How would you keep an AI recommendation useful when live business data is unavailable?

Instruction: Design the data freshness, error, and fallback behavior, making clear which claims the assistant may still make.

Context:

A fictional sales assistant uses approved product documentation and a live inventory API to advise employees. Documentation refreshes nightly; inventory must be no more than five minutes old when an employee promises stock. The API sometimes times out. The assistant may draft a customer reply but cannot reserve stock or send messages. A cached inventory result is 40 minutes old.

Updated

Official answer available

Read the opening below, then unlock the full answer and practical guidance.

I would distinguish stable product facts from time-sensitive availability. The assistant can still describe a documented product feature if its source is approved and current enough for that purpose. It cannot turn a 40-minute-old stock result into a current availability promise. Each retrieved result should carry its source, version, retrieval time, and relevant business timestamp...

Upgrade to view official answer

Related Questions