The EHR Era Pushed Nurses from the Bedside. Will AI Do It Again?

Editor’s note: The author is Chief Nursing Officer at EBSCO and Editor in Chief of Dynamic Health. EBSCO’s Dynamic Health, DynaMed, and DynaMedex include AI features (Dyna AI Mode). Nurse.org received no compensation for this article.

Health systems continue to pour dollars into AI investment to help reduce administrative burden, ease burnout and address persistent nursing shortages. The promise to nurses is certainly compelling: less time spent documenting and fewer repetitive tasks leads to more time for patient care.

In practice, AI tools are introduced without accounting for how nurses actually work. Even on existing technology, a recent Carrington College nursing survey found that nearly 60% of nurses don’t feel fully trained on the workplace technology their jobs demand. The technology intended to support the workforce could ultimately push more experienced nurses away from the bedside, and healthcare has seen this pattern before.

When electronic health records (EHRs) were first deployed, systems that promised massive efficiency gains often introduced more rigid workflows, duplicative documentation and new administrative demands nurses bore the burden of. For some experienced nurses, the transition became the final straw that prompted them to retire or leave bedside practice. In a 2024 KLAS Research survey of nurses, 40% of those who cited EHRs as a burnout factor said they were likely to leave their organization within two years. As AI moves deeper into clinical care, hospitals risk repeating that history unless they treat adoption as a clinical redesign effort rather than another software rollout.

Want to see more Nurse.org articles in your Google results? Add us as a preferred source.

Electronic health records brought meaningful improvements to healthcare, but many early systems were designed around billing and compliance rather than the realities of clinical practice. Instead of improving inefficient processes before digitizing them, health systems often automated the processes they already had. Alerts, hard stops and new documentation requirements were layered into already complex workflows, requiring clinicians to complete more steps while creating the perception that the system trusted their judgment less. A mixed-methods study of acute and critical care nurses, published in the Journal of the American Medical Informatics Association (JAMIA), found that data redundancy, cumbersome entry and poor workflow navigation contributed to documentation burden. The lesson? Automating a process does not improve it when the underlying process remains misaligned with clinical work.

AI begins from a more promising position. Many emerging tools are being developed around narrower, purpose-driven use cases. They can give clinicians something useful back, whether that is a draft note, a synthesized answer or relief from a repetitive task. AI does not have to become one enormous system attempting to automate every aspect of care at once.

Yet that advantage will disappear if health systems layer disconnected AI tools onto an already fragmented technology environment. A new tool may perform its individual function well and still add burden if it requires another login, another device, another workflow or another output for a nurse to verify. The industry cannot evaluate these technologies in isolation from the full clinical environment in which they will operate.

Ambient AI demonstrates why discipline-specific design matters. These tools have largely gained traction in physician workflows, where information required for documentation emerges through a verbal exchange with the patient. The physician asks questions, the patient responds and the conversation becomes the foundation of the clinical note. Physicians then add their impressions and treatment plan notes to complete the patient encounter. Historically, physicians utilized dictation to ease the clinical note burden, and now ambient listening AI is a natural extension of dictation with integration into the electronic health record (EHR).

Nursing scope of practice is uniquely different. A care encounter is often 8-12 hours of conversation with the patient, with initial and ongoing assessment throughout a nurse’s shift in addition to care interventions and patient/family education. Much of a nurse’s assessment happens silently through observation, touch and continuous clinical reasoning. Nurses do not typically narrate every step of a head-to-toe assessment by announcing that they are checking a patient’s pupils, cognition, skin, breathing or heart sounds. They observe, listen, compare findings, interpret subtle changes and continuously reprioritize care.

This is the difference between workflow and nursing thought flow. Workflow captures the visible tasks a nurse completes and the sequence in which they occur. Thought flow is the critical reasoning that connects those tasks (like noticing a change in condition, interpreting it in context, determining whether it alters the plan of care and deciding when escalation is necessary). An ambient tool may appear to fit the workflow while still interrupting the thought flow.

If the technology requires nurses to verbalize observations they have historically processed silently, using it changes the bedside interaction itself. The tool may eventually save time, but the transition can initially create new work through narration, correction and verification.

The early experience of nurses with ambient AI therefore may not mirror the results seen among physicians. That does not necessarily mean the technology has failed. It just means developers and health systems have underestimated how significantly nurses must change their behavior to make it work. Physician ambient listening AI tools have been designed with the physician workflow and thought flow in mind. Nurses can learn from their physician colleagues on how to advocate for tools that support their full scope of accountabilities. Nurses will benefit most from and adopt AI technologies if they are designed with clear intention, similar to how physician AI tools have been developed.

>>Listen to The Latest Nurse News Podcast

Every major technology transition also creates an adoption threshold, and much like during EHR implementation, some nurses will decide that learning another system is one transition too many to handle. The potential loss of experienced nurses should be especially concerning. These clinicians carry the pattern recognition and judgment needed to identify when an AI-generated note, recommendation or alert does not fit the patient. If poor implementation pushes them away, health systems will lose the very expertise needed to oversee the new technology safely.

This tension is often framed as nurses fearing that AI will replace them. The more immediate concern is responsibility without authority. Nurses may be required to use tools they did not select, test or help design while remaining professionally accountable for identifying inaccurate, irrelevant or unsafe outputs. Adoption will remain fragile when the professional carrying the clinical risk has little influence over the technology entering their practice.

The challenge is also larger than any single tool. Nurses already work across numerous systems, devices and platforms during a shift. Even a useful AI application can become one more source of cognitive load if it does not integrate with existing technology and organizational policy. The question is not simply whether an AI tool works. It is whether the tool improves the total experience of delivering care.

Health systems need a more rigorous definition of success. A pilot should not be considered effective solely because notes are generated faster or usage rates are high. Leaders must examine whether nurses spend less time correcting outputs, whether documentation becomes more clinically meaningful, whether after-shift work decreases and whether the technology creates new administrative or cognitive work elsewhere.

Nursing expertise must also be embedded at the beginning of product development and purchasing decisions, not requested after a tool is largely built. That requires both nursing informatics leaders and direct-care nurses who can evaluate how the technology performs during a real shift, when priorities change, interruptions are constant and multiple systems already compete for attention.

Most importantly, decision-making authority must reflect clinical accountability. Nurses need clear mechanisms to question, override and report unsafe or irrelevant results. Frontline users should help establish desired outcomes before a pilot begins and should have a meaningful voice in determining whether the technology scales. Choosing not to expand a tool that creates burden is not a failure of innovation; it is evidence of responsible governance.

Nurses are not broadly resistant to technology. They have spent decades adapting to new devices, documentation requirements and models of care. What nurses resist is technology that makes it harder to see the patient, exercise clinical judgment or complete essential work safely. The EHR era showed healthcare what happens when technology reaches the bedside without enough attention to the people expected to use it. AI offers the industry a chance to apply that lesson before another generation of nurses bears the cost. Success will depend on understanding both nursing workflow and thought flow, testing technology in the reality of bedside care and giving nurses the authority to shape what comes next.

🤔 Nurses: what’s one tool at your facility, AI or otherwise, that was supposed to save you time but didn’t? Share your experience in the comments.