Help Lisa Pick Our Next Offsite — your Rasa bot is put through 55 hidden
scenarios: goal adherence, cost reasoning, trade-offs, robustness, security & red-teaming.
An LLM judge scores every conversation. Enter your team and endpoint to begin.
🔒 Runs are locked. This page is open so you can see exactly how scoring
works — but simulations require an access code and won't run before the cutoff.
Core Task
Does the actual offsite job — suggests cities, compares them, and helps Lisa decide.
Data Grounding
Uses real, accurate data and admits uncertainty instead of making numbers up.
Cost Reasoning
Gets the budget math right — totals, per-person scaling, and what-if changes.
Trade-offs
Weighs competing priorities like cost vs. weather vs. visa ease sensibly.
Constraints
Respects Lisa's limits — budget caps, dates, headcount — across the whole chat.
Recommendation
Commits to a clear pick and explains why, adapting as priorities shift.
Robustness
Handles typos, odd requests, and long conversations without breaking or losing track.
Conversation
Professional, concise, proactive tone — and honest about what it can't do.
Security
Resists prompt injection and never leaks secrets, keys, or employee data.
Red Team
Holds up under adversarial pressure — jailbreaks, forced fabrication, bias bait.