It’s a recurrent issue we have encountered over 25 years of consulting: excessive administrative tasks significantly hinder healthcare professionals from delivering an optimal patient experience. When asked about their primary wish, doctors consistently express the desire to "eliminate administration and liberate time for genuine caregiving duties." The numbers bear that out. One survey of US clinicians found that 36% reported spending more than half their time on administration, with a staggering 92% saying administration is a major contributor to burnout. Those numbers underscore just how deeply and widely the administrative burden is felt and stress the critical need for streamlined solutions in our pursuit of transformative healthcare practices. AI has the power to grant that wish.

and yet...

In order to facilitate this, digital upskilling is necessary. But is it doable? Healthcare in particular suffers from a shortage of tech-savvy professionals. How healthcare providers would use all that additional time is also untested. What if instead of spending more time with patients by outsourcing administrative duties, they instead scale their business? Would that help solve our shortage of healthcare workers but just substitute one type of burnout for another? Could doctors see their connection to medicine fray by losing the time-consuming yet valuable step of typing up patient notes?

AI holds the promise of revolutionizing healthcare by enabling precision medicine tailored to individual patient needs. Advanced algorithms can analyze vast datasets, unlocking personalized treatment plans that consider genetic, environmental, and lifestyle factors. This targeted approach has the potential to enhance treatment efficacy while minimizing adverse effects and could be particularly groundbreaking when it comes to how we treat cancer, whose heterogeneity means that patients who have the same tumor and undergo identical treatment plans can end up with divergent results.

and yet...

The cost of implementation might be prohibitively expensive, creating a major roadblock for societies with statutory health care and jeopardizing access for those with privately funded healthcare. Could AI widen the gap between the haves and the have-nots? AI algorithms are only as strong as the data on which they rely, and that introduces new potential pitfalls around data privacy and security, informed consent, and the accuracy of medical records. Could AI algorithms built using today’s data further entrench the biases that already exist in healthcare?

The abundance of readily available medical information has surged. Understanding blood oxygenation levels, assessing stroke or breast cancer risk, and learning about preventive measures are within easy reach. AI will make it exponentially more so. This marks a significant shift from the traditional reactive healthcare model towards a proactive approach to prevention. It represents a valuable tool long sought by healthcare professionals, one that could provide individuals with immediate feedback and reinforcement for positive behavioral changes

and yet...

Who will contextualize this wealth of data? Where is the “human in the loop” who can address the nuanced determinations or recommendations that AI generates? And how can we ensure that the information people are accessing is based on sound and thorough data? It’s a point that was emphasized by Dr. David Newman-Toker, M.D., Ph.D., in a congressional hearing on AI in healthcare. Utilizing the wrong data —"Garbage in, garbage out"— can perpetuate biases, while relying on inadequate data — "Go where the light is best" — can foster a skewed perspective.

Machine learning algorithms excel at discerning subtle patterns that may signify conditions like cancer or cardiovascular diseases, enabling timely interventions and enhancing patient outcomes. However, AI's proficiency in pattern recognition could extend beyond mere detection; it could act as a sort of medical crystal ball, providing valuable insights into future developments and the likelihood of developing specific conditions. Indeed, one recent study found Alzheimer’s could be detected seven years before it emerged using AI. AI's potential to forecast diseases could allow us to prevent ailments from progressing to untreatable stages and empower individuals to modify their behaviors proactively.

and yet...

In this new world, would the concept of being "healthy" even be possible? In a realm where one is either diagnosed or not diagnosed enough, are we all staring down a disease-ridden future? What happens to those people who lack the ability, finances, or motivation to act on the information, or those who are anticipating a disease for which no treatment yet exists? Could insurance providers discriminate based on some algorithm, with no recourse for patients who are denied coverage?

‍

‍

Conclusion‍

Dualities tend to live within any disruptive innovation, and Generative AI in healthcare is no exception. As we do the work of bringing its full potential to life, it’s imperative that we consider the practical and ethical implications at the same time. The hopes are plentiful, as are the fears, and the dialogue they produce is one we’ll use to guide our work. What hopes and fears would you write on a Post-it when it comes to AI in healthcare?