What conversational analytics actually means
Conversational analytics is asking a question of your data in plain language and getting the answer back. Not a query builder. Not a dashboard someone has to configure first. Not a chatbot that hands you SQL and wishes you luck.
The term gets stretched, so it's worth separating three things that often share the name:
Text-to-SQL turns your question into a query. That's genuinely useful, but the output is a query, not an answer — and you need to read SQL well enough to know whether to trust it before you run it.
Conversational BI is a chat box on top of a dashboard tool. It answers questions about metrics someone already modelled. Ask it something outside that model and it has nothing to say.
Conversational analytics is the whole path: read the schema, write the query, run it against live data, return the result, and hold enough context that your next question can build on the last one.
GenViz does the third one, against the databases your business actually runs on.
The warehouse is the part nobody mentions
Most tools in this category share an assumption they don't advertise: the data is already in a modern warehouse. Snowflake, Databricks, BigQuery — pick one. The demo is impressive because the hard part happened before the demo started.
For a lot of companies that hard part is a months-long ETL project standing between them and their first question. You need pipelines, a place to put them, someone to own them, and a security review for the whole arrangement. The analytics tool is the cheap part.
GenViz starts from the other end. It connects to the database where the data already is — the operational one, the one with today's orders in it — and asks the question there. There is no copy of your data to build, maintain, or explain to your auditor.
What it connects to
Seven database types today — PostgreSQL, MySQL, Microsoft SQL Server, MongoDB, Google BigQuery, IBM Db2 (LUW) and IBM i / AS/400. Some run straight from the desktop app, some through a connector you host, and some either way.
Rather than repeat the details here and let them drift, we keep one page current: supported databases and connections.
If your data lives on an AS/400, that's worth its own page — we wrote one: IBM i and AS/400.
What an answer looks like
A table.
Not a screenshot in a Slack thread, not a link that expires, not a chart with opinions baked into it. A result set your team can open, sort, and check against the source. Larger results become an extract you can download as Parquet and hand to whoever needs it in a real tool.
Answers are shareable, which is the part that usually matters. Someone asks a question on a laptop; the person who needed the number opens it on a phone. No KPI pipeline, no dashboard to publish, no warehouse in between.
Where it runs, and who holds the keys
GenViz is a desktop app. Your connection credentials are stored encrypted, and queries run from your machine straight to your database — or from a connector you host, if you use one.
There is no bulk copy of your database: no ETL, no replica, no warehouse. What does leave your network is the results you asked for. When an answer becomes an extract, we store that Parquet file in our cloud storage so it can be opened on another device, shared with a colleague, and read on a phone. That is a real step off your network and we would rather say so plainly than imply otherwise.
Queries are read-only by construction. Write-shaped statements — INSERT, UPDATE, DELETE, DROP, ALTER, TRUNCATE, GRANT — are refused before they ever reach the driver. That's
not a setting someone can forget to turn on.
On the Team plan you bring your own LLM API key, so inference is billed to you by your provider at their price. You pay us for the workspace, not a markup on tokens. See pricing for the full picture.
What it isn't
It's not a SQL editor. If you want to hand-write queries against a hundred engines, that tool exists and it's good. This is for the question you'd otherwise have to ask someone else.
Frequently asked
Do I need a data warehouse? No. GenViz connects to operational databases directly. If you already have a warehouse, BigQuery is one of the supported types — but nothing has to move there first.
Does my data leave my network? Your database doesn't — there is no bulk copy, no ETL and no replica, and credentials are stored encrypted. Query results do, in one specific case: when an answer becomes an extract, the Parquet file is stored in our cloud storage so it can be opened on another device and shared. Queries themselves run from your machine, or from a connector you host inside your own network.
Can it write to my database? No. Write-shaped statements are rejected before reaching the driver.
Is this text-to-SQL? It uses text-to-SQL as one step. The difference is that you get the answer, not the query — schema inspection, execution, and the result are all part of the same request.
Which LLM does it use? Anthropic, OpenAI, or OpenRouter models. On the Team plan you supply your own API key and pay your provider directly.
What if my data is on an AS/400? That's a supported connection type and something of a specialty. See IBM i and AS/400.
Start asking your own data questions.
GenViz is in early access. Join the waitlist and we'll get you set up — or download the app for macOS, Windows, or Linux and have it ready.