Skip to content
Log inStart free
AI data analyst

Anyone on your team can ask a question in plain English and get a chart, a dashboard, or a straight answer. Every result runs on your governed definitions and shows the SQL behind it.

Start freeBook a demo

14-day trial. No credit card required.

Trial-to-paid by signup source

Postgres

Which signup sources convert best from trial to paid this year?

Referral converts best, at 32% in September and 29% across the year. Partner signups have climbed every month since March, from 15% to 27%.

Trial-to-paid conversion · governed metric

Trial-to-paid by signup source

Jan to Sep 2026 · monthly

ReferralOrganicPartnerPaid search

Postgres · trials, users, subscriptions · 36 rows

Powering analytics for 5000+ teams worldwide

Exa logoTiger Data logoPurpose logoGumloop logoComposio logoFullscript logoShiftrx logoDrata logoExa logoTiger Data logoPurpose logoGumloop logoComposio logoFullscript logoShiftrx logoDrata logo

Governed definitions

Ask about churn or NRR and Basedash uses the metric your data team already defined, the same one your dashboards use. When nothing is defined, it says which definition it picked and gives you the main alternative.

Explore the context engine

Models/subscriptions

Verified

Measure

Customer churn

customer_churn_rate · Percent · Monthly · Data team

Share of customers active at the start of the month who cancel during it. Excludes internal and test accounts.

count(distinct case
  when canceled_at between month_start
    and month_end then customer_id end)
/ count(distinct customer_id)

What was churn last month?

Customer churn was 2.4% in September, using the Customer churn definition from Models. Revenue churn, the other common definition, was 1.1%.

Customer churn · governed metric

Measured accuracy

On BI Bench, the benchmark we run and publish for AI data analysts, Basedash scored 94.8% on hard questions against a real, complex schema, with a 32.6-second average response. That's the highest of the 11 agents tested.

See the BI Bench results

Basedash on BI Bench

94.8%

Accuracy

32.6s

Avg. response

Accuracy on hard questions

11 agents

  1. 1Basedash94.8%
  2. 2Codex90.9%
  3. 3Hex80.6%
  4. 4Claude Code78.0%
  5. 5TextQL64.7%

From answer to dashboard

Ask a follow-up to change the grain, the breakdown, or the chart type. When an answer is worth watching, add it to a dashboard, where it refreshes on its own with the same SQL behind it.

Explore dashboards

Show it weekly and split it by plan

Weekly active teams by plan

EnterpriseGrowthStarter

191 this week

Same SQL, refreshed with the dashboard

KPIs

MRR

$412k

Churn

2.4%

Add a chart

Wherever you ask

Mention @Basedash in a Slack thread and the answer comes back with its chart. Add Basedash to your iPhone Home Screen and ask out loud, or connect Claude over MCP. Every path runs on the same definitions and permissions.

Explore the MCP server

Under the hood

What a data leader asks before rolling AI analytics out to the whole company, answered by how the product works.

Start free

  • Real queries only

    Every number comes from a query run in that conversation.

  • SQL on every answer

    Open the steps to read, copy, or edit the query.

  • Your definitions first

    Models and saved context shape every answer.

  • Asks when it matters

    One focused question when a gap would change the answer.

  • Your access rules

    Answers follow who can query each source, in Slack too.

  • Across sources

    Combine your database, warehouse, and SaaS data in one answer.

  • Read-only by default

    AI write queries need explicit approval before they run.

  • Never trains on your data

    Customer data is never used to train AI models.

How answers are checked, who can ask, and what Basedash can query.

How can I check that an answer is right?

Open the steps under any answer. Basedash shows the definition it used, the tables it searched, and the SQL it ran, and every number in the reply comes from a query run in that conversation rather than from memory. You can copy or edit the SQL, rerun it, and look at the raw rows. The team reruns internal evaluations with every new AI model release, and on BI Bench, the public benchmark Basedash runs for AI data analysts, it ranks first for accuracy.

What happens when a question is ambiguous?

Basedash fills routine gaps from context and tells you what it assumed: "last quarter" means the most recent one, and a metric your team has defined uses that definition. When an ambiguity would still change the answer, most often a business metric nobody has defined, it asks one focused question before running anything. If it has to choose, say between customer churn and revenue churn, it names the definition it used and gives the other number too when the two differ.

Who on my team can use the AI data analyst?

Anyone who can describe what they want to know. Sales, finance, product, support, and leadership ask in plain English without writing SQL, while analysts can read and edit every query it writes. Access follows your workspace settings: admins choose which groups and members can query each data source and can hide sensitive schemas, tables, or columns from AI. In Slack, people are matched to their Basedash accounts by email, so the same rules apply there.

What data can the AI data analyst query?

Anything you connect: databases like Postgres, MySQL, and SQL Server, warehouses like Snowflake, BigQuery, and Databricks, and SaaS tools like Stripe, HubSpot, and Salesforce synced into Basedash Warehouse. It can combine sources in one answer, for example product usage from your database with revenue from Stripe. It also reads your Models and AI context, so it knows which tables matter, what your columns mean, and how your team calculates each metric.

What kinds of analysis can I ask for?

Quick lookups like top customers by revenue, and deeper work like cohort retention, funnel conversion, segmentation, and explaining why a number moved. Follow-up questions keep the context, so you can drill from a total to a segment to the accounts behind it without restating anything. Very long-running jobs suit scheduled automations better than a live chat, and analysts can always take a generated query into the SQL editor.

Can I edit the SQL or the chart behind an answer?

Yes. Every chart in an answer keeps its query, and you can edit the SQL directly or ask in plain English to change the chart type, the time grain, the breakdown, or the formatting. The chart updates in place instead of starting over. When it's right, save it to a dashboard, share it to a Slack channel, or export the underlying data as CSV, JSON, or Excel.

Can I ask questions outside the Basedash app?

Yes. Mention @Basedash in a Slack channel or DM to get the answer and its chart in the thread, and conversations started in Slack sync back to the app. On an iPhone, add Basedash to your Home Screen and ask out loud. Claude, ChatGPT, Cursor, and other MCP clients can connect to Basedash too. Each surface uses the same definitions and permissions as the app.

Related features

All features

Connect a source and get an answer you can check, in plain English.

Start freeBook a demo