🏔️ Summit 2026 Feature
Let Slate AI answer questions with data from approved queries.
🏗️ Provisional document
This feature is pending release, and this document may change over time. Check What’s New for the latest releases.
Slate AI Query Knowledge lets administrators make selected query outputs available as knowledge sources for Slate AI. Use query knowledge sources when Slate AI should answer questions from structured, permissioned query data.
Creating a query knowledge source
Go to Database → Bots.
In the Knowledge Sources section, select New Source.
Set Type to Query.
Choose the query type, category, and base that Slate AI should use.
Add context that explains what the query represents and how Slate AI should use it.
Select Save.
Configure the exports and filters that should be available to Slate AI.
Open Edit Permissions and grant Access via Slate AI to the appropriate grantee.
Ask Slate AI a test question that should use the query knowledge source. When testing, use a user who has the grantee selected for Access via Slate AI, or impersonate a user with that access.
Designing the query knowledge source
A query knowledge source can include many exports because Slate AI uses the available query data that is relevant to the question. Include the data points that Slate AI should be able to answer from, and omit data points that should remain outside the source.
The query provides both context and a data boundary. For example, if the query includes an export such as Lifetime Giving, Slate AI can answer from that predefined calculation instead of inferring how your database defines lifetime giving. If a data point is not included in the query or another query knowledge source available to the user, Slate AI cannot use that data point in its answer.
For complex, calculation-heavy, or wide query knowledge sources that do not require real-time results, consider building the source from a materialized view or query base view. Materialized views store point-in-time query results and refresh on a schedule or on demand, which can improve response performance when the source depends on complex calculations.
Examples
Create a query knowledge source for current funnel counts so Slate AI can answer staff questions about applicants by round or population.
Create a query knowledge source for assigned students so advisors can ask Slate AI about their own caseload.
Create a query knowledge source for lifetime giving and campaign participation so gift officers can ask data-driven questions before outreach.