Analytics
Analytics with the SQL attached.
Ask your data a question in plain language — the top ten customers by net revenue this quarter — and the answer comes back as a formatted table, the statement that produced it attached, every source named.
There is no dashboard to assemble here and no report on a schedule. This is Ask pointed at your data: one question, one answer, the evidence attached.
Plain language in, a ranked table out.
The landing asks the demo workspace for its top ten customers by net revenue. What comes back is not prose about the data — it is the table itself, ranked, with its work shown.
You ask
In your own words, against the connections your workspace granted. Nobody writes a query to get an answer — naming what you want to know is enough.
Datis writes one statement
The planner turns the question into a single scoped, read-only statement against the schemas you connected. The database role behind the connection carries SELECT and nothing else, so a bug in Datis cannot widen what a question may touch.
The table comes back with its work
Formatted and ranked, with the SQL that produced it attached and the sources it read named. You check the arithmetic instead of trusting a summary.
Derived live, and it says so.
An answer you act on has to say where it came from and what it refused to do. Every answer here does both.
The live system, not a copy
Answers are derived from your connections at the moment you ask — the agent cites the live system it read, not an export prepared in advance. Databases are read against a read replica when you have one.
The statement travels with the table
The SQL is attached and the sources are named, so the answer can be checked instead of believed. An answer without its evidence is not an answer here.
A runaway question fails small
Every statement runs under a row cap and a timeout. A question that would drag a whole table across the wire fails and says so — it does not dump the table, and it does not page anyone.
One connection feeds all five verbs.
The database that answers a revenue question is not plugged into a separate analytics product. It is connected to Datis once — ask across it, watch it, fix what breaks, build on it.
Connect once
Twenty-seven providers, one way in — every one connects the same way. A data question is one use of a connection you already made, not another copy of your data to keep somewhere.
Access is a policy
Every connection starts read-only, and asking never widens it. What Datis may do with a connection is a policy you set — not a property of the provider, and not a side effect of a question.
Every read is recorded
The audit log keeps every read, every approval, every apply. Which sources a question touched is part of the record, not a thing to reconstruct afterwards.
4m 13sfrom the first bad signal to the merged fix — the run the landing replays, in the same demo workspace that ranks those ten customers.
Connect a database. Ask it something real.
The beta is where this runs against your stack. Every request is read by a person, and design partners from the waitlist step in first.