The AI Hype: Separating Fact from Fiction
“I, for one, welcome our new computer overlords.” Ken Jennings (former Jeopardy! champion on the February 16, 2011 episode, in which he conceded defeat to IBM’s Watson)
Let’s ask ourselves a question: “Does the hype surrounding AI within healthcare merit attention from investors?” The key word is hype. We do see a lot of reports in trade journals and in the mainstream media about the potential benefits of AI tools within the healthcare and life sciences industries. But a lot of that reporting is really about potential and not about tools that are currently available or business strategies that could support their deployment.
That isn’t to say that there aren’t AI-empowered organizations out there proving their worth. They absolutely exist, but the hype around AI in healthcare IT tends to focus on the benefits that futurists like to talk about. Those discussions have more to do with what is possible, not what the first step of AI integration with existing systems and processes looks like.
It’s exciting to speculate about the different kinds of uses that integrated health systems technology may be able to support, but those visions are not based on the current reality of the industry. This does not mean, however, that AI technology is not being deployed effectively in the healthcare and life sciences industries.
Reality Behind the Hype
There are three key hurdles in the industry that any potential AI technology needs to navigate in order to be meaningfully implemented.
First, a lot of the technology is simply not as advanced as some people think it is. Some AI technology deployed in the market has failed to meet expectations, and some development efforts are taking longer than expected. As with all technological advances, there’s great potential built into AI, but a lot of it is still very much in the early stages of development.
Second, and more significantly, healthcare technology must work within business models that make sense. This means that a solution needs to be financially viable for the organization that uses it, and it must function without creating regulatory risk for its developer or users. Right now, a lot of the technological solutions that we like to speculate about don’t really fit within traditional workflows or business plans. So they have a hard time being deployed in an effective and economically attractive manner.
The unique economic and regulatory environment of the healthcare and life sciences industries create a challenging business environment. A tool may have very attractive functionality, but when deployed in the real world, it might not be supported by a business model that makes sense, or it might immediately create compliance risk for its users. In addition, because of the complicated and diffuse reimbursement environment, it may take a long time for a tool to get traction and scale and create real economies to sustain business. Even worse, a system may not get off the ground because there is no business model that makes sense in the highly regulated healthcare market.
Third, anything that involves patient care needs to be proved in some way. Physicians are trained, certified, and licensed. Medical devices are cleared by the FDA. Healthcare provider organizations are accredited to participate in reimbursement programs. When it comes to some AI solutions, particularly with healthcare delivery, how we evaluate those tools remains open ended.
The FDA has done a lot in terms of thinking about how it may approach AI, but licensing boards, accreditation agencies, and malpractice carriers are beginning to consider the extent to which they should play a role in the “approval” process. This is a process that is just beginning, and until it is completed, questions will remain about what standards apply and what risk may exist.
These three major factors act as lenses that cut out the glare of science fiction and help developers, investors, and users focus on actionable opportunities in the healthcare industry.
Currently Usable AI
There are workflows within the healthcare and life sciences sectors in which AI solutions can currently be deployed effectively. Pharmaceutical research and development can benefit from AI solutions. The amount of lab work, testing, and dead ends encountered in R&D results in very long timelines to get products to market. AI solutions can make the processes significantly more efficient.
Healthcare delivery can leverage basic AI chatbot tools to manage patient intake. Various companies deploy chatbots in customer service. So-called dynamic questionnaires may be powered by an AI system designed to elicit key information and deliver it to an individual who can provide a higher level of service.
This same process can streamline and improve clinical-information intake. AI systems are being used to gather basic but very meaningful data about patients and their conditions before they ever interact with healthcare professionals. The healthcare professionals can then use that data in their discussions with the patients. That takes time off the doctors’ schedules and gives them information that they don’t have to get from the patients.
AI solutions can also make a difference in a diagnostic setting. Clinicians working on diagnosing patients can benefit from a tool with perfect memory and flawless associations between symptoms and possible diseases. These tools are demonstrating the ability to read imaging data at the same level as professional physicians and can be used as an imaging diagnostic aid to those professionals.
AI Will Continue to Advance
The hype around AI in healthcare is significant, and although advances in technology will improve AI’s usability in the life sciences and healthcare sectors, it is important to ground expectations in the reality of the industry. Investors in these sectors must be cautious when confronted by an “AI-powered” solution. AI is exciting technology, but it is only technology.
Investors can discern actionable ventures from futurist dreams by making sure that the technology is effectively deployed (to solve for a problem in a way that is otherwise unavailable), reliable (has been developed using good data sets lawfully obtained), viable (supported by a detailed business model tuned to working in healthcare), and verified (licensed, certified, and proved according to available frameworks).
Current opportunities where these tools are being deployed exist throughout a wide range of healthcare and life sciences operations from revenue cycle management to diagnostics. The field only widens from there as the industry experiences success and satisfaction from these advances.
Dale C. Van Demark is a Partner with McDermott Will & Emery, a sponsor for KLAS DHIS19.
Photo Cred: Shutterstock, TippaPatt