This is more of a fluff piece than actual advancement. Here is the meat of the article:
Hippocratic says its Constellation model outperformed real nurses 79% to 63% in identifying a medication's impact on lab values; 88% to 45% in identifying condition-specific disallowed over-the-counter medications; 96% to 93% in correctly comparing a lab value to a reference range; and 81% to 57% in detecting toxic dosages of over-the-counter drugs.
Basically, these are low hanging fruit things, not the primary job of those "expensive" nurses. We need to look at this AI against scheduling appointments, determining whether the complaint is critical to go to ER, can wait a few days, or routine and can be handled over the phone.