Product design question

A multi-location nonprofit provider wants Sierra voice and chat agents to help patients find in-network specialists, check availability, and book appointments. How would you design the experience when provider data, scheduling rules, and specialty taxonomy vary across clinics? What would you standardize across locations, what would you customize, and where would you require human handoff?

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 the ability to design a multi location product experience that balances standardization for scale against necessary local customization.

Standardize the core conversation flow and intent structure, finding a specialist, checking availability, booking, across all clinics for consistency and maintainability.
Customize the underlying data layer per clinic, provider directories, scheduling rules, and specialty taxonomy, since these genuinely vary and cannot be forced into one schema.
Build a normalization layer that maps each clinic's local taxonomy into a shared internal structure, so the conversation flow does not need to know about local variation.
Require human handoff for any scheduling edge case the normalization layer cannot confidently resolve, such as an unmapped specialty or unclear provider availability.
Pilot with one or two clinics first to validate the normalization approach before rolling out across all locations.
Set a data quality bar each clinic must meet, accurate provider directory and scheduling rules, before their agent goes live.

Separates what should be standardized, conversation flow, from what must be customized, underlying data, rather than forcing one uniform approach across variable clinic data.
Proposes a normalization layer as the technical solution to reconcile standardization and customization, which is the concrete design choice this question is testing.
Sets a data quality gate per clinic before launch, since inconsistent source data was named as the core challenge in the prompt.

Trying to fully standardize scheduling rules and specialty taxonomy across clinics that genuinely differ, causing real booking errors.
Launching a clinic without validating its underlying data quality first, given how directly errors there would affect patients.

How would you handle a clinic whose data quality never reaches the required bar?
What would trigger human handoff versus letting the agent attempt to resolve an ambiguous specialty match itself?