AI is already changing how care is delivered and offering a glimpse into where it may lead next. As its capabilities continue to evolve, Large Language Models (LLM) and multi-functional AI agents have the potential to streamline workflows, improve decision-making and expand access to care. LLM’s designs have advanced to “Chain-of-Thought” and combined mini-LLM models through fundamental steps such as research, prediction, sentence variety, and human modification. Practical frameworks were introduced to ensure careful planning was implemented, keeping AI’s transformative promise safe and ethical.
The Good: Addressing Challenges Through Innovation
- Chart Summarization and Note Generation: AI helps to reduce documentation burden, freeing up clinician time and alleviating burnout. For reference, 54% of doctors say they are burned out and 88% are moderately to severely stressed.
- AI-assisted Patient Intake Tools: AI captures relevant medical history across languages and literacy levels, potentially increasing accuracy and comfort for patients. To achieve accurate sources of “truth,” a common solution is Retrieval Augmented Generation (RAG) where models pull knowledge solely from documents you give it.
Early studies suggest that AI-generated patient communications, such as follow-up messages, may outperform clinician-written ones in clarity and empathy. As AI improves in synthesizing clinical evidence, it could enhance decision support and speed the adoption of best practices.
The Bad: Areas of Concern and Caution
- Inaccurate Outputs: There is a possibility that AI may generate incorrect information; if left unchecked, this could lead to misguided clinical decisions.
- Biases in Training Data: Learning procedures may perpetuate disparities, particularly for underrepresented groups. With clinical decision support, the judgment a provider makes for a patient may not translate into inputs to the AI mode. Without the ability to hear, see, or sense, LLM “hallucinations” increase.
- Ethical and Legal Uncertainties: AI’s growing role raises complex questions, especially when its recommendations influence clinical judgment. It’s important to note that current generation decision support is poorly accepted by physicians.
Over time, reliance on AI could lead to a lack of accountability, weakened clinical reasoning, negative patient sentiment, and disrupted traditional training. If AI substitutes human interaction, it may contribute to a deterioration of the patient-provider relationship and undermine the trust that supports it.
The Likely: What’s Coming in the Near Future
- AI agents, which combine tools, memory, and reasoning, will likely become central to routine healthcare interactions within three to five years. Agents combine models, tools, persistent memory, and interaction with people or other agents to carry out tasks.
- Virtual AI providers could expand access to care, particularly for underserved populations, with clinicians overseeing escalated cases.
- Medical training will evolve to reflect AI’s growing presence, requiring new competencies and redefining roles across specialties. As AI decision support evolves more comprehensively in its insights and more authoritatively in the way it communicates information, uptake will increase.
Implications for the Field
The adoption of AI will not be confined to technical upgrades. It will reshape workflows, redefine clinical roles, and prompt systemic change. As AI tools increasingly support core functions such as documentation, communication, and decision-making, healthcare organizations must invest in new governance structures, training models, and ethical safeguards to guide responsible implementation.
