HCA and Palantir's AI Nurse Scheduler Reaches 130 Hospitals. Nurses Say It's Compromising Patient Care.

On a recent critical-care night in Florida, Amber Retzloff counted the experienced nurses on her floor and realized she was the only one. The AI scheduler had given her four junior colleagues; she spent the shift deciding which of her sickest patients she could reach in time.

Retzloff has worked at the bedside for ten years and leads the local chapter of National Nurses United. She is not anti-AI. She is one of five named nurses who told WIRED the same thing: that the tool HCA Healthcare built with Palantir is producing understaffed, exhausted shifts in the hospitals where it runs. The story is a test case for whether AI in critical care can be adopted without eroding the frontline agency that keeps the work safe, and the answer so far is visible in the Sunday coverage gaps, the appeal logs, and a federal complaint now on the public docket.

HCA Healthcare is the largest for-profit hospital chain in the United States, with 190 facilities. The tool at issue is Timpani, an in-house nurse-scheduling system that HCA co-developed with Palantir and has rolled out to roughly 130 of those 190 hospitals since 2023. The company calls the deployment a workforce-modernization program. The nurses on the floor call it something else.

Over a four-month window, Retzloff says, the tool reassigned her to a different shift on more than half of the 50 specific 12-hour shifts she had requested, several of them stacked back-to-back-to-back. The pattern, she says, is not random. The scheduler optimizes for fill rate, getting a body into every slot, and treats seniority, specialty certification, and fatigue as constraints to be solved around rather than as hard floors. The result, on her unit, is a critical care floor where the only experienced clinician is the one who drew the short straw.

Four other nurses at separate HCA hospitals told WIRED they see the same pattern: thin senior coverage, junior-heavy skill mix, and Sunday shifts that reliably come up short. The complaint they describe is not that Timpani makes scheduling more algorithmic. It is that the tool treats nurse preference and patient acuity as soft variables while it treats headcount as the only hard one. Workers have also posted similar gripes on Reddit.

The nurses' allegations are now in court. Russell v. HCA Healthcare, Case 3:26-cv-00983, filed in the Middle District of Tennessee in 2026, puts the deployment on the public docket. The complaint alleges that the AI scheduler has produced understaffed units, unsafe skill mixes, and shifts that violate state nurse-to-patient ratios. HCA has not commented on the specific claims in the filing.

The company has, however, published its own framing. In a July 2026 strategic memo on the HCA Healthcare Today blog, chief medical officer Michael Schlosser describes the AI program as a workforce-modernization push to route more clinical hours to the bedside and reduce administrative toil. That is the version of the story Palantir's Foundry platform is sold on: better data, fewer handoffs, less paperwork. The nurses' account is the same product seen from the other side of the screen, where workforce modernization can mean a third consecutive 12-hour shift on a unit you do not normally staff.

The constructive question is not whether AI can schedule nurses. It obviously can. The question is what guardrails a deployment in critical care has to carry, and whether the vendor and the operator are willing to encode them as hard constraints rather than as preferences. The override test is concrete: when a senior nurse flags a skill-mix problem on a critical care shift, does the system respect the override, or does it log the override and reschedule around it on the next cycle? Retzloff says the answer today is the second. The appeal logs, the back-to-back runs, and the Sunday gaps are the visible evidence that the override is being treated as noise.

There is a version of this rollout that goes well. A scheduler that treats seniority, specialty, and fatigue as constraints with teeth, that logs overrides, audits them, and feeds them back into the model, would be a genuine improvement. A scheduler that treats them as preferences to be optimized around is not a workforce tool. It is a cost tool that wears scrubs.

The HCA deployment is now the largest at-scale test of that question in US healthcare. The federal complaint, the on-the-record nurses, and HCA's own AI memo are all on the public record. The next move belongs to the company: publish the override rate, publish the skill-mix data, and tell the five named nurses why the deployment is safe in their units. So far, it has not.