You’ve most likely had an experience like this:
One day, you wake up with a hint of a headache and a sniffle. You sneeze. What do you do next? You Google it. You eventually land on WebMD. Was that sneeze a common cold or a bout of seasonal allergies? Or are you in fact dying from a rare form of nasal cancer? You determine you should see an allergist. But which one? How do you find a good one? Does your health plan require a referral? How much is this going to cost you? You text a friend who is a nurse, and another who works at the local hospital, hoping they’ll guide you in the right direction.
If so, you’re not alone. 1 billion health-related queries are submitted to Google every day, and 7% of overall questions asked to Google are health-related. WebMD is frequented by over 180 million people every month. It turns out that a broad set of the population turns to the internet and friends to figure out what their health condition might be, and which doctor to go see for it. In industry speak: the internet and your friend are today’s “front door to healthcare.”
While the Google –> WebMD –> Friend protocol may be a common way to enter into the healthcare system, it is certainly not an optimal one. Most of us are not well-equipped to self-diagnose from a list of possible conditions, and our friends who work in healthcare often don’t have the exact expertise we need. But when we turn to this protocol, what we’re effectively looking for is:
We’ve long imagined a version of the healthcare future in which every human has an all-knowing, empathic healthcare companion who fulfills these criteria based on personalized data—like a real life Baymax, the healthcare robot from the movie Big Hero 6.
But this isn’t entirely in the realm of fiction. The aforementioned tasks are what large language models (LLMs) are well poised to be tuned to do—e.g. exhibit college graduate-level smarts, follow some rules, send programmatic instructions to third-party software systems to take an action, and help provide empathetic, patient guidance to someone who is likely feeling stressed and confused due to uncertainty about a health issue.
Furthermore, they can escalate to a trained professional if they encounter a high-risk use case, or based on preference of the user. Hospitals and provider groups already hire hoards of their own call center agents to triage inbound consumers and make sure they are getting to the right site and type of care—because getting this wrong could mean wasting scarce appointment slots by sending the patient to the wrong type of doctor.
As such, we are optimistic about the opportunity for a specialist LLM to serve as a far more effective future front door to healthcare, integrating into a healthcare marketplace to enable people to book appointments and pay for their care.
From a patient perspective, the experience would also be much smoother:
One day, you wake up with a hint of a headache and a sniffle. You sneeze. What do you do next? You turn to your Baymax-like app, input your symptoms, and after a few follow up questions, it predicts— given your current location, the weather, your recent sleep scores, your diet, and your personal trends—that you’ve got allergies. It offers you a same-day appointment with a nearby allergist covered by your insurance to confirm the diagnosis. In the meantime, it recommends you try an over-the-counter allergy medication, offering to have it delivered to your house. It orders extra tissues for you, for good measure.
Not only is this business a great opportunity to improve human health, patient experience, and provider experience, but the business case is also very clear.
First, the traditional businesses mentioned above are quite lucrative today. Health information sites like Google and WebMD make the majority of their revenue through ads; in 2016, the year before WebMD was taken private, it did $561 million in advertising revenue. The healthcare and pharmaceutical industry spends an estimated $18 billion a year on digital ads alone, usually advertising on condition-specific searches that indicate a consumer’s intent to engage with a specific healthcare product or service.
B2B revenue is also quite large for these businesses; WebMD offered “private portal” health information services products to health plans and employers that generated $114 million of revenue even back in 2016. Accolade, a comparable employer- and payor- sponsored navigation business, generated $300 million revenue in 2022.
Add to this the opportunity to monetize transactions once the LLM integrates into a marketplace of services that can be directly booked, and numerous revenue streams worth billions of dollars in aggregate become possible.
While a big, white, puffy healthcare robot like Baymax is a possible manifestation of the future front door to healthcare, perhaps a more realistic one is an LLM tied to a healthcare services marketplace. We envision a not-so-distant future in which the daunting task of navigating the healthcare maze is significantly simplified for both patients and providers, with a viable business model to boot. When that happens, you’ll no longer need to self-diagnose, and your healthcare friends will thank you.
Daisy Wolf is an investing partner on the Bio + Health team, focused on consumer health, the intersection of healthcare and fintech, and healthcare software.
Julie Yoo is a general partner on the Bio + Health team at Andreessen Horowitz, focused on transforming how we access, pay for, and experience healthcare.