Nurses Say Palantir Is Making a Mess of It

Artificial intelligence was supposed to take tedious administrative work off the shoulders of doctors and nurses.

Instead, some nurses say they are spending more time fixing the decisions made by the machines.

A growing controversy surrounding Palantir Technologies is bringing together several debates that normally exist separately: artificial intelligence, hospital staffing, patient safety, healthcare data, worker burnout and government surveillance.

At the center of the workplace controversy is Timpani, an AI-powered scheduling platform developed by HCA Healthcare with Palantir.

The idea sounds logical: predict how many patients a hospital will have, determine the staffing needed, consider employee preferences and let software construct an efficient schedule.

But nurses interviewed by WIRED say the reality can look very different. They described unwanted shifts, consecutive 12-hour workdays, ignored scheduling preferences, staffing imbalances and additional time spent trying to trade or appeal assignments.

And that is raising an uncomfortable question:

What happens when an algorithm designed to optimize efficiency begins affecting the humans responsible for keeping patients alive?

The Palantir-Powered Scheduler

HCA Healthcare began rolling out Timpani in 2023.

The system is now operating in more than 130 hospitals and over 1,200 nursing departments, according to HCA. It uses AI and forecasts of patient demand to help construct nursing schedules.

HCA reports significant benefits.

The company says scheduling that previously took managers roughly 8 to 15 hours per month can now take approximately two to three hours per scheduling cycle. HCA also says more than 98% of shifts produced under the system contain a mix of experience levels and that departments using it have reported lower turnover and reduced reliance on contract labor.

Those numbers make Timpani look like exactly the kind of AI application hospitals have been searching for.

But some frontline nurses are describing a much different experience.

When Optimization Meets a 12-Hour Nursing Shift

One Florida critical-care nurse told WIRED that she requested 50 specific 12-hour shifts during a four-month period.

According to her account, more than half were scheduled differently.

She said the resulting schedule sometimes placed her on three consecutive days, leaving her exhausted. Other nurses interviewed described being scheduled on requested days off, receiving inconvenient combinations of shifts and encountering teams without what they considered an adequate balance of experienced nurses.

That distinction matters.

A computer can determine that five nurses equal five nurses.

A hospital floor doesn’t necessarily work that way.

Five relatively inexperienced nurses aren’t always operationally equivalent to a team containing several veterans. A schedule can therefore look perfectly staffed on a spreadsheet while being considerably more difficult in practice.

Nurses interviewed by WIRED argued that these staffing combinations can force experienced workers to simultaneously care for their own patients while assisting less-experienced colleagues.

That turns scheduling from an HR problem into a potential patient-safety problem.

HCA Pushes Back

There is another side to the story.

HCA disputes the idea that an algorithm is simply running its hospitals’ nursing schedules without human intervention.

A company spokesperson told WIRED that nursing leaders make the final scheduling decisions—not Timpani. HCA says the objective is to have caregivers with the appropriate skills available when patients need them while still considering employee preferences.

The company also says only about 1% of requested days off are ultimately scheduled as working days under Timpani.

HCA’s own published assessment describes Timpani as a success, arguing that automation reduces administrative work, improves the initial quality of schedules and decreases dependence on expensive temporary labor.

So there are essentially two competing pictures of the same technology.

Management sees efficiency.

Some nurses see exhaustion.

And both could technically be true.

An algorithm can reduce scheduling costs and management hours while simultaneously producing schedules employees consider worse.

That’s one of the biggest challenges businesses will face as AI spreads through the workplace: efficiency and human experience aren’t always measuring the same thing.

More Than 1,000 Nurses Take the Fight to the Streets

The scheduling dispute isn’t happening in isolation.

Palantir is simultaneously facing a much larger backlash from organized nurses over its government work.

More than 1,000 nurses and supporters marched in San Francisco on September 30, targeting Palantir’s relationship with U.S. Immigration and Customs Enforcement and Customs and Border Protection. The demonstration was organized by the California Nurses Association and National Nurses United as part of the broader People vs. Palantir Global Week of Action.

National Nurses United had already organized demonstrations in eight U.S. cities in August.

The union says Palantir manages systems associated with more than one-fifth of U.S. hospital beds and has called on hospitals, universities, pension funds and other institutions to sever relationships with the company.

The protests connect healthcare technology to something far larger than scheduling.

Nurses are objecting to Palantir’s work with immigration authorities while simultaneously questioning how much influence its technology should have inside hospitals.

Palantir rejects many of the protesters’ claims.

The company told Becker’s Hospital Review that it is a software company rather than a data company and said it does not collect or store data itself, conduct surveillance or determine immigration-enforcement policy. Palantir has also pointed to healthcare applications it says help hospitals place patients into beds faster and improve outcomes.

The Controversy Has Crossed the Atlantic

Palantir’s healthcare debate isn’t limited to the United States.

The company is also deeply involved with Britain’s National Health Service through a £330 million contract for the NHS Federated Data Platform.

Healthcare workers and campaigners have raised concerns about privacy, ethics and patient trust.

At a recent North Yorkshire health scrutiny meeting, nurse practitioner Naomi Wigley warned that distrust surrounding the platform could have consequences extending beyond technology.

Her argument was straightforward: if people stop trusting how healthcare systems handle their information, some may stop sharing information—or potentially avoid care altogether.

The local NHS board stressed that its system currently does not contain identifiable patient-level primary-care information and said future uses would undergo legal, information-governance and data-protection reviews.

Nevertheless, opposition is growing.

More than 44,000 people have filed legal objections concerning NHS England’s use of the Palantir-operated platform.

And another significant development arrived this week: a Financial Times analysis found that more than 20 NHS trusts had stopped using Palantir waiting-list tools, raising questions about previous claims surrounding the platform’s effectiveness.

The Real AI Problem May Be the Objective

The most interesting part of the nursing controversy isn’t whether AI can create schedules.

Clearly it can.

The bigger question is:

What exactly has the algorithm been told to optimize?

A hospital could optimize for labor costs.

It could optimize for coverage.

It could optimize for minimizing overtime.

It could optimize for employee preferences.

It could optimize for experience levels.

It could optimize for patient demand.

But maximizing one variable can negatively affect another.

The cheapest schedule isn’t necessarily the safest.

The mathematically cleanest schedule isn’t necessarily the one that produces the least burnout.

And the schedule requiring the fewest contract workers isn’t automatically the schedule nurses want to work.

That is where human judgment becomes critical.

AI Should Be the Assistant—not the Charge Nurse

There is enormous potential for artificial intelligence in healthcare.

AI could identify staffing shortages days earlier, predict emergency-room demand, reduce paperwork, detect dangerous scheduling patterns and give nurse managers better information.

But the Palantir controversy illustrates the danger of crossing the line between decision support and decision replacement.

The strongest model may ultimately be relatively simple:

AI builds the first schedule. Humans build the final one.

Algorithms should identify problems—not create new ones.

Nurses should have transparent ways to override automated decisions.

Hospitals should measure fatigue and employee satisfaction alongside cost savings.

And when a scheduling algorithm affects patient care, hospitals should be able to explain how that decision was produced.

BuckBully Bottom Line

Palantir’s healthcare controversy represents a much bigger story about the next phase of artificial intelligence.

We spent the first AI boom asking:

What can AI do?

The next phase will force companies to answer:

What should AI be allowed to decide?

Scheduling nurses might sound like routine administrative work.

But those schedules determine who is standing beside a hospital bed at 3 A.M., how rested that person is, how experienced the team around them is and whether enough people are available when something goes wrong.

That’s not just optimization.

That’s healthcare.

AI can crunch the numbers.

But when people’s health—and potentially their lives—sit on the other side of those numbers, human judgment still needs the final shift.