A product team is designing a fraud detection agent for a bank. The agent analyzes transactions in real time…

Question

A product team is designing a fraud detection agent for a bank. The agent analyzes transactions in real time and flags suspicious activity. False positives block legitimate transactions, causing customer frustration. False negatives allow fraud through. Currently the agent flags 30% of transactions, of which 90% are false positives. What analysis framework should be applied?

Accepted Answer

B. Use a tiered confidence system: high-confidence fraud → block immediately, medium-confidence → add friction (2FA), low-confidence → allow with monitoring. Each tier has different precision/recall tradeoffs

Answer

A. Lower the sensitivity threshold to reduce false positives

Answer

C. Remove automated blocking entirely and rely on post-transaction review

Answer

D. Train a separate classifier to pre-filter the agent's results

Question 280