Three rounds. The same three fictional clients you have seen throughout this site. Somewhere in a shared knowledge base, sorted by topic instead of by client, is the exact rule you need. Find it, pick it, and see how your time compares to Agent First matching the call to a Client Profile and the active rule, automatically.
One shared knowledge base
Four topics, three clients each: twelve lines total, sorted by topic the way most internal documentation actually is, not sorted for you.
This is not a live AI guessing
What races you here is the same deterministic client-and-topic match described throughout this site, animated at an illustrative simulation speed, not a measured product performance number. Nothing invented, nothing hidden.
Speed and accuracy, tracked separately
Answer first with the wrong client's rule and you still lose the round on accuracy. Both get counted, then shown to you honestly at the end.
Works best on a laptop or desktop screen. Nothing is recorded or sent anywhere. This runs entirely in your browser.
Round 1 of 3
Reset each round: a fresh scenario, a fresh timer, the same shared knowledge base.
Customer on the line
The knowledge base below is grouped by topic, not by client. Expand the right topic, then pick the line for this client.
You
0.0s
Agent First
--
1
FindMatching call to Client Profile...
2
InterpretLocating the active rule...
3
ActRule applied.
The rule itself stays hidden until you answer, so this cannot just hand you the result. Only the timing shows.
Every round, Agent First's side resolved the same way: match the call to a Client Profile, match the request to the active rule, apply it. No search, no scrolling, no chance of grabbing the wrong client's line by accident. See what that looks like against your own rules.