A critical care nurse in Florida requested 50 specific shifts over four months. More than half the time, the scheduling software assigned her elsewhere, often stacking consecutive days in ways she describes as mentally exhausting. Her experience is not an isolated complaint. It is, according to nurses across one of the largest hospital systems in the United States, a pattern.
HCA Healthcare, the largest hospital chain in the US, has deployed an AI-powered scheduling tool called Timpani at roughly 130 of its 190 locations since 2023. Built on Palantir Foundry, a data integration and analysis platform developed by Palantir Technologies, Timpani uses algorithmic forecasts of patient volumes and staffing needs to generate nursing schedules automatically. The premise is straightforward: replace time-consuming manual scheduling with software that can optimize across many variables at once. The reality, nurses say, is considerably messier.
What the Complaints Actually Describe
The nurses who spoke with WIRED do not simply object to an inconvenient app. They allege that Timpani creates conditions that compromise patient care. Specific concerns include shifts staffed with too few nurses, an insufficient mix of experienced and junior staff, and a recurring problem on Sundays in particular. One nurse with a decade of bedside experience described shifts where all four of her colleagues were junior nurses, leaving her as the sole senior clinician and forcing her to delay care for the most critical patients in order to guide less experienced staff through lower-acuity cases.
Nurses also report that the tool frequently disregards stated preferences for days off, known internally as “red days.” HCA has acknowledged that Timpani schedules nurses for 1 percent of their requested days off. That figure may sound small, but nurses say it represents a meaningful change: being assigned to work on a red day was previously described as essentially unheard of at some facilities. Now, according to one nurse in Missouri, virtually every nurse at their location has experienced it.
The downstream effects are practical and cumulative. Nurses describe spending more time trying to trade shifts or appeal assignments than they did before Timpani arrived. Some colleagues, they say, are increasingly using paid or unpaid time to skip shifts, a pattern that carries its own risk: doing so more than a few times per year can lead to termination.
The Gap Between Official Framing and Reported Experience
HCA’s position is that nursing leaders, not Timpani, make final scheduling decisions. A company spokesperson stated that the tool is not designed to reduce staffing at the expense of patient care, and that HCA continues to refine Timpani based on nurse feedback. The company’s top innovation executive wrote in July that the tool has substantially reduced the hours managers spend on scheduling, lowered dependence on costly contract nurses, and improved retention. He also noted that over 98 percent of generated schedules include a mix of skill levels and experience.
That framing sits in direct tension with what nurses describe. The gap is not simply a matter of perception. A former HCA data science manager filed a lawsuit in July alleging she was fired in retaliation for raising concerns about the tool. Her suit claims that HCA routinely deleted data Timpani uses to build schedules, which she argued prevented any meaningful audit of how well the system was actually performing and may have violated healthcare laws. HCA had not formally responded in court at the time of reporting.
Palantir, for its part, has described its software as a tool that presents information more usefully, and has emphasized that customers bear responsibility for data, policies, and decisions. A separate case involving Rayus Radiology, a large US imaging chain, adds another data point: after introducing a scheduling tool also built on Palantir Foundry, Rayus received complaints about errors including patient details linked to wrong profiles and tests that did not match doctors’ orders.
Why This Case Matters Beyond One Hospital Chain
The Timpani situation illustrates something that gets lost in most coverage of AI in the workplace: the difference between what a system optimizes for and what the people using it actually need. Scheduling software can be designed to minimize cost, reduce manager workload, or maximize shift coverage. Those are legitimate goals. But in a clinical environment, the quality of a schedule is not just an operational metric. It affects whether a patient in critical condition receives timely attention from someone qualified to provide it.
This is also a story about oversight, or the absence of it. When managers scheduled shifts manually, nurses report that overrides of their preferences were communicated in advance and came with explanation. Timpani operates with what nurses describe as minimal human review, and the feedback loop between the tool’s outputs and its design appears to have been limited. The lawsuit’s allegation that relevant data was deleted before it could be analyzed points to a broader problem: AI systems in high-stakes environments require robust mechanisms for auditing their own performance, and those mechanisms are not always built in from the start.
Palantir’s quarterly sales have nearly doubled to almost $2 billion, driven in part by healthcare providers eager to manage soaring costs and staffing shortages. The commercial logic is clear. The harder question, which the Timpani case puts in sharp relief, is what happens when efficiency gains at the system level translate into degraded conditions at the point of care.
In Short
Timpani is a real-world test of what it means to automate a high-stakes human process. The nurses raising concerns are not objecting to technology in principle. They are pointing to a specific failure: a system that optimizes for measurable outputs while producing outcomes that are difficult to measure until something goes wrong. In healthcare, that distinction is not abstract. It is the difference between a schedule and a care plan.
Based on reporting from Wired.