How grounding works
What the assistant reads, what it cannot read, and how it cites its sources.
What is it?
Every AI answer shows its sources — the exact records it read to produce it. A draft campaign points to the segment and the tone it used; a lapse-risk answer points to the guest and the signals behind the prediction.
The assistant can never read across businesses. Your data stays walled off from every other Stampede customer — enforced by the system, not by policy.
Grounding is what keeps the AI honest: every piece of data it reads is logged, and every change it proposes needs your explicit approval.
Element features
Every response names the guest, visit and segment records it read.
Reads can never cross from one business to another.
Everything the AI reads is logged for later audit.
No profile or workflow change happens without an explicit human sign-off.
Grounding reads the current profile, not a stale snapshot.
FAQs
Every response cites the exact records it read. If a claim has no source, it wasn't grounded in your data.
No. Reads can never cross from one business to another, so one group's data never reaches another's responses.
No. Reads happen automatically, but any write to a profile or workflow needs explicit approval first.
Yes. Every read is logged, so you can audit exactly what informed any response.
See How grounding works running on your venues
Thirty minutes is enough. Book a demo or send a question — both routes land on the same team.
