Cut the Hype about AI - Cover

Cut the Hype about AI

AI is an exciting topic of healthcare. Two years ago, most healthcare organizations were wondering about the impact and advantages of AI. They wondered how real AI was. Today, there is almost no doubt about the benefit of incorporating AI into the healthcare ecosystem, so major health systems are starting to move forward with their AI pilots and programs.

KLAS interviewed 57 organizations that are closely partnering with their AI vendors and delivering outcomes. In this blog post and this video presentation, I would like to use those organizations’ responses to share not only the impact AI has made on hospital operations, care deliveries, and customer experiences but also the lessons and best practices that customers and vendors have learned throughout this journey.

The Expansion of AI

If you ask 10 people to define healthcare AI, you will probably get 10 different answers. Based on the current status of AI applications in healthcare, KLAS’ research has focused mainly on software that provides machine learning or NLP capabilities. It is also critical to know the key offerings from AI products.

We have observed an increasing amount of investment from vendors into areas of explainable AI, data preparation, and change management. As AI continues to progress in the market, we anticipate that success in those areas will eventually be accomplished by vendors and providers.

So how well is AI being adopted in healthcare? In partnership with CHIME, KLAS surveyed 483 progressive provider organizations and asked them about their adoption of AI solutions. Between 5% and 7% of these organizations reported having fully deployed AI solutions and achieving desired outcomes. That is a relatively small percentage compared to those who are just starting to use or implement AI or who are still evaluating AI.

An Immature Space

Our research has found that AI is truly helping healthcare organizations drive outcomes today; some outcomes are quite diverse in clinical, financial, and operational areas. To be exact, we have found 37 unique use cases across the aforementioned 57 organizations we interviewed.

Although we were pleasantly surprised by this diversity, most of these organizations are just dipping in their toes and trying to discover the real impact of AI. So it is still too early to say whether these use cases can be scaled across a broader customer base. The most validated use cases are actually in the clinical areas, but we have seen far fewer financial and operational use cases. As the data becomes more available and normalized, we anticipate that AI applications will make a much larger impact on hospitals’ operational efficiency.

Here is another reality check: all the customers we have interviewed in our research not only are closely partnering with their AI vendors but also have seen use cases and outcomes in the space. Health Catalyst's and IBM’s AI applications are being most widely adopted across 11 outcome areas, but their customer bases are still very limited today. Jvion offers models with prebuilt healthcare content, and they have the deepest adoption in population health and clinical decision support compared to other vendors.

What It Will Take to Succeed

Outcomes are definitely achievable through leveraging AI technology. But how much effort does it take to achieve the Quadruple Aim? And what does the journey look like? Based on our conversations with organizations currently using AI, the journey is definitely not a walk in the park.

The common misconceptions and the lessons being learned represent the work that both vendors and customers need to put forth to drive outcomes. Often, organizations with highly successful AI projects say that building the AI model is actually the easiest part. For AI to be successful in healthcare, we need to overcome many challenges, including incorporating accurate data into the models, understanding the amount of time and effort required, monitoring and managing the model’s performance over time, and building an AI-ready culture and workforce.

To really drive AI adoption and deployment, it is critical that both vendors and customers collaborate very closely. The success or failure of the project really comes down to change management and operations, not necessarily the technicality of the AI. That entails embedding AI in the workflow for clinicians and staff members, establishing an interdisciplinary team from the project’s start to its finish, and turning the insights delivered by AI into real actions.

Room for Growth

A lot of the AI products are still immature, and there is a general lack of vendor expertise with operations, deployments, and clinical integrations. Many implementations are taking longer than expected, and all these factors have contributed to a lower satisfaction of operations and product areas.

But considering the customer experience in the AI market today, customer satisfaction is high, and customers have an overall positive experience working with their AI vendors. The vendors’ cultures, values, and relationships that they have with the providers have really driven high customer loyalty. Many AI vendors are very customer-centric, and they partner with the providers to help them succeed.

If you are interested in learning more insights, I would love your questions and feedback through Twitter or LinkedIn. You can also email me at

     Photo credit: Adobe Stock, ReisMedia