Product design question

A top-10 bank wants Sierra to launch an AI agent for card-fraud support. The agent could lock cards, collect dispute details, issue provisional credit, and escalate sensitive cases to human specialists. Walk me through how you'd discover requirements across operations, compliance, risk, and support; define the MVP; and decide which workflows should be fully automated, human-in-the-loop, or human-only at launch.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Tests requirements discovery and risk based scoping for a sensitive financial services use case with real regulatory and safety stakes.

Interview operations, compliance, risk, and support leads separately to surface conflicting priorities before proposing scope, since fraud has real regulatory exposure.
Define the MVP around the lowest risk, highest volume workflow first, such as collecting dispute details, rather than starting with card locking.
Classify each capability as fully automated, human in the loop, or human only based on reversibility and regulatory sensitivity.
Keep card locking and provisional credit issuance human in the loop or human only at launch, since these are hard to reverse and regulated.
Automate low risk, high volume steps like collecting dispute details and verifying identity, freeing human specialists for judgment calls.
Define escalation criteria explicitly, for example any case involving a disputed amount above a threshold routes straight to a human.

Separates capabilities by reversibility and regulatory risk, not just technical feasibility, which is the core judgment this question is testing.
Starts the MVP with a lower risk workflow like dispute detail collection instead of jumping straight to card locking.
Involves compliance and risk as MVP stakeholders from day one rather than treating them as a post launch review gate.

Automating provisional credit issuance at launch without human review, given the regulatory and financial exposure.
Treating requirements discovery as a formality instead of surfacing real conflicts between operations speed and compliance caution.

How would you handle a case where the agent is unsure whether to lock a card automatically?
What evidence would convince the bank to expand automation to card locking later?