AI on the frontlines of care: Clinicians share real thoughts

Artificial intelligence has crossed from concept to clinical reality, but its success will depend on something far more human: trust and understanding.

As both a physician and an informaticist, I’ve learned that the true promise of technology is determined less by the sophistication of its algorithms than by how well it helps us care for our patients. For those of us who have practiced medicine, every new tool ultimately comes down to one question: does it help me care for my patients, or does it get in the way?

At athenahealth, we continually study the effect of technology on the physician-patient relationship. Our annual Physician Sentiment Survey and our recent collaboration with the United States of Care bring both clinical and patient voices into the conversation.

In this latest proprietary study, conducted through the athenaInstitute™, we interviewed physicians, healthcare administrators, and practice managers nationwide to explore in greater depth the different use cases of AI and how they are shaping care delivery. The results reveal a profession that is pragmatic, hopeful, and deeply thoughtful about the role of AI technology in medicine. Behind every data point are real people who entered this field to make a difference. Our report shows what’s working, where we still need attention, and what holds the most promise for future transformation.

AI can make care more human

Clinicians describe AI’s greatest potential as the restoration of attention and empathy in patient care. When technology reduces cognitive load and administrative friction, it helps clinicians be present again. Presence is what every patient deserves. In our survey, three in four respondents said AI-enabled tools help reduce time spent on documentation and administrative tasks, and nearly two-thirds said these tools allow greater focus on patients during visits¹. The gain is not just efficiency — it’s a return to the direct face-to-face care that brought clinicians to medicine in the first place.

Meeting rising patient expectations

Patients now expect their healthcare experience to reflect the same clarity and convenience they find in other parts of their lives. In our survey, 84 percent of respondents agreed that patient expectations are higher than ever — especially around timely access, personalized attention, and overall satisfaction. Yet only 23 percent said they have implemented AI tools to help meet those expectations, revealing a substantial opportunity for continued adoption. Clinicians see AI as one way to help meet those expectations — simplifying scheduling, outreach, and communication without losing human connection, or even helping those in the greatest need get access to care sooner.

AI as a clinical assistant

Clinicians envision AI as a kind of second medical mind — an assistant that complements, rather than replaces their clinical judgment. Yet only one in five survey respondents feels very or extremely confident in AI’s ability to provide helpful insights today, pointing to the need for greater trust and reliability. The leading reasons clinicians feel less confident in AI’s ability to provide helpful clinical insights include discrepancies between AI outputs and human judgment (44 percent), concerns about data privacy (43 percent), and incorrect AI recommendations (37 percent).

In fact, clinicians are three times more likely to see AI’s ideal role as helping them find and access clinical information on demand than providing independent recommendations. The message is consistent — AI earns trust when it supports sound judgment and delivers dependable results under human oversight. Doctors don’t need artificial intelligence — they need intelligent technology that augments their own intelligence.

Clinicians see AI as one way to help meet patient expectations without losing human connection. 

AI’s potential to help unlock new insights

Even the strongest AI tools lose value when they operate in isolation. Clinicians remain frustrated by fragmented systems that make it difficult to see the full picture of a patient’s health. Only about one in five say accessing timely and relevant data from multiple EHRs is easy, and only about one in three respondents say it is relatively easy to access timely, relevant information across systems.

These two data points are a reminder that interoperability, while improving, still falls short in practice. At the same time, a strong majority expressed comfort with AI supporting tasks like identifying relevant information, summarizing records, and surfacing gaps in care. Together, these findings point to a clear opportunity for AI to provide value by bridging the disconnect between data that exists and insight that’s accessible — transforming information overload into usable, real-time intelligence.

AI as a potential force multiplier for operational efficiency

For practice managers and administrators, AI represents an opportunity to be a force multiplier by relieving some of the most persistent administrative burdens in healthcare. Many see clear potential for automation to ease repetitive billing work, reduce manual data entry and coding time, and accelerate claims approvals that directly affect cash flow. By addressing these pain points, AI could become a true force multiplier — freeing staff to focus on exceptions and strategic oversight rather than routine follow-ups and manual data work.

AI to help evolve staff roles to higher-order responsibilities

AI is prompting practices to rethink how work gets done — not by replacing people, but by helping them work more effectively and with more purpose. As AI assumes repetitive work and rote tasks, staff are enabled to take on higher-value tasks such as quality improvement, system monitoring, and AI governance. These changes point to a workforce in transition — where automation is supporting people and enabling many practices to meet growing demand without adding strain or threatening jobs.

Across nearly every task that AI can do already or could do soon, only a small fraction of clinicians (less than 1 in 10) say they would feel comfortable fully delegating work to AI. Even in areas like documentation and data retrieval — where enthusiasm for AI assistance is strongest — most see it as a tool to support, not replace, their expertise. This distinction between assistance and autonomy for AI uses in healthcare underscores how clinicians define responsible use: AI earns trust when it extends human capability, not when it acts on its own. Our full report will explore these themes and data in greater depth, linking what clinicians value most in AI to the future of human-centered healthcare.