Chapter 1
Obtain executive commitment to activate and scale AI and GenAI
Leaders should strive for a comprehensive vision and not abandon AI strategies entirely if a pilot fails.
A holistic AI-at-scale strategy is necessary, so if the organization needs to recalibrate during an initial AI project, it doesn’t lead to a loss of organizational buy-in around realizing the power of GenAI for health. “Adoption of bold new innovations is often challenging, especially with health care professionals whose overriding concern is, first and foremost, patient safety. Nonetheless, strong leadership can overcome these challenges through proper education and proof points of clear clinical benefit to patient care,” said Dr. Mault of BioIntelliSense.
More than 40% of health care CEO respondents to the EY CEO Outlook Pulse survey said they had already established an AI task force, with direct line to the C-suite, responsible for the firm’s vision and strategy.
First, leadership must bring clarity and strategic vision around the future AI- and GenAI-infused operating model. Planning for this future state, executives must consider how the organization will mature its capabilities around operating and maintaining a portfolio of algorithms. They must plan for how they will monitor underlying data supporting the algorithms, support testing, training and change management around each algorithm and watch for changes in data elements, care models or procedures that may affect the relative effectiveness of each algorithm. One can imagine a future wherein algorithms will be dependent on other algorithms and hence the change management capabilities, roles and skills across the enterprise must evolve to appropriately manage such complexities and risks. As such, executives must consider these future health care operating model impacts and what steps need to be taken now to plan, invest, hire, train and protect the health organization.
Health care CEOs are taking steps to shape their AI strategies
What you can do now: start with a quick win — an internal-facing use case that is low-risk, low-cost and will allow executive stakeholders to understand the algorithms, governance process and change management needed before venturing into higher complexity use cases.
Focus for the future: executives should be key players in championing new projects, and in providing strategic communications and project oversight.
Chapter 2
Build confidence in your AI strategy with appropriate governance
As demographics shift and new treatments and technologies emerge, data governance and AI performance management is critical.
As AI continues to evolve globally, regulators are scrambling to navigate this new environment, especially when it comes to integrating the technology into medical devices and clinical workflows. To guard against the risks of biased algorithms and shifting datasets impacting patient care, health organizations must be vigilant about performance monitoring and change management.5
Governance must be the anchor to ensure secure, sustainable, responsible and transparent AI.6 “Explainable AI is important so that physicians can understand that the information presented is based on clinically accepted treatment protocols, and can make better informed decisions,” said Femi Ladega, Group Chief Digital Officer of Dedalus, a global digital health care and diagnostic solutions company. “An organization’s AI governance should be flexible to meet maturing needs of AI models in the health care enterprise.”
In the excitement over AI and GenAI, as health organizations turn to external parties to help, danger exists in leaning on dozens of point solutions, which could become unmanageable and expensive. Health organizations need policies in place to properly vet and assemble the ecosystem of partners and solutions that they may choose in this rapidly changing environment.
What you can do now: harmonize your AI governance with existing organizational and data governance. AI must be plugged into existing processes so that the organization can better embrace and understand it.
Focus for the future: establish continuous feedback loops that monitor for regulation changes, risks, and biases across the AI portfolio. The feedback loop should be part of the governance structure to be constantly overseeing and improving all algorithms.
Chapter 3
Build the right data infrastructure to power your AI strategy
Craft a data infrastructure that is rooted in standards and can flex to the needs of the health care system in the future.
One of the key obstacles for health organizations with AI is data infrastructure. In integrating five of its care systems, London’s health organizations adopted a single health information exchange infrastructure to securely share records at the point of care across all its organizations. This is called the “London Care Record,” and helps ensure frontline staff have the information they need about a person when they need it, wherever they are working in the city. Nationally in England, a process is underway to implement a federated data platform in support of consistent approaches to local use of data for multiple purposes. To enable AI now and in the future, health systems must craft a data infrastructure that can bend and flex to future needs.
“The London health data strategy was saying actually, we make all this data available — not just six months data, but near to real-time in a linked way across all our patient contacts and patient care settings,” said Luke Readman, Director of Digital Transformation for NHS England, London.
