COVID-19 Is Accelerating AI in Health Care. Are Federal Agencies Ready? 

Artificial intelligence is rapidly expanding its foothold in health care, including at many federal health agencies such as Veterans Affairs and Health and Human Services departments and the Defense Health Agency.  

The ongoing coronavirus pandemic is demonstrating the power of AI-enabled capabilities for private and public sector health care organizations responsible for responding to today’s health care challenges. 

For example, the pandemic has catalyzed numerous AI-enabled development efforts for vaccines. After scientists decoded the genetic sequence of SARS-CoV-2—the virus causing COVID-19—and publicly posted the results on January 10, the race was on. Based on that data, firms began using AI-enabled methods to rapidly develop potential vaccines, some of which are already proceeding to clinical trials. By comparison, traditional non-AI drug development processes take many months, if not years, to proceed to human clinical trials. 

Likewise, federal health agencies are also incorporating AI-enabled responses. The Centers for Disease Control and Prevention, for example, is hosting an AI-driven bot on its website to help screen people for coronavirus infections as a way to reduce the numbers of patients flocking to increasingly overwhelmed urgent care facilities. 

Additionally, the Food and Drug Administration recently approved use of an AI-driven diagnostic for COVID-19 developed by behold.ai. The tool analyzes lung x-rays and provides radiologists with a tentative diagnosis as soon as the image is captured, reducing time and expense. 

But there is an important caveat to this activity: we don’t yet know whether these and other non-AI related efforts will produce the long-term impact we are all hoping for.

In a milestone report published in December titled “Artificial Intelligence in Health Care: The Hope, the Hype, the Promise, the Peril,” the National Academy of Medicine noted the many ways in which AI is revolutionizing health care. However, they also warned that careful planning and implementation is required to avoid the risk of a backlash—or an “AI winter,” as some refer to it—that can occur when hyped AI solutions fail to deliver expected performance or benefits.  

Federal and defense health care agencies will be expected to mobilize more quickly and ensure that AI solutions produce results. So how can federal health care agencies improve their odds of success? How can they implement and scale AI projects and, more importantly, try to realize AI’s vast potential to improve healthcare while lowering costs? 

Based on successful public and private sector AI implementation, federal and defense agencies can achieve greater success in their AI deployments if they:

Have a strategic plan for AI. Select the purpose and focus of initial efforts with care and clearly define business challenges warranting AI adoption. That means, in part, identifying use cases that provide a significant return on investment. Moreover, the plan should also provide a means to address the agency’s readiness to leverage AI as there are several dimensions to readiness. For example, technology readiness refers to having the needed tools, technical infrastructure and data management strategies and capabilities in place. Workforce readiness refers to having needed talent recruitment and development, training, incentives, communications and change management structures and programs in place to successfully launch and sustain AI.

Understand your requirements and then phase solutions, from simpler to more complex. Amid the variety of AI “solution types”—such as task automation, pattern recognition or contextual reasoning—organizations will need to investigate the requirements of different user groups or use cases, technical and analytic complexities, and the ability to scale and sustain solutions across the enterprise. For example, robotic process automation is a relatively easy AI solution to implement to complete time-consuming, repetitive tasks such as data entry, data capture and data transferal from one source to another source. RPA can then serve as an easy gateway for the organization to tackle more advanced automation leveraging AI. 

Use an agile approach and develop iteratively. Such an approach can strengthen efforts to engage users and build trust, and there has to be a level of risk tolerance for this approach to work. Agile methodology can be helpful for facilitating collaboration and adoption. Central to this approach is adaptive planning, evolutionary development, early delivery, continuous improvement, and rapid and flexible response to change, which inherently allows for a fail fast element to quickly identify success or failure.  

AI is a human endeavor. People must bring needed leadership, accountability, motivation and expertise to the project both before and after it becomes operational and, later, as it scales. Having humans in the loop ensures better integration into work processes, builds trust, and creates accountabilities for the performance of AI solutions. 

There are many factors that contribute to a project’s success, but these considerations can be key as agencies strive to harness AI more fully in support of their missions. 

Philip Dietz, MBA, is a principal at Booz Allen Hamilton leading data science and analytics.