Insights use case

Build an AI data agent that answers questions over your own data

Ask in plain language, and your agent queries the warehouse, validates the numbers and returns a chart-ready dashboard — with the SQL it ran attached.

Works with the tools your team already uses

GmailGoogle DriveGoogle CalendarGitHubLinearJiraAsanaZoomOutlookTrelloAirtableGmailGoogle DriveGoogle CalendarGitHubLinearJiraAsanaZoomOutlookTrelloAirtable

10x

faster ad-hoc analysis

0

dashboards to maintain by hand

100%

answers with a source trail

How the data analysis agent works

1

Ask a question

Type the question the way you'd ask an analyst — no SQL, no dashboard hunting.

2

Locate the data

The agent maps your question to the right tables, sheets or connector records.

3

Query and validate

It runs the query, sanity-checks row counts and reconciles totals before answering.

4

Deliver a dashboard

You get a shareable HTML dashboard, the underlying rows and the query that produced them.

What makes it different

Grounded answers only

Every figure traces back to a record the agent actually read — no invented numbers.

Charts, not CSV dumps

Results render as dashboards with KPIs, tables and trend charts your team can share.

Runs code when it needs to

The built-in sandbox lets the agent transform, join and aggregate data before reporting.

Scheduled reporting

Turn any question into a recurring Monday-morning report delivered to Slack or email.

Ask it anything

No workflow builder to learn. Talk to the agent the way you'd brief a colleague, and watch each step of its work in the thread.

What changed in revenue week over week, and which segment drove it?

Build a dashboard of pipeline by stage and owner for this quarter.

Find rows in this sheet that don't reconcile with Stripe payouts.

Summarise the top 10 accounts by expansion potential and why.

Everything the agent needs, built in

Isolated connections

Each user connects their own accounts — nothing is shared unless you say so.

Model choice

Pick fast, balanced or deep-reasoning models per agent and switch any time.

Triggers and schedules

Run on a schedule, on a new email or from a webhook — no babysitting.

Governed by default

Role-based access, audit logs and usage limits across the workspace.

Common questions

Does it need access to production data?

No. Connect a read-only warehouse user or a sheet; each user's connections are isolated.

Can I see the query it ran?

Yes — every step, tool call and query is shown inline in the thread.

Explore other use cases

Build your data analysis agent

Connect your tools, start from a template and have it running in an afternoon.