The most reliable data in the business is usually the hardest to reach: it needs SQL, the schema, and knowing which of four hundred tables holds the thing you want. So requests queue behind whoever writes the queries. The database connector is part of askFinz Knowledge, which is still rolling out; here is what it will do, and what it will not.
Ask a question instead of writing a query
Ask "which accounts churned after a support escalation" and get an answer drawn from the tables you exposed, with the rows it used named so an analyst can verify them.
Search the text columns nobody indexes
Descriptions, reason codes, free-text notes and comment tables hold the explanation behind a status change. Those are matched by meaning rather than exact wording, which is exactly what LIKE '%...%' cannot do.
Expose a defined slice, not the whole schema
You choose the tables, and the views. Exposing a curated view rather than a raw table is the normal way to do this, and it is the recommended one.
What is read, and what is never written
What crosses, and what stays in the database
Read
rows from the tables and views you list, on a schedule you set.
Written back
nonenothing. The connection uses a read-only account; no inserts, updates or schema changes.
Stored
a searchable copy of the columns you exposed, plus the key needed to point back at the row.
Never read
tables you did not list, and columns you excluded.
What you will be granting
A read-only database user, scoped to the tables and views you name — the same thing you would give a reporting tool, and nothing that can change data.
Setting it up
Planned — not available yetWhat connecting it will look like once the connector arrives.
- 01
Create a read-only user, or reuse the one your reporting stack already has.
- 02
List the tables and views to include, and exclude any sensitive columns.
- 03
Decide the mode: read a full refresh on a schedule, or only rows changed since the last pass.
Limits, honestly
- Read-only, one direction. Nothing is written back, ever.
- Answers reflect the last refresh, not the live table.
- Joins across databases are not resolved for you; expose a view that already joins them.
- Very large tables should be exposed as a filtered view rather than in full.
- Blob and binary columns are skipped.