- OverviewUse the assistant
How grounding works
The rule that keeps the AI honest — every claim cited, and a clean refusal when it can’t be.
What is it?
Grounding is the rule that keeps the AI trustworthy. Every claim in an answer points to a real detail on the profile — a visit, a spend, a prediction. If the AI can’t back a claim, it refuses rather than guesses.
Grounding is the rule that keeps the AI Engine trustworthy: every claim in an answer points to a real detail on the guest profile — a visit, a spend, a prediction — shown as a citation you can check. If it cannot back a claim, it refuses rather than guesses.
Citations carry the time and confidence behind a value and link through to the schema, so you can always see where an answer came from. The assistant reads your live data, so there is no drift from a snapshot you cannot see.
- ✓Every claim points at a real detail on the profile.
- ✓Fresh sources — each citation carries the time it was last updated.
- ✓No drift — the AI reads your live data.
- ✓Refusals come with a reason you can act on.
- ✕Inferring details that aren’t on the profile.
- ✕Softening “I don’t know” into a confident-sounding maybe.
- ✕Citing something that doesn’t exist.
- ✕Drifting from a snapshot you can’t see.
Why did Marcus stop coming?
I can confirm Marcus Chen hasn’t visited Edinburgh since 14 March. I can’t answer “why” — that isn’t on the profile.
What I can tell you: he’s opened your last two emails but not clicked, and his churn risk is 64 (medium-high). A targeted SMS might surface a reason — or a manager note on his record would give me something to work with next time.
The AI refuses the unanswerable part and offers next steps instead of guessing.
Connected pages
FAQs
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.