Artificial intelligence is not going to replace nurse practitioners, but it is already changing the way they provide care, and NPs need to be AI literate and play a leading role in the choice, implementation and development of clinical solutions, according to presenters at the 2026 American Association of Nurse Practitioners National Conference.
AI tools are being adopted quickly. Many NPs already are using it. Richard Ricciardi, Ph.D., of George Washington University, and DNP Stephen Ferrara, a past president of AANP, said NPs’ voices are needed within their organizations and at the federal and state levels. NPs can serve as clinical champions, workflow design leaders, educators of staff and patients, and innovators.
The important thing is for NPs to not just be end-users but to be active in deciding and guiding where AI goes from here. “Because right now we’re in the midst of it, we don’t know where it’s going to go,” Ricciardi said. “But we know it’s going to go.”
Areas where AI can save time for NPs
DNP Kaneez Odgers of Ramapo College of New Jersey said her presentation came out of her interest in the potential of AI to help nurse practitioners. When it comes to AI, some clinicians are early adopters, some may be opposed, and some are curious but scared. “I was in the ‘curious but scared,’ but then decided to just jump in and see what I could do,” Odgers said.
What she found was that the areas where AI has the most potential to help NPs include the tasks that can cause the greatest amount of burnout and moral injury. About half of NPs have reported moderate-to-high levels of burnout, and administrative burdens that add hours to the workday often are cited as a cause.
The good news is that what eats up the most time in a nurse practitioner’s day – clinical documentation, a full email inbox, wrangling prior authorization requests – are among the places where research indicates AI can help. AI’s weaknesses? Its work must always be reviewed, and accountability and responsibility fall on people, not machines.
Where AI is being used in clinical practice
Amber Vermeesch, Ph.D., of Western Carolina University and DNP Sun Jones of the University of Phoenix said AI is being used in clinical practice for clinical decision and diagnostic support, patient monitoring and prediction, and documentation and workflows.
In primary care diabetes management, for example, AI might be used to identify patients at high risk of complications or emergency department visits; to detect patterns in continuous glucose monitoring data; to provide chatbot support and health coaching; or to optimize medications. In an NP clinic, AI may help identify skin lesions that need an urgent dermatology referral, and help assess remote-monitoring data for patients with heart failure.
“The point we’re trying to make is that AI tools are available to really help us manage those patients well,” Jones said.
The importance of transparency
“I did use AI help with this presentation,” Odgers said at the beginning of her talk. “So, if you like it, thank you,” she added. (And if not) “we’re going to have to talk to Claude.”
Odgers brought up this opening disclosure later to underline the importance of transparency in clinical settings. Patients react positively when AI is broached as something that can improve their care, research has found.
Disclosing the use of AI – and documenting that disclosure in visit notes – is good for both the patient and clinical practice. Transparency is the basis for the NP-patient relationship, and that’s true with or without AI, Odgers said. In turn, patients may be better able to share their full histories when they feel informed and respected.
What AI is, what AI is not
AI, Odgers said, can be a capable assistant that recognizes patterns and creates first drafts of letters, notes, forms and instructions. It is a tool suited for repetitive tasks, a way to reclaim time that works alongside your clinical judgment.
AI is different than past technological leaps, Ferrara said. It has a persona, and it has real limitations. It’s important to understand how it works to be able to best use it.
AI cannot replace clinical judgment and should never restrict clinical judgment. It is not a decision-maker. The results of AI are not without bias. Sometimes the data it was been trained on is incomplete, out of date or not representative, and sometimes it reflects back the biases and limits of the people behind it. When AI is trained on unrepresentative or incomplete patient populations, that can undercut the effectiveness of machine learning-based clinical tools and open the door for error.
Creating better tools requires improving how AI is developed and trained. Some dermatology tools, for example, don’t work well for patients with dark skin, Vermeesch said. The Mind the Gap initiative in the UK is seeking to expand training data to help address that. But it’s up to clinicians to understand their own biases and the biases of their tools, and work to mitigate them.
In short, AI is not perfect. It will hallucinate misinformation with a confident air. Every sentence produced by AI needs review. “I always think of chat as my friend,” Odgers said. Specifically — the sort of friend who may tell you what it thinks you want to hear, even if it’s wrong.
“AI tools are very overconfident,” Jones said. “They want to tell you what you want to hear.”
Public vs. enterprise tools
Public AI platforms – think ChatGPT and Claude, among others – cannot be used for private health information and are not Health Insurance Portability and Accountability Act compliant. They are for general use, such as drafting a standard form letter. As a rule of thumb, if you create a user account for an AI option you find on the Internet, that platform is public and not suitable for patient health data or other information that’s not for public consumption. Anything you enter will stay there forever and may be used for training future models.
In contrast, clinical AI platforms are enterprise software that should operate under HIPAA-compliant business associate agreements with healthcare organizations. Clinicians should always know whether any AI platform they use operates under a business associate agreement and understand what that means for how they can use it.
How can AI help NPs?
Odgers outlined four areas where AI may make a difference in helping NPs go home sooner and spend less time on repetitive tasks:
Clinical notes. Ambient AI can provide a first draft of clinical notes and has been shown to save time. Research indicates that notes that take 20 minutes without AI can be produced in eight minutes with ambient note-taking. Those 12 minutes can add up to a couple hours or more a day.
Prior authorization. AI tools can be used to draft justification and appeal letters, generate reminders and updates, track the status of requests, and retrieve insurer criteria from clinical notes.
Clinical decision support. “The keyword there is ‘support,’” Odgers said. AI can flag drug interactions and surface relevant clinical guidelines.
Inbox management. Some AI tools can help prioritize messages and medication refills based on clinical importance, and generative AI can be used to draft forms for common responses. While every line needs to be checked, not having to type out the first draft can make a difference for NPs’ well-being.
A few questions to ask about AI tools for clinical practice
- Is a business associate agreement in place to protect patient information?
- Has the tool been validated in peer-reviewed studies?
- Does the training data match my patient population?
- Who is reviewing this before it reaches patients?
- Is there a plan for continuous evaluation and improvement?
- Can I explain it to my patients?