-only database role for analysis.

Is my company data used for training?

Not on ChatGPT Business or Claude Team by default. Anthropic lists training as opt-out on individual plans, and OpenAI says Free, Go, Plus and Pro chats can be used for training unless you opt out.

How do I check an AI analysis is right?

Rerun the query yourself and compare totals. Claude's /validate command reviews method, accuracy and bias, but it does not replace your own check.

Which SQL skills work with Claude?

See SQL Optimization Patterns for query tuning and bigquery-basics for BigQuery.

Sources

Docs, issue trackers and discussions this page is built from.

Show all 6
  1. Data Analyst plugin · github.com
  2. ChatGPT for data science and analytics · chatgpt.com
  3. Claude pricing · claude.com
  4. ChatGPT for finance (Excel add-in) · chatgpt.com
  5. ChatGPT plans and pricing · chatgpt.com
  6. ChatGPT data controls · help.openai.com

Updated Nov 3, 2026.