When venture capitalist Deena Shakir invests in health care startups that utilize artificial intelligence, she looks for opportunities to inject a digital solution that will make care more efficient, effective, and equitable.
As an example, Shakir invested in AI fertility technology startup Alife initially in 2020 and again two years later when she co-led a $22 million Series A. Secular tailwinds supporting these investments include data that shows mothers are having their first child at an older age than past generations and that more LGBTQ+ parents are turning to in vitro fertilization to have children. As many as 180 million couples worldwide are affected by infertility.
“Although the foundation of IVF is grounded and one of the greatest inventions of our time, there has been very little digital adoption in the actual full stack clinical treatment process,” says Shakir, a general partner at New York–based VC firm Lux Capital. Alife’s machine learning model assists clinicians, embryologists, and clinic managers with insights that optimize the IVF process.
Alongside women’s health, Lux Capital has been a prolific and early investor in software tools for clinical research. A lot of Shakir’s investments over the past five years have specifically homed in on the digital health space and the intersection of AI. One question Lux Capital is sure to ask founders: Are they aiming to solve a technology problem or a clinical adoption problem? Because if a startup doesn’t have the right business model in mind, it doesn’t matter how incredible the technology is. It just won’t get adopted.
Health spending in the U.S. reached $4.5 trillion in 2022, and billions upon billions are wasted each year. And experts say that health care is behind almost all other industries in its AI adoption journey. A big issue is data, as in, there’s too much of it and it isn’t connected across a highly fragmented industry. Health care is the world’s largest data source, at 30% of annual production, but 80% of that data is unstructured, according to Deloitte.
“The health care industry is excellent in generating reams of data,” says Simon Gisby, managing director and the global leader of life sciences and health care at Deloitte. “But as an industry, we’re not surprisingly that good at analyzing and drawing correlations out of that data.”
Gisby sees the health sector’s AI investments focused on three buckets: early stage clinical trials, clinical diagnosis, and back-office tasks. “Today we are a sick care industry,” says Gisby. “Can we use [generative] AI to help us maintain our health and intervene before we get sick?”
Noam Solomon is the CEO of Immunai, a biotech unicorn that is using AI to map the human immune system to help develop more effective treatments. The problem, as Solomon sees it, is that millions are dying of cancer each year and just as many are suffering from it. “And the question is,” he asks rhetorically, “do we believe that fundamentally there is a way to give more precise medicine to patients?”
He says yes and that the answer is to create the right data. With as much as 90% of clinical drug development failing, Immunai is now forging partnerships with pharmaceutical and biotech companies to help influence decision-making for drug development. AI can help determine the right treatment mix or optimal doses given to patients during clinical trials.
“The companies that are able to create a new type of data modality and then apply AI to create value, this is going to allow them to succeed,” says Solomon.
Natural language processing and machine learning have gained traction across health care, with use cases ranging from virtual physician assistants to clinical trial patient recruitment. Three out of every four leading health care companies are experimenting with generative AI or attempting to scale it across their business, Deloitte says.
But a lot of barriers remain. The cost of health care has marched steadily upward even as life expectancy in the U.S. has declined. A vast majority of clinicians have reported burnout and labor shortages are persisting for both doctors and nurses. Profits are under pressure from rising operational costs and shrinking reimbursement rates.
“What we should be doing as a society is invest in preventative health care, because it is cheaper for all of us, including the insurance companies,” says Sundeep Peechu, a general partner at California-based VC firm Felicis. He has led investments in AI-powered drug discovery startup Genesis Therapeutics and Prenuvo, which provides full-body MRI scans for early detection of hundreds of cancers and diseases.
Dr. Taha Kass-Hout, chief technology and science officer at GE HealthCare, takes a similar view in asserting that AI can make treatments more predictive. To treat cancer, medical advancements have given patients more treatment options today beyond chemotherapy and radiation therapy. But with new treatments comes more complex decisions for health practitioners.
GE HealthCare worked with Vanderbilt University Medical Center to scour medical record data and use AI-powered applications to help predict how a patient would respond to more precise cancer immunotherapies. This could help avoid potentially damaging and ineffective treatments, while also saving costs.
“We built a predictor with about 70% to 80% accuracy, to predict whether … there’s going to be some kind of adverse event to their therapy, and then we went and validated that,” says Kass-Hout. He adds that “the future is really bright in leveraging these technologies. Generative AI has the ability to scour large volumes of unstructured information and be able to structure it in a way that you can derive the right insights.”
Dr. Kingsley Ndoh found inspiration to improve cancer-patient care outcomes in new middle-income countries after his beloved aunt died of colorectal cancer in Nigeria. And after reading Jerry Kaplan’s Humans Need Not Apply: A Guide to Wealth and Work in the Age of Artificial Intelligence, he also started to think more seriously about how AI advancements could result in greater disparities in patient outcomes globally.
“If I manufacture an ultrasound machine in America, I can use it in Brazil or Nigeria and it works fine,” says Ndoh. “But if I build an AI algorithm or a clinical decision-making algorithm, it probably won’t work well in a place like Nigeria.”
That led Ndoh to create Hurone AI, a precision-care startup that allows doctors and patients to communicate through auto-generated text message prompts about symptoms. Not only does the technology help overworked oncologists, Hurone AI’s models predict the likelihood of treatment dropouts and severe side effects.
“When a clinician monitors their patient, they can just click a button and it generates a message for the patient. It’s contextual for that health care system, and it follows the regional treatment guidelines to basically tell the patient what to do,” says Ndoh.
Hurone AI is one of more than 200 organizations that received cloud credits and technical expertise from the Amazon Web Services (AWS) Health Equity Initiative, which has committed $60 million to help solve gaps in health equity.
“AI has become a tool that will continue to be leveraged to identify those disparities and to be more responsive to patients and health care providers to reduce those disparities,” says Danielle Morris, who leads global health equity at AWS.
Amy Brown, founder and CEO of Authenticx, says her past experience in social work and state government exposed her to huge sets of unstructured data from the conversations that health insurance companies record with their customers.
“I saw a huge opportunity to listen at scale, using AI, and applying AI to a kind of archaic data source that has been sitting on a shelf for decades,” says Brown.
Authenticx takes recorded customer conversations ranging from calls to emails to chatbot exchanges and uses AI and natural language processing to generate insights that could help boost the quality of service agents provide as well as compliance.
“What I encourage venture capital—and what I’d encourage purchasers of AI—to really think about is what is the business problem that we are trying to achieve with AI?” asks Brown. “And let’s make sure we understand the right tool for this job and what is our intended outcome.”
