Nursing informatics leaders say the most meaningful patient safety improvements tied to AI in nursing workflows so far have come from mature, predictable decision-support tools — while more experimental applications, including generative AI, remain largely unproven at the bedside.
Marc Benoy, BSN, RN, chief nursing information officer at Summa Health in Akron, Ohio, first cautioned that the term “AI” is often applied too broadly, obscuring critical differences between traditional predictive analytics, embedded machine-learning models and generative AI — each with distinct risk profiles, governance needs and levels of clinical maturity.
At his organization, generative AI is not currently operationalized in bedside nursing workflows. Any measurable safety gains have instead come from established decision-support tools and predictive risk scoring embedded in the electronic health record.
“When implemented well, they can support safer care by reinforcing consistency, reducing variation and nudging standardized actions in safety-sensitive workflows,” Mr. Benoy said, emphasizing that such tools remain supplements to, not replacements for, clinical judgment.
Because these systems behave predictably, he said, they can be validated, monitored and governed over time — a key requirement in evidence-based nursing practice. By contrast, he warned that opaque or poorly understood AI tools can unintentionally shift cognitive burden back onto nurses, introducing new safety risks rather than reducing them.
He also pointed to operational constraints, noting that successful implementation requires staffing, informatics capacity, capital investment and sustained governance — resources that many health systems lack, particularly when returns on newer AI initiatives remain uncertain.
Colleen Mallozzi, BSN, RN, senior vice president and CNIO at Jefferson Health in Philadelphia, said early safety gains have emerged in shift handoffs — a critical transition point in inpatient care.
“Generative AI-supported handoff tools are improving consistency around critical safety elements such as falls risk, lines and tubes, wounds, and high-risk medications, while reducing noise and variability in reports,” she said.
Rather than making clinical decisions, Ms. Mallozzi noted, these tools help nurses maintain shared situational awareness so that essential safety information is less likely to be lost during shift changes.
Both leaders emphasized that safety gains depend less on the sophistication of algorithms and more on how well tools are integrated into real-world nursing workflows.
“Safety does not come from models; it comes from workflows,” Mr. Benoy said, adding that without strong training, adoption strategies and monitoring, new tools risk becoming another source of alarm fatigue instead of a meaningful safety intervention.
For now, the clearest patient safety improvements stem from technologies that are transparent, validated and operationally supported — while broader AI-driven transformation in nursing remains a longer-term promise rather than a present-day reality.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
