Two custom topics exist for the group's service agent: "Lab Result Follow-Up" and "Appointment Management." Owners asking "my dog's bloodwork came back, when should I bring him in" are routed to Appointment Management about half the time. What should the specialist change first?
Select an answer to reveal the explanation.
Short Explanation
When two topics both look like the right answer, the reasoning engine is guessing, and a coin flip is exactly what you get. Fix the boundary, not the symptom. Write the classification descriptions so a human reading them could tell you instantly which one owns "recheck after bloodwork."
Full Explanation
Inconsistent routing between two topics is a classification problem, and classification is driven entirely by the classification descriptions. The durable fix is to redraw the boundary in language: one description should explicitly claim recheck scheduling that follows a lab result, and the other should explicitly exclude it. Adding actions to the losing topic does nothing, because actions are only considered after a topic has already won classification. Merging everything into one topic is a real anti-pattern: a single sprawling topic with many actions gives the engine more ways to pick the wrong action, and it destroys the ability to scope instructions and guardrails per job, so it trades a routing problem for a worse action-selection problem. The guardrail option is the closest-looking trap because it does address the symptom the owner sees, but a guardrail constrains behaviour once a topic is active; it cannot send the conversation to the other topic, so the owner still gets a refusal instead of a recheck booking. Exam-worthy caveat: after any classification change, re-run the affected cases in Testing Center, since sharpening one description frequently steals traffic from a third topic you were not looking at.