What you can do now: review your data strategy and data governance including existing metadata, data lineage, data ownership and infrastructure. Leveraging data standards is key. Ingest data and map the various standards to each other. Moving from on-premises to the cloud allows for more scalability and flexibility. Create a semantic layer of data to make the information consumable and exposed via APIs. Determine the key infrastructure components that are necessary for the desired business outcomes.
Focus for the future: focus on building a scalable, flexible infrastructure that can withstand a portfolio of AI algorithms. Be strategic with procurement decisions as it can become costly quickly when running a suite of GenAI and AI algorithms at once at scale enterprise-wide.
Chapter 4
Equip and upskill your workforce with AI training
Health organizations should build support for the entire workforce – from the AI-averse to GenAI leaders.
Health organizations will have employees either starting to use AI and GenAI on their own, or who are making decisions based off the outputs, so it’s important to think through the training needed to help workers recognize bias and understand how to monitor performance. Working with clinicians when implementing AI or GenAI tools is integral to successful integration. Transparency around the data and models that feed the outputs is necessary from the beginning so that clinicians know exactly how these new tools might impact their future state.
What you can do now: when it comes to AI education in general, many prominent universities are incorporating AI courses to prepare the future workforce. Individuals delivering care should receive AI literacy training to prepare them for the incorporation of AI and GenAI in their daily lives.
Focus for the future: establish specific roles for more AI-literate or AI-interested health care providers to work with those developing and maintaining AI algorithms to continually create experiences that satisfy patients and to optimize care delivery.
Chapter 5
Prioritize use cases that match your AI maturity
As organizations build the needed AI governance and capabilities, they can unlock the potential of AI to transform health care.
Health organizations will want to make sure their investment will pay off in terms of value equation – whether it be financial value, or to the clinician and patient experience. To produce value, the right type of AI must be applied based on the situation, what is most clinically applicable, cost effective and, importantly, sustainable. Each use case might not revolve around just GenAI or one type of AI, but rather a mixture of robotic process automation (RPA), machine learning and GenAI together in order to be cost-effective and sustainable. The key is architecting these tools together in a way that is manageable and has the appropriate oversight to produce ethical, valuable algorithms at scale.
Cost is certainly a driver when it comes to maintaining algorithms at scale. Data sets and trends in algorithms may change over time, meaning they will continually need to be optimized. As algorithms become more complex and grow in scale, the operating model of how to manage them changes drastically. Further, from an enterprise level, there will be a portfolio of AI algorithms consistently needing to be overseen, maintained, traced and regulated, which requires specialist skills, data infrastructure and significant oversight.
But value is not just financial, says Dr. Stewart of Johns Hopkins Medicine. Value can come in the form of increased clinician satisfaction through reducing EHR clicks, or taking giant steps forward to improve the world’s health. “If I could address health care disparity in the application of best practices across all people for hypertension, diabetes, congestive heart failure and breast cancer, and eliminate the health care disparities for those diseases across all people, then I almost really wouldn't care how much it cost,” he said.
Here's what you can do now: determine your organization’s strategic objectives and goals when using AI. Is the goal cost reduction, revenue growth, operational efficiency, customer satisfaction or innovation? Start with a low-risk, easy-to-implement use case that fits your objectives.
Focus for the future: once your organization has tested its governance on a few use cases and as the infrastructure and skill sets mature, it can take on more complex, transformative use case. Continually measure the use cases that are implemented in order to determine if they are meeting the desired outcomes or if they need to be optimized further.
Special thanks to the following individuals who contributed greatly to the development of this point of view:
Sezin Palmer, EY Global Health Sector AI Leader
Kayla Horan, EY Global Smart Health Analytics Solution Leader
Crystal Yednak, EY Global Health Senior Analyst
Rachel Dunscombe, CEO, OpenEHR
Summary
In order to build a foundation for AI in health care that both accelerates transformation and adds future value, executives must strategize current work while architecting for the future.