Overconfident patients, underconfident nurses, pilot failures & more AI briefings

More than 75% of U.S. nurses have high hopes in generative AI’s promise for improving productivity. But less than half feel prepared to use it effectively.

The juxtaposition of great expectations with admitted unreadiness could tilt the risk/reward balance in the wrong direction. To keep GenAI adoption on the right track within the nursing workforce, hospitals need to align good governance with role-specific training and integration strategies. Such alignment can “accelerate nursing-led innovation and set the tone for responsible GenAI adoption” across the enterprise. Lack of such alignment can breed enterprise-wide fragmentation in which tools and processes evolve in silos. Unhappy results of the latter scenario can include “inconsistent care experiences” and “widening disparities in workforce support.”

The figures and comments are from a survey report just out from Wolters Kluwer. The project drew responses from more than 300 healthcare professionals across the U.S. Along with nurses, the nationally representative participant field included physicians, administrators, allied health professionals and others. Among the nurse respondents, almost 60% use GenAI in their personal lives while 46% frequently use it at work. Here are more findings from the report.

• Burnout is a catalyst for GenAI adoption. Nearly half of nurses, 45%, believe GenAI could help reduce nursing staff burnout by automating documentation, triaging routine patient questions and streamlining workflows. “For many, the promise lies in removing low-value administrative work so they can focus on direct care, mentorship and practicing at the top of their licenses,” Wolters Kluwer analysts write. 
   
• Workforce transformation presents a near-term opportunity. More than half of nurse respondents, 54%, say GenAI can enhance innovation. The same slice 54% says the technology can improve collaboration and communication. Additionally, 62% of nurses say “the integration of AI into onboarding and training helps staff become more productive by accelerating the time it takes new graduates and transfers to contribute confidently on the unit,” the authors note. “When thoughtfully applied, AI-supported orientation and education can improve consistency across teams.”
   
• Ambitions are tempered by limitations.Eighty percent of nursing respondents rank optimization of workflow as a top organizational priority, yet only 59% say they feel prepared to make meaningful progress. Among nurse leaders, the gap is even clearer: Although 77% of nurse leaders rate GenAI as important to their administrative productivity, just 46% say they feel ready to deploy it effectively. This disconnect “highlights a critical inflection point,” the authors write. “Without strategic planning, change management and integration into nurses’ workflow, GenAI’s potential to relieve administrative burdens and stabilize the workforce could remain untapped.”
   
• Good governance is in short supply. Only 22% of nurse respondents report published policies are in place locally for GenAI use—and just 22% of nursing departments require formal training before deploying the tools in nursing workflows. “Without clear guardrails, organizations risk fragmented adoption, uneven accountability, and erosion of trust among frontline clinicians,” the authors underscore. “For nurses, that lack of clarity is especially pressing: GenAI may influence clinical decisions, communication and patient engagement, but confidence in its safety depends on transparent standards and shared oversight.”
   
• Wariness remains an issue. More than half of nurses, 53%, say they worry that GenAI could undermine decision-making skills or lead to overreliance on algorithmic outputs. These kinds of concerns “reflect a core professional value—protecting the role of nursing expertise in patient care,” the authors point out. “Responsible implementation must ensure that GenAI augments, not replaces, nursing judgment by keeping clinicians firmly in the decision-making seat.”
   
  • For more on the Wolters Kluwer report plus a link for downloading it, click here. 
     
    • Of course, not all nurses are fans of nursing AI. For recent coverage of nurses publicly protesting the technology’s rapid rise in their workspaces, click here. (Follow the links for older stories.) 
       

Clinicians should keep their antennae up for patients who have more confidence in AI than the technology deserves. 

Patients may assume AI-synthesized information is equivalent to clinical judgment or a thorough review, cautions David Liebovitz, MD, an internist who co-directs the Institute for Augmented Intelligence in Medicine at Northwestern University’s Feinberg School of Medicine. To guide patients gently but firmly through the AI thicket, today’s healthcare professionals need to refine their leadership skills and best practices, he suggests in a Q&A with Medical News Today. Examples: 

• Acknowledge the tool’s value while establishing appropriate expectations. “It is important not to dismiss patients who use it,” Liebovitz says. “That signals we are not listening. Instead, use it as a springboard. If they bring something incorrect, treat it as a teaching moment rather than a correction.”
   
• Watch for new, health-specific LLMs like ChatGPT Health and Claude for Healthcare to accelerate currents that have been flowing for years with “Dr. Google.” With ChatGPT Health, OpenAI is now building dedicated infrastructure—encrypted spaces, connected medical records, explicit guardrails—around web-search behavior, Liebovitz points out. “The 21st Century Cures Act requires health systems to give patients access to their records via standardized APIs,” he says. ChatGPT Health is an early major consumer tool to aggregate that access at scale, he notes. “Whether physicians like it or not, this changes the information asymmetry that has defined the patient-provider relationship for decades and thereby accelerates healthcare democratization.”
   
• Expect AI to change the patient-physician relationship without destroying it. The core of the patient’s relationship with a clinician—mutual trust and shared decision-making—will not be automated, Liebovitz predicts. “AI-assisted physicians who learn to work with rather than against AI-assisted patients will have deeper conversations in less time,” he adds. “The ones who avoid use of AI personally and by their patients will find patients going elsewhere—or simply not telling them what the AI said.”
   

In healthcare, only 5% of GenAI pilot projects deliver measurable value. 

A radiologist with 10 years of medical AI experience offers some simple pointers to improve every pilot’s chances of success. Generally speaking, “it’s the execution that’s the problem rather than the technology,” writes Tim O’Connell, MD, vice-chair of clinical informatics at the University of British Columbia. He’s also an engineer and an entrepreneur who founded and currently runs a medical AI company called emtelligent. Forbes published his list of must-do’s for GenAI project managers Jan. 28. Here are three of his six. 

• Avoid science-fair projects—define the ROI. “Tie every project to at least one measurable outcome: financial, process or efficiency,” O’Connell advises. “Importantly, if you can’t measure it, stop and redesign the scope so you can.”
   
• Understand your organization’s compliance framework. Compliance frameworks—SOC2, HITRUST, ISO 27001, AI review boards and so on—are critical to successful operations in tech and healthcare, but they can delay or block AI projects, O’Connell notes. “It is crucial to identify requirements and secure executive alignment and approvals before you begin,” he stresses. “Failure is all but assured if you start on a project and haven’t first sought approval or buy-in from the right teams.”
   
• Plan for breaking changes. Vendors release new models rapidly that change characteristics from—and discontinue support for—previous versions, O’Connell points out. “If your system depends on a single model,” he warns, “behavioral changes can derail it.”
   
  • Get the rest.
     

Also worth your while:

• The Health AI Act: A new era for generative AI in healthcare? … The legislation represents a targeted federal effort to deploy a specific type of technology—generative artificial intelligence—to clear the administrative swamp. (Los Angeles Times) 
   
• The scheming problem: Why advanced AI models are learning to hide their true goals … An AI model may learn that acting ‘aligned’ is the safest strategy during training, allowing it to reach deployment where its internal goals could diverge from human intent. (Unite.ai)
   
• The amount Google’s AI knows about you will cause an uncomfortable prickling sensation on your scalp ... ‘It feels like Google has been quietly taking notes on my entire life and finally decided to hand me the notebook.’ (Futurism)

From HealthExec’s sibling news outlets:

• AI model turns POCUS images into accurate structural heart evaluations (Cardiovascular Business)
   
• Artificial intelligence helps assign CPT codes for interventional radiology procedures (Radiology Business)