Why GenAI will unleash the long-awaited tech transformation of healthcare

The technology is tailor made to address root causes of runaway costs and clinician burnout, creating a bright future for startups with the grit to work through the complex realities of this highly-regulated sector.

The healthcare sector has a well-deserved reputation as a digital technology laggard – except when it comes to AI. Healthcare providers routinely use machine learning tools to interpret MRIs and other clinical images, make sense of physicians’ scribbled prescriptions and assist in robotic surgery. The US Food and Drug Administration (FDA) has authorized more than 500 AI-based medical devices and applications. Tempus, an NEA portfolio company, sells machine learning-based software that can spot rare genetic mutations in malignant tumors, giving doctors important information to choose the best treatment. 

Healthcare’s record as an early and effective adopter of older types of AI makes us confident that the industry will be among the biggest beneficiaries of generative AI. We see a future in which chatbots can gather and analyze patient information more seamlessly than overworked medical staffers, freeing up time so doctors can focus on providing patients with quality care rather than pushing papers. Powerful diagnostic and therapeutic tools will help doctors improve outcomes, often by providing round-the-clock monitoring and recommendations rather than infrequent office visits. The result will be faster, more convenient, streamlined interactions that provide better healthcare - and maybe even help put a dent in U.S. healthcare costs that make up 18% of gross domestic product (twice the average for 36 other market-based economies).

This vision may take a decade or more to materialize. Change happens slowly in healthcare, due to the high level of regulation, complicated value chains and the understandable conservatism of providers in applying new technology to life-or-death decisions. But having invested in healthtech companies throughout our 45-year history, we know that these transitions don’t just happen. They require bold, patient and pragmatic entrepreneurs to drive systemic changes. 

Generative AI in Healthcare: Unlocking a $265 Billion Productivity Dividend

Why is generative AI such a good match for healthcare? Because it gives healthcare researchers and clinicians the ability to quickly gain insights from massive amounts of both structured and unstructured data, with no programming expertise required. Pharmaceutical companies large and small are experimenting with GenAI to radically shorten the time it takes to identify and produce new medicines and to run clinical trials.

But the biggest long-term impact of GenAI may be in the delivery of care – assuming AI companies move ahead in an incremental, deliberate way. The lowest hanging fruit is to use this technology to streamline administrative, non-medical processes, such as issuing prescriptions, coding treatments for insurance reimbursement, and following up with patients after operations to make sure there are no complications. Besides improving the patient experience, such projects could unlock 25% efficiency gains worth $265 billion, according to McKinsey & Co.

When it comes to using AI to improve care outcomes, technology providers should focus on applications that help doctors and other clinicians do their jobs, rather than replace them outright. For example, a GenAI system could handle questions normally asked at intake, either online or at the doctor’s office, and provide the information in a familiar, useful format to the doctor. This frees up doctors so they can spend more time with patients, which is often associated with improved outcomes for many types of health problems.

This deliberate, targeted approach has two main advantages. For starters, it allows AI companies to tackle the most beneficial, least complex applications first. With a smaller surface area for things to go wrong, the likelihood of adoption being slowed by negative sentiment among stakeholders is greatly reduced.

Why data is key to unlocking GenAI’s potential

For example, Radiology Partners (an NEA portfolio company) has built a vast database containing all manner of medical imaging. The company’s tools use machine learning to examine roughly 50 million scans each year from more than 3,000 hospitals and imaging centers, far faster than radiologists could ever do. The software quickly surfaces those scans that suggest cancer so the radiologist can prioritize important cases. Evidence suggests these AI-based analyses are just as accurate – and in some cases more accurate – than the opinions provided by general practitioners who may not be up to speed on how to interpret MRIs, ultrasounds and other types of images. 

But Radiology Partners has been careful to adapt its technology to the complicated realities inside hospitals and clinics. It named a Chief Medical Officer in 2020, and has focused on providing clinicians with recommendations and options – for example, to flag potentially concerning scans for radiologists – rather than entirely new processes that would disrupt their operations. As a result, the 3,600 radiologists that use the company’s tools typically see up to a 20 percent increase in their efficiency and accuracy, according to the company.

We believe this deliberate approach employed by Radiology Partners will resonate with regulators, leading to protocols that will enable even better use of GenAI. For example, privacy and other regulations prohibit the consolidation of various kinds of data, preventing a GenAI system from getting a holistic view of a patient's needs. Yet Tempus recently won FDA approval for a test that uses a range of data types to spot rare tumors. The goal is both to help doctors administer personalized treatment and to provide a robust dataset that researchers need to drive future innovations. 

For this to happen more broadly, regulators will need to go beyond current legal frameworks to ensure interoperability between different types of software systems, allow research on de-identified health information, and permit more data-sharing between healthcare providers. The result will be a much richer corpus of data to refine and improve the performance of AI models, ultimately enabling improved patient outcomes.

It’s an Rx for a future in which AI systems will advise doctors on certain tasks, particularly time-consuming ones that contribute to burnout and offer little room to add value. A GenAI-based diabetes monitoring service could track a patient’s glucose levels, weight and other signals and generate easy-to-understand instructions on how to adjust medication dosages. 

Embracing Generative AI: Balancing Ambition with Pragmatism in Healthcare

For now, most healthcare providers are understandably skittish about using generative AI in ways that come even close to regulatory lines. Just one instance of an AI system that shows bias to a particular class of patients or, worse, makes poor care decisions that lead to complications or death, could result in this technology’s upside potential being dwarfed by reputational, legal and financial ramifications. And regulators will be even more cautious. 

That’s why we’re on the hunt for healthcare startups with the particular set of skills and perspectives to work through these knotty problems. It’s not just about coming up with breakthrough technology. It’s also about being willing to invest substantial time and money to acquire and hone datasets to improve the efficacy of products, and to establish a track record to win over overworked regulatory agencies. It means having a deep appreciation for the complexities healthcare providers face, and a determination to make sure products fit easily into those organizations’ existing workflows. It means understanding that there are no overnight successes in healthcare, but that success is worth waiting for, both financially and in terms of the real impact these companies can have on people’s lives.

In our experience, such understanding only comes when teams include experts from both the technology and healthcare realms. That’s not always the case in venture-backed companies, but it’s how we’ve always operated at NEA. For decades, the leaders of our technology and healthcare practices have worked closely together – not only to make investment decisions, but on various occasions to incubate companies from scratch, including Radiology Partners, Strive Health, Curana and Trivalence. 

So if you have an idea for using GenAI to improve healthcare, please contact us. We share your enthusiasm for the technology’s potential, and believe we have a good sense of how to convert that enthusiasm into great companies.