By Yossi Shahak, CEO and Founder of Maverick Medical AI
From finance to manufacturing and beyond, AI is already being leveraged to automate key workflows and processes, enhance customer service via sophisticated chatbots, help organizations convert large quantities of unstructured data into helpful business insights, and more. With a growing number of organizations adopting AI into their business operations, however, some may be shocked to learn a majority of them haven’t achieved the intended ROI.
Analyzing healthcare’s relationship with AI, including its struggles, can provide important insights and lessons on how AI can be used effectively to improve care as well as provide other organizations and industries with a blueprint for improving their ROI.
The healthcare industry is notorious for its slow and cautious adoption of technology, despite a growing array of cutting-edge tools capable of augmenting and assisting both clinical and operational healthcare tasks. This is largely due to the industry having to contend with regulatory factors, patient safety concerns, and workflow disruptions caused by replacing and updating legacy systems. We also can’t discount obligations to ensure these technologies have undergone strict checks to confirm their effectiveness.
AI isn’t immune from these factors. In fact, implementing any of the many AI tools and solutions in healthcare often presents even more challenges for an organization to account for. Despite divided industry attitudes toward AI use in many settings, its application within the sector is becoming more prevalent. Nowhere is this more apparent than with the large language models (LLMs) which aren’t typically designed for use in healthcare due to lack of data from hospitals.
A major reason why we are seeing more AI adoption in healthcare is due to care providers across the U.S. and globally coming under intense occupational pressure. Recent massive ransomware attacks threaten credibility and cut into budgets while shortages of physicians, nurses, and other administrative personnel impact the quality of care and exacerbate operational inefficiencies—and these problems will only worsen with time.
Can healthcare AI not cause a commotion?
That answer depends on a few factors.
What makes healthcare’s adoption of AI more unique than most industries is the sheer complexity of its systems. Healthcare organizations have to navigate a broad range of disciplines in addition to focusing on all areas of care. This includes staying on top of advancements in medicine, technology, and regulation while balancing financial considerations, demographic and epidemiology changes, secure IT systems, and collaborations with other organizations.
Therefore, when an organization, be it a small outpatient practice or an integrated delivery network consisting of numerous hospitals and clinics decides to explore AI, there must be participation from the widest range of stakeholders. This includes department heads, physicians, administrators, IT specialists, and even patients themselves.
These stakeholders must clearly define where they want AI to help—optimizing workflows, streamlining patient intake, analyzing medical images, augmenting medical coding—and set clear objectives of what they expect to ensure the models are accurate and provide value to the customer. This will help organizations allocate assets and lead to better time, effort, and resource investment.
Auditing
When assessing an entire IT ecosystem’s readiness to integrate AI, leaders need to gain a comprehensive understanding of their data quality, structure, governance policies, and strategies including their entire cloud computing infrastructure.
Likewise, a comprehensive review of all internal operational and workflow processes is just as essential to understanding how an AI tool will fit into existing systems. This process is crucial in identifying gaps where AI can support or supplement current mechanisms—or maybe where it isn’t worth using.
During this stage, it’s important to be aware of AI’s challenges. This requires all stakeholders to acknowledge common shortcomings like generative AI’s black box and hallucination problems alongside possible ethical and regulatory concerns. Healthcare leaders must understand there will almost certainly be initial growing pains with deploying AI in any of its forms and this can manifest itself in workflow disruptions and slumps in operational efficiency. It’s important to put mechanisms in place to work around this learning curve period.
Selecting the right AI vendor and post-deployment support
Once the issue AI is poised to solve or improve has been defined and the right people are overseeing its implementation, the next step involves researching and pinpointing the right technology vendor to work with. It helps to have a dedicated AI specialist to take charge during this segment of the AI journey.
When evaluating AI vendors, they must demonstrate a level of experitse with the healthcare domain and regulatory requirements. This process isn’t meant to be quick and simple. It demands healthcare organizations review references and case studies, evaluate the robustness and scalability of its algorithms, the quality of data processing, and the model’s ability to learn and adapt to new trends and updated regulatory activity. All these factors directly impact the long-term sustainability of an AI investment.
The best vendors to partner with are the ones who guide organizations through the process of workflow improvements and documentation changes required. As such, most organizations will require a vendor capable of providing support during pre-deployment and validation testing periods as well as post-deployment. Selecting a vendor that provides transparent, or glass box, use of AI is equally vital, including explaining the technical components of its algorithms and its governance process to ensure stakeholder confidence in the technology.
Whether a healthcare provider turns to AI to optimize their revenue cycle management, help with diagnosis prediction and disease progression, interpret medical images, or streamline clinical trials, stakeholders must find the right vendor who provides the highest degree of education, training, implementation, and technical support. This is vital to ensure staff can effectively use the tools, and if it’s a generative AI solution, interpret its outputs and integrate it into their workflows.
Other vital post-deployment support functions include regular updates, audits, and performance monitoring to ensure the system is running smoothly, securely, and optimized for performance—all crucial to achieving ROI. This support involves software patches, enhancements based on end-user feedback, usage indicators, and analytics and insights to further improve efficiency.
Additionally, healthcare providers must work with AI vendors who assist with regulatory compliance, provide a customer success manager, and allow for the AI system to adapt and align with the organization’s specific workflow needs.
While AI can’t solve every problem in healthcare, failed projects in the field stem from individuals and organizations not recognizing the importance of the implementation process, including what they want the AI to achieve. Overcoming the challenge of implementing complex and powerful AI systems in healthcare demands the highest degree of teamwork to ensure all voices and concerns are heard and the most thorough and thoughtful plan is in place.
Healthcare organizations must think of an AI investment as a down payment on the future and understand AI in healthcare isn’t a one-size-fits-all saving grace. Without addressing the complexities around implementation, there won’t be any ROI.
About the author:
Yossi Shahak brings over 25 years of combined experience in the world of digital health and artificial intelligence sectors—with a specialty in AI medical imaging. Prior to founding and leading Maverick Medical AI, Shahak served as Vice President of Sales and Business Development at Zebra Medical Vision. Additionally, Shahak has a wealth of experience with startups and Fortune 15 companies, including serving in executive leadership roles for McKesson. As Founder and CEO of Maverick, he is fully dedicated to enhancing the productivity and efficiency of the revenue cycle management process for healthcare providers and payers.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.