Despite adoption hurdles, healthcare is all-in on Generative AI

Generative AI (GenAI) is having a renaissance, but few industries are experiencing this like healthcare. As early adopters, everything from hospital operations and administrative duties, to clinical trials and drug discovery are being impacted by the technology. Despite the rosy outlook,  it doesn’t paint the full picture of GenAI in healthcare. The 2024 Generative AI in Healthcare Survey, however, does a better job at that.

Highlighting both the successes and challenges of GenAI in healthcare, the survey uncovers some interesting trends. As a leader in the field, the results could be indicative of bigger shifts to come as it relates to AI use in the enterprise. This article will explore the main findings.

1) GenAI budgets are growing exponentially

Adoption of GenAI varies significantly across roles and company sizes. Technical leaders are at the forefront, demonstrating higher adoption rates and driving budget increases. While 34% of all respondents reported a 10-50% budget increase for GenAI, 22% witnessed a 50-100% rise. Among technical leaders, nearly one-fifth reported a staggering 300% budget growth, underscoring their advocacy and investment in GenAI.

Large companies, with their abundant resources, are leading the charge in evaluating GenAI use cases. Medium-sized companies are actively experimenting with and developing AI models, while small companies, often constrained by resources, show the highest percentage not actively considering GenAI.

2) Task-specific language models reign supreme

The healthcare sector’s unique needs are driving a preference for custom-built, task-specific language models. Healthcare-specific small models are utilized by 36% of respondents, reflecting a clear trend towards targeted solutions over general-purpose large language models (LLMs). This is not surprising given the high stakes of real patient outcomes, the sensitive nature of healthcare data, and a host of regulatory standards to adhere to.

What was interesting was technical leaders showing broader interest in all types of language models, indicating a willingness to explore with various options. For those getting started on their GenAI journey, it makes sense to focus on healthcare specific models, while practitioners with more experience test out other methods.

3) Use cases vary by technical experience and company size

GenAI’s applications in healthcare are diverse, with the most common uses being answering patient questions (21%), medical chatbots (20%), and information extraction/data abstraction (19%). Technical leaders, on the other hand, prioritize information extraction and biomedical research, indicating a strategic focus on gleaning data-driven insights and advancements.

Respondents foresee GenAI having the most significant impact on transcribing doctor-patient conversations, medical chatbots, and answering patient questions over the next few years. Smaller companies, in particular, have high expectations for these technologies, likely due to their agility and drive to gain a competitive edge.

4) Human intervention remains necessary 

Accuracy, security, and privacy are paramount when evaluating LLMs, with Technical Leaders placing even greater emphasis on these criteria. The survey reveals that cost is the least important factor, suggesting a willingness to invest in high-quality, reliable models. Major roadblocks to adoption include concerns about accuracy, legal and reputational risks, and the technology’s alignment with industry-specific needs.

Human oversight remains crucial, with “human in the loop” being the most common strategy to test and improve LLM models. This approach ensures quality and performance while addressing biases and inaccuracies. Different company sizes prioritize various testing requirements, from fairness and private data leakage in large companies to bias and freshness in smaller ones. But whichever way you slice it, humans will remain a necessary component of responsible GenAI programs.

It’s clear technical leaders are spearheading GenAI in healthcare, as reflected by significant budget increases and a deep understanding of the technical advantages. However, challenges remain, particularly around accuracy, industry-specific requirements, and ethical considerations for all groups and company sizes who have already deployed or are considering deploying GenAI.

As healthcare organizations continue to explore and implement GenAI solutions, tailored strategies and collaborative efforts between technical experts and domain specialists will be essential. While considerable oversight is critical to success, when done right, GenAI has the power to improve patient care, streamline operations, and accelerate research. We’re excited to see what the next year will bring!