In response to questions around which use cases leaders thought were most important or exciting, seven out of ten brought up generative AI and Natural Language Processing (NLP) models, including chatbots and scribing. This consistent response indicated the group’s optimism toward generative AI improving the workforce and clinical efficiency. Those who were already exploring and implementing generative AI solutions mentioned improvements to client engagement, higher efficiency in simple IT tasks, and reduced employee response times.

So what? Leaders see a real potential for generative AI to improve employees' quality of life, automate rote tasks, and increase productivity. Improvements to these metrics have the potential to increase retention rates and decrease stress in healthcare systems.

Based on our interviews, health systems are proactively seeking where they can take advantage of certain AI technologies, rather than waiting on their existing vendors to adopt AI. However, many healthcare leaders reiterated the difficulty in navigating the high-growth market and knowing how to evaluate the rising number of vendors available for pinpoint solutions. The interviewees see an increased need to develop better tools to evaluate vendors. Ideas ranged from revamping vendor evaluation and intake forms to simply increasing networking between CIOs to make sure that the solutions are not over-engineered for their needs and can be easily integrated into their existing systems.

So what? To prevent risky investments and loss of resources, it will become increasingly important for healthcare leaders to take a problem-driven approach to evaluating AI opportunities. Healthcare leaders collaborate with vendors based on trust; small vendors should invest in collaboration with large-scale vendors (Microsoft, Epic, etc.) to promote word-of-mouth marketing and build a trusted reputation. 

Most healthcare leaders agreed that with the current pace of change, AI solutions will eventually become a commodity, and the real value will be the data. High-functioning AI solutions require well-structured, unbiased, representative, and protected data. If the input is faulty, then the output will amplify these biases and result in unusable results. All the interviewees mentioned the need for standardized data testing and evaluation to prevent AI outputs from perpetuating discrimination in healthcare settings. It is evident from our interviews that having a strong IT infrastructure and department enables a strong culture of using data, and a better understanding of AI principles and strategies that can eliminate data challenges.

So what? Healthcare leaders should bring data evaluation frameworks to the forefront before investing in AI solutions. Also, enhanced AI literacy can increase the transparency of AI model training, the generalizability of the output, and a better understanding of the AI lifecycle.

Health leaders recognize the opportunities of AI but have valid concerns around equity, quality, accuracy, privacy, and transparency. Specifically, many of the healthcare leaders we interviewed identified a lack of standardized AI regulation and governance as a primary obstacle in the implementation of new use cases. However, a few leaders identified strategies they have used to mitigate these barriers, including using independent data analytics/IT departments, having a communicative and supportive leadership, building good relationships with existing vendor partners, and enhancing AI literacy.

So what? New AI solutions are here to stay and are becoming part of how we operate as a healthcare industry. Dismissiveness around the challenges and decision paralysis will only cause healthcare organizations to fall behind in this competitive market.