The premise
Your ability to provide accurate answers depends wholly on how healthy your information is.
Accuracy isn’t a property of the model. It’s a property of the corpus. A fluent answer built on a figure that changed two quarters ago is still wrong — it’s just wrong faster, and with more confidence. Everything you produce is only as current as the knowledge it came from.
Health is upstream of accuracy. Everything else is downstream.