HCA Healthcare runs 190 hospitals across the United States, making it the largest hospital chain in the country. Since 2023, it has been rolling out an AI scheduling tool called Timpani, built with the data company Palantir, and it now covers roughly 130 of those hospitals.
The pitch was simple: let software build nurse schedules automatically, factoring in predicted patient volume, staffing needs, and each nurse's stated preferences. HCA's own leadership said it would save time, cut costs, and remove favoritism from shift assignments.
The nurses using it tell a different story. Several described shifts packed with inexperienced colleagues and almost no senior staff, frequent last minute reassignments, and requested days off that get ignored far more often than when managers handled schedules by hand. One nurse said she had to delay care for her sickest patients because she was the only experienced nurse on a shift otherwise staffed entirely by juniors.
HCA disputes the characterization and says nursing leaders, not the software, make the final call on every schedule. The company points to a number it tracks internally: nurses only get scheduled on their requested days off about one percent of the time. Nurses counter that this used to happen essentially never, so even a small percentage change feels like a real shift in how often their time gets overridden.
The harder problem is that nobody outside HCA can easily check who is right. A former HCA data scientist has sued the company, claiming it deleted the records Timpani used to build schedules, which would make it impossible to audit how well the tool actually performs over time. National Nurses United, the union representing about 10,000 of HCA's 100,000 nurses, says HCA has refused its requests for detailed information about the system.
This pattern is not unique to HCA. Rayus Radiology, a large imaging chain, built a separate scheduling tool on the same Palantir platform and ran into a different version of the same issue: the software linked test orders to the wrong patients and invented details that were not in the original paperwork.
Palantir's business is booming regardless. The company says its software now touches roughly one fifth of all hospital beds in the country, and its commercial sales have been growing at triple digit rates this year, making hospital systems eager customers at a time when staffing shortages and rising costs are two of healthcare's most painful problems.
But Palantir also carries baggage that most scheduling software vendors do not. The company's contracts with immigration enforcement and the military have made it a target for protests, and nurses in multiple cities have demonstrated against hospitals using Palantir products at all, treating the scheduling complaints and the company's government work as part of the same objection.
For any business considering automated scheduling or staffing tools, the real lesson is not about Palantir specifically. A tool can perform well on the metrics a company chooses to track while still making daily work worse for the people living with it, and that gap only shows up when workers have a real channel to complain and someone outside the company can check the underlying data. HCA built the tool with sophisticated technology, but it did not build an equally sophisticated way to prove it is working.