Find the people who can take over when the AI gets stuck
Candidates inherit broken AI conversations with frustrated, high-value customers, then call the partner to close the service on the customer’s behalf. Gemini scores every message as they go, and every resolution becomes material that makes the AI better.
15 min. Read a failed AI chat, write the first message, diagnose the failure.
20 min. Gemini plays the customer. Backend tools, a mid-chat curveball, live quality scoring.
12 min. Call the partner on the customer’s behalf, get past policy, secure a reference, close the loop.
10 min. Document decisions end to end and propose the fix for the AI.
10 min. Two written questions on judgment and closing the loop with the AI team.
How the AI improves
Every resolution record, from candidates and from real cases your concierges log, becomes a structured improvement: failure category, root cause, the ideal AI response, a proposed fix and a regression test. Reviewers approve the good ones, then export them as tuning data, eval cases or prompt rules.
Candidates: use the personal link from your invitation. The password box above is for the hiring team and concierges.