Duke Launches AI Training for Nurses Focused on Safety and Clinical Judgment

By Suad Seferi · Aug 25, 2026

Nurses are increasingly encountering artificial intelligence in hospitals and clinics, often without having been formally trained to understand what those systems are doing. Duke University School of Nursing is trying to address that gap with a new three-course programme focused specifically on the safe and responsible use of AI in clinical practice. The programme, called AI in Nursing Practice: Foundations for Quality and Safety, was announced on August 24 and is available through Coursera. It is designed for nurses and other healthcare professionals rather than programmers or data scientists. No technical or coding background is required. The emphasis is not on teaching nurses how to automate their work. It is on helping them understand when an AI recommendation makes sense, when it may be unreliable and when professional judgement should take priority. AI is already entering everyday clinical work Healthcare AI is no longer limited to experimental research projects. Terry McDonnell - Senior Vice President and Chief Nurse Executive Hospitals are using algorithms for tasks such as predicting patient risk, prioritising cases, generating clinical alerts, assisting with documentation and helping staff identify patients who may need faster attention. That means nurses can interact with AI-generated information even when they are not the people who selected or built the system. Duke says many clinicians have received little formal education about how such tools work, what their limitations are or how bias can enter an AI-generated recommendation. The new programme is intended to give nurses enough AI literacy to assess those tools critically rather than simply accepting whatever appears on a screen. Participants begin by learning how AI systems appear in clinical environments. Later courses move into evaluating AI-generated outputs, recognising possible bias and using AI-informed information in patient-care scenarios. Case studies cover acute care, outpatient treatment and telehealth. The full specialization can be completed in around one month, with each course requiring approximately three to four hours of study. The nurse still makes the decision One principle runs throughout Duke's programme: AI can provide information, but clinical responsibility stays with healthcare professionals. Dr Michael P. Cary Jr., who leads the project, said nurses need to understand both the strengths and limitations of the technology while keeping professional nursing judgement at the centre of patient care. That distinction matters because medical AI can produce outputs that look authoritative even when the underlying data is incomplete, biased or poorly matched to the patient in front of a clinician. A risk score, automated alert or AI-generated recommendation may help a nurse notice something important. It can also be wrong. The practical skill therefore becomes less about learning to operate an AI tool and more about knowing what questions to ask when one is already part of the workflow. Where did the recommendation come from? What information was used? Could the system work differently for different groups of patients? Does the result match what the nurse is seeing clinically? And when should the recommendation simply be ignored? From AI adoption to AI literacy Duke began developing the training earlier this year as part of a continuing nursing education initiative supported by Johnson & Johnson. When that project was announced in June, the university said its goal was to prepare practising nurses to integrate AI into patient care ethically and effectively. The newly launched Coursera specialization turns that work into a programme that can now be taken outside Duke itself. The approach reflects a broader problem emerging as AI enters workplaces. Organisations can deploy new technology considerably faster than they can prepare the people expected to use it. Healthcare makes that gap particularly serious because mistakes do not only affect productivity. They can affect decisions about real patients. For nurses, understanding AI therefore becomes part of understanding the clinical environment around them. A useful model beyond Duke The same issue will eventually reach healthcare systems across Europe and the Balkans. Hospitals do not need every nurse to become an AI specialist. But as clinical software begins incorporating more prediction models, automated documentation and decision-support systems, staff need to understand what those systems can and cannot tell them. Training also needs to arrive before technology becomes routine, rather than after problems appear. Duke's programme is relatively modest: three short courses rather than another promise to reinvent medicine. That may be precisely what makes it useful. AI is already finding its way into healthcare. The next question is whether the people working closest to patients are being prepared to challenge it when necessary.