How would you reduce shipment-status enquiries?

Example Answer

I'd first improve the freshness of the portal status. The evidence points to delayed information, and a chatbot using the same stale source would still give a stale answer.

I'd confirm the feed's reliability and whether the portal can safely consume the more frequent updates. Then I'd propose a small integration change with a visible last-updated time and an explicit unavailable state if the feed fails. That would directly address the mismatch between the available data and what customers see.

Before rollout, I'd measure status-related contacts per shipment and the current delay from carrier update to portal display. After a controlled rollout, I'd compare those measures while checking for incorrect status changes and new support problems. Raw contact counts would be misleading if shipment volume changed.

I would keep the chatbot idea as a separate hypothesis. If customers still need help interpreting exceptions after the information is current, we could test whether a guided interface or an assistant improves that specific task. I wouldn't add a conversational layer just to satisfy the original label of the request.

What the interviewer is assessing

Essential: Explains the nightly-versus-fifteen-minute mismatch, recommends using the available feed, and measures status enquiries per shipment rather than raw contacts.

Stronger: Specifies a last-updated indicator, feed-failure behavior, and a controlled comparison that also checks incorrect status changes. A combined interface proposal is acceptable if it still fixes freshness and justifies the additional work.

Red flags: Builds a chatbot before inspecting its data source or claims the integration must eliminate all contacts.

Follow-up

How would you explain this to a sponsor who wants a visible AI initiative?

Example response: “The fastest path to the customer outcome is to expose the reliable information you already receive. We can show the reduction in avoidable contacts, then test AI on remaining interpretation problems with a clear baseline. That makes any later AI investment easier to justify.”