Contoso HR is building a Copilot Studio agent for employee leave requests. The agent must reliably open the Leave Request topic when users say things like "I need time off," "book PTO," or "schedule vacation." A maker notices that "can I take Friday off" sometimes falls to the fallback topic instead. Which change most effectively improves intent coverage for this topic without broadening it to unrelated HR questions?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Trigger phrases are the agent's "ears" for a topic — more natural variations of how people ask for leave help the NLU land on the right topic. Broadening coverage with good samples beats tanking the confidence threshold or killing fallback.
Full explanation below image
Full Explanation
Copilot Studio topics are activated by trigger phrases (and generative orchestration descriptions when generative mode is on). Coverage improves when makers add diverse, realistic utterances that reflect how employees actually speak — informal ("Friday off"), formal ("request annual leave"), and synonym-rich ("PTO," "vacation," "time off").
Option A is correct. Expanding trigger phrases/sample utterances and validating in the Test pane is the supported way to improve intent recognition for a specific topic without global side effects.
Option B is wrong because an extremely low confidence threshold increases false positives: unrelated HR questions (payroll, benefits) may incorrectly fire Leave Request.
Option C is wrong because disabling Fallback removes graceful handling for unrecognized intents and does not fix Leave Request matching; users get stuck or receive poor responses.
Option D is wrong because Copilot Studio does not use raw regular expressions as the primary intent model for classic trigger phrases, and keyword matching on "off"/"leave" would over-match ("sign off," "leave this open").
Exam tip: Prefer richer trigger phrases and testing over lowering confidence or removing system topics.