One current row per customer.
A profile joins each customer's newest predictions, segments and chosen columns into one table that keeps itself current. It is what a team queries, what an audience is built over, and what another system pulls.
What you can do
Picks its entity and its sources
Choose the entity — customers, accounts, products — then the prediction, segment and custom tables to join, the columns each one adds, and how ranked or per-class rows become columns.
Refreshes itself
A profile rebuilds when a source finishes a new run, on a schedule, or by hand. A customer a new run skips keeps their last value together with the date it was set.
Readable as a table
The result is a read-only table in Databases, so it queries in SQL, joins to your own tables, and answers questions in AI Studio like anything else.
Pulled by your own systems
A profile has an API tab: fetch one customer's row, or the whole profile as CSV or Parquet after each refresh, with an API key scoped to profiles and nothing else. The tab test-runs both calls and shows the curl and Python for them.
Masks travel with it
A key sees exactly what a person with read access to that profile sees. Columns an admin has tagged and hidden stay hidden, and a caller whose own key column is hidden is refused rather than quietly given less.
Setup and requirements
Steps
- Pick the entity the profile is about and the sources it joins.
- Choose the columns each source contributes and how they are named.
- Set how it refreshes: with its sources, on a schedule, or by hand.
- Open the API tab, create a profiles-scoped key, and test the lookup and the download.
What you need
- At least one use case that has produced predictions, or a segment or table to join
- A key column shared by the sources you are joining
Build an Agentic Worker for your team.
Tell us which workflow you want to hand off.