Ask your data in plain words. Get the number, and its proof.
Converse connects to your database with a login that can only read, learns how your business counts things, and answers with the number and the query behind it. Your team sees the real names; the model sees them masked.
Give it a login that can only read. It learns your definitions.
Converse refuses a login that can write and runs every query read-only, so it can’t change your data. Then it learns your database: what each table holds, how they join, and what your business means by an order, a customer or churn.
When a definition is wrong, your team corrects it once and the answers that follow use the fix. Analysts approve what viewers suggest.
Where it connects. Converse works with the databases and warehouses you already run: PostgreSQL, SQL Server, Oracle and MySQL; BigQuery, Snowflake, Databricks and Redshift.
Average order value
Net revenue per completed order, leaving out test orders.
Leave out test orders
Leave out gift-card orders
Ask in plain words. Every answer comes with its proof.
Ask the way you would ask a colleague, and follow up the same way. When a question could mean two things, Converse asks rather than guesses.
Each answer shows the query it ran, the sources and definitions it used, and the date your data runs through. A month that is only partly in is labelled partial, not drawn as a drop. If it can’t answer, it says so.
How are orders trending?
Completed orders peaked in September at 48,210, up 4.1% on August.
October is partial, so it is not compared.
October · partial · to 21 Oct
select date_trunc('month', ordered_at) as month,
count(*) as orders
from orders
where status = 'completed'
group by 1 order by 1;Every answer has two readers. Only your team sees the names.
To write an answer, the model sees your table and column names, your question and a capped slice of the result, with personal data masked first. Your team sees the real names and emails.
Card numbers and other secret-grade data stay masked for everyone, exports included.
What we keep. The answers your team sees, names included, are saved in conversation history and pinned answers on Nectorq’s side. The query itself runs read-only in your database.
Q · Who were our top 5 customers last quarter?
The top customer led last quarter with 214 orders.
| Customer personal | Email personal | Card secret | Orders public |
|---|---|---|---|
| Lena Fischer | lena@example.com | •••••• | 214 |
| Omar Haddad | omar@example.com | •••••• | 188 |
| Grace Liu | grace@example.com | •••••• | 171 |
| Diego Alvarez | diego@example.com | •••••• | 160 |
| Hannah Price | hannah@example.com | •••••• | 152 |
read-only · your database · data through 21 Oct 2026
Ask why a number moved. It investigates.
Ask why orders fell in March and Converse tests the likely causes against your data. You get the finding, how much of the change each cause explains, and what it ruled out.
An analyst starts each investigation and can steer it. Investigations mask personal data the same way answers do.
Why did completed orders fall 8.4% in March?
Tacoma Mall closed for a refit, 11–24 March61% of the drop
Spring promotion moved from March to April28% of the drop
Price rises on top productsprices flat month on month
Ruled outMore stock-outs than usualout-of-stock days in the usual range
Ruled outStarted by Priya Raman, Analyst
Pin the numbers you check every week. They wait on your home page.
Pin an answer and its number joins Your numbers, with the date its data runs through. Open one to refresh it against live data, or present them all as a deck. When a number moves unusually, ask why in one click.
- Your numbers, presented as a deckLive
- Scheduled insightsComing soon
- Alerts on pinned numbersComing soon
Bring your database and a real question. Check the proof yourself.
Connect with a read-only login, let Converse learn it, and read the query behind every answer